Home

ROGERS-THESIS - OAKTrust Home

1. Webcam Hall sensor IFB OO sy 0 DIO USB Motor IFB S Robot IRI Left ser he Hall server P J counter DIO CH apto 4 bit input Wireless gt 9 MC right E Voltage Lj DPDT control 4 bit 5 V O O switch output Power supply Mouse i O oo MC left ime O ae o I Wireless ina au Two chanels o el Motor Desktop computer Jsic Joystick A Right gear motor client system Hall sensor Q 0 IFB 5 V Figure 3 8 Hardware system overview diagram The motor control system of the PML robot is composed of two control systems that are completely independent and separate This system is distributed through the hardware modules of the IFB motor control block the laptop the two MCs and the two gear motors An additional capacity for manual control of the robot exists by the use of a serially connected joystick through the joystick interface circuit JSIC An overview of this system can be traced in Figure 3 8 A complete description of the motor control system is found in Chapter IV 32 The optical capability of the robot is performed by a Lego s webcam which has been mounted to a two degree of freedom DOF bracket at the front and center of the robot This bracket allows for both vertical and horizontal positioning It does not allow for rotation along the camera axis The webcam data is sent directly to t
2. This is the equation round 0 05x 5 5 and is more aggressive TurnDelta Round 0 05 Abs PRatio 5 5 Control equation for P Control This is the equation round 0 022x 4 and is less aggressive TurnDelta Round 0 022 Abs PRatio 4 Control equation for P Control This is the equation round 0 011x 3 and is the least aggressive TurnDelta Round 0 011 Abs PRatio 3 Control equation for P Control If TurnDelta lt gt OldTurnDelta Then OldTurnDelta TurnDelta updates for next HallsCompute sub Call MotCont recomputes Turn Variables and writes new speed to MC won t call MotCont in the While loop End If TurnDelta lt gt OldTurnDelta Print 1 TurnDelta TurnDelta RE ENABLE BLOCK part of P Control block This block should be just inside and at the end of the P Control block 154 If Abs DegError lt 5 Then Print 1 Inside I Control Re enable block OldDist ActDist Update the distance old for next test PControl False unflag so next loop w goto I Control MotInc 11 upon return to I Control motinc will reset TumDelta 2 Basic condition for normal I Control Recompute Turn Variables w FwdSpeed and TurnDelta as inputs Call MotCont write to MCs with reset TurnDelta End If Abs DegError lt 5 End If end of PControl True ossi eee eee END OF P CONTROL BLOCK
3. Logic Input Logic Input Single switch Single switch system diagram y g einen modular diagram h sw 1 HCF4093B gt i Out In y BOND Out CD4066BC i Switch R I 2 2 Swe Rx Y CD4066BC Switch Drawing A Drawing B Figure 4 7 Single resistor circuit of voltage controller Figure 4 8 is a complete circuit diagram of the voltage control system that shows all of the pin connections between the four chips the resistors the trim pot the 5 V power supply the DIO input and the Vrap output voltage Two elements shown in this figure should be discussed the first is the 5K trim pot shown in the circuit diagram and the 44 second is the clock input from the DIO As previously discussed the 5 kQ trim pot is only installed on the right side voltage controller for the purpose of making small and accurate speed adjustments The clock input from the DIO is a pulsed signal that is necessary to control the latch or flip flop chip When the latch receives the clock pulse it reads the 4 bit logic values then writes and maintains this value at the output side of the chip If the input values change the output will remain the same until the flip flop receives the next clock pulse The 5 VDC power input to both of the IFB voltage controllers is switched by the side switch that is located next to the left side resistor block 4 2 4 Voltage controller summary There is
4. i FOR EXCELL PRINT OUT USE DOC 1 csv IN EXCEL XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Print 2 time2 Tab 10 ActDeg Tab 17 OffError Tab 28 Turn 10 Tab 37 TurnDelta 10 XKXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX If ActDist gt Distance Then NewLine True Passes command back to TABLEREAD for Excel read End If Print 1 End HallsCompute End Sub HallsCompute sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX CorrectCompute This block is called repeatedly during the blue while loop the purpose of CorrectIndex is to schedule events based on met conditions Each if block index is only done once This sub will output turn values for the MotCont Need to update the OldDist same as HallsCompute XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub CorrectCompute enters when ActDist gt OldDist CorrectIndex 1 block first Index block decodes the 6 digit number 155 If CorrectIndex 1 Then Print 1 Tab 12 inside corblk 1 Index CorrectIndex this is working CorX Int numData 0 0001 get Ist pair of digits Print 1 Tab 12 numData numData CorX CorX CorY Int numData 0 01 CorX 100 get 2nd pair of digits Print 1 Tab 12 CorY
5. 95 A simple way of understanding the use of the DIO is to consider the two DIO command statements below These two commands are written in VB 6 code and will either write voltages to a channel or read voltages from a channel These are the only two DIO commands used by the ROS e Error DIOWRITE Device Channel Input Value e Error DIOREAD Device Channel Output Value The Error value is either a one 1 indicating an error or a zero 0 indicating a successful command outcome The two functions of the DIO are described by using either the DIOWRITE or DIOREAD The Device is a fixed number that identifies the specific PCMCIA card Since there is only one card used there will only be one device number throughout all of the code The Channel is the same as the Logical Channel entry in Table 7 1 It determines which channel will be read from or written to by the DIO Finally the value will be either an input if the argument is a write command or an output if the argument is a read command The number of bits per channel will determine the range of both the input and output values 7 2 2 ROS wireless capability The next peripheral to be described is the wireless connectivity used by the robot to communicate between the client and the server As previously discussed the client side program has the primary responsibility of processing the image of the optical mark and acting as an interface for the u
6. 113 lt cew START Torgus Torque i i Direction of travel Torque impulse i application and duration I I I ie 100 200 300 400 500 600 Top View of robot cm and course of travel In both experiments the robot returned to straight Il i path travel at approximately this point Figure 7 9 Course layout for SP controller experiment Figure 7 10 shows the results from the SP controller CW impulse experiment The impulse began after 10 5 s of travel and lasted for 0 9 s Figure 7 11 shows the results from the CCW experiment In this case the torque impulse began at 10 9 s and lasted for a total time of 0 8 s Several observations can be concluded from both sets of experimental data The objective of the controller is not just to return the robot to a straight path but is also to correct for the offset error caused when the impulse forced the robot to veer from its intended course Both results show that after the initial angle has been corrected the robot heading will be approximately 5 in the direction opposite to the angular error caused by the impulse This opposite heading will cause the robot to return to its intended path In the CW impulse experiment the robot returned to its normal path and direction after 4 4 s and 115 cm From this point on the oscillation was limited to 1 114 e In the CCW impulse experiment the robo
7. An immediate objective is to travel a path of a reasonable distance of approximately 5 m and then optically correct for any positional errors at the end of each path segment Since any errors in dead reckoning will be minimized by the optical correction the total positional errors will not accumulate as each segment path has been completed The result of joining an optical camera able to read a fixed error mark with a dead reckoning capability should allow the PML robot to travel any desired distance and end with a total error that is kept within a specific error bound In order to achieve this objective the robot must have an accurate dead reckoning capability Initial experiments on the IPRV indicated that the Hall sensors that were used to determine the position had significant errors It was discovered that some of the pulses were not counted which caused an inaccurate estimation of the distance traveled Because of the relatively low 1 cm accuracy of the Hall sensors it is essential that there are no lost counts since the effect of even a single lost count has a significant effect on the position accuracy of the robot The successful completion of this research will be a working hardware model that will demonstrate the ability to navigate a path with minimal ending positional error of approximately 5 cm This capability requires the robot to have a map of the environment in its database In the case of this research the map
8. CorY CorAlpha Int numData CorX 10000 CorY 100 Print 1 Tab 12 CorAlpha CorAlpha PRINT COMMANDS IstInput AddItem CorX IstInput AddItem CorY IstInput AddItem CorAlpha Print 1 Tab 12 set slow turn speeds for course correction Right R 7 necessary to prevent MC from stopping w Turn 0 during Right_L 6 Case Blue Left_R 6 Left L 7 Maen nnn nnnnnnnnnn REPLACE THE 4 CONTROL VARIABLES Convert the Image process values CorX CorY CorAlpha into the 4 control variables used in Case Green All computation and variable relations should be dealt with in the code below Print 1 Replace the 4 Control Variables Distance Int 36 CorX 50 0 2 2 45 in pulses or 36 inches OffsetRef CorY 50 0 2 23 OffsetRef is in Pulses FwdSpeed 7 Slow speed for correction DirAngle CorAlpha 50 DirAngle is the angular correction required based on the image angle CorAlpha nen nnnoo OUTPUTS OF CONVERSION TxtDist Text Distance TxtFwd Text FwdSpeed TxtDeg Text DirAngle TxtOff Text OffsetRef Print 1 Distance Distance FwdSpeed FwdSpeed Print 1 DirAngle DirAngle OffsetRef OffsetRef Print 1 end of correct compute blk 1 Print 1 156
9. Position Sensors Chapter 5 Chapter 7 Chapter 7 User Commands Wireless OCs G R Y Chapter 6 Chapter 7 Chapter 7 Figure 7 1 The ROS and it peripherals 7 2 Peripheral Modules This section will discuss all the peripherals that have not yet been explained in earlier chapters The first subsection will explain how the DIO is used to read the position sensors and send speed commands to the two motor controller systems 7 2 1 DIO device The digital input output DIO device used by the PML robot is manufactured by SuperLogics and can be seen in Figure 7 2 The Personal Computer Memory Card International Association PCMCIA card is installed in one of the slots of the laptop and connects to the CP 1037 cable as shown in the figure 21 The CP 1037 cable converts the PCMCIA 33 pin output to a D 37 male connector which is then connected to a D 37 female connector permanently mounted to the electronics housing box 93 Individual control wires are connected to the output side of the female D 37 connector grouped into channels and then connected to the interface boards channel connections SuperLogics Simplifying Data Acquisition PCM Series i Figure 7 2 PCMCIA card and CP 1037 cable One of the advantages of using the SuperLogics PCMDIO device is its ability to have its channels defined by the user Figure 7 3 shows an image of the configuration utility that is i
10. eee eeeeeeseceeeeeeeesseceeseeeenaeeeenes ROBOT OPERATING SYSTEM 3 c ccccnssvsesssssesevtsersaoscossestevonsvensensoens Dek MAME UN ON r i e ace ge dra SOI an Daa cee wba Sa Slt 7 2 Peripheral Modules csi sescccasissicacisunsaasveidocedaseasae eders TZ WUC dEV C sect facet atti ial cia one nere heel ies cea Aes 7 2 2 ROS wireless capability cacwicsheceled aa dieeietee dates 7 23 Opertatonal data Sige ya ener 7 24 Pathdataba sensin dian dyan cet ccabengton shang lated a inss 1 3 Program Design ic ssiaiivereadndiscsaciswrsdesvesdocees a A ate ant 7 3 M n prosta ounan a a N 7 3 2 Subroutines used by the main program 2 T3 3 MOJUlES isiin aieiaiei annist 7 4 Straight Path Controller W u u X sseneveerere renere rsererrenn reen eren renerne 7 4 1 Definition of units used in ROS and controllers 7 4 2 SP controller error signal computation cece eeeeeeeee 7 4 3 One to one map between OffsetRef and RefAng 7 4 4 Supplemental small angle oscillation controller 7 4 5 SP controller experimental results and conclusion TILA CONTO iini eeit ie BT ae deat Deel Gt ae eevee esau edna ise 7 5 1 Design of LA controller oc eeeceeececssececsscceessecesneeeesnes 7 5 2 LA control conclusion sa cs pe ceece eg stick codes oad et poeta eases TAO S WIT ab of ROSA ne ole aa it Agathe eases ix Page 63 63 64 66 67 67 68 69 69 71 73 75 79 81 85 87 89 CHAP
11. use the first name for the form Not the one in Call Load frmMC_camera frmMC_camera Show vbModal gintlogicaldevice openDevice Debug Print LogicalDevice amp gintlogicaldevice errUnknown Call unknownErrorMessage End Sub Public Sub pemdioError ByVal LogicalDevice As Integer ByVal ErrorCode As Integer Call errorMessage intStatus This is the one with the problem 170 Call PCMCloseDeviceVB gintlogicaldevice End Sub 171 APPENDIX B CLIENT SIDE PROGRAM The client side program is a Visual Basic 6 project installed on the remote computer and is used as the primary interface with the operator As explained in this thesis this code is also responsible for processing the webcam image and returning the error value to the ROS This project consists of a VB 6 form and one module 1 Camera client frm This form provides the primary interface with the user reads the Excel results and controls the wireless connection This program is found in Appendix B 1 2 Modulel bas This module provides function used by the webcam and can be used for any additional sub programs that may be needed in the future This program is found in Appendix B 2 B 1 Camera client frm Camera Comm frmClient Option Explicit Dim ExcelLoc As String use this as the location of the excel doc use between computers Dim Echo As String Use variant because sometime Echo somtimes Color Dim EchoPrevious As String
12. 4 5 6 7 8 130 REFERENCES G Moore Cramming More Components onto Integrated Circuits Electron Mag vol 38 no 8 pp 114 117 1965 R Homji 2005 Intelligent Pothole Repair Vehicle M S thesis TX A amp M College Station TX R House Random House Webster s Unabridged Dictionary 2nd ed New York Random House Ref 2001 H Sahin and L Gtivenc Houshold Robotics Autonomous Devices for Vacuuming and Lawn Mowing IEEE Control Systems Magazine vol 27 no 2 pp 20 23 2007 S M Engineers 2004 Apr Mobile Robots Fit Well into Flexible Manufacturing Systems Society of Manufacturing Engineers Online Available http www sme org cgi bin get press p1 amp amp 20041092 amp TE amp amp SME R Sharma and H Sutanto A Framework for Robot Motion Planning with Sensor Constraints Trans Robot Autom vol 13 no 1 pp 61 73 Feb 1997 N Pears B Liang and Z Chen Mobile Robot Visual Navigation Using Multiple Features EURASIP J on Appl Signal Process vol 2005 no 14 pp 2250 2259 2005 S Martins and J C Alves A High level Tool for the Design of Custom Image Processing Systems presented at the 8th Euromicro Conf on Digital Syst Design Porto Portugal Jul 2005 9 10 11 12 13 14 15 16 131 The Mathworks Inc 2007 Sep Video and Image Processing Blockset Users Guide Online Available http
13. End If IstInput AddItem strData lists the commands on the server list box IstInput AddItem numData Case red This means STOP blnRun False Unload Me calls Neutral close doc close server Case green This means GO Print 1 Tab 12 CorCarryDeg CorCarryDeg inside green ActDeg ActDeg DirAngle DirAngle Print 1 Tab 12 Turn Turn OldTurn OldTurn Print 1 frmServer BackColor vbGreen Call Initialize2 Used to reset conditions variables for subsequent green calls shouldn t hurt the Ist green run 139 DoEvents If NewLine True Then only do when newline Call ReadTable End If Call HallsRead do every loop First trip through ActDeg CorCarryDeg If bInRun False Then need this to prevent the DIO from writing to the latches for some GoTo 1000 reason after Neutral and the lt Read Table the following commands below inputs 1000 End If Halls Compute and MotCont the DIO alters the latch If ActDist gt OldDist Then do with new distance info Call HallsCompute End If If Turn lt gt OldTurn Then only do when new command Call MotCont End If If timel gt CamTime 2 Then send a pic on an interval Call tepServer SendData picture Call StreamCam End If DoEvents this is the GOTO label for the if block above Wend blnrun is true Call RunCamera automatically starts camera at end
14. Private Sub Form_Unload Cancel As Integer Call Neutral tcpServer Close Close 1 waitTime 1000 End End Sub OC Oe eo AI AAAA KAR CMD FORMAT _ XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Private Sub cmdFormat Click Call SendMessage hCap WM CLOSE 0 0 hCap is defined in the Cam module hCap capCreateCaptureWindow Take a Camera Shot WS CHILD Or WS VISIBLE 0 0 PicWebCam Width PicWebCam Height PicWebCam hWnd 0 If hCap lt gt 0 Then Call SendMessage hCap WM_CAP_DRIVER_CONNECT 0 0 Call SendMessage hCap WM_CAP_SET_PREVIEWRATE 66 0 amp Call SendMessage hCap WM_CAP_SET_PREVIEW CLng True 0 amp End If temp SendMessage hCap WM_CAP_DLG_VIDEOFORMAT 0 amp 0 amp End Sub Oe ees SEERE CMD UNLOAD XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Private Sub cmdUnload Click blnRun False Unload Me End End Sub E E KK KOGER KOOKS SS SS SS SS E S S XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX 138 Private Sub tcpServer DataArrival ByVal bytesTotal As Long On Error GoTo error Call tcpServer GetData strData SINGLE CALL TO DOWNLOAD THE SENT INFO Print 1 Tab 12 data arrived strData num Val strData equals to zero if input string otherwise If num lt gt 0 Then input numData num only updates numData when is sent from client strData blue when is sent make dummy val for string data
15. and heading 90 N gt 152 cm 60 in Robot path Figure 7 16 Course layout for 90 turns The data output indicates that both the SP controller and the LA controller are working together effectively at controlling the 90 turns Once the large angle turn is completed the SP controller takes over and drives the robot to a 0 heading angle The Offset Error is seen to be constant during the large angle turn and then changes its value once the SP controller takes over 124 ActD ctDeg g0 L 60 f 40 F 20 F 0 i 0 1 2 Offset 10 Error 5 0 Turn Delta 7 6 5 4 3 2 1 3 4 Time s Figure 7 17 Results from 90 course with right turn Offset Error 0 5 Turn Delta NUAR NO 3 4 Time s Figure 7 18 Results from 90 course with left turn 125 7 6 Summary of ROS This chapter has presented a detailed explanation of the ROS used by the robot Several of the peripheral modules such as the pathway database the wireless hardware and the user command interface are described from the point of view of the ROS The basic function of the ROS is to act as a manager for all of the peripheral elements necessary for the proper functioning of the robot Section 7 3 was dedicated to the description of the main program design with the intent of describing the logic behind it The remaini
16. record of previous Echo Dim Excl_X As String row and column of Excel output 172 Dim Excl_Y As String __ processed Y output from Image data Dim Excl_AAs String Image rotation angle w no process direct measurement Dim Excl_Cor As String encoded value created by Excel Dim XError As Single numerical output from Excel program Dim YError As Single error from Excel Dim AError As Single _ Angular error from direct measurement of image Dim TotError As Single output error in code Dim Basic_X As Single computed X value as if no angular displacement Dim Basic_Y As Single computed Y value Dim Basic AAs Single computed Angle value this should be closer to the actual rotation Dim CorOut As Single Correction Output 6 digit that is coded signal sent to Server Private Sub cmdConnect_Click Print 1 above error in connect On Error GoTo error Print 1 below error in connect frmColorBar BackColor vb White If tepComm State lt gt sckClosed Then tcpComm Close End If Call tepComm Connect robot_server 10101 remote host and remote port Exit Sub error End Sub Private Sub cmdRed_ClickQ On Error GoTo error 173 Call tepComm SendData red RED Stop IblEcho Caption red Exit Sub error End Sub Private Sub cmdGreen_Click On Error GoTo error Call tepComm SendData green green START IblEcho Caption green this is a text box frmColor
17. time2 0 62 ActDist 4 ActDeg 0 timel 0 63 inside HallsRead time2 0 63 ActDist 4 ActDeg 0 timel 0 64 inside HallsRead time2 0 64 ActDist 4 ActDeg 0 timel 0 65 inside HallsRead time2 0 65 ActDist 4 ActDeg 0 timel 0 66 inside HallsRead time2 0 66 ActDist 4 ActDeg 0 188 189 timel 0 67 inside HallsRead time2 0 67 ActDist 4 5 ActDeg 1 Start HallsCompute 10 OffsetX 0 OffError 0 DegRef 0 DegError 1 SpecTurn 0 2 LTum 2 RTum 2 MotInc 9 Left Ratio 0 2222222 Right Ratio 0 2222222 Tum 0 OldTurn 0 timel 0 67 inside HallsRead time2 0 67 ActDist 4 5 ActDeg 1 Start HallsCompute 10 OffsetX 0 OffError 0 DegRef 0 DegError 0 SpecTurn 0 LTum 2 RTum 2 MotlInc 10 Left Ratio 0 2 Right Ratio 0 2 Turn 0 OldTurn 0 timel 0 69 inside HallsRead time2 0 69 ActDist 5 ActDeg 0 timel 0 71 inside HallsRead time2 0 71 ActDist 5 ActDeg 0 timel 0 72 inside HallsRead 190 time2 0 72 ActDist 5 ActDeg 0 timel 0 73 inside HallsRead time2 0 73 ActDist 5 ActDeg 0 timel 0 74 inside HallsRead time2 0 74 ActDist 5 ActDeg 0 timel 0 75 inside HallsRead time2 0 75 ActDist 5 ActDeg 0 timel 0 76 inside HallsRead time2 0 76 ActDist 5 5 ActDeg 1 Start HallsCompute 10 OffsetX 8 726195E 02 OffError 0 DegRef 0 DegError 1 SpecTurn 0 2 LTum 1 RTurn 1 Motlnc 1 Left Ratio 1 Right Ratio 1 Turn 1 OldTurn 0 End HallsCompute MOTCONT Inside MotCont Tur
18. vi ACKNOWLEDGMENTS I would like to take this opportunity to thank my fellow students at A amp M They helped shed light and bring perspective during my time here I would especially like to mention Reza S Sheridon H Ruzbeh H and Ali S for their help friendship and academic support Thanks to G Wayne and Suze for all the gourmet meals during my stay in College Station All the time I spent at the THRC helped me focus and study those long hours Thanks to Randall L for the discounts Mostly I would like to thank my Amy for her unflagging support and constant cheer Vii TABLE OF CONTENTS Page ABSTRACT ar ata E ates sto teearcattalgeceante teaches iii DEDICATION isinai nera ea a sees gr A EEEE TE ERTAS v ACKNOWLEDGMENTS ensin re EE eae vi TABLE OF CONTENTS ini sess ere rear n a a a vii BEES EOE FIGUR S E A A ETE E I T xi LIST OF TAB LES sussie see hero de A as XIV CHAPTER I INTRODUCTION ais i eiior iaee aaasta inated a ieS 1 TIS OLY a sesconddnraede ds aa a SES a ERE Se SE SE OSSE NERE 2 EX THESIS Objectives cireni i E born 3 1 2 1 First thesis ODjECUVE vejicdssssccesanscceastseeuadtesnaccteeneseseneanesacteses 3 1 2 2 Second thesis objective iana 4 1 2 3 Third thesis Object Veres nien i is 5 I LITERATURE REVIEW 232 sonnerie dysse es aes 7 2 1 Possible Developmental Topics cccceeseceesseceesteeeesteeeenseeeenes 7 Dela Household POD OIS iapucsdicetiudsagectaceatengtonsvagtete issis 8 Del 2 Factory rObOtS sn nnn dern
19. will be a programmed path constituting a priori information about the environment In order to navigate it is necessary that the robot be able to minimize the linear and angular errors from the defined path At the hardware level errors in data acquisition and calculation should be shown to be less than a specified maximum A successful demonstration will show that the ending error after the obstacle course has been run will be within specified bounds 1 2 3 Third thesis objective Once the determination was made to build a robust robot base and a robot that could be used to traverse any distance with accuracy by using the above path segment method it was necessary to determine what improvements needed to be made in order to accomplish this goal The earliest experiments were designed to investigate if the capabilities of the IPRV robot were sufficient to meet the above requirements When the robot was run at a walking speed these experiments showed that the laptop in combination with the DIO did not have a fast enough sampling rate It was concluded that there were serious limitations in the hardware Although the laptop had a 451 MHz processor which appeared to be fast enough the speed of the entire system was not high enough to capture every pulse of each of the Hall sensors Because of the low precision of the positional sensor a single error would have a dramatic effect on the dead reckoning capability of each segment Therefore the
20. Chapter IV gave a complete description of the 4 bit voltage controller which as a module has the most complicated circuitry in the robot This chapter will cover two of the items in the above list the 5 V power supply and the 4 bit Hall counter input Section 5 2 will outline the 5 V power supply and Section 5 3 will discuss the development the circuit diagram and the design logic for the two 4 bit Hall counter circuits The DIO connection and wiring will be discussed in Chapter VII as part of the discussion on software and data acquisition Figure 5 1 shows the block diagram of the two interface boards the connections between both boards and the external connections to the DIO and sensors 5 2 Five VDC Power Supply When the 4 bit Hall sensor counter system was being developed there were consistent problems with its accuracy The cause of the inaccuracy was traced to more than one source however the primary and most severe cause was found to be the overuse of the voltage regulators Previously a single LM7805 voltage regulator powered both the Hall sensors and the position counting circuit 54 4 Bit DIO Z Clear 2 Channels l Position Counting Circuit 4 Pairs of Resistance Clock e 4 _ Power e L _ 4Bit signal e _ 6 5 VDC Resistor 4 Pairs of Block Resistance 8 Voltage a Regulators 5 K ohm 3 A Resistor Variable Resistor Block
21. Tab 12 DirAngle DirAngle OffSetRef OffsetRef Print 1 If FwdSpeed 7 Then 7 is the slowest speed Right R 7 Left L 7 Else FwdSpeed is less than 7 Right_R FwdSpeed 1 example if speed 6 then slow wheel goes 7 Left_L FwdSpeed 1 End If Eventhough in I Control the TurnDelta should always 2 this is in case of future change that uses TurnDelta for control Right_L Right_R TurnDelta Left_R Left_L TurnDelta eee INDEX Indx Indx 1 Increments the index used to read the Excel table NewLine False won t read this block again till distance is reached If Distance 10000 Then do 10000 end command use this for interval stops blnRun False Call Neutral waitTime 500 for the robot to stop then get distance timeCarry time2 update the cumulative time at very end of Case Green runs End If 145 If Distance 12000 Then do 12000 end command use this for final stop blnRun False Call Neutral Unload Me End If End Sub ends the READ TABLE sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub HallsRead Print out 3 in order to debug the Hall skip problem this doc used to track the ripple and the hall counts intStatus singleDigitalInput gintlogicaldevice R Hall HallCntRight If intStatus lt gt 0 Then Call pemdioError gintlogicaldevice intStatus errPemdioError End I
22. The error distance from the target center is easily computed once the Xp and Yr coordinates are known The center of the optical target defines a point in the image space the procedure above will convert this point into a real space Cartesian location Several assumptions in this procedure were made that need to be verified For example both assumptions of independence should be tested and the functional relationship between the image and real variables should be determined It should be recalled that this procedure was 79 developed with a set equal to zero and it is not yet able to evaluate the rotation of the optical target 6 3 5 Decoupling of input variables As mentioned in the previous subsection in order for the above procedure to work it is necessary to establish the fact that the input variables to the data conversion equations are decoupled If this decoupling is true then both equations 6 1 and 6 2 will also be true If this decoupling was not true then the function that defines 6 1 would require two input variables 6 Rp The existence of the two decoupling conditions will mean that the input to the function of both 0 and R will be a single input variable The result of this implies that the following two equations 6 1 and 6 2 will define the functional relationship between the image space and the real space Or fOD 6 1 Rr g RD 6 2 The first condition of decoupling is that Ry is not a function o
23. Write faster speed lower number to the right MC block intStatus singleDigitalOutput gintlogicaldevice MC_R Left_R This writes the 4 bit value to the right MC latch If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If Write a slower speed higher number to the left MC block intStatus singleDigitalOutput gintlogicaldevice MC_L Left_L This writes the 4 bit value to the left MC latch If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If Pulse both right and left latches Call PulseMC Write DIO to both latches and then Clock the speed to the MC End Sub Left_Turn XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Write same speed as fwdspeed to the right MC block intStatus singleDigitalOutput gintlogicaldevice MC_R FwdSpeed This writes the 4 bit value to the right MC latch If intStatus lt gt 0 Then 163 Call pcmdioError gintlogicaldevice intStatus End If Write the same speed to the left MC block intStatus singleDigitalOutput gintlogicaldevice MC_L FwdSpeed This writes the 4 bit value to the left MC latch If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If Pulse both right and left latches Call PulseMC Write DIO to both latches and then Clock the speed to the MC End Sub Neutral Turn XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
24. measurements from a previously known position The PML robot is able to achieve dead reckoning because it is able to measure and record each wheel s rotation Each of these measurements is processed to yield two measurement derivatives The first derivative is the average of both wheels rotation which will yield the distance traveled by the center of the robot START OBSTACLE Maximum positional error diameter and angle End of path position and angle Actual result compared to desired result Figure 3 1 Single segment of dead reckoning path 19 The second measurement is the difference between both wheels rotation which will yield an angular measurement based upon the fixed geometry of the robot s physical size The basic angle and distance measurements are the primary inputs that the control algorithm uses for dead reckoning Figure 3 1 illustrates a typical obstacle avoidance path performed by the robot during its dead reckoning function The robot s dead reckoning capability will allow it to avoid an obstacle provided that the robot has an a priori map of the obstacle s size and location It should be noted that in order to avoid the obstacle as indicated in Figure 3 1 the robot must make four turns in order to get back on the original path Each of these turns will induce angular and positional errors As can be seen by the call out at the end of the path there will be some er
25. pe ices IFB wire bus Single conductor cables oP Multiple conductor cables Figure 5 1 Block diagram of the two interface boards This system was redesigned in order to distribute the demand for power over several of these regulators As seen in Figure 5 1 a 12 VDC supply from the main batteries is connected to a bus that feeds eight LM7805 voltage regulators The 5 VDC output of this block is cabled to a six element switch block The switch block controls the power that is supplied to individual modules on the IFB and in the electronics housing box It should be noted that having eight voltage regulators represents an over capacity in the ability to provide five volts of DC power On the underside of the IFB a power and ground bus provides 12 VDC supply to the eight voltage regulators At present only six of the voltage regulators are installed with room for another two as needed for future 55 expansion The switch block is made from a cut out section of proto board that is attached to the IFB main board by one inch standoffs Each switch on this board is rated for 2 A of current and at present only the first five switches are connected Another switch is located next to the right side resistor block on the IFB and is used to switch power to both the MC voltage control IC blocks Table 5 1 outlines the connections between the voltage regulators that produce 5 V the switches that control the power and the modules that
26. programming language and has the ability to access the API code segments that are 68 available to the Windows OS An outline of this code was originally found on the internet and has been re written and modified to support the robot s use of the webcam 18 This subroutine controls the webcam by using the API code in the following manner When the camera is first started an alias named Capture Window is defined as a function based on the avicap32 dll file VB 6 then creates a variable handle called hCap used for all references to this Capture Window alias Within the Windows OS there is a messaging service used by all programs that access these DLL files The Run_Camera subroutine controls the webcam by using the message service to call commands from the Windows OS An example of the code used to call one of these commands is seen below e Call SendMessage hCap operation to be done variable list 6 2 2 Server side optical correction data return Figure 6 3 shows that there are two parts of the server side program code The second responsibility of the server side optical correction code is to receive the 6 digit correction number sent from the client side program The correction code datum is always sent in the form of a single number All other commands from the client are sent as colors When the server side program receives the 6 digit number it decodes the error number and makes the results available as input to the ROS
27. remains incomplete The benefit of running the robot while manually extracting data has provided valuable information This data based upon experience can be used to specify the accuracy of the automated data extraction module One of the objectives of this research has been the development of a robust base capable of supporting future development Possible future development of the robot could be the following e The addition of an advanced sensor and environmental mapping capability Subsection 2 2 5 discussed a paper written about the Simultaneous Localization and Mapping procedure that could be used as a model for future development 14 151 e The development of the dynamic model of the vehicle that could be used for trajectory planning 8 3 Conclusions With the exception of the automated data extraction all of the goals of this thesis have been met The relative accuracy of the robot during its demonstrations given the lack of resolution of both input and output capabilities is encouraging At the end of Homji s research thesis he stated the following three future work items 2 The following is proposed as future work to enhance the functionality of the IPRV and mitigate the current limitations Use of a second processor to send live streaming images to the client computer Use of a 1 MHz external clock to sample sensor signals This can lead to a sampling period of 1 us which is a dramatic improvement to the cu
28. timel 0 92 inside HallsRead time2 0 92 ActDist 7 ActDeg 0 timel 0 93 inside HallsRead 192 193 time2 0 93 ActDist 7 ActDeg 0 timel 0 96 inside HallsRead time2 0 96 ActDist 7 5 ActDeg 1 Start HallsCompute 10 OffsetX 0 2617859 OffError 0 DegRef 0 DegError 1 SpecTurn 0 2 LTum 1 RTurn 1 Motlnc 5 Left Ratio 0 2 Right Ratio 0 2 Turn 1 OldTurn 0 End HallsCompute MOTCONT Inside MotCont Turn 1 OldTurn 0 TIME SKIP Distance 150 FwdSpeed 6 DirAngle 84 OffSetRef 0 timel 4 41 inside HallsRead time2 4 41 ActDist 53 ActDeg 84 Start HallsCompute MOTCONT Inside MotCont Turn 1 OldTurn 1 Right_R 7 Right_L 6 Left_R 6 Left_L 7 end motcont Inside Initial block of P Control 194 TurnDelta 1 Right R 7 Right L 6 Left R 6 Left L 7 incPC 0 incPC2 1 Test PTime 4 41 DegHistory incPC 84 PRatio 84 DegHistory incPC2 0 pratio MOTCONT Inside MotCont Turn 1 OldTurn 1 Right_R 7 Right_L 4 Left_R 4 Left L 7 end motcont TurnDelta 3 End HallsCompute timel 4 45 inside HallsRead time2 4 45 ActDist 54 ActDeg 84 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 46 inside HallsRead time2 4 46 ActDist 54 ActDeg 84 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 47 inside HallsRead time2 4 47 ActDist 54 ActDeg 84 Start HallsCompute
29. true for both Green Blue flagCamera True set true for longer initial wait time or to only start cam one time at beginning of Case Green CaseBlueFlg False true disable EMO false enable EMO for case green Print 1 Inside Initialize2 before camstart timeint timeInt Call SendMessage hCap WM_CLOSE 0 0 hCap is defined in the Cam_module hCap capCreateCaptureWindow Take a Camera Shot WS CHILD Or WS_VISIBLE 0 0 PicWebCam Width PicWebCam Height PicWebCam hWnd 0 If hCap lt gt 0 Then Call SendMessage hCap WM CAP DRIVER CONNECT 0 0 Call SendMessage hCap WM CAP SET PREVIEWRATE 66 0 amp Call SendMessage hCap WM CAP SET PREVIEW CLng True 0 amp End If Print 1 just below camstart time Timer initiate fwd motion OldDist 0 5 so initial call into HallsCompute OldTurn 10 resets turn so initial movement takes place From Initialize Sub above OffsetX 0 reset I Control error each new Case Green CorrectIndex 1 use to sequence the Case Blue commands this call is at beginning of Case Green TurnDelta 2 2 is the basic condition for I Control variations done in P Control set back to 2 at I Control re enable End Sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Private Sub Zero Print 1 Tab 12 Inside sub Zero RESETS THE 191 COUNTER XXXXXXXXXXXXXXXXXXXX
30. www mathworks com access helpdesk help pdf_doc vipblks vipblks pdf S J Julier and H F Durrant Whyte On the Role of Process Models in Autonomous Land Vehicle Navigation Systems JEEE Trans Robot Autom vol 19 no 1 pp 1 6 Feb 2003 F Chaumette and S Hutchinson Visual Servo Control Part 1 Basic Approaches IEEE Robot Automat Mag vol 13 no 4 pp 82 90 Dec 2006 D Bank A High Performance Ultrasonic Sensing System for Mobile Robots presented at ROBOTIK 2002 Ludwigsburg Germany Jun 2002 N Falc n C M Travieso J B Alonso and M A Ferrer Image Processing Techniques for Braille Writing Recognition presented at EUROCAST 2005 Palmas de Gran Canaria Spain Feb 2005 H Durrant Whyte and T Bailey Simultaneous Localization and Mapping Part 1 IEEE Robot Automat Mag vol 13 no 2 pp 99 110 Jun 2006 H Durrant Whyte and T Bailey Simultaneous Localization and Mapping SLAM Part 2 IEEE Robot Automat Mag vol 13 no 3 pp 108 117 2006 Germany s Federal Ministry of Education and Research 2003 Nov 11 MORPHA Online Available http www morpha de php_e morpha_ projekt php 3 17 18 19 20 21 22 23 132 G F Franklin J D Powell and A Emami Naeini Feedback Control of Dynamic Systems 4 ed Upper Saddle River NJ Prentice Hall 2002 A Hussain 2007 Mar 03 How to Capture Picture Using Webcam in VB6 0
31. 55 ActDist 55 5 ActDeg 83 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 56 inside HallsRead time2 4 56 ActDist 55 5 ActDeg 83 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 57 inside HallsRead time2 4 57 ActDist 55 5 ActDeg 83 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 57 inside HallsRead time2 4 57 ActDist 55 5 ActDeg 83 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 58 inside HallsRead time2 4 58 ActDist 56 ActDeg 82 TIME SKIP Start HallsCompute Start HallsCompute TurnDelta 3 197 198 End HallsCompute timel 4 64 inside HallsRead time2 4 64 ActDist 57 ActDeg 82 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 65 inside HallsRead time2 4 65 ActDist 57 ActDeg 82 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 65 inside HallsRead time2 4 65 ActDist 57 ActDeg 82 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 66 inside HallsRead time2 4 66 ActDist 57 5 ActDeg 81 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 67 inside HallsRead time2 4 67 ActDist 58 ActDeg 82 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 67 inside HallsRead time2 4 67 ActDist 58 ActDeg 82 199 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 68 inside HallsRead time2 4 68 ActDist 58 ActDeg 82 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 6
32. Configuration 2 PML Modified Configuration Figure 5 2 Modification to the Hall sensor module The robot s operating system ROS works by constantly repeating a basic cycle composed of several computational tasks Some tasks are repeated every cycle while others occur less frequently Early experiments indicated that the ROS cycle rate would slow down when doing the more complicated computational tasks When performing the simplest tasks the ROS would operate at its fastest cycle rate and complete a cycle in approximately 3 ms When the program was required to perform the more difficult tasks the cycle time would increase to as much as 5 ms The ROS is structured so that it will read both position sensors only once during each cycle Therefore the position count frequency is directly dependent upon the ROS cycle rate 57 5 3 1 Minimum sampling frequency for position count Now that the slowest actual sampling frequency is known an evaluation of the required sample rate should be undertaken The actual rate should then be compared to the theoretical rate in order to determine if there will be any errors in the count The loss of just one position pulse from one Hall sensor has a significant effect on the accuracy of the dead reckoning operation Based on the distance between the drive wheels it is computed that a single pulse count error on one wheel will create an angular error of 1 Equation 3 1 indicates that a 1 angular error ove
33. EmoDist EmoDist EmoTime EmoTime 1010 goto statement if EMO occurs during acceleration skips the above estimation End If timel EmoTime txtActDeg Text ActDeg prints ActDeg on the form txtActDist Text ActDist Print 1 Print 1 timel timel inside HallsRead 148 Print 1 time2 time2 Tab 16 ActDist ActDist ActDeg ActDeg Print 1 End Sub Halls Read XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX If ActDist gt OldDist Then is the condition for this sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub HallsCompute Print 1 Print 1 Start HallsCompute I CONTROL BLOCK If PControl False Then if true then skip the I Control program block eee ee ee INCREMENT COUNTERS AND RESET VALUES MotInc MotInc 1 If MotInc gt 11 Then MotlInc 1 LTurn 1 RTurn 1 End If UPDATE VARIABLES OFFSET ERROR This block outputs the Offset Error and updates the OldDist used to access the Halls Compute subfunction call DelDist ActDist OldDist Compute delta distance based on 191 skip OldDist ActDist Update the distance old for next test DelX DelDist Sin ActDeg 180 3 14159 Incremental Offset error Units of DelX OffsetX OffsetX Del
34. Once a basic image processing system that is capable of gaining data about the environment has been implemented the next step will be to process the large amount of incoming data Simultaneous Localization and Mapping SLAM is a procedure that is about 10 years old and has been developed for this purpose This procedure is generally a recursive method for relating a large amount of sensor data in such a way that an accurate map of the local environment is created and the real time position of the robot is determined A paper by Durrant Whyte and Bailey discussed the history and basic premise of the SLAM procedure 14 15 However as this paper suggests there remain 14 significant real world hurdles that need to be overcome before this method is robust enough for general robotic use If this procedure were implemented into the PML robot it would be possible that the robot would have the ability to be placed in an unknown environment and then independently create a map of the environment In addition if a general map of the region were a priori knowledge the robot would be able to match the two maps and determine its exact position within the larger general map and framework One advantage of using the SLAM procedure is that there are several open source and freeware resources and many of these programs are written in the Matlab and C languages 14 15 2 2 6 Communication between a human and a robot Unlike a software program th
35. The client side program is responsible for encoding the three positional data values see Subsection 6 3 3 for an explanation of the encoding scheme The server side will decode the correction number by using the reverse of the encoding scheme and pass the three correction values to the ROS The server side data return code is located in the first index block of the Case Blue section of the ROS 69 6 3 Client Side OCS Manager The client side optical correction system resides on the client computer and is designed to minimize the computational effort of the robot s main computer A stand alone VB 6 program titled Camera Client performs the function of a manager by controlling the programs or utilities used by the client For example this program manages the wireless device controls the Excel program and acts as an interface with the user The client side program fulfills the two main tasks data extraction and data conversion for the optical correction system An outline of these tasks can be seen on the right side of Figure 6 3 This program manages the flow of correction data by first sending the image data to the data extraction module and then sending the output to the data conversion module After the data have been converted mapped from the image space to the real space the client side OCS manager returns the data to the server side program With the return of the error data to the robot the ROS will begin its autonomous movement alon
36. TurnDelta 3 195 End HallsCompute timel 4 47 inside HallsRead time2 4 47 ActDist 54 ActDeg 84 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 48 inside HallsRead time2 4 48 ActDist 54 ActDeg 84 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 49 inside HallsRead time2 4 49 ActDist 54 ActDeg 84 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 49 inside HallsRead time2 4 49 ActDist 54 5 ActDeg 83 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 5 inside HallsRead time2 4 5 ActDist 54 5 ActDeg 83 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 51 inside HallsRead time2 4 51 ActDist 54 5 ActDeg 83 196 Start HallsCompute PRatio Block incPC 1 incPC2 2 Test PTime 4 51 DegHistory incPC 83 PRatio 83 DegHistory incPC2 0 pratio TurnDelta 3 End HallsCompute timel 4 51 inside HallsRead time2 4 51 ActDist 54 5 ActDeg 83 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 52 inside HallsRead time2 4 52 ActDist 55 ActDeg 84 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 53 inside HallsRead time2 4 53 ActDist 55 ActDeg 84 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 54 inside HallsRead time2 4 54 ActDist 55 ActDeg 84 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 55 inside HallsRead time2 4
37. amp error_descr amp vbCrLf End Sub B 2 Modulel bas Cam module Option Explicit Public hCap As Long Public Const WS CHILD As Long amp H40000000 Public Const WS VISIBLE As Long amp H10000000 Public Const WM_USER As Long amp H400 Public Const WM_CAP_START As Long WM_USER Public Const WM CAP DRIVER CONNECT As Long WM CAP START 10 Public Const WM CAP DRIVER DISCONNECT As Long WM CAP START 11 Public Const WM CAP SET PREVIEW As Long WM CAP START 50 179 Public Const WM CAP SET PREVIEWRATE As Long WM CAP START 52 Public Const WM CAP DLG VIDEOFORMAT As Long WM CAP START 41 Public Const WM CAP FILE SAVEDIB As Long WM CAP START 25 Public Const WM CLOSE amp H10 Public Const WM CAP GRAB FRAME As Long 1084 Public Const WM_CAP_EDIT_COPY As Long 1054 Public Declare Function capCreateCaptureWindow _ Lib avicap32 dll Alias capCreateCaptureWindowA _ ByVal lpszWindowName As String ByVal dwStyle As Long _ ByVal X As Long ByVal Y As Long ByVal nWidth As Long _ ByVal nHeight As Long ByVal hwndParent As Long _ ByVal nID As Long As Long Public Declare Function SendMessage Lib user32 _ Alias SendMessageA ByVal hWnd As Long ByVal wMsg As Long _ ByVal wParam As Long ByRef lParam As Any As Long Public Sub CamToBMP Picturel As PictureBox On Error GoTo error Call SendMessage hCap WM_CAP_GRAB_FRAME 0 0 Call SendMessage hCap WM_CAP_EDIT_COPY 0 0 Picturel Picture Cli
38. be used to narrow the scope of a research topic One goal of this thesis is to integrate a visual processing capability onto a mobile robot and investigate robotic vision and navigation Due to the complexity of robotics in general and vision processing in specific the purpose of a good literature review will be to set real bounds on the development of the PML robot This section is organized into two parts the first part is a discussion of potential areas of development and the second part reviews the specific topics and skills that need to be mastered in order to succeed with these development objectives 2 1 Possible Developmental Topics After a search in the literature several potential subjects seem worthy of investigation The PML robot could serve as a base for research into several areas that may not seem at first to be practical For example it could be used to develop a lawn mowing robot s visual navigation capability The ability to cut grass is not as important to the development goal as is the efficiency of the visual signal processing and navigation algorithms The PML robot could serve as a test bed for this type of research One possible area of development would be a robotic wheelchair In the case of someone who is not able to use a joystick to navigate effectively or someone who has limited vision the goal of developing a robotic wheelchair that can intermittently interact with its user is a requirement For example
39. certainly more than one way to implement the voltage controller circuit in the previous section However there are advantages to the current design For example control over the ranging of each speed can be accomplished by varying the step size of the resistance multiple Another advantage is that each of these chips is available from the TAMU Physics supply shop which makes replacement of any of the components and troubleshooting of any problems easy to do 45 Vcc In O Zz CD4066BC a HE Figure 4 8 IFB complete voltage controller circuit diagram 4 3 Joystick and JSIC Circuit The next part of the motor control system is the manually controlled joystick and the joystick interface circuit JSIC The JSIC is also manufactured by Diverse Electronics 46 and is shipped attached as an add on module to the MC 7 motor controller For its application on the PML robot the JSIC has been removed from its original soldered mount on one of the MC 7 controllers and is now installed inside the electronics housing box Because of this removal the JSIC must be powered by an external source In addition the soldered connection served as an electrical signal connection to the attached MC 7 controller and this connection had to be replaced An external 5 V power supply has been provided by the voltage regulator bank and switch block that is on the IFB The supply voltage for the JSIC is now controlled by switch 5 on the switch block IFB pow
40. commands to be straight commands Table 7 7 lists the motor turn fraction that the small angle controller uses to affect the small angle control For purposes of programming one more variable is defined Spectrum turn SpecTurn is defined to be the fraction of motor control turn commands This variable is a dimensionless percentage that defines a ratio of turn to straight path commands Table 7 7 Functional relationship between DegError and SpecTurn DegError SpecTurn degrees 1 20 2 60 3 100 gt 3 100 7 4 5 SP controller experimental results and conclusions In order to test the effect of the SP controller the following experiment was designed A straight course of approximately 6 m was run and a torque impulse was imparted to the robot body at the 2 m mark The impulse was a torque of approximately 15 N m for one second of duration This amount of torque was applied by physically grabbing the robot at the same location on the frame each time With practice the amount of force that could be applied was consistent enough that the robot s maximum deviation in heading was 10 with less than 1 of difference for each run By reading the Primary txt output data it is possible to determine the amount of time that the force was applied Figure 7 9 shows the layout of the physical course Two different experiments were run where one torque impulse was clockwise CW and the other was counter clockwise CCW
41. designed that consisted of a 45 turns The objective of the course was to have the robot end its path at a fixed point hit the target and also end with an accurate heading of 45 Each experiment was run at least three times to insure that the course objective was met In each case the heading angle and stop position was consistently achieved The results from this experiment are shown in Figure 7 14 and Figure 7 15 121 45 Track to the right 45 Track to the left 121 cm 48 in 121 cm 48 in p 4 a 7 N Robot position Robot position y 3 and heading and heading N L N Robot path 5 7 45 i 45 X Robot path 7 183 cm 7 N 183 cm 72 in gi Ng 72 in 7 N i Point of turn Point of turn initiation initiation 43 cm 43 cm 17 in 17 in Start Start Figure 7 13 Course layout for 45 turns As previously discussed in Subsection 7 4 1 the units of each graph are ad hoc and were defined in Tables 7 5 Table 7 6 and 7 4 The results from the 45 experiment show that after the turn is initiated the heading angle settles down quickly The TurnDelta plot shows that the LA controller is functioning Once the angle of the turn reaches the controller hand off point the SP controller takes over and drives the robot to a 0 heading When the SP controller begins its operation the OffError variable can be seen to change its value A more detailed analysis of the ActDeg data shows th
42. have been used to convert the image data however there were two main reasons for choosing Excel for this purpose e Experience in using the client side manager to operate other programs e Development of the algorithm is graphically based see Subsection 6 3 3 In the current development of the PML robot the user inputs the data into the Excel program The client side OCS manager takes the output result of the Excel program and sends it to the server side program as indicated in Figure 6 3 The basic responsibility of the data conversion module is outlined in Figure 6 5 It is a two step process that begins after the image data has been extracted The module first converts the three image values into three real values The next step is to merge these real values into one encoded 6 digit number The encoded number is then sent to the client side OCS and finally returned to the robot to be used for course correction Note that the Excel program is responsible for both of these tasks The next two subsections will describe in detail how both of these programs are designed 73 DATA CONVERSION MODULE 6 Digit X Real Encoded Mapping Algorithm a Real Encoding Excel Program Excel Program Number Y Image Image to Real Error Data From Client side To Client side OCS Data Extraction Figure 6 5 Data conversion module responsibilities 6 3 3 Development of the data conversion map
43. less than 5 pulses per 0 5 seconds then motion is stopped the while loop is exited red is sent to the client s colorbar A picture will be made upon exiting the Case Green loop so nothing extra needs to be done to have a image sent If timel gt EmoTime 0 5 Then 0 5 is the 1st parameter for the EMO command 147 If timel lt 1 1 Then GoTo 1010 End If If ActDist EmoDist lt 3 Then 5 is the 2nd parameter for the EMO command DISABLE CONDITIONS If Distance 10000 Then This is done to prevent the EMO from activating GoTo 1010 when a 10000 command camera causes a stop inside End If the main while loop If CaseBlueFlg True Then if true then disable EMO GoTo 1010 End If Call Neutral This stops the robot Call tcpServer SendData red changes the color bar on Client to red frmServer BackColor vbRed change Server color to red waitTime 700 because the red was not being sent blnRun False kicks out of Case Green While loop at end of cycle OldDist ActDist keeps out of HallsCompute and a possible restart of motion OldTurn Turn keeps out of MotCont and a possible restart of motion Print 1 Inside of the EMO block delta ActDist EmoDist End If ActDist EmoDist updates EMO variables Print 1 EMO delta ActDist EmoDist EmoDist ActDist EmoTime timel Print 1 update EMO vars
44. mm in width by 127 mm in height The X and Y position on the image of each intersection point was measured The X distance is measured from the lower edge of the image and the Y distance is measured from the left edge of the image Table 6 1 is the look up table presents the data extracted from the webcam image of the grid 6 3 4 Correlation of image data to real position A first attempt to correlate the data from Table 6 1 by using just the measurements themselves was not successful The data from this table are considered as Local measurements in that they are defined in a Cartesian coordinate system that is limited to the image itself It was observed that the radial lines could be traced back to a Virtual point of origin at some distance outside of the image Figure 6 7 shows how the radial lines were extended to a hypothetical intersection point that defines the origin of a polar coordinate field Table 6 1 Local measurement data from webcam image of grid REAL POINTS R18N20 R21N20 R24N20 R27N20 R30N20 R33N20 R36N20 R39N20 R18N10 R21N10 R24N10 R27N10 R30N10 R33N10 R36N10 R39N10 IMAGE POINTS 49 61 72 81 88 95 102 29 43 56 49 48 47 46 45 44 26 24 21 REAL POINT R18P20 R21P20 R24P20 R27P20 R30P20 R33P20 R36P20 R39P20 R18P10 R21P10 R24P10 R27P10 R30P10 R33P10 R36P10 R39P10 R18P0 R21P0 R24P0 R27P0 R30P0 R33P0 R36P0 R39P0 IMAGE POINTS 50 62 13 82 89 96 103
45. position sensors processes this information and then outputs the speed commands to the motor control system The purpose of the SP controller is to control the robot s motion during straight path travel The SP controller should be able to prevent oscillation and keep the robot from drifting off of the path programmed by the PDB 106 7 4 1 Definition of units used in ROS and controllers A decision was made early in the PML development to use ad hoc units for all programming variables Using variables that are defined by the fabrication of the hardware has several advantages for the designer For example the definition of distance is defined as the number of pulses from the Hall sensor Several times during the development of the robot this one to one correspondence between the data and the physical construction made troubleshooting much easier The first two ad hoc variables to be created are the left Hall sensor count LHail and the right Hall sensor count RHaill Subsection 5 3 3 gives a basic description of the output of both Hall sensors as a single digit hexadecimal number The output of each Hall sensor is stored in each of the two variables LHall and RHall The value of these two variables will always be an integer in the range from 0 to 15 Table 7 5 lists the first four ad hoc variables used by the ROS and the controllers The four variables below are defined from the physical dimensions of the robot and the number of positio
46. requirement of reading every pulse from both of the wheels became a mandatory specification and served as a constraint for future design One possible solution to the above sensing problem would have been to replace the hardware with more accurate and expensive equipment In the early stage of this research a decision had to be made about the degree to which the hardware would be replaced The author had conversations with his advisor and was encouraged to keep the main hardware such as the laptop the Windows XP operating system OS and to continue to use the existing SuperLogics DIO The choice to keep the existing hardware meant that the course of the research would be affected In effect the poor capability of the existing hardware presented design limitations that unexpectedly lead to another research goal This goal could be described as the weak link in the chain theory of design The application of this theory would imply that there is no reason to have excessive capability in a processor if the operating system was responsible for slowing down the sample rate In addition if the positional resolution of the wheels was higher the sensor bandwidth would have to be wider and the signal processors would have to be faster and potentially more expensive Each of these areas of capability is considered to be a module CHAPTER II LITERATURE REVIEW Because of the complexity in robotics a literature review is an important tool that can
47. robot can be localized by measuring from the origin point under the camera center Figure 6 7 shows the origins for all four coordinate fields In order to simplify the development of the algorithm a rotation of the optical target was initially set to zero Once the algorithm was developed and tested with no rotation the addition of a non zero a was included in the data mapping conversion procedure and the algorithm was modified An outline of the initial procedure with a set to zero is as follows e First the optical mark s local image position is measured in the local Cartesian coordinates X and Y This will yield data similar to Table 6 1 The local coordinates are then converted into a global coordinate pair e Next the global Cartesian point is converted into a polar coordinate system of two variables R 0 The R refers to the radius distance from the camera origin the 9 refers to the camera radius angle and the subscript J indicates image data In Figure 6 7 the position of point P is uniquely described by the polar coordinate variables Ri 1 e The first assumption is that Ry is independent or decoupled from 0g The second assumption is that 0 is decoupled or independent from Rr Next functions relating Rj to Rr and dj to Op are developed Finally once Rg and Oz are known the unique point in the real space is also known This point in the polar coordinate system is then transformed into a Cartesian coordinate point
48. simplicity will allow it to be easily repaired or in the worst case it could be replaced at a reasonable cost The IFB Voltage Control circuit has a robust design and is easy to repair and troubleshoot One of the main advantages of this motor control system is that it allows for any configuration of speeds to be selected The result of the motor control system as described in this chapter is the dual control of both wheel speeds by the use of two 4 bit logic channels output by the DIO Table 4 3 presents experimental data on the IFB control of the wheel speeds Each of the columns is described below e Column 1 The DIO speed variable is an integer between 0 and 15 It is used as an input and determines the speed of the motors e Column 2 The binary equivalent is included for clarity and indicates what is really going on from the point of view of the 4 bit controller e Column 3 The Vap to ground resistance is the value of the variable resistance used to alter the output voltage e Column 4 The wheel speed is the final result of the output speed of the wheel The speeds of the right and left wheels will not be identical for each DIO input value since each of these motor controllers is independent Therefore Table 4 3 has the response of both right and left wheels The wheel speed was found by raising the robot on its stand engaging the motor gear drive and running the motor control system at each speed variable Each of the DIO speed vari
49. the PML robot is a project that will have a succession of multiple owners or developers it is important that each owner improve the robot in a manner that will add to its total capability After several owners the goal would be the development of a robust robot that possesses a high degree of autonomy and an ability to demonstrate complex behavior 16 CHAPTER III PML DESIGN OVERVIEW 3 1 Introduction The second objective of this research as described in Chapter I defines the main goal of the design effort The implementation of the design should yield a robot prototype that is able to go a long distance by combining dead reckoning path segments with an optical correction capability used to reduce total position errors The purpose of this chapter is to describe the overall design of the PML robot and its subsystems The first section of Chapter III describes the system overview and the subsystems from the point of view of their function The second section provides an overview of the hardware components The third section briefly discusses the robot s operating system ROS Chapter III does not contain specific information such as the pin diagrams of the IFB motor controller or the algorithm of the controllers A more detailed description of specific topics will be discussed in later chapters 3 1 1 Examples of redesign The current design of the PML robot is a result of correcting several earlier designs that failed to meet the desire
50. the VB6 robot OS that acts as a software driver for the camera After the image is taken it is posted to a directory that is shared between the laptop server and the desktop client The image is now available to the client for processing The third step is the processing of the image This step can be broken into two parts processing the image to get the image position and the data extraction from the image position The image processing part is responsible for determining the fractional 22 position of the mark s location in the image in relationship to the entire image As seen in Figure 3 4 the cross is closer to the top and right of the image frame oo 70 Y image 50 X image 50 Y image 60 X image X axis Y axis 0 0 Figure 3 4 Webcam picture of optical target If the image were exactly in the center its position would be described as being at 50 of the X axis and 50 of the Y axis The 0 0 location is the lower left of the image frame If the robot were in the position indicated above the cross s position on the screen would be defined as being 60 of the X axis and 70 of the Y axis In addition to the X and Y percentage numbers the rotation of the mark is also measured in degrees These three numbers are the outcome of the image processing step 23 The current development of the PML robot does not have an automated image processing capability Instead this is curr
51. the color bar to yellow on the Client End If With PicWebCam Visible True End With With picStore Visible False End With Call SendMessage hCap WM CLOSE 0 0 hCap is defined in the Cam module hCap capCreateCaptureWindow Take a Camera Shot WS CHILD Or WS VISIBLE 0 0 PicWebCam Width PicWebCam Height PicWebCam hWnd 0 If hCap lt gt 0 Then Call SendMessage hCap WM_CAP_DRIVER_CONNECT 0 0 Call SendMessage hCap WM_CAP_SET_PREVIEWRATE 66 0 amp Call SendMessage hCap WM_CAP_SET_PREVIEW CLng True 0 amp End If If flagCamera True Then waitTime 3000 wait 3 seconds Ist shot allows camera to set light levels flagCamera False Else waitTime 500 shorter wait after 1st shot End If On Error Resume Next Call CamToBMP picStore Kill App Path amp C Documents and Settings All Users Documents robot_lap temp bmp Call SavePicture picStore Picture C Documents and Settings All Users Documents robot_lap temp bmp 158 Call SavePicture Picture1 Picture C Documents and Settings All Users Documents robot_lap temp bmp waitTime 500 wait 1 second Call SendMessage hCap WM_CLOSE 0 0 Dim temp As Long temp SendMessage hCap WM_CAP_DRIVER_DISCONNECT 0 amp 0 amp tcpServer SendData New Pic waitTime 500 wait 1 second With PicWebCam Visible False End With With picStore Visible True End With End Sub RunCamera XXXXXXXXXXXXXXXXXXX
52. the robot to follow a predetermined path is dependent upon both the sensing of its position and the accurate control of the wheel speeds The specifics involving path following feedback control will be discussed in Chapter VII The motor control system is divided into the following three parts IFB Voltage Controller Input e Joystick Output JSIC Input MC Gear Motor Output of result These three parts will be discussed in detail in the following sections 4 2 IFB Voltage Controller Figure 4 2 illustrates the basic function of the IFB voltage control module It is designed to accept two 4 bit logic level inputs and output two voltage values Tweet bee IFB Voltage Control Te channels voltages Figure 4 2 Basic function of IFB voltage controller The IFB voltage control works by using a resistance based voltage divider Figure 4 3 expands on Figure 4 2 by showing the method used to produce the voltage output from the 4 bit inputs Each of the blocks in Figure 4 3 is capable of an n bit number of 38 resistance values For example the 1 bit block has two 2 resistance values while the 3 bit block has eight 2 resistance values The total number of possible speeds with this 4 bit system would be 16 however as indicated in the Figure 4 3 the most significant bit MSB has been reserved to control the run and stop commands of the robot This leaves 3 bits to control the forward speeds T
53. there is a desire to simplify the video signal algorithm A related paper by Martins and Alves presented a technique for using a field programmable gate array FPGA chip in order to deal with some of the front end processing requirements 8 One of the goals of the PML robot is to make a relatively simple and cheap video capture method and then send the pre processed image wirelessly to a remote computer for final processing The idea of Martins and Alves could be applied to the PML robot in that a FPGA chipset would be coupled with a camera that would help to minimize and simplify the data image sent to the main computer Specifically the FPGA would be responsible for altering the contrast of the image the pixel depth and other basic arithmetic operations involved with signal processing The PML robot s remote computer would use a Matlab Simulink toolbox such as the Video and Image Processing Blockset to further process the simplified data steam 9 2 2 2 Distributed architecture control systems One area of potential development that would benefit the PML robot would be the use of a distributed architecture control system The lowest layer would be responsible for path planning and motion stability The level above would have higher level responsibilities that would perhaps involve collision avoidance or mapping capabilities Sahin and Guvenc discuss a distributed architecture for a household robot that uses a common communication networ
54. this would have allowed the original wheelchair designer to use different polarities for some reason not discovered by this author 4 4 2 Gear motor description The two gear motors are the original equipment that came with the wheelchair A thorough search was not able to discover any specific information about them What is known is that each has a label that says Invacare 3 3 amp minimum speed 110 rpm The original motor was powered by a 24 V PWM supply It has been observed that the reduction gear has a 32 1 ratio This fact is important in calculating the rotation distance of the wheels as will be discussed in the next chapter Experiments were conducted using both a 12 and 24 V input to test the motor response speed It was determined that there was no significant difference between the two voltages under normal loading conditions The lack of information about the gear motor has not presented significant difficulties to the design or operation of the PML robot This is mostly due to the fact that both the gear motors and the MC 7 controllers are operated well below their maximum capability 4 5 Summary of Motor Control System The choice of the MC 7 as the replacement motor control has proven successful The uses of the PML robot as a laboratory controls instrument require a robust motor 51 controller The MC 7 is less susceptible to damage from errant voltages than was the original controller If this controller is damaged its
55. time2 0 46 ActDist 2 ActDeg 0 timel 0 47 inside HallsRead time2 0 47 ActDist 2 5 ActDeg 1 Start HallsCompute 10 OffsetX 8 726195E 02 OffError 1 DegRef 0 DegError 1 SpecTurn 0 2 LTum 0 RTurn 2 Motlnc 6 Left Ratio 0 Right Ratio 0 3333333 Tum 1 OldTurn 0 187 End HallsCompute MOTCONT Inside MotCont Turn 1 OldTurn 0 timel 0 49 inside HallsRead time2 0 49 ActDist 3 ActDeg 0 Start HallsCompute 10 OffsetX 8 726195E 02 OffError 1 DegRef 0 DegError 0 SpecTurn 0 LTum 2 RTurn 2 Motlnc 7 Left Ratio 0 2857143 Right Ratio 0 2857143 Turn 0 OldTurn 1 End HallsCompute MOTCONT Inside MotCont Turn 0 OldTurn 1 timel 0 51 inside HallsRead time2 0 51 ActDist 3 ActDeg 0 timel 0 53 inside HallsRead time2 0 53 ActDist 3 ActDeg 0 timel 0 53 inside HallsRead time2 0 53 ActDist 3 ActDeg 0 timel 0 55 inside HallsRead time2 0 55 ActDist 3 ActDeg 0 timel 0 56 inside HallsRead time2 0 56 ActDist 3 ActDeg 0 timel 0 57 inside HallsRead time2 0 57 ActDist 3 ActDeg 0 timel 0 58 inside HallsRead time2 0 58 ActDist 4 ActDeg 0 Start HallsCompute 10 OffsetX 8 726195E 02 OffError 1 DegRef 0 DegError 0 SpecTurn 0 LTum 2 RTurn 2 Motlnc 8 Left Ratio 0 25 Right Ratio 0 25 Turn 0 OldTurn 0 End HallsCompute timel 0 58 inside HallsRead time2 0 58 ActDist 4 ActDeg 0 timel 0 6 inside HallsRead time2 0 6 ActDist 4 ActDeg 0 timel 0 62 inside HallsRead
56. to its operation although some of the code resides on the robot s laptop computer The ROS is responsible for stopping the robot s movement at the end of the dead reckoning path where it is assumed that there is a positional error that needs to be corrected The ROS then requests the OCS for correction data The robot remains stopped until it receives the correction data from the optical system and is able to correct its positional errors Note that the ROS is responsible for maintaining the wireless link between the robot s server computer and the remote client computer 64 Optical Correction System Capture image ROS requests for webcam on robot correction data Send image ROS from Server to Client Stops robot gt Waits for data b ring Controls wireless Return corrected data OCS returns from Client to Server correction data lt Figure 6 1 Functional diagram of OCS 6 1 1 Camera hardware Figure 6 2 shows the front view of the mounted webcam This camera is a Lego s Mindstorm camera and came with a driver used for simple experiments in video processing and tracking The actual camera inside the Lego s package is a Logitech QuickCam Web camera The existing Lego s camera drivers were inadequate and had to be replaced by a custom VB 6 program designed for the PML robot s requirements As shown in the figure the webcam is attached to a hand formed bracket composed of a 1 in aluminum bar The camera is mounte
57. 0 timel 0 08 inside HallsRead time2 0 08 ActDist 0 ActDeg 0 timel 0 1 inside HallsRead time2 0 1 ActDist 0 ActDeg 0 timel 0 1 inside HallsRead time2 0 1 ActDist 0 ActDeg 0 timel 0 1 inside HallsRead time2 0 1 ActDist 0 ActDeg 0 timel 0 12 inside HallsRead time2 0 12 ActDist 0 ActDeg 0 timel 0 12 inside HallsRead time2 0 12 ActDist 0 ActDeg 0 timel 0 12 inside HallsRead time2 0 12 ActDist 0 ActDeg 0 timel 0 14 inside HallsRead time2 0 14 ActDist 0 ActDeg 0 timel 0 16 inside HallsRead time2 0 16 ActDist 0 ActDeg 0 timel 0 17 inside HallsRead 182 183 time2 0 17 ActDist 0 ActDeg 0 timel 0 17 inside HallsRead time2 0 17 ActDist 0 5 ActDeg 1 Start HallsCompute 10 OffsetX 8 726195E 02 OffError 1 DegRef 0 DegError 1 SpecTurn 0 2 LTum 0 RTum 0 MotInc 2 Left Ratio 0 RightRatio 0 Turn 1 OldTurn 0 End HallsCompute MOTCONT Inside MotCont Turn 1 OldTurn 0 timel 0 19 inside HallsRead time2 0 19 ActDist 0 5 ActDeg 1 timel 0 21 inside HallsRead time2 0 21 ActDist 0 5 ActDeg 1 timel 0 21 inside HallsRead time2 0 21 ActDist 0 5 ActDeg 1 timel 0 23 inside HallsRead time2 0 23 ActDist 0 5 ActDeg 1 timel 0 24 inside HallsRead time2 0 24 ActDist 0 5 ActDeg 1 timel 0 25 inside HallsRead time2 0 25 ActDist 0 5 ActDeg 1 timel 0 25 inside HallsRead time2 0 25 ActDist 0 5 ActDeg 1 timel 0 26 inside HallsRead time2 0 26 ActDist 0 5 ActDeg 1 timel 0 2
58. 000 The specifics involving the actual values of the parameters will be discussed later in this chapter and in the sections that cover the SP and LA controllers 7 3 Program Design The purpose of this section is to discuss the organization of the main program from the point of view of the VB 6 code design This section is divided into three parts The first part covers the main program while the next two parts discuss the subprograms and the program modules Figure 7 6 shows the basic design of the main program and lists the five separate code modules that are available to the main program In this figure a specific reference to the subprograms is not shown 102 MAIN PROGRAM Cam Module Responsible for driving the webcam Copies Dimension Variables Load Form the image to a common directory PCMData Defines variable arrays and global constants used by the DIO device Manual Operation Remote Operation Gat Q lt PCMDrvFun Defines 40 functions used by the DIO device User Controled PCMMain User written function that controls program start up PCMUserFunc Written by the user to access DIO read and write commands Figure 7 6 Basic design of main program with modules 7 3 1 Main program The basic design of the main loop program structure has both a manual operation and a remote operation part The manual loop is used to test the parameter values of a s
59. 3 4 Robot Operating System OvervieW ssesssesseersssessserssrrsseeeseee 32 IV MOTOR CONTROLS Y STEM 0 seeren renere renerne senere rene 36 ArT Introduction scssi neue Ge avad een ee 36 4 2 IFB Voltage COM Olle tian eis i ed aeesueysandaie seayundedind shes tateSoawtensauss 37 4 2 1 Voltage controller output range eeeeeceeeeeeeseeeeeteeeeees 39 4 2 2 Voltage controller system design 1 u u u u sseseereeerrrerrreree 40 4 2 3 Voltage controller circuit design u u sseceereererrerrrr ere 42 4 2 4 Voltage controller SUMMATY ceeeceeeeeceeeteeeeeteeeeneeeenes 44 4 3 Joystick and JSIC Circuit cs Sia yeca sed snae sac dacodessgsesvecaseceecasveasecousse 45 AA NIGH 7 aid Gear Motoi ressesie a 47 4 4 1 MC 7 motor controller cece eeeceeeeececeeececeeeeeeeeeeeseeeeenes 48 4 4 2 Gear motor description 0 0 eeeeeeceeeeceseeecseeeeceeeeeesaeeeenes 50 4 5 Summary of Motor Control System c cc eeecceeeseeeesseeeesteeeenes 50 V POSITION SENSING AND 5 VDC POWER SUPPLY 008 53 SE tro UC HON ass a a gcdhcs ius ta lescn dias 53 5 Five VDC Power Suppl ys aen o a a EA A 53 5 3 Four Bit Position Counter Buffer System eee eeeeceeeeteeeeeee 55 5 3 1 Minimum sampling frequency for position count 57 5 3 2 Position sensor diagram sessssseesssesssssereseressseesseesseess 58 5 3 3 Application of position buffer GW u u u senveereererrrersereerenee 59 5 3 4 S
60. 36 50 63 12 82 90 97 03 31 45 59 123 125 129 131 134 135 136 138 99 76 77 It should be noted that for the entire development of the mapping algorithm the image was kept at the same arbitrary size of 154 mm in length and 127 mm in height At this image size the distance from the bottom edge of the image to the camera radius point was measured at 200 mm At this point in the investigation of the mapping algorithm the final strategy used to convert image data to real data was conceived Webcam image of grid c 127 mm R Jb Radius L mee a 0 0 Local Cartesian i origin y H y Camera radial angle 200 mm Image radius and ee measurements are id FA Notto scale i i i ie Common origin point for both a global Cartesian field and i the image and real i polar coordinate system f 0 0 Figure 6 7 Development of polar coordinate system If the coordinate field could be defined as the image field of the camera it would be likely that the data conversion would be less mathematically demanding The primary benefit of this strategy is that the polar image field and the polar real field are in one to one correspondence with a common origin The virtual point of origin for the image field coincides with the physical origin or physical point just under the plumb bob The 78 result of this is that when the real theta Op and the real radius Rp are known the real position of the
61. 8 inside HallsRead time2 0 28 ActDist 0 5 ActDeg 1 timel 0 28 inside HallsRead time2 0 28 ActDist 1 ActDeg 0 Start HallsCompute 10 OffsetX 8 726195E 02 OffError 1 DegRef 0 DegError 0 SpecTurn 0 LTum 0 RTurn 2 Motlnc 3 Left Ratio 0 Right Ratio 0 6666667 Turn 0 OldTurn 1 End HallsCompute MOTCONT Inside MotCont Turn 0 OldTurn 1 184 185 timel 0 31 inside HallsRead time2 0 31 ActDist 1 ActDeg 0 timel 0 32 inside HallsRead time2 0 32 ActDist 1 ActDeg 0 timel 0 33 inside HallsRead time2 0 33 ActDist 1 ActDeg 0 timel 0 35 inside HallsRead time2 0 35 ActDist 1 ActDeg 0 timel 0 37 inside HallsRead time2 0 37 ActDist 1 5 ActDeg 1 Start HallsCompute 10 OffsetX 0 1745239 OffError 1 DegRef 0 DegError 1 SpecTurn 0 2 LTum 0 RTum 2 Motlnc 4 Left Ratio 0 Right Ratio 0 5 Turn 0 OldTurn 0 timel 0 38 inside HallsRead time2 0 38 ActDist 1 5 ActDeg 1 Start HallsCompute 10 OffsetX 0 1745239 OffError 1 DegRef 0 DegError 0 186 SpecTurn 0 LTum 0 RTurn 2 MotInc 5 Left Ratio 0 RightRatio 0 4 Turn 0 OldTurn 0 timel 0 39 inside HallsRead time2 0 39 ActDist 2 ActDeg 0 timel 0 4 inside HallsRead time2 0 4 ActDist 2 ActDeg 0 timel 0 41 inside HallsRead time2 0 41 ActDist 2 ActDeg 0 timel 0 42 inside HallsRead time2 0 42 ActDist 2 ActDeg 0 timel 0 44 inside HallsRead time2 0 44 ActDist 2 ActDeg 0 timel 0 46 inside HallsRead
62. 9 inside HallsRead time2 4 69 ActDist 58 ActDeg 82 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 69 inside HallsRead time2 4 69 ActDist 58 5 ActDeg 81 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 72 inside HallsRead time2 4 72 ActDist 58 5 ActDeg 81 Start HallsCompute PRatio Block incPC 3 incPC2 4 Test PTime 4 72 DegHistory incPC 81 PRatio 81 DegHistory incPC2 0 pratio TurnDelta 3 End HallsCompute timel 4 73 inside HallsRead 200 time2 4 73 ActDist 59 ActDeg 80 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 74 inside HallsRead time2 4 74 ActDist 59 ActDeg 80 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 75 inside HallsRead time2 4 75 ActDist 59 5 ActDeg 81 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 76 inside HallsRead time2 4 76 ActDist 60 ActDeg 80 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 78 inside HallsRead time2 4 78 ActDist 60 ActDeg 80 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 79 inside HallsRead time2 4 79 ActDist 60 5 ActDeg 79 Start HallsCompute TurnDelta 3 End HallsCompute 201 timel 4 8 inside HallsRead time2 4 8 ActDist 60 5 ActDeg 79 Start HallsCompute TurnDelta 3 End HallsCompute timel 4 81 inside HallsRead time2 4 81 ActDist 61 ActDeg 80 Start HallsCompute Tu
63. 91 ActDist 63 5 ActDeg 77 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 91 inside HallsRead time2 4 91 ActDist 63 5 ActDeg 77 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 92 inside HallsRead time2 4 92 ActDist 64 ActDeg 76 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 93 inside HallsRead time2 4 93 ActDist 64 ActDeg 76 Start HallsCompute PRatio Block incPC 0 incPC2 1 Test PTime 4 93 DegHistory incPC 76 204 PRatio 7 DegHistory incPC2 83 pratio MOTCONT Inside MotCont Turn 1 OldTurn 1 Right_R 7 Right_L 3 Left_R 3 Left L 7 end motcont TurnDelta 4 End HallsCompute timel 4 97 inside HallsRead time2 4 97 ActDist 65 5 ActDeg 75 Start HallsCompute TurnDelta 4 End HallsCompute timel 4 98 inside HallsRead time2 4 98 ActDist 65 5 ActDeg 75 Start HallsCompute TurnDelta 4 End HallsCompute timel 4 99 inside HallsRead time2 4 99 ActDist 65 5 ActDeg 75 Start HallsCompute TurnDelta 4 End HallsCompute timel 4 99 inside HallsRead time2 4 99 ActDist 66 ActDeg 74 Start HallsCompute TurnDelta 4 End HallsCompute Name Address Email Address Education 205 VITA Adam G Rogers Department of Mechanical Engineering 3123 TAMU College Station TX 77843 3123 ar44815 neo tamu edu B S Engineering Technology Texas State University Sa
64. Bar BackColor vbGreen Exit Sub error End Sub Private Sub cmdSendData_Click On Error GoTo error frmColorBar BackColor vbBlue Excl_X R amp CStr 61 amp C amp CStr 2 Excl_X processed X output With txtX This is the format for reading a cell in an Excel sheet LinkMode vbLinkNone LinkTopic ExcelLoc LinkItem Excl_X don t use paraenthesis around the String cell LinkMode vbLinkAutomatic End With XError txtX Text Print 1 Tab 12 XError XError 174 Excl Y R amp CStr 63 amp C amp CStr 3 Excl_Y processed Y output With txtY This is the format for reading a cell in an Excel sheet LinkMode vbLinkNone LinkTopic ExcelLoc LinkItem Excl_Y don t use paraenthesis around the String cell LinkMode vbLinkAutomatic End With YError txtY Text Value in Excel C62 Print 1 Tab 12 YError YError Excl_A R amp CStr 50 amp C amp CStr 7 Excl_A direct measure rotation angle With txtA This is the format for reading a cell in an Excel sheet LinkMode vbLinkNone LinkTopic ExcelLoc LinkItem Excl A don t use paraenthesis around the String cell LinkMode vbLinkAutomatic End With AError txtA Text 2 txtA Text AError Print 1 Tab 12 AError AError Print 1 Tab 12 Basic A Basic_A Basic X Basic X Basic Y Basic Y 175 Excl Cor R amp CStr 65 amp C amp CStr 3 Wit
65. CorrectIndex 2 increment for the next command block EndIf first index block for the CorrectCompute subroutine INDEX 2 If CorrectIndex 2 Then Print 1 start of correct compute blk 2 Call HallsCompute END DISTANCE REACHED If ActDist gt Distance Then Print 1 Inside CorrectCompute End Distance blnRun False exits the Blue Loop OldTurn Turn AvoidMotCont block upon return to Case Blue Call Neutral stop the robot waitTime 500 wait for full stop then measure position Call HallsRead read the position CorCarryDeg ActDeg load the final position angle Call tcpServer SendData white changes the color bar on Client to white at end of Blue run timeCarry time2 use to update the time constant for multiple runs in Case Blue End If Print 1 end of correct compute blk 2 End If End Sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX RunCamera_ XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub RunCamera Print 1 Print 1 Tab 12 Inside run cam ActDist ActDist Print 1 157 Call tepServer SendData yellow in case EMO then keep color bar red If frmServer BackColor lt gt vbRed Then frmServer BackColor vbYellow Allows for
66. I control inside of HallsCompute sub mostly Dim PControlInitial As Boolean use for first trip through P Controller Dim FlgPDirection As Integer value 1 right left indicator Dim DegHistory 4 As Integer use to store the 5 previous ActDeg values 136 Dim PTime As Single time of last write to DegHistory Dim incPC As Byte increment P Control DegHistory array Dim incPC2 As Byte use to extract oldest data from DegHistory array Dim PRatio As Single diff in ActDeg 5 ActDeg 0 over 1 2 sec i EMO VARIABLES Dim EmoTime As Single __ this will be in a calculation w timel Dim EmoDist As Single this will be in a calculation w ActDist RunCamera sub Dim flagCamera As Boolean Dim NeutralMC As Byte This works for both Right and Left MCs Dim CorrectIndex As Byte Dim CorX As Long Dim CorY As Long Dim CorAlpha As Long Dim CorCarryDeg As Integer Dim flagDB1 As Integer Dim flagDB2 As Integer XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX FORM LOAD UNLOAD XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Private Sub Form_Load tcpServer LocalPort 10101 tcpServer Listen put all the initial values here Close 1 Otherwise there is an error when the doc is on the desktop from the previous instance Close 2 use for graph purposes Close 3 use for Hall skip error 137 Call Initialize
67. OR CONTROL SYSTEM 4 1 Introduction Chapter III gave a brief overview of the motor control system its relationship to the IFB and how the 12 V power is supplied to the MCs The purpose of this chapter is to give a complete description of the motor control system Figure 4 1 is a functional block diagram of the motor control system and shows the relationships between the inputs and the outputs The control input signals are either two 4 bit DIO command voltages from the laptop or two channels of variable resistances from a standard joystick The outputs of the motor control system are the two independent speeds of the wheels Motor Control System ee ce ee ERE Two 4 bit channels IFB 5 V IFB VO I power supply voltage control Control Inputs Q Q eee MC left o 2 output voltages D O O BEDT 1 to 3 Volts O switch O MC right Joystick Se 12 V PWM y A o VE Two variable resistance ee ee pe J channels Wheel output speed Figure 4 1 Block diagram and basic description of motor control system 37 Because the PML robot motor control system is made up of two completely independent motor controls robot motion control is similar to a bulldozer In this case turns are initiated by speeding up or slowing down one wheel when compared to the other The ability of
68. Online Available http www developerfusion co uk forums threads 150967 The Mathworks Inc 2007 Sep Signal Processing Toolbox 6 Online Available http www mathworks com access helpdesk help pdf_doc signal signal_tb pdf The Mathworks Inc 2007 Sep Image Processing Toolbox 6 Online Available http www mathworks com access helpdesk help pdf_doc images images_tb pdf SuperLogics Inc 2003 Mar PCMDIO 24 Channel Digital Input Output Users Manual Online Available http www superlogics com Netgear Inc 2004 May User Manual for the NETGEAR 54 Mbps Wireless USB 2 0 Adapter WG11 Version 2 0 Online Available http kbserver netgear com downloads_support asp Netgear Inc 2004 May User Manual for the NETGEAR 54 Mbps Wireless PC Card WG511V2 Online Available http kbserver netgear com downloads _support asp 133 APPENDIX A SERVER SIDE PROGRAM This program is found on the PML s computer and functions as the principle code for the operation of the robot All of the code presented here is in the Visual Basic 6 programming language The conventions of this programming language use a modular programming structure An outline of these modules is presented below 1 frmMC cam3 frm This form is the primary program used by the robot as its operating system For a complete explanation of this program see Chapter VII This code can be found in Appendix A 1 2 Camera bas This module desc
69. PRECISION MECHATRONICS LAB ROBOT DEVELOPMENT A Thesis by ADAM G ROGERS Submitted to the Office of Graduate Studies of Texas A amp M University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE December 2007 Major Subject Mechanical Engineering PRECISION MECHATRONICS LAB ROBOT DEVELOPMENT A Thesis by ADAM G ROGERS Submitted to the Office of Graduate Studies of Texas A amp M University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Approved by Chair of Committee Won jong Kim Committee Members Yoonsuck Choe Daejong Kim Head of Department Dennis O Neal December 2007 Major Subject Mechanical Engineering iii ABSTRACT Precision Mechatronics Lab Robot Development December 2007 Adam G Rogers B S Southwest Texas University Chair of Advisory Committee Dr Won jong Kim This thesis presents the results from a modification of a previously existing research project titled the Intelligent Pothole Repair Vehicle IPRV The direction of the research in this thesis was changed toward the development of an industrially based mobile robot The principal goal of this work was the demonstration of the Precision Mechatronics Lab PML robot This robot should be capable of traversing any known distance while maintaining a minimal position error An optical correction capability has been added with the addition of a webcam and the appropriate image p
70. Print 1 end motcont End If if PControl True ur CONTROL BLOCK use for both P and I control If Turn 1 Then Call Right_Turn Else If Turn 0 Then Call No Turn straightens out a previous turn Else Call Left_Turn End If End If OldTurn Turn everytime especially when too fast on change End Sub MotCont sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub Right Turn RTurn RTurn 1 increment the motor counter for Turn Block Write slower speed higher number to the right MC block intStatus singleDigitalOutput gintlogicaldevice MC_R Right_R This writes the 4 bit value to the right MC latch If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If Write a faster speed lower number to the left MC block intStatus singleDigitalOutput gintlogicaldevice MC_L Right_L This writes the 4 bit value to the left MC latch If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If Pulse both right and left latches Call PulseMC Write DIO to both latches and then Clock the speed to the MC 162 End Sub Right Turn XXXXXXXXXXXXX XX XXXX XXX XXX XXX XXX XXX XXX KK KK KK ESS SEES DEER LEFT TURN XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub Left Turn LTurn LTurn 1 increment the motor counter for Turn Block
71. TER Page VII CONCLUSIONS AND SUMMARY GGsm sssueeeeeeeeeeeeeeesee enes ese eee eee enenenee 126 8 1 Specific Accomplishments W G u ssesesereree rer eeen reen ener ne renere en rnee 127 8 2 Limitations and Future Work Xu u u ssseneerevevev evnen er enn eernenee 128 9 3 CONCIUSIONS A E E E E EAEE Ene RER 128 REFERENGES Sia NNE NANA De NER no anaa aan unli 130 APP PND A n A a a aa a t 133 APPENDIX Bacco Tends ee e e n o eaa tek e eaa ses ea ae aaa Toas 171 aN d A AI D DE OEE E E S 181 FIGURE 1 1 3 1 3 2 3 3 3 4 3 5 3 6 3 7 3 8 3 9 4 1 4 2 4 3 4 4 4 5 4 6 4 7 4 8 4 9 4 10 4 11 5 1 5 2 5 3 5 4 LIST OF FIGURES Front view of PML robot ssssseseesessesesessessessrerrreseesresreeseesseseesseeseesereseses Single segment of dead reckoning path eeeeeeeeeeesseeeeeteeeesteeeenaeeees Printed cross used as an optical target Gu u u sssnseeereeereserenrene ren rrenenee Top view of robot in relation to optical mark eeeseseeeseeeeesseresresssereeresss Webcam picture of optical target 1 2 32 4 426 ead GA a eet dns Screen image with X and Y and error result eee eeseeeeeeeseeeeeenaeeees Course correction path description ceecceeseeeceseeeceeeeecseeeeseseeeseeeeenaeeees System diagram of the 12 VDC supply ececeeeececsscesseeeeesteeesneeeenaeeees Hardware system overview diagram ceesccceeeeeceeececseeeeceeeesseeeenseeees Operating system overview CiagQr
72. The ground to the motor controller is the common ground or case ground for the robot Motor Positive Left MC T p Right M Motor Negative z Battery Common T3 C7 12V o Voltage 10 A U fuze Disconnected 22 uF Capacitor Lo oHe elo oh Kill Switch ol o LM7805 ol 2LED Battery Indicators Negative Controller input from IFB or JSIC Three heat sinks and power supplies 5 VDC Figure 4 11 Modular diagram of the MC 7 and its connections 50 It should be noted that the right MC 7 is configured to go in the forward direction T4 connector while the left MC 7 is configured to go in the reverse direction T5 connector The reason for this difference in configuration is probably caused by the manufacturer s original design of the two electric gear motors During the installation of the MC 7 motor controllers and the connection to the gear motors the motor positive outputs were connected to the red or positive lead on each of the motors After the installation was completed it was discovered that the motors went in opposite directions In order to get both motors to turn in the same direction the polarity of the right motor controller is opposite from the left motor controller It should be noted that the right and left motors are not interchangeable
73. This controller is responsible for accurately following the desired RefAng correction response The degree error DegError variable defines the angular difference between the robot s actual heading and the desired heading determined by the one to one map in Figure 7 8 The DegError variable is computed by using 7 2 DegError ActDeg RefAng 7 2 Since the SP controller is able to control the heading angle to a level of 2 The small angle controller will begin to operate when the degree of error is less than 3 The effect of the small angle controller is to hold the robot to the desired angle of the RefAng output Chapter V discussed how the ROS had a variable cycle rate and how this caused problems with the Hall position counters Under normal conditions the ROS will cycle every 30 ms and is able to issue a new motor command ever time it cycles The small angle controller works by varying the fraction of turn commands in a proportional manner For example when the of error in the heading is equal to 2 a left turn is required to correct the path to 0 When the MC 7s are commanded into a left turn the result is a 112 slight overshoot and the beginning of the oscillation behavior The small angle controller is able to smooth the control response by commanding a fractional left turn With a 2 error input the small angle controller will command 60 of the next 10 turn commands to be left turns and 40 of the next 10 turn
74. VDC supply and was never designed to be adjusted or modified This motor controller was not a good fit for a robot wheelchair intended to be used in a laboratory and research setting The IPRV command circuitry was directly connected to the Ranger II motor controller and twice transient voltages burned out the motor controller One of these accidents happened during Homji s research and another motor controller was burned out more recently by the current author Because of the expensive nature of medical equipment such as wheelchairs the replacement cost for this identical controller is 2000 The modified command inputs to the motor controller could not be adequately isolated from the circuitry of the motor controller and the replacement of the existing motor controller would not solve the basic problem Both for cost reasons and a desire to make the motor control system more robust the existing motor controller was replaced by a hobby type pulse width modulation PWM motor controller One advantage in the use of this particular brand of motor controller is the wide voltage range of the PWM output which ranges from12 to 36 VDC Because the wheelchair was designed to potentially carry the weight of a heavy person in an environment that could have ramps or other elevations it had a surplus of torque and power that are not required for its current application as a laboratory robot Experiments indicated that the response of the gear motors whe
75. X want to accumulate Offset all the time OffError Int 10 OffsetX OffsetRef a OffError means a left turn for correction OffsetRef comes from Excel table Print 1 10 OffsetX 10 OffsetX OffError OffError txtOffVal Text Int OffsetX 10 output live to form txtOffErr Text OffError offset error 149 Determine the abs OffError If OffError lt 0 Then DegRefFlg True if true then RefDeg a negative Else DegRefFlg False End If absOffError Abs OffError because of the equation outside of bounds If absOffError gt 100 Then it may go negative Keep it within the absOffError 100 range End If CONTROL LAW This block has a 3rd order function used to relate the OSE to the Reference degree The function comes from a 3rd order curve fit based on desired points for response The output of this block is the Degree Error based on the desired response that will eliminate the OSE DEGREE REFERENCE UPPER LOWER BOUND 3E 05x3 0 0017x2 0 1128x 0 7021 DegRef Round 0 00003 absOffError 3 0 0017 absOffError 2 0 1128 absOffError 0 7021 If DegRef gt 20 Then DegRef 20 End If lower end cutoff If absOffError lt 5 Then DegRef 0 End If MIRROR DegRef IF OffError WAS NEGATIVE If De
76. XXXXXXX 166 intStatus singleDigitalOutput gintlogicaldevice Count CLR 0 Sets the Z clear value to high voltage If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If Call waitTime 1 creates a pulse of 1 milli second intStatus singleDigitalOutput gintlogicaldevice Count_CLR 1 resets the Z clear to low voltage If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX RESETS THE RIPPLE COUNT AFTER EACH LINE OF THE PATH TABLE XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX RRip 0 LRip 0 PrePulseR False PrePulseL False timeInt Timer _ this resets the time value everytime Zero is called End Sub END Sub Zero XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX This is used to write neutral values to the MC so it won t move XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub Neutral intStatus singleDigitalOutput gintlogicaldevice MC R NeutralMC writes neutral speed If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If intStatus singleDigitalOutput gintlogicaldevice MC_L NeutralMC to left MC If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus 167 Call PulseMC a Clocks the DIO write to the Latches and then the MC Print 1 Inside Sub Neutral End Sub Ends the NEUTRAL sub ae eee eee Creates a lms p
77. XXXXXXXXXX Public Sub Initialize Open C Documents and Settings Supreme Overlord Desktop excel txt For Output As 1 Open C Documents and Settings All Users Documents robot_lap primary txt For Output As 1 Open C Documents and Settings All Users Documents robot_lap graph txt For Output As 2 Open C Documents and Settings All Users Documents robot_lap BadHall txt For Output As 3 Print 2 timel Tab 9 ActDeg Tab 17 OffError Tab 27 Turn Tab 33 TurnDelta Print 3 This data is collected to troubleshoot the Hall skip problem nanan nnnnnn anna nn anna VALUE ASSIGNMENTS SpecTurn 0 this will allow for NoTurn at stact up MotlInc 1 must not 0 for divide error LTurn 0 poth values used in TURN BLOCK RTurn 0 i flagCamera True set true for longer initial wait time blnRun True Allows while loop to function NewLine True used to show new path table line 164 NeutralMC 15 value used for no movement Indx 1 Reads the first line of table intStatus 0 used for dummy value for pcmdio error calls OldDist 0 used for first distance in Direction sub ActDeg 0 used for Ist loop OffsetX 0 used for Direction sub no initial error CorrectIndex 1 found in CorrectCompute PControl False allow for I control by default DirAngleOld 0 use the 2 DirAngles for Case Blue runs w o Case Green DirAngle 0 this will be re
78. XXXXXXXXXXXXXXXXXXXXXXXXXXXXX StreamCam timel gt CamTime 2 is the condition for this sub to work XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub StreamCam Print 1 Inside StreamCam timel time1 CamTime timel update interval time With PicWebCam Visible True End With With picStore Visible False End With 159 Call SendMessage hCap WM CLOSE 0 0 hCap is defined in the Cam module hCap capCreateCaptureWindow Take a Camera Shot WS CHILD Or WS VISIBLE 0 0 PicWebCam Width PicWebCam Height PicWebCam hWnd 0 If hCap lt gt 0 Then Call SendMessage hCap WM CAP DRIVER CONNECT 0 0 Call SendMessage hCap WM CAP SET PREVIEWRATE 66 0 amp Call SendMessage hCap WM CAP SET PREVIEW CLng True 0 amp End If If flagCamera True Then waitTime 3000 wait 3 seconds 1st shot allows camera to set light levels flagCamera False Else waitTime 500 shorter wait after Ist shot End If Call CamToBMP picStore Xxx Call HallsRead Print inside stream cam just below HallsRead ActDist ActDist Kill App Path amp C Documents and Settings All Users Documents robot_lap temp bmp Call tecpServer SendData picture sent as string to code subsequent data Call tepServer SendData picStore Picture Call SavePicture picStore Picture C Documents and Settings All Users Docu
79. a turn is implemented l INCREMENTS L R TURN WHEN TURN IS CONSTANT If Turn OldTurn Then If Turn 1 Then RTurn RTurn 1 End If If Turn 1 Then LTurn LTurn 1 End If Print 1 LTurn LTurn RTurn RTurn MotInc MotInc Print 1 Left Ratio LTurn MotInc Right Ratio RTurn MotInc SECONDARY PWM FOR I CONTROL BLOCK Turn 0 if specturn 0 or if turn ratio is under limit If SpecTurn lt 0 Then If LTurn MotInc lt Abs SpecTurn Then Turn 1 End If 151 End If If SpecTurn gt 0 Then If RTurn MotInc lt SpecTurn Then Turn 1 End If End If Print 1 Turn Turn OldTurn OldTurn Print 1 s Ene If If PControl False Output of this block is Turn value END OF I CONTROL BLOCK XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX OR DIVIDE THE HALLS COMPUTE SUB XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX START OF P CONTROL BLOCK If PControl True Then primary input to P Control ActDeg DegError ActDeg a error is an error to the right for P Control since RefDeg 0 this is true INITIAL CONDITION BLOCK If PControlInitial True Then use for only 1 trip through at start Pas False reset after first t
80. ables was input and both right and left wheel speeds were measured Table 4 3 Summarized motor control results LEFT SIDE MOTOR CONTROLER DIO Binary Resistance Voltage Wheel Speed Equivalent V tap to at Viap Speed Variable Ground Q V RPM 15 1111 870 0 57 0 7 0111 2850 1 46 8 6 0110 3610 1 73 16 5 0101 4360 1 95 22 5 4 0100 5110 2 15 28 6 3 0011 5850 2 30 33 3 2 0010 6610 2 46 36 1 1 0001 7360 2 60 42 9 0 0000 8120 2 71 46 5 RIGHT SIDE MOTOR CONTROLER DIO Binary Resistance Voltage Wheel Speed Equivalent Viap to at Viap Speed Variable Ground Q V RPM 15 1111 816 0 56 0 7 0111 2800 1 49 8 6 0110 3570 1 79 15 5 5 0101 4320 2 00 21 6 4 0100 5080 2 19 27 1 3 0011 5810 2 34 31 7 2 0010 6580 2 50 36 1 1 0001 7330 2 63 40 0 0 0000 8090 2 76 43 5 52 53 CHAPTER V POSITION SENSING AND 5 VDC POWER SUPPLY 5 1 Introduction The purpose of this chapter is to describe both the position sensor system that reads the rotation of the two wheels and the 5 VDC power supply system for the PML Both of these systems are installed on the IFB The IFB is composed of four parts which are listed below 4 bit voltage controller MC output Described in Chapter IV 5 VDC power supply Described in Chapter V 4 bit Hall counter input Described in Chapter V DIO connection and wiring Described in Chapter VI
81. am ceeeeceeeseceseseeeesteeeesteeeeseeeenaeeees Block diagram and basic description of motor control system Basic function of interface board voltage controller eeeeeeeseeeeeeeeees Expanded functional description of IFB voltage controller Voltage controller OulpurTange cone aoe System diagram of n bit variable resistance blocks 00 0 eee eeeeeeeeeeeeees Image of right side resistor DIOCK eeccecesneeceeeeeceeeeeceeeeeceeeessteeeenaeeees Single resistor circuit of voltage controller eee eeseceseeeseeeeeenaeenes IFB complete voltage controller circuit diagram 2 00 0 eee eeeeeeeeeeeeeereees JSIC connection and wire identification 0 ee eeeeeseeeeseecseeceeeeeeeeeeneees Connection diagram between the MC and gear MOtOD ee eeeeeeeeeereees Modular diagram of the MC 7 and its CONNECHIONS eee eee eeeeeeeeeeeeees Block diagram of the two interface boards 00 eee eee eseeeneeeeeeeeeeeeneees Modification to the Hall sensor module eee eeeeeeseecneeeeteeeeeeeeneee Modular and circuit diagram of position sensing SySteM eeeeeeeeeees Signal output from right Hall Sensor ooo ee eee eeeeeeseecseeceseeeeeeeeneeeaeenes Xi Page xii FIGURE Page 5 5 Signal output from right NAND ssssssssesssessssseesseesseessessseresseeessresseesseessees 61 6 1 Functional diagram of OCS eachsccavcsdeuersquadeevesyetazasguenaainsudedacdavedaeareyacsawegens 64 6 2 Front view o
82. am that is pertinent to the current task The one physical constraint applied to their model was the assumption that the robot operated on a flat and level floor The authors note that a future goal of robotics will be a computer system that has a computational framework able to combine features and cues in a contextual way They attempt to achieve the lesser goal of manually determining which set of features are used in advance Using this manual method they successfully developed an algorithm to divide the visual image of the ground plane into areas of increasing distance from the robot As the robot moved and subsequent images were taken it possible to gain an understanding about the height of images in the monocular image frame 7 This is a significant accomplishment since there is no straightforward way to determine the sizes of an object in a single image taken by a non stereoscopic and un calibrated camera The methodology presented in this paper could potentially be used by the PML robot especially since the manual selection of multiple features could be varied with the intention of gaining additional understanding about the surrounding environment In a paper by Falcon et al a description of a machine algorithm used to read Braille was described 13 It is conceivable that this method could be adjusted to read a positional error mark on the floor by the PML robot that would correct its positional error 2 2 5 Mapping and localization
83. ance from bottom edge of image as a percentage of the total image size are measured manually The rotational error is also determined by measuring the difference in the angle between the optical mark s horizontal line and the true horizontal It should be noted that parallax causes the vertical line of a non rotated optical mark to appear to be rotated The horizontal line of the optical mark remains nearly true to the horizontal as the image is moved in the Y direction and is the better choice for determining the angle of rotation The vertical line should never be used to measure rotation 71 In order for the OCS to determine the physical location of the robot relative to the desired stopping point the optical mark s position within the image frame as a percentage of image width and height must be known If the webcam image is always sized the same total X and Y distances are always the same the user can measure the optical mark s position from the edge of the image and input these two values into the Excel program This constant size of the image will allow the relative percentage of total distance to be computed from the linear measurements of the optical mark When the operator first opens the image for data extraction care should be taken to ensure that the image is expanded to the same size each time In Section 1 2 2 the second thesis objective stated that the goal was to develop the ability for the robot to travel autonomously any
84. ar buffer is used to compute the angular rotation Every 100 ms the most recent ActDeg value is loaded into the buffer After each datum is taken the index is incremented The index acts as a marker and controls which one of the five positions in the buffer is active at any time When the index is equal to 5 the next increment will reset the index and set it equal to 1 The buffer is initially filled with zeros and takes 0 4 s to fill the first five angular values The first four calculations of rotational rate are therefore inflated and not accurate This lag of 0 4 s presents a minor limitation in the LA controller algorithm Table 7 8 Design of angular buffer Time 1 2 3 4 5 Index LA Ratio 0 0 0 0 0 0 0 1 0 0 0 1 0 05 0 0 0 2 0 0 0 2 0 05 03 0 0 3 03 0 0 3 0 05 03 04 0 4 04 0 0 4 0 05 03 04 05 5 05 0 5 06 05 03 04 05 1 06 82 0 6 06 07 03 04 05 2 07 0 0 7 06 07 Os 04 05 3 Og 04 Equation 7 3 describes how the LA Ratio is computed Because of the way the data is written into the buffer the result of this equation will give the difference in angular direction over a period of 0 4 s The difference between these two angles yields an accurate average of the rotational rate over the previous 0 4 s The units of LA Ratio are in degrees per 0 4 s For purposes of control it is not necessary to have the LA Ratio specifically equal to degrees per second Therefor
85. are powered by them Table 5 1 Power supply configuration Power Source Switch ff Power destination 1 1 R and L Hall sensors 2 Counter block NAND and resistor block 3 3 191 Left counter chip 4 4 191 Right counter chip 5 Side MC voltage control IC block Switch 6 5 JSIC 7 NA Not installed 8 NA Not installed 5 3 Four Bit Position Counter Buffer System Homji designed and installed a position sensing system for the IPRV that consisted of several magnets attached to a plate installed on the drive rotor The original sensor configuration of the IPRV can be seen on the left side of Figure 5 2 The image on the left does not show the signal and power wires connected to the Hall module The original Hall sensors and connection wires were not compact enough to allow the motor covers to be installed This lack of protection placed the wires and the sensitive Hall sensors very close to the ground and exposed them to possible damage from the environment during the robot s movement One of the modifications completed for the PML was the reduction in size of the Hall sensor module and connection wires and its 56 encapsulation in a plastic coating material This reduction in size of the sensor module and the improved routing of the wires allows the original covers to be reinstalled Figure 5 2 compares the previous sensor system of the IPRV to the current sensor system on the PML IPRV Original
86. as described In this section an overview of what the robot is will be explained The approach taken will be to divide the hardware into modules and show how these modules interact Figure 3 8 shows the signal and control interactions that exist between these modules The power distribution was previously shown in Figure 3 7 An IFB has been hand built out of two integrated chip IC perforation boards and functions as the center of hardware control for the PML robot The IFB stands between the laptop and the input output of the motor system Both the input and the output have a 4 bit capability The IFB is composed of three basic functional systems Motor control interface e Hall counter interface 30 5 V power supply and switch block The function of the motor control interface is to take two 4 bit outputs from the laptop s digital input output DIO and convert it into two voltages These two voltages are routed through a double pole double throw DPDT switch and then go to the two Diverse Electronics PWM motor controllers MC 7 The output of each of the MC 7s then powers one of the two gear motors For purposes of clarity it is important to distinguish between the two Diverse Electronics MC 7 and the entire motor control system Throughout this paper these two PWM motor controller modules will be referred to as MCs whereas the motor control system refers to the entire system that is responsible for controlling the spee
87. at exists in the virtual space of a personal computer a robot by its very nature has the ability to interact with physical space As robots continue to gain the ability to process information about the environment they will be expected to take on tasks that more frequently involve interaction with people A very good resource on this field is the MORPHA project This project was conducted by a consortium of German companies and was completed in 2002 The central idea and statement of this project is stated below The core idea of the MORPHA project is to equip intelligent mechatronic systems particularly robot assistants or service robots with the capability to communicate interact and collaborate with human users in a natural and intuitive way 16 Two systems were selected for investigation the first is the manufacturing assistant and the second is a homecare assistant One of the main goals of both systems is to develop the ability of the robot to be taught by its human co worker with a minimum of programming input For example the ability of the robot to be taught a repetitive task 15 by a factory worker and not by programmed instructions would mean that this type of robot would have a wider application to a larger market of users 2 3 Summary of Literature Review Autonomous robotics and visual processing are very complicated subjects Therefore setting a development goal for the PML robot should not be taken lightly Since
88. at the rotation rate varies at 0 4 s after the turn has been initiated This is caused by the buffer lag and does not present a problem with the overall accuracy of the robot In each experimental run the positional error from the target was less than 1 5 cm and the deviation in the ending heading angle was not noticeable ActDeg Offset Error Turn Delta Time s Figure 7 14 Results from 45 course with right turn ActDeg 0 10 Offset Error 5 10 Turn Delta RD QW RBR MW OV T 3 Time s Figure 7 15 Results from 45 course with left turn 122 123 The next set of experiments was run on a course with a single 90 turn Figure 7 16 shows the layout of the two courses Similar to the 45 experiment there were two objectives the first was to hit a target with consistent accuracy and the second was to end with a heading of 90 Each experiment was run several times to verify that the controllers were in fact capable of meeting these two objectives In all cases the positional errors were less than 1 5 cm and the heading error was not visible 90 Track to the right 152 cm 60 in Robot path gt gt d 90 45 om Alm Robot position 17 7 in initiation and heading Start Start 90 Track to the left 45 cm At turn 17 7 in initiation Robot position
89. axis value of each point of the line is determined as the average of the eight measured theta values for each real radial line of the grid at that specific real 0 value For example the left most point of the line has a positional value of 12 23 20 This data was derived using the method discussed for Table 6 2 The other four points of the line are similarly computed The linear relationship between the image theta and the real theta values has a correlation coefficient R value that is near one Image Theta Real Theta 30 20 T 2 3 10 y 1 698x 0 6297 2 E Rf 0 9999 T 10 20 1 7 15 00 10 00 10 00 15 00 0 0 00 5 00 Image Theta degrees Figure 6 10 Graphical evaluation of theta functional relationship The next step in the development of the data conversion algorithm is the functional relationship between the radius evaluated from the image and the radius that corresponds to a physical real distance Table 6 3 Average image radius compared to real radius Theta Radial Arcs dagen Wie 21m 7h WH 20 248 273 282 290 297 303 10 235 250 262 273 282 289 296 303 0 236 250 263 273 282 290 297 303 10 236 251 263 274 283 290 297 304 20 264 ra 283 291 298 305 Avg mm 235 262 273 282 84 The co
90. b 12 test print inside initialize sub End Sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXK i DATA ARRIVAL XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX 177 Private Sub tcepComm DataArrival By Val bytesTotal As Long On Error GoTo error Call tepComm GetData Echo Print 1 inside data arrival echo Echo If Echo yellow Then frmColorBar BackColor vb Yellow End If white means that the Case Blue run is complete Red means EMO condition If Echo white Then frmColorBar BackColor vbWhite End If If Echo red Then frmColorBar BackColor vbRed End If EMO CONDITION If EchoPrevious red Then This IF block is used when EMO Case Yellow picture frmColorBar BackColor vbRed End If lblEcho Caption Echo this is a label picReturn Picture C Documents and Settings me Desktop temp bmp waitTime 1000 picReturn Picture LoadPicture Robot_server robot_lap temp bmp EchoPrevious Echo updates the previous echo marker 178 Exit Sub error End Sub Private Sub tepComm_Error ByVal Number As Integer Description As String ByVal Scode As Long ByVal Source As String ByVal HelpFile As String ByVal HelpContext As Long CancelDisplay As Boolean Dim error_num As Integer Dim error_descr As String error_num Number error_descr Description lblErrorOut Caption Error num amp error num amp vbCrLf
91. be overwritten Table 7 3 Operational data file types Index Name Purpose 1 Primary General data on program operation 2 Graph Only used for the conversion of data into an Excel graph that represents the run data see Figure 7 9 3 BadHall Used to investigate an ongoing intermittent proble m that remains unresolved 7 2 4 Path database The last peripheral to be discussed is the path database PDB The PDB is an Excel spreadsheet that exists on the server computer and is accessible to the ROS through commands written in VB 6 The operator is responsible for inputting the correct parameter values into the DB that will command the robot to follow a particular path and take actions at the appropriate location Table 7 4 shows an example of the database and the values used to command the robot to perform a two segment run Each row of the DB table is read individually and acts as 100 a single command line Each of these lines of command has four parameters used to control the robot s motion Each command line of the database is in force until its distance has been reached When the robot travels the distance indicated a new row is read and the four parameters are updated The four parameters are described below e Distance This variable determines the distance traveled by the robot in centimeters e Speed This variable is the speed command that is written to the motor control system see Table 4 3 for t
92. broutine followed by a parenthesis that includes any program or function called by the subroutine At the end of each line is a brief explanation of the purpose or responsibility of each subroutine ReadTable Neutral waittime responsible for reading the Excel table line commands e HallsRead SingleDigitalInput reads 4 bit output from both position counter chips HallsCompute MotCont contains the LA and SP controllers CorrectCompute SendData Neutral decodes error data manages Case Blue program RunCamera SendData SendMessage driver for webcam MotCont Right_Turn Left_Turn No_Turn computes motor speeds variables and calls the turn commands e Right Turn SingleDigitalOutput writes motor speed variables to MC that will cause a right turn to occur Left Turn SingleDigitalOutput writes motor speed variables to MC that will cause a left turn to occur e No Turn SingleDigitalOutput writes motor speed variables to MC that will cause a path to go straight 104 Initialize Zero sets initial variables at start up of first path Initialize2 Zero sets initial variables on subsequent paths Zero SingleDigitalOutput resets the 4 bit position counter to zero Neutral SingleDigitalOutput PulseMC when called this subroutine writes a neutral speed to both motor controllers PulseMC SingleDigitalOutput waittime creates a 1 ms pulse onto MC latch chip e tcpSe
93. ction data from the client side program Both of these tasks are described in the two subsections below 6 2 1 Server side webcam driver A webcam can be used to capture one image at a time or to function by constantly streaming images to its web server For the purposes of the PML robot only one picture at a time is required Therefore in order to conserve computational resources the webcam is operated only when it is needed The first responsibility of the server side optical correction program is to power on the webcam in order to capture a single image A code subroutine titled Run Camera is part of the server s VB 6 program its three main functions are listed below e Start cam Set file e Close cam These functions are supported by application programming interface API code The Run Camera subroutine see Appendix A 1 uses the Windows messaging capability and the API code to access the dynamic link library DLL files used to support the webcam s operation A basic explanation of how the Run Camera subroutine controls the web camera is necessary in order to explain how the optical correction system works Within the Windows OS there are DLL directories that have many program files or scripts used to control all of the functions that Windows may be required to run One of these files avicap32 dll is used by the Windows OS to control any webcam that is attached to the computer The Run_Camera subroutine is written in the VB 6
94. d then it must be able to find its re charger In addition a lawn mower has to cover the entire space to be mowed Therefore a region filling algorithm is used to program a more efficient path 4 Autonomous vacuum cleaner One of the examples Sahin and Guvenc discuss is a cleaner that uses three layers of sensor based navigation 4 The first layer of sensors controls specific hardware events such as the power requirements The second layer involves the dynamic sensing of path and movement and has the ability to make turns or follow a straight line The third layer is a task based navigation system used to learn about the local environment and map the local area 2 1 2 Factory robots Previously factory robots performed pick and place type tasks and were fixed in location Typical tasks were manufacturing jobs such as welding However in recent years due to the improvement in the intelligence level of robots their use and application in manufacturing has fundamentally changed Recently Wyeth president of the Australian Robotics and Automation Association said Particularly in the manufacturing sector we are starting to see new products coming out now which are delivering goods intelligently around the factory without the need for laying down guide wires or guide cables machines now can intelligently go from place to place and collect parts and take them to the appropriate work cell which opens up a new way of structuring the ma
95. d to the MC 7 controllers This will cause the robot to either increase or decrease its rotation rate and will maintain a constant rate of rotation 120 The selection of the points and the resulting line used to map the TurnDelta output is based upon the following reasoning On the left side of the graph the line has a negative slope that was selected to achieve stable rotation The curvature of the line occurs when the LA Ratio approximately equals 70 This point of curvature was selected to insure that an unstable or runaway condition would not occur On the right side of the graph the line has a positive slope The reason for this is to increase the rate of rotation during the first 0 4 s when the LA Ratio is inflated due to the buffer lag A second order polynomial was chosen as a compromise between the increase in the computational requirement that comes with higher order equations and the benefit of achieving a one to one map with the appropriate response The primary attribute of this map was determined to be the point of curvature therefore a second order equation was sufficient It is less important to have the polynomial fit the exact points selected as long as the rotational response of the robot is consistent 7 5 2 LA control conclusion Experiments were run in order to test the effectiveness of the LA controller Two sets of experiments were conducted The first can be seen in Figure 7 13 In this experiment two courses were
96. d of both motors The function of the Hall counter interface on the IFB is to count the binary pulses from the Hall sensors and convert this binary number to a 4 bit number For example the output of each of the Hall sensors is either a 1 or a 0 Two 4 bit counter chips tally the binary count and output a 4 bit number Therefore the output of the Hall sensor interface block is two 4 bit numbers between 0 and 15 The laptop reads the 4 bit number and interprets this number as the amount of rotation of both the right and left wheel Once the output value gets to 15 it rotates its value back to 0 The IFB does not have a capability of a ripple counter or a second digit in its hexadecimal number scheme The function of the 5 V power supply and switch block located on the IFB is to take a 12 V source and convert it to a 5 V output The switch block is an 8 cm by 5 cm prototyping board that is attached to the IFB by using standoffs It has six switches that control most of the 5 V power requirements required by other hardware modules Another switch is attached to the left IFB board next to the resistor block and is used to power the resistor blocks on both of the interface boards Currently the output of six power supplies are being used and are passed through six switches There are two 31 unused spaces that are available for the future expansion if more 5 VDC supply is needed
97. d specifications The most serious problems requiring significant rebuilding and redesign are discussed below The number of voltage supplies on the IFB was increased in order to resolve a counting error in the 4 bit 74LS191 counter chips The earlier IFB had fewer voltage supplies that would overheat due to an excessive demand for current Analysis of this problem indicated that an errant signal was created when the power supply chip would self regulate its current flow in order to prevent burnout The counter chips misinterpreted this very fast on and off cycling as a signal from the Hall sensors The IFB boards were rebuilt when it was discovered that a resistance of less than approximately 10 MQ between some of the pin connections would cause 17 signal cross talk in the counter chips This error was found when the robot was run on its stand Even with only one wheel running both counters would show an increase in count To correct this problem most of the board had to be re soldered with special attention given to the resistance values between the pins of the counter chips e The design of the robot s OS was significantly changed in order to maximize the speed of the program Due to the inefficiencies of Win XP and VB6 it was discovered that the earlier program was not fast enough to capture all of the position counts even with the assistance of a 4 bit counting buffer The program was redesigned emphasizing the importance of its ti
98. d to the bracket by two bolts in a direct horizontal line with its lens This allows the vertical center of the webcam image to be adjusted with minimal impact upon the image parallax The bracket holding the camera is attached to the electronics housing box at the approximate midpoint between the wheels At both the vertical and horizontal attachment points there are rubber grommets that allow the position of the camera to be moved to a point of adjustment and then remain fixed by friction The effective zero point for all measurements of distance and angle is defined as being directly under the camera In order to measure this point a 65 cylindrical brass weight is suspended directly under the lens of the web camera This weight functions as a plumb bob used to mark the dead center of the webcam Since the webcam is intended to accurately measure the positional error of the robot it must be adjusted accordingly Figure 6 2 Front view of webcam and mount The robot is moved to a point near an optical mark on the floor such that the plumb bob is exactly 61 cm 24 in distant from the center of the mark The camera is moved in both degrees of freedom yaw and pitch until the webcam image shows the optical mark 66 at the exact center of the image When this condition is met the webcam can be used to determine the position of the robot in relation to its intended stopping point of 24 inches distance from the optical mark The camera
99. de sections are described 1 The Case Green code reads the Excel database line by line and retrieves four numerical values that determine the path that the robot will follow These four numbers describe the distance traveled the speed traveled the angle of travel and the offset value to be defined later When the robot is moving forward the Case Green code also reads the position of both wheels and computes the distance and orientation of the robot A proportional and integral PI controller is used to keep the robot on its intended path During its run the robot measures key metrics and stores them in a text file These saved files can be used after the run is complete to display the results of the run in a graphical form When the robot gets to the end of its intended dead reckoning path it will exit Case Green and begin the Case Yellow code 2 The Case Yellow code functions as the software driver for the webcam This code is activated in one of two ways first at the end of every Case Green code and second when the color command is sent from the client In this case the responsibility of the webcam driver is to start the webcam which allows it to set the focus capture the image and save it to a shared file After these tasks are completed the webcam is then turned off and the Case Yellow code relinquishes its control 3 The Case Blue program can only be activated by the client s command With a few exceptions this code
100. designed to prevent the turning speed at which P controlled turns take place This will hopefully prevent the robot from swinging wide and crashing into the door Option Explicit Dim num As Single Dim numData As Long Dim strData As String Dim temp As Long Dim CaseBlueFlg As Boolean use to disable the EMO during Case Blue deceleration Dim NewLine As Boolean used for the logic on a new line of the path table Dim binRun As Boolean Dim Indx As Integer Dim TrigCol As Integer Used to read the distance and direction from the path table Dim TrigFwdCell As String Used to read the Forward value from the Excel table Dim TrigSpdCell As String Used to read the Speed value from the Excel table Dim TrigLRCell As String Used to read the Left Right value from the Excel table Dim Distance As Integer Three numerical values from the table Dim FwdSpeed As Integer Dim DirAngle As Integer Dim DirAngleOld As Integer Use to effect large angle turns Dim OffsetRef As Integer Offset Reference desired offset value from data table use for turns Dim Right R As Byte The computed values sent to the turn subs so only one computation Dim Right L As Byte called Turn Variables 135 Dim Left R As Byte Dim Left L As Byte Dim timel As Single Dim time2 As Single use for cumulative time accounting over multiple green runs Dim timeCarry As Single use for cumulative time accounting Dim timeInt As Single Dim CamTime As Si
101. distance through an obstacle course and end with a minimal position error of approximately 5 cm With the exception of the manual operation of the data extraction module the robot is completely autonomous Once this module is automated the PML robot will be able to traverse its programmed path with no human interaction The client side OCS manager is designed to support the future development and easy integration of an autonomous data extraction module Preliminary research has indicated that one possible solution would be to use Matlab and its image and signal processing toolboxes to extract the position data 19 20 Operating the robot while manually extracting the image data has proven to be beneficial in terms of learning about the robot s performance and parameter choices Experience gained during repeated operations has determined the level of accuracy needed by the data extraction module 6 3 2 Client side data conversion module An Excel program was implemented on the client computer and performs the function of the data conversion module The client side data conversion module is responsible for converting the datum point extracted from the image into a datum representing the real 72 error Because of the nature of this task the conversion is a one to one map from the image position to an associated position on the floor near the optical mark Once the physical position is known the true error can then be calculated VB 6 could
102. e 105 PCMUSERFUNC This module was also written by SuperLogics and has the highest level functions used to control the DIO device The user controls the input and output capability of the device by calling the listed functions in this module For example SingleDigitalOutput is a function found in this module This function is used to output a voltage to any of the defined channels PCMDATA This module was written by SuperLogics and is used to support for the VB 6 drivers for the PCMDIO device There are no subprograms or functions in this module This module describes the basic variable arrays and global constants necessary to control the PCMDIO device 7 4 Straight Path Controller The PML robot has two distinct direction and path control algorithms One is referred to as a Straight path SP controller and the other a large angle LA controller The next two sections will fully discuss the development and design of these two motion controllers The physics of a wheelchair based mobile robot makes it a multiple input multiple output system This is due to the fact that both the position sensing and the motor speed of each wheel are independent of each other Once the MC 7s and the position sensors were installed and determined to function within specification the next step was the development of the motion controllers In this thesis the word controller is defined as the software code that accepts input data from the
103. e In the future the course correction path could be combined with the main path segment run For the purposes of troubleshooting and ease of course layout these two path following functions are kept separate 3 3 Hardware Overview This section includes a basic description of the elements of the wheelchair from which the robot is constructed and the 12 VDC battery system that powers it Also presented is an overview of the robot s hardware from a modular point of view A working definition of a hardware module is a physical object that is easily separated from the rest of the assembly and can be treated as an individual part A description based on modules will allow the entire system to be more easily understood 27 3 3 1 Wheelchair and power supply The base of the PML robot is an Invacare Ranger II Electric Powered Wheelchair 2 It is constructed of a tubular frame and has four wheels The two front wheels are 30 cm 12 in diameter drive wheels The two rear wheels are 15 cm 6 in diameter solid castor wheels The main wheels are 56 cm 22 in apart and the castor wheels are 43 cm 17 in apart The wheelbase the distance between the castor wheels and the main wheels is 24 inches Sub section 3 2 1 introduced the two measurement derivatives of distance and angle The measurement derivative of angle is affected by the geometry of the wheelchair Specifically the distance between the two drive wheels has an effect upon the orientat
104. e bes 9 2 2 Developmental Chall nS eS issic i s s9 dcaccecssstislonceansaseaeaeastacasaneceinas 9 2 2 1 Hardware based front end image pre processoT 10 2 2 2 Distributed architecture control systeMS eeeeeeeeeeeee 10 2 2 3 Multiple sensor control system ceeseceesseceesteeeeneeeenes 11 2 2 4 The development of an inexpensive secondary sensor syste M teat sa ss arsen ease Reesen SD 12 2 2 5 Mapping and localization ceceeecceesseceeseeceesteeeenseeeeees 13 2 2 6 Communication between a human and a robot 14 2 3 Summary of Literature Review 5 65 osic is coc sess sacnaddagnocedeaseveeneconses 15 vili CHAPTER Page II PME DESIGN OVERVIEW 63 onkg lenker esiga tgs ases 16 DoE Intro du CHON sanni ennan e e a N 16 3 1 1 Examples of redesign sed cis ssiycasedadnynucnnatelonadoase seedaapheveaedeaanes 16 3 2 System OyervieW saiessestevasvensarepnieeetassaveadesousndeashancdueseeveetersnnacesens 17 3 2 1 Dead reckoning OVErVIEW cecccceecceeseeceesteceesteceenseeeeees 18 3 2 2 Optical correction OVervieW ssseesseessesssessseressseesseessees 19 3 2 3 Course correction overview ssseesssesseesseeeseresseeesseessees 24 3 3 Hardware Overview ssesssesssseseeserssesreesresreseresressesererressereresees 26 3 3 1 Wheelchair and power supply sssessssssssssesssesssssresseesseess 27 3 3 2 System hardware overview cccescceeseeceesseceesteceesseceeees 29
105. e of error data by robot to correct for positional error Subsection 3 2 3 gave a explanation of how the error data is used to correct for the positional error The course correction software in the ROS will first convert the real Y error into an offset error value It then uses the I control program and the offset error to correct for its lateral or Y axis error See Appendix A 1 and the CorrectCompute subprogram for the program code that utilizes these error numbers 6 4 Summary of Optical System An important point to realize about the OCS is that it can be treated as a modular capability of the ROS Even though the OCS resides on two computers and is composed of multiple programs it functions as a single element in the PML robot s operation This chapter has presented a detailed description of the four main functions of the OCS It has explained how the webcam image is captured and sent to the client side OCS It has given a full description of the two steps in the image processing procedure along with an explanation of how the data is exchanged between the client and the server The 90 actual output of the optical correction system are the three error data values X2 Y2 and a 91 CHAPTER VII ROBOT OPERATING SYSTEM 7 1 Introduction The purpose of Chapter VII is to discuss the design and implementation of the ROS used by the PML robot The primary responsibility of the ROS is to manage the peripheral modules used by t
106. e the LA Ratio is in fact a ratio of degrees and time ActDeg 0 4 s 118 LA Ratio 0 6 7 3 Each newly measured angle will overwrite the oldest angle and the index will then be incremented by one This will yield a new calculation of the rotation rate every 100 ms Within the program the computation of the LA Ratio takes place after the latest 0 value has been measured There is no significant amount of time between the latest angle measurement and the LA Ratio difference The significance of this is that the time of the measurements between the two values never significantly varies from 0 4 s The absolute value of the difference in speed between the two wheels defines a new LA 99 control variable named TurnDelta This value is expressed in 7 4 In this equation the term right DIO speed is the same as the DIO speed variable found in Table 4 3 The DIO speed variable will be an integer between 0 and 7 From the definition in 7 4 it can be seen that the TurnDelta variable will also be an integer The minimum value this variable will take is limited to 2 which will command the slowest rate of turn The highest value is set to 7 which will command the fastest rate of turn TurnDelta right DIO speed left DIO speed 7 4 Now that the two primary variables TurnDelta LA ratio have been described the remaining description of the LA controller can be given It should be understood that as the Tu
107. ection cable of the JSIC and power is applied a red light emitting diode LED will be lit indicating that the JSIC is operating The wire identifications for the series connector have been tabulated in Figure 4 9 The ID number identifies the location of the wire in the ribbon cable with the red wire identified as wire number one The color refers to the wire that runs between the series connector and the JSIC connection Note that it is important to have the kill switch in the un powered position during the connection of the joystick to the serial port because a transient resistance may briefly occur if the connector is attached at a slight angle This will cause the MCs to engage briefly and an erratic and unplanned movement may occur 4 4 MC 7 and Gear Motor The last part of the motor control system is the MC 7 motor controller and gear drive motor The point of view of this discussion is to assume an input control voltage and an output wheel speed as indicated in Figure 4 10 48 MC 7 Motor 12V Controller gear motor 0 to 3 5 V O U Input control PWM driving voltage voltage Output wheel speed RPM Figure 4 10 Connection diagram between the MC and gear motor 4 4 1 MC 7 motor controller The MC 7 motor controller is manufactured by Diverse Electronics and is used to power a DC motor by producing a pulse width modulation PWM power supply voltage The MC 7 is a robust controller and is a good match for the PML robo
108. een erne ne ennen 106 7 6 Basic design of main program with modules ee eescceceseceeeteeeeeteeeeees 102 ii Error signal used by SP controller s 40542655 2 RG 107 7 8 One to one map offset error to reference angle u u u ssenvereerrrerrererrrnee 110 7 9 Course layout for SP controller experiment eee eeeeseeceteceeeeeeeeeeneees 113 7 10 Results from CW torque experiment 2 425 cises incu shnevensvnrs secvbegstabeenie So cuneye vane 114 7 11 Results from CCW torque experiment cceeeccecssececesececeteeeeeteeeeneeeenes 115 7 12 One to one map between LA ratio and TurnDelta cccccceccceeteeeeeteeeeees 119 7 13 Course layout for 45 turns cas ciics caaiwes nad ane dtaa pupedudn sess achounstaatesebncrsabedeees ase 121 xiii FIGURE Page 7 14 Results from 45 course with right turn ssessseeessseesseesseesseessseessresseesse 122 7 15 Results from 45 course with left turn sssssoessseesesseesseesseesseessseesseessresse 122 7 16 Course layout for 90 Waris sacs an r a E aaa 123 7 17 Results from 90 course with right turn accisc cose oe eau ats lela 124 7 18 Results from 90 course with left turn oo cc cceseeeseecceccesesseeeeeeceeees 124 TABLE 3 1 4 3 5 1 6 1 6 2 6 3 7 1 7 2 7 3 7 4 7 5 7 6 7 8 XIV LIST OF TABLES Page Data input output sources 5 cscisgssecicghavervaderraccessnsaevasgeesdeadeonaceesacseoeaeenaantsons 35 Summarized motor control results ssss
109. election of magnets 5 02 5 sasainssduteoncenncacscseoadenstaacteecsad asunes 59 5 3 5 Use of the NAND IC in the position sensor circuit 60 5 4 Summary of Electronics System cceesccecsseceeseeceesteceesteeeeees 64 CHAPTER VI VII OPTICAL CORRECTION SYSTEM wy svctsacsacostsvacsvsvecsectyusstennbngsaeaains GL MIMO UC TOD sparring oct a 6 1 1 Camera hardware nes fo ek das det ctiics gcncen ot ci ee nens ennen enes 6 1 2 Organization of this Chapter ceccceeeseceesseeeesteeeesteeeeees 6 2 Server Side Optical Correction Program eseeeseeeseeeeeeeeeees 6 2 1 Server side webcam driver sic cs0 vieaeetasa Nias pies iecsene ee 6 2 2 Server side optical correction data return 1 6 3 Client Side OCS Manager s ccsccaccpsnccdacesacacnceesnctoacs ceedacthessneas anes 6 3 1 Client side data ex tracnoWsc ie dai ceieiann hindi ek 6 3 2 Client side data conversion module 0 0 eee eeteeeeeeeeeee 6 3 3 Development of the data conversion mapping algorithm 6 3 4 Correlation of image data to real position 6 3 5 Decoupling of input variables 2 0 0 0 eeeeceeseeceesteeeenteeeenes 6 3 6 Development of the one to one map equation W W 1 6 3 7 Development of one to one map with a non zero 6 3 8 Error data encoding 5 sssesdies Gansiecancdanes zens deceateostevncn ennen eee 6 3 9 Use of error data by robot to correct for positional error 6 4 Summary of Optical System
110. ently done by an operator who is responsible for manually measuring the cross s image position on the screen The measurement is then input into the data extraction program This lack of capability is the only missing component that prevents the PML robot from achieving complete autonomy of action in completing the goal of the thesis objective This capability should be the next area of development for the robot It should be understood that the PML robot represents multiple developers and remains a work in progress For the purposes of this thesis the current level of development is sufficient to prove the intended concept Once the fractional position of the mark s image is known the three numbers that describe the robot s position are given to the data extraction algorithm for further processing Currently this processing is done by an Excel program This program maps the two position marks and one rotation mark onto a representation of the robot s physical space The error in the robot s position is the difference between the ideal image center and the mapped distance to the image s position Figure 3 5 shows the result of the Excel data extraction program 24 X Mapped X position error in inches Y Mapped Y position error in inches ye a Center of screen ideal position et NP a X axis Y axis Figure 3 5 Screen image w X and Y and error result The fourth and final st
111. ep is to return the correction data to the robot The image processing and data extraction step above yields the error position in inches The client VB6 program takes the three numbers that describe the error data and encodes them into a single 6 digit number This number is sent to the robot server where it is de coded back into the three numbers that describe the positional and rotational errors The result of the optical correction function is that the robot now knows its physical position relative to the fixed mark on the floor What remains to be done is for the robot to act on the error data and correct its position before beginning the next dead reckoning run 3 2 3 Course correction overview The third sub function of the robot is the course correction task This function takes the error data and builds a short run that is 36 inches in length The course correction 25 algorithm translates the Y axis error value into an offset error OSE value and takes the X axis error value and adds or subtracts this amount to the 36 inches of path so that the course correction ends at the correct point Figure 3 6 shows two waypoints and their relationship to the other positions of the course correction path Each waypoint is both a beginning and an end and in order to prevent confusion it is important to identify the waypoints by name The end of a main path segment is the beginning of the course correction path The end of the course correct
112. er supply PIC chip Series connector Not used 6 Pp i 3 Joystick J S C Not used Red Left Orange Right Red Right Serial connection ie abla DPDT Wire ID Color Black Left switch Ground 1 Green Black Left Red Right Fwd Rev 6 Red IFB voltage L RE 3 Black Figure 4 9 JSIC connection and wire identification Since the robot only moves in the forward direction a permanent voltage has been provided to both of the MC 7 s T 5 connectors Therefore joystick control through the JSIC is no longer able to go in reverse Figure 4 9 shows the system diagram of both the 47 joystick and JSIC combination and how the control input can be switched from the manual joystick to the autonomous IFB voltage control The control output of both JSIC and IFB are selectively switched to the MC 7s by a DPDT switch that is located at the front of the robot on the electronics housing box A standard joystick works by outputting two channels of variable resistance in a range between 0 and 100 kQ The channels are defined to be the right left channel and the forward reverse channel When both of these channels is at the 50 kQ value the joystick is in its neutral position The JSIC functions by mixing the two channels from the joystick and outputting a converted signal that is composed of a voltage and a polarity to each of the MC 7s When the joystick is first connected to the serial port conn
113. f intStatus singleDigitalInput gintlogicaldevice L Hall HallCntLeft If intStatus lt gt 0 Then Call pemdioError gintlogicaldevice intStatus End If Print 3 HallCntLeft HallCntLeft HallCntRight HallCntRight Print 3 If HallCntRight gt 10 And PrePulseR False Then Print 3 PrePulseR PrePulseR PrePulseR True Print 3 PrePulseR PrePulseR End If 146 If PrePulseR True And HallCntRight lt 5 Then Condition statement for Ripple incease RRip RRip 1 PrePulseR False Print 3 After increment RRip RRip End If THEN THE LEFT SIDE If HallCntLeft gt 10 And PrePulseL False Then Print 3 PrePulseL PrePulseL PrePulseL True Print 3 PrePulseL PrePulseL End If If PrePulseL True And HallCntLeft lt 5 Then LRip LRip 1 PrePulseL False Print 3 After increment LRip LRip End If time Int 100 Timer 100 timeInt 100 time2 Round timel timeCarry 2 Cumulative time used for graph restrict to 2 decimal places ActDist HallCntRight RRip 16 HallCntLeft LRip 16 2 ActDeg RRip 16 HallCntRight LRip 16 HallCntLeft CorCarryDeg DirAngle note that ActDeg is zeroed at the DirAngle value Print 3 Timel timel Print 3 o Print 3 EMERGENCY OVERRIDE COMMANDS if the robot s progress ActDist is
114. f Og Figure 6 8 demonstrates that this first condition is true Each of the eight lines on this figure is derived from Table 6 1 data The purpose of this figure is to relate the real 0 position from the grid to the measured radius from the image The X axis of Figure 6 8 displays the known positions from the grid Each arc of the grid has five intersections at 10 intervals therefore there are five points on each line in this figure The Y axis of this figure displays the polar radius measurements on the image of each intersection point As an example the point on the graph that relates to the R18N20 intersection can be found at the left most point of the Radius_18 line in the figure Note that the image radius values use the 200 mm extension and represent the polar measurement value The determination of decoupling is based on the horizontal track of each of these lines 80 Another way of expressing this is that as the real angle is varied the radius on the image remains constant Radius Independent of Thetag E E n 2 ne G cc o D G E 230 Real Radius inches 250 Radius_18 amp Radius 21 Radius 24 Radius 27 210 Radius 30 Radius 33 a Radius 36 A Radius 39 200 25 20 15 10 10 15 20 25 5 0 5 Real Theta degrees Figure 6 8 Graphical proof of first condition for decoupling The second condition for decoupling is that 6 is not a function
115. f webcam and MOuM eee eesceeseceeeceeeeeeneecaeceseenseeeeaeees 65 6 3 Data flow diagram of OCS 1 s cccss asacsossesawes scadsesdstay iuccaadeucegaateaeedadegueranaaedeee 66 6 4 Derivation of the three error values from the image eeseeeeeeeeeeeee 70 6 5 Data conversion module responsibilities eee eeeeeeeeseeeeeeeeeeeeeeeeeaeees 73 6 6 Webcam image of polar grid 9 33 sci seyscefeacecaseesyaaasave shes reen renerne eotsceeseoaees 74 6 7 Development of polar coordinate system eeeesecseeceeeteeeeeeeeeeeteeeeaees 77 6 8 Graphical proof of first condition for decoupling ce eeeeeeeeeeeeeteeeeneees 80 6 9 Graphical proof of second condition for decoupling e ce eeeeeeeeseeeeetees 81 6 10 Graphical evaluation of theta functional relationship ee eeeeeeeeeeeees 83 6 11 Graphical evaluation of radius functional relationship 0 0 0 0 cesses 85 6 12 Strategy used to convert non zero error ata cece esse eeeceseeeeeeeeneees 86 6 13 Error data encoding scheme isJ lt 3 5 4c0 doysesentevdsus Gacausde Seca ioemeeeieetnee 88 Tel The ROS andit peripherals ecreis BE eee an cede nde a ai 92 7 2 PCMCIA card and CP 1037 cable i404 SeenON 93 7 3 PCMDRIVE configuration utility 0 0 0 eeeececeeceecssececseeeeceeeeeeeeeeenaeeeenas 94 7 4 Client and server wireless devices 0 testes teaes const pda fug eee renen en cata 96 7 5 Client form for user commands u ssssenveereereerer eee renen eee r
116. future and more advanced studies in the sensing and mapping of the environment The objective is to build a permanent base that can accept input from any additional sensors or controllers without having to be significantly modified Homji s development of the IPRV was useful when it came to proving the basic concept of a wheelchair robot used as a test vehicle for the specific investigation of road repair However after conducting performance experiments problems with the accuracy of the Hall sensors and motor controllers were discovered These experiments showed that the basic positional sensing and motor control should have high level repeatability and accuracy Without this level of performance it will be difficult to isolate any problems that occur during the development phase of a more advanced navigational system 1 2 2 Second thesis objective The second objective of this research is the development of the capability for the PML robot to travel any distance with accuracy It should be demonstrated that the robot has the ability to travel at least 30 m 100 ft avoid obstacles and complete its course of travel with an insignificant positional error The basic design of the robot will use a combination of a dead reckoning capability supplemented by an optical sensor The definition of dead reckoning is The calculation of one s position on the basis of distance run on various headings since the last precisely observed position 3
117. g the preprogrammed path segment The next two subsections describe how the two modules of the client side system work 6 3 1 Client side data extraction The data extraction module of the client side OCS is responsible for evaluating the webcam image and determining the distance of the optical mark from the image center This module produces three numbers that represent the optical mark s X error Y error and rotational error Figure 6 4 illustrates the desired result of the data extraction module The three error values are defined from the optical mark s distance and rotation from its ideal center position This figure is similar to Figure 3 5 but includes the addition of the angular error and percentage of image size 70 100 75 50 Center of screen Ideal Position lt 4 0 50 68 100 t k F y X Axis Image Screen Edge X Mapped X position Y Mapped Y position Angular difference error in mm error in mm from true horizontal Y Axis Figure 6 4 Derivation of the three error values from the image Currently this step is performed manually After the client side program updates the VB 6 form on the client s computer the operator opens the image for data extraction The operator places a millimeter ruler on the screen and measures the optical mark s position from the lower left corner of the image The Y error distance from left edge of image and the X error dist
118. gRefFlg False Then If False then absOSE gt 0 if error is then DegRef DegRef 1 DegRef and correction should be negative End If DegError ActDeg DegRef a error is an error to the right Print 1 DegRef DegRef DegError DegError TURN SPECTRUM Output of this block is the SpecTurn value that is between 1 and 1 in aQ 1 increment steps The SpecTurn value determines the of turn initiated 150 by the MCs Performs a secondary PWM function 0 25x 0 2667 is the more aggressive response 0 4x 0 2 is the less aggressive response DegSign Sgn DegError holds the sign of DegError due to equation below if DegError positive then SpecTurn should be to the left SpecTurn Sgn DegError Round 0 25 Abs DegError 0 267 1 more aggressive SpecTurn Sgn DegError Round 0 4 Abs DegError 0 2 1 less aggressive If Abs SpecTurn gt 1 Then Bounds SpecTurn to within 1 SpecTurn Sgn SpecTurn sgn outputs a 1 ora 1 End If Print 1 Print 1 SpecTurn SpecTurn l TURN BLOCK The output of this block is the TURN value of 1 0 1 It increments the trip counter for each pass through the HallCompute subroutine therefore when command is passed to the next line in the WHILE loop if the TURN value is different
119. ght and left Hall effect sensors The DIO also wrote the speed and direction commands to the wheelchair s motor controller This author worked with Homji during the end of his thesis project in order to gain familiarity with what had been accomplished and to take over the project A different definition of the IPRV was in order that would meet with the author s desire to change the development objective of this project from a pothole repair vehicle to a more generic robot in an industrial setting that could be used to study various issues in robotics To that end the name of the robot was changed to the Precision Mechatronic Lab PML robot Figure 1 1 shows an image of the current configuration of the robot Figure 1 1 Front view of PML robot 1 2 Thesis Objectives The primary objective of this thesis is to build a mobile robot that can be used to explore elements of visual processing Using the visual processing capability as a navigation aid is a natural choice 1 2 1 First thesis objective The first objective of this thesis is the building of a robust lower level hardware base As will be discussed in the literature review a useful architectural design for robotic control is a multi level navigation system The lowest level of this type of navigation system is responsible for basic movement and directional control Specific hardware and software improvements are intended to create a robust robot base that can be used for
120. gnal that is to be sampled 17 That would mean that the time between samples should be less than 6 7 ms Comparing the theoretical sampling rate to the slowest actual sample rate proves that the original position counting system would not work 17 Actual sample period slowest 50 ms and normal 3 ms Theoretical sample period less than 6 7 ms The comparison above shows that when the ROS is cycling at its normal speed it is fast enough to sample the position sensor signal However when the ROS is operating at its slowest speed the sample rate is too slow and will miss position counts Therefore a method must be developed to store the position counts until the ROS takes the sample 5 3 2 Position sensor diagram The method chosen to resolve the problem of the slow ROS cycle rate is the use of a 4 bit counter that acts as a buffer for the position sensor Figure 5 3 shows the circuit diagram of the position sensing module installed on the IFB The output of the Hall sensor is a binary signal as shown in the figure The result of passing these two signals through the IFB is that they are converted into two 4 bit signals that are then input into the DIO 59 Right wheel 5 V input IFB Position Sensing Module 1 0 see oar 4 Bit counter I i 4 74LS191 DIO Hall sensor gt ias ana 7 Output signal ZTO NAND Schmidt 4 Bit wheel count WMA HCF4093B 2 channels 4 Bit co
121. h to avoid damage in case of a collision The slowest speed was selected in order to make it easy to observe the details of motion during experimental runs Once the fastest and slowest speeds were determined the corresponding voltages were found by connecting a variable power supply to the MC 7 input setting the robot on a stand engaging the gear drive and running each wheel at each of these speeds Once the upper and lower voltage values were determined the voltage range was divided into seven equal parts by using eight equally spaced voltages Figure 4 4 illustrates how the voltages are distributed Seven equal divisions Stop N a ee S 0 1 2 Volts 3 4 5 Slow Fast Figure 4 4 Voltage controller output range 40 The speed experiment determined that a 2 7 V signal input ran the wheels at the fast speed and a 1 5 V signal input ran the wheels at the slow speed Once the maximum and minimum voltages were determined this range was then divided into seven parts requiring eight resistances Any voltage less than 1 V was found to not turn the wheels at all 4 2 2 Voltage controller system design Figure 4 5 is a system diagram that shows how the IFB voltage control works by cutting in and cutting out four resistors in a series connection This figure expands on Figure 4 3 by showing the internal system diagram of the n bit variable resistance blocks When the appropriate resistor values are selected output vol
122. h txtInput This is the format for reading a cell in an Excel sheet LinkMode vbLinkNone LinkTopic ExcelLoc LinkItem Excl Cor don t use paraenthesis around the String cell LinkMode vbLinkAutomatic End With TotError txtInput Text Print 1 Tab 12 TotError TotError Print 1 Tab 12 CorOut CorOut txtInput Text CorOut waitTime 1000 Call tepComm SendData txtInput Text sends the 6 digit coded correction signal IblEcho Caption txtInput Text Exit Sub error End Sub Private Sub cmd Yellow_Click Call tepComm SendData yellow YELLOW run the camera IblEcho Caption yellow _ this is a text box frmColorBar BackColor vbYellow End Sub Public Sub Form_Load Close 1 prevents error if doc is open when program starts 176 Call Initialize ENE LOCATION OF THE EXCEL DOCUMENT That would be Excel pipe backslash etc ExcelLoc Excell ADAM Flash_bu_nodate Research V B_laptop Adam_active MC_camera MC_c amera_client R_eval xls POSITION location is on the school desk computer ExcelLoc Excell Documents And Settings adam Desktop Research_bu vb_desktop Adam_active MC_camera MC_Camer a_client Image_eval xls POSITION End Sub Private Sub cmdUnload_Click Unload Me End Sub Public Sub InitializeQ Open C Documents and Settings adam Desktop MC_cam_client txt For Output As 1 Print 1 Ta
123. has a manual focusing ring used to focus the image allowing the focus plane to be set to the center of the optical mark The webcam connects directly to the laptop via a USB cable 6 1 2 Organization of this chapter Chapter VI is organized in a manner similar to Figure 6 3 which shows how data are controlled and evaluated by the optical correction system In the following two sections of this chapter the server side and client side optical correction programs are fully explained Figure 6 3 will be referred to in this chapter as an example of how the OCS manages data flow Shared Directory Se Webcam Webcam Driver Controls the webcam and transmits the image to the shared directory Receive Data Receives and decodes numerical data Makes available to ROS Server Side Optical Correction Data Extraction Manual or automatic extracts the location of the optical mark in the image Data Conversion Maps the mark location from image space to real space Encodes and sends data to server Client Side Optical Correction Figure 6 3 Data flow diagram of OCS 67 6 2 Server Side Optical Correction Program As seen in Figure 6 3 the VB 6 server side program has two primary responsibilities When the ROS requests for correction data the server side program will first function as the software driver for the web cam The second responsibility of the server side program is to receive the corre
124. he exact values Angle This variable is the direction angle taken by the robot in degrees e OffsetRef This variable stands for offset reference and is the amount of offset from the straight path that the SP control will seek The operator must follow a specific format when creating the PDB For example the first row of data should occur on the 11th index row of the Excel worksheet The selection of which row to start with is an arbitrary programming choice that allows comments to be written above the database and still have the least significant digit start with 1 Column B through column E must be used for the four parameters as indicated in the table below Column A is not used by the program and could be used for comments or other identifiers In addition to the column formatting constraints there are two specific numerical flags used in the database When the ROS reads a command line that has 10000 in the distance it flags the program to stop and take a picture When the distance is flagged as 12000 it causes the robot to shut the program down after coming to a complete stop 101 Table 7 4 Example of path database Offset Ref O 0 E E O 0 The first four lines of the PDB make up the first path segment the robot will travel This segment is defined by the four rows that begin with index 11 and continue until index 14 The second segment is made up of the next four rows and ends when the distance is equal to 12
125. he functional relationships can be determined it is necessary to discuss the intermediate steps in the derivation of the image theta 0r and the image radius R1 The data in Table 6 2 represent the evaluated data of the 20 line The X local and Y local numbers are read directly from Table 6 1 The image X global values are the individual X local values plus 200 mm Each Image Y global value is the horizontal measurement of the optical mark from the vertical centerline of the image The image theta 07 is the result of the inverse tangent of the X global and Y global pair as seen in 6 3 The image radius Rr is the hypotenuse of the orthogonal pair Yc XG and is computed by using 6 4 82 0i tan Yg XG 6 3 R Y X 6 4 The eight 64 and Ry entries in Table 6 2 are evaluated by using the intermediate transformation steps above The eight image 0 values from each line on the grid are averaged and the result is presented at the bottom row of this table This procedure is done for each of the five radial lines The next step in the development of the data conversion algorithm is the functional relationship between the image theta 6r and the real theta Ok Table 6 2 Evaluation of 20 line on polar grid Real Points degrees R18N20 R21N20 R24N20 R27N20 R30N20 R33N20 R36N20 R39N20 Average degrees 83 Figure 6 10 shows the graphical relationship between the image and the real 9 The X
126. he laptop via a universal serial bus USB cable A complete description of the optical sensing system and the webcam can be found in Chapter VI The final hardware module is the laptop and the DIO The DIO is manufactured by SuperLogics and fits into the laptop s personal computer memory card international association PCMCIA slot on the laptop One end of a cable is connected to the PCMCIA card and the other end is directly connected to the IFB board s cable connector The laptop functions as the brain of the robot and inputs signals and images while outputting control commands Section 3 4 presents an overview of the OS and the code design while a detailed explanation of the design and theory can be found in Chapter VII 3 4 Robot Operating System Overview This section presents an overview of the robot s operating system OS As discussed in Section 3 2 the robot s capability for long range movement is broken into three functional parts The OS reflects this division by also having three main sections of code that parallel the division of the robot s functionality Since the robot is a mobile device it is useful to have a visual indicator that can be seen from a distance by the operator The VB6 user interface form on both the desktop client s computer and the robot s server has a color bar that is visible at a distance When the robot is performing one of its functions the appropriate color will be displayed on bo
127. he robot is moving the ROS motor control system uses the four control variables Distance Speed Angle and OffsetRef as defined in Subsection 7 2 4 to control the path taken by the robot When the distance the robot travels ActDist equals the Distance variable value the ROS will read the next command line in the PDB and load four new control variables If the difference between the previous angle variable Angle and the new angle variable Angle is more than 5 the LA controller will be activated The reason that 5 was chosen to be a threshold is due to the fact that the cutoff value had to be less than any potential turn while at the same time allowing the PDB to have some minor variability in consecutive paths of travel Since most turns will not be less than 45 this cutoff value of 5 was determined to be sufficient Early experiments showed that computation of the robot s velocity had large fluctuations in magnitude The cause for this was discovered to be the variable cycling rate of the ROS as previously discussed in Subsection 5 3 1 and the poor resolution of the Hall sensors In order to control for the rate of rotation of the robot body it is necessary to be able to measure the rotation rate 117 Realizing the objective of accurately measuring the rotation rate of the robot begins with the creation of a 5 position buffer used to store the angle of the robot ActDeg every 100 ms Table 7 7 shows how the angul
128. he robot to accomplish its tasks For the purposes of this thesis a peripheral module is defined as a separate program a hardware device or a directory that exists external to the VB 6 main program Figure 7 1 shows how the ROS interacts with its peripheral modules The arrows that connect the blocks that represent the modules show the direction of data flow The two grey arrows indicate that the ROS interacts with its external modules through a hardware device The primary responsibility of the ROS is to control the robot s motion in such a manner as to minimize the positional error from the programmed path In order to accomplish this task the ROS has a large angle LA and a straight path SP controller used to minimize its positional and angular errors As seen in Figure 7 1 the LA and SP controllers are the main focus of the operating system The focus of earlier chapters has been to explain in detail the modules used by the robot The intent of this chapter is to limit itself to the description of the design of the VB 6 main program and its interaction with its external modules The previous chapters were responsible for discussing the more complicated peripheral modules whereas this chapter will complete the description of the remaining modules 92 Robot Operating System ROS Excel Database Motor LA Control Control Chapter 7 Chapter 4 DIO Chapter 7 SP Control
129. herefore the robot has eight forward speeds The box in the upper left of Figure 4 3 illustrates the voltage divider rule that is used to determine the value of the output at the Vap point in the circuit In this circuit R is a constant resistance and R is an adjustable resistance Depending on the resistance value of Ra the output voltage Viap will be divided according Equation 4 1 The actual value of Ra is a sum of all of the resistance values that are between the Viap point in the circuit and the ground potential It includes the 5 KQ trim pot the 1 bit and 3 bit variable resistance blocks Right motor control has the trimpot 5KQ trim pot V tap Equation 4 1 Stop T gt 1 Bit Run Voltage Divider Rule MSB 8 4 bit 3Bi Speeds DIO 3 Bits 7400 I Figure 4 3 Expanded functional description of IFB voltage controller 39 Rc 4 1 Vtap 5 p Rc Ra 4 2 1 Voltage controller output range The voltage controller is able to produce 16 different voltages although as indicated above only nine of these values are actually used The fastest and slowest speeds of the robot were chosen based upon practical considerations The wheelchair s maximum speed is far too fast for an autonomous vehicle intended to be run in a laboratory environment where damage to expensive equipment may occur The maximum speed of the robot was set to a speed slow enoug
130. ies of the robot prior to any implementation of the selected research objective The objective of this thesis was to successfully demonstrate the development of the PML robot in order to meet with the above goals 8 1 Specific Accomplishments The development of the PML robot occurred in three phases e First phase Motor control system and position sensor improvement Second phase Development of the ROS and support modules e Third phase Webcam installation and image processing capability The first phase of development was intended to improve the basic functionality of the robot and its ease of use As discussed in previous chapters the hardware has been completely reworked with the emphasis placed on robustness of operation The previous motor controller has been replaced with one that is more acceptable for use as a laboratory instrument used for controls research The power supply has been reworked with the intent of maintaining safe operating conditions The second phase involved the development of the SP and LA controllers The goal of using both of these controllers was to establish the ability to perform dead reckoning over a moderate distance of approximately 20 ft The third phase was the design and implementation of a webcam based wireless optical system used to correct the robot s positional errors 128 8 2 Limitations and Future Work The automated data extraction of the optical mark s position is the development goal that
131. ingle path Once a single path has been tested it can then be appended with other paths in order to create a complete course made up of path segments Every time the manual operation completes a single loop the robot will travel one path segment When the manual operation is complete the operator will generally be required to move the robot back to its starting position in order to run another test The remote operation code will autonomously complete as many loops as there are path segments As shown in Figure 7 6 each time a loop is completed the Green Yellow and Blue color codes will operate one time The colors indicated in the code loops are the same colors as defined by Table 7 2 These colors represent code segments that 103 exist as part of a case structure in the Halls Compute subprogram Appendix A 1 In the case of the manual operation the code is a copy of the Case Green code that operates in the command start subprogram The manual operation program loop is initiated when the operator clicks the Start button on the Server form on the screen of the laptop In the case of the remote operation loop the program is initiated when the robot receives a wireless color command from the client 7 3 2 Subroutines used by the main program There are 15 subroutines written specifically for the ROS All of these programs are found in the second half of the main program and are listed below Each line begins with the name of the su
132. ion angle of the robot For example when the right wheel advances 1 cm more than the left wheel the robot will rotate counter clockwise by approximately 1 During the IPRV construction the chair was removed and an electronics housing box was attached to the tubular frame in its place With the removal of the chair the height of the PML robot including the laptop is now 23 inches This means that the robot has a very low profile and is very stable The power system of the wheelchair consists of two 12 VDC marine gel batteries The original power configuration for the wheelchair had the two batteries configured in series which yielded a 24 VDC output The power supply was rewired to a 12 VDC system The primary reason for doing this was a concern for the safe operation of the robot In order to run the laptop for an extended period a 12 VDC voltage converter was attached to one of the batteries In addition a component called the IFB also had a 12 VDC input Both of these systems were connected to only one battery During normal operations this would cause this one battery to discharge at a faster rate than the other battery Because the batteries were connected in series the recharge current would be the same through both of them The battery with the lesser discharge would then be overcharged creating explosive hydrogen gas and leading to an unsafe condition 28 The original motor controller that came with the wheelchair required a 24
133. ion path is the beginning of the next main path segment The two waypoints are called the Main path ending waypoint located at the left of Figure 3 6 and Main path beginning waypoint In Figure 3 6 note that the optical mark is placed 61 cm from the Main path ending waypoint The entire length of the course correction is 91 cm 36 in which is 30 cm 12 in further than the position of the optical mark The reason for this is that during the development phase it was determined by experiments that the optimal distance for the camera was 24 inches However this distance was too short for many of the course correction runs to correct for offset position errors When an entire course layout is designed it is necessary to provide a 36 inch segment between every one of the main path segment runs for the purpose of course correction 26 x Mapped X position Y axis error in inches y Mapped Y position error in inches X axis Main path end Optical Lt mark Main path beginning waypoint Main path ending waypoint I l Image distance I 24 inches La gt l Course correction distance 36 inches Robot direction of rt Note that distances are not to scale Figure 3 6 Course correction path description This section describes a sequence of three functions that can be repeatedly run in order to meet the research objective of navigating an obstacle course of any distanc
134. is follows the style of IEEE Transactions on Automatic Control The only sense that the child uses for this task is vision This simple human experiment is extremely difficult for a robot to reproduce For a non mobile robot to catch a ball thrown in an unpredictable manner by using the exclusive sense of vision presents a very complicated engineering problem A basic understanding of this fundamental limitation in the capability of robots is essential in choosing a field of study that will yield a meaningful result In recent years the improvement in processor speeds has allowed advanced research to make substantial inroads in the capability of vision processing systems The ability to process optical information would have many benefits for a mobile robot and would present a good choice for graduate level study 1 1 History This thesis advances the research of Ruzbeh Homji whose master s thesis was titled Intelligent Pothole Repair Vehicle IPRV 2 Homi originated the idea of using a wheelchair that was modified to demonstrate the concept of an autonomous road repair vehicle that would be used to fill potholes This vehicle had a wireless connection to the Internet and could be operated remotely The demonstration model created by Homji used an electric wheelchair that was controlled by a laptop which ran Visual Basic 6 VB6 The laptop had a digital input output DIO card installed that was responsible for reading both ri
135. is similar to the Case Green algorithm After the client computer has evaluated the webcam image this code takes the six digit number output by the client and decodes it to get the error position data It then uses the I controller and the reference error data to correct its position as described earlier At the end of the Case Blue correction the robot s position error should be at its minimum 35 4 The Case Red code is input by the user from the client and will immediately stop the robot if the user becomes concerned that the robot may crash into something The four sections of code that have been identified by a color as just described have the ability to control the VB6 software in order to read and write to the robot s peripherals Figure 3 9 illustrates how the robot OS shown with a heavy dotted line is organized and communicates with its environment There are five different sources for the reading and writing of data Table 3 1 shows what these external sources are if they read or write and which case color is communicating Table 3 1 Data input output sources SOURCE READ WRITE CASE COLOR DATA DATA USED Excel Path Data None Green Webcam Image Bitmap Camera Yellow Commands DIO IFB Hall Count IFB Motor Green Blue Control Wireless Command Colors Image and Echo Receives All Connection Feedback Sends Yellow Text File None Performance Green and Blue Data 36 CHAPTER IV MOT
136. k such as a controller area network CAN that is modular and inexpensive and is used to integrate the controller sensors and actuator nodes 4 11 2 2 3 Multiple sensor control system The use of multiple sensors in a control system would in general represent a high level of controller development From the point of view of the PML robot a control strategy that involved sensor fusion would represent a significant accomplishment that would be worth investigating by a future user Julier and Durrant Whyte conducted a theoretical and empirical study of a vehicle navigation system VNS and the effect of the vehicle model on the navigation system performance 10 They use an extended Kalman filter to fuse the sensor suite In their paper they show a significant portion of their mathematical development of the control system The demonstration vehicle was an autonomous car driven at a maximum speed of 40 mph This study presents some of the methodology that could be used to develop a multiple sensor suite control algorithm for the PML robot Industrial robot arms that are fixed in position and used for industrial tasks seem to have the most sophisticated capability in the use and control of multiple sensors Because of the complexity of these system models it is doubtful that there could be a direct application to the PML robot However it is instructive to consider the more advanced control models in order to get a sense of what may be p
137. le the point indicated on Figure 6 6 is R18P10 During the process of working out the algorithm for this program it was discovered that this webcam has optical distortions and in fact presents an image that is warped This makes the process of the mapping conversion more difficult An example of this warping can be seen in Figure 6 6 Great care was taken to ensure that the measurements of the arcs seen in this figure were accurately drawn In addition the robot and the webcam were accurately placed and adjusted in order to ensure that the image field would yield precise data Note that the R21P30 intersection can be seen in the above image but the R21N30 intersection cannot be seen This demonstrates that 75 even when the camera is accurately adjusted there is a warp in the image This was the first indication that the image field was not exact During the development of the mapping algorithm a review of the data showed that the image field had warping in the lower left corner Once the data conversion methodology was developed it was determined that the errors caused by the warping were tolerable and the camera did not have to be replaced The obvious solution to this problem would be to replace the existing camera with one that did not have a warped field Figure 6 6 was used to generate data that relates the intersection positions in real space to a position in the image space In order to create this data set the image was sized to 154
138. lt choice Choosing a unique topic of research in the field of robotics and visual processing proved to be very challenging A review of robotics literature with the primary attention focused on visual navigation reveals a field of research with a high degree of development and specialization The design and building of an artificially intelligent robot capable of autonomous movement is typically conducted by an engineering team composed of specialists from several areas of study It would not be reasonable to assume that this thesis could compete with this level of hardware cost and technical specialization The novel idea that this research presents is the concept that an inexpensive mobile robot could be used in a factory type setting This robot would have the following hardware requirements low cost low resolution position sensors low cost low resolution vision sensors low cost onboard processors e low cost wireless communication capability 127 This robot would be able to self correct for its positional errors by sending an optical image at a low frame rate to a central processor This remote computer could have a high level of processing capability and sophisticated mapping algorithms and would be responsible for returning the processed error data to the robot An early evaluation of the IPRV showed that there were serious limitations in its basic mobility Considerable time and effort was spent upgrading the basic capabilit
139. lumn elements of Table 6 3 are the image radius measurements for each of the eight arcs The average image radius for each arc is displayed on the last row of the table Eight data pairs are created from the table and are shown on the graph of Figure 6 11 Note that the result is non linear and the equation that is the best fit for the data is a second order polynomial Similar to the 0 relationship the function for the radius also has a R value that is close to one The significance of this correlation is that it is possible to map the image radius to the real radius with the expectation that the accuracy in the transformation will be close 85 Image Radius Real Radius 42 39 36 33 oO P 30 3 ke amp 27 f y 0 0023x 0 9278x 110 19 24 gt R 0 9996 21 18 15 T T T T T T 1 220 230 240 250 280 290 300 310 260 270 Image Radius mm Figure 6 11 Graphical evaluation of radius functional relationship Now that Rp and Og are known the unique point in real space that represents the center of the optical target is also known This point in the polar coordinate system is then transformed to the Cartesian coordinate system The error distance from the desired target center of Xr equal to 61 cm 24 in and Yr equal to 0 cm can then be easily computed 6 3 7 Development of one to one map with a non zero a Until this p
140. me efficiency instead of other design criteria The variable that the SP controller uses to track the desired path was changed when it was discovered that systematic errors could not be eliminated in the earlier controller design This change required a significant alteration in the structure of the program 3 2 System Overview The intent of section 3 2 is to present an overview of the PML robot s entire system The approach taken is to focus the discussion on what the PML robot does from the point of view of its functions The following two sections will discuss how the PML robot performs these functions by presenting an overview of both the hardware system and the robot s software operating system As previously discussed the PML robot is designed to have the ability to travel a long distance and end with an acceptable positional error If this objective is viewed as a function or capability of the robot it can be broken down into three sub functions or parts e Dead reckoning function Optical correction function 18 e Course correction function 3 2 1 Dead reckoning overview In order to maintain its accuracy over the entire long distance course of travel the robot divides the whole path into several main path segments In each individual segment the robot uses its dead reckoning capability to follow the intended path As discussed in Chapter I dead reckoning is the ability to extrapolate the current position by taking
141. ment One possible sensor system that would seem fairly easy to implement in the PML robot would be an ultrasonic based system discussed in a paper by Bank 12 The idea presented by this paper used 24 ultrasonic transmitter receivers mounted symmetrically in a 360 ring around the robot The advantage of Bank s system is that multiple sensors gather different elements from the reflection of sound echoes off of the object being measured These multiple measurements allow for better resolution of the object The multiple signals are sent through a DSP board that serves as a front end image pre processor as discussed in Section 2 2 1 and a multiplexer that outputs a singly composed sensor signal to the main processor for final signal representation 12 This system could potentially be used by the PML robot as its secondary sensor input for all subsequent programming based upon the raw environmental data An ultrasonic sensing method is one possibility for gaining knowledge about the environment Another possibility is the use of a video camera A paper by Pears Liang and Chen presented a novel method of processing a raw video signal that comes from an un calibrated monocular camera 7 Ideally the visual image would be passed through a hardware based front end pre processor prior to the image being sent to the main 13 processor The authors describe their system as an algorithm that uses all of the information that is in the image stre
142. ments robot_lap temp bmp Call SavePicture Picture1 Picture C Documents and Settings All Users Documents robot_lap temp bmp waitTime 500 wait 1 second Call SendMessage hCap WM CLOSE 0 0 Dim temp As Long 160 temp SendMessage hCap WM CAP DRIVER DISCONNECT 0 amp 0 amp waitTime 500 wait 1 second With PicWebCam Visible False End With With picStore Visible True End With tepServer SendData stream send keyword to client so image is displayed End Sub XX XXXX XX XX XXX XX XXX XXX XXX XXX XXX XXX XXX XXX XX XX XXXX MotCont If Turn lt gt OldTurn Then is the condition for this sub to work XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub MotCont Print 1 MOTCONT 3 Print 1 Tab 12 Inside MotCont Turn Turn OldTurn OldTurn Print 1 If PControl True Then if so recompute the turn command variables Only use when RECOMPUTE THE TURN VARIABLES GIVEN FwdSpeed and TurnDelta If FwdSpeed 7 Then 7 is the slowest speed Right_R 7 Left L 7 Else FwdSpeed is less than 7 Right R FwdSpeed 1 example if speed 6 then slow wheel goes 7 Left_L FwdSpeed 1 End If Right_L Right_R TurnDelta Left R Left L TurnDelta Print 1 Right R Right R Right L Right L Print 1 Left R Left R Left L Left L 161
143. mp H10 Public Const WM_CAP_GRAB_FRAME As Long 1084 Public Const WM_CAP_EDIT_COPY As Long 1054 Public Declare Function capCreateCaptureWindow _ Lib avicap32 dll Alias capCreateCaptureWindowA _ ByVal lpszWindowName As String ByVal dwStyle As Long _ ByVal X As Long ByVal Y As Long ByVal nWidth As Long _ ByVal nHeight As Long ByVal hwndParent As Long _ ByVal nID As Long As Long Public Declare Function SendMessage Lib user32 _ Alias SendMessageA ByVal hWnd As Long ByVal wMsg As Long _ ByVal wParam As Long ByRef lParam As Any As Long Public Sub CamToBMP Picturel As PictureBox On Error GoTo error Call SendMessage hCap WM CAP GRAB FRAME 0 0 Call SendMessage hCap WM_CAP_EDIT_COPY 0 0 Picture1 Picture Clipboard GetData Exit Sub 169 error End Sub A 3 PCMMain bas Option Explicit Option Base 0 Global Const L_Hall 0 4 bit input left Hall SW Global Const R_Hall 1 4 bit input right Hall SW Global Const Count_CLR 2 I bit output Z clear for 191 Global Const Count MM 3 2 bit input Max Min from 191 Global Const MC_L 6 4 bit output MC IFB left module Global Const MC_CLK 5 1 bit output MC clock both modules Global Const MC_R 4 4 bit output MC IFB Right module Global Const Solenoid 7 1 bit output Solenoid Public intStatus As Integer Public gintlogicaldevice As Integer Public flag As Byte Private Sub main Dim LogicalDevice As Integer Dim ErrorCode As Integer
144. n If incPC 4 Then determines the second array increment value incPC2 0 End If PTime timel closest time when ActDeg was measured DegHistory incPC ActDeg write the current orientation into the data store PRatio ActDeg DegHistory incPC2 find diff in orientation over 1 2 sec Print 1 PRatio Block dj Print 1 incPC incPC incPC2 incPC2 Print 1 Test PTime PTime DegHistory incPC DegHistory incPC Print 1 PRatio PRatio DegHistory incPC2 DegHistory incPC2 Print 1 pratio i Control the array index incPC incPC incPC 1 If incPC gt 5 Then run incPC from 0 to 4 incPC 0 End If 153 End If Time gt PTime PTURN BLOCK STARTS this block inputs the PRatio to the output of TurnDelta Use a linear function VARIOUS EQUATIONS OF ROTATION CONTROL LAW i TURN DELTA This is the 2nd order equation designed to correct turns gt 85 having low response and will go 3 lt 30 0 0011x2 0 1489x 7 1691 TurnDelta Round 0 0011 PRatio 2 0 15 Abs PRatio 7 17 This is the 3rd degree inflexion equation TurnDelta Round 0 00004 Abs PRatio 3 0 006 PRatio 2 0 26 Abs PRatio 6 Control equation for P Control we nenn nn ne anne nnn nen LINEAR EQUATIONS
145. n 1 OldTurn 0 timel 0 78 inside HallsRead time2 0 78 ActDist 5 5 ActDeg 1 timel 0 8 inside HallsRead time2 0 8 ActDist 6 ActDeg 0 Start HallsCompute 10 OffsetX 8 726195E 02 OffError 0 DegRef 0 DegError 0 191 SpecTurn 0 LTum 1 RTurn 1 Motlnc 2 Left Ratio 0 5 Right Ratio 0 5 Turn 0 OldTurn 1 End HallsCompute MOTCONT Inside MotCont Turn 0 OldTurn 1 timel 0 81 inside HallsRead time2 0 81 ActDist 6 ActDeg 0 timel 0 82 inside HallsRead time2 0 82 ActDist 6 ActDeg 0 timel 0 83 inside HallsRead time2 0 83 ActDist 6 ActDeg 0 timel 0 85 inside HallsRead time2 0 85 ActDist 6 ActDeg 0 timel 0 85 inside HallsRead time2 0 85 ActDist 6 5 ActDeg 1 Start HallsCompute 10 OffsetX 0 1745239 OffError 0 DegRef 0 DegError 1 SpecTurn 0 2 LTum 1 RTurn 1 Motlnc 3 Left Ratio 0 3333333 Right Ratio 0 3333333 Turn 0 OldTurn 0 timel 0 87 inside HallsRead time2 0 87 ActDist 6 5 ActDeg 1 timel 0 89 inside HallsRead time2 0 89 ActDist 6 5 ActDeg 1 timel 0 9 inside HallsRead time2 0 9 ActDist 7 ActDeg 0 Start HallsCompute 10 OffsetX 0 1745239 OffError 0 DegRef 0 DegError 0 SpecTurn 0 LTum 1 RTurn 1 Motlnc 4 Left Ratio 0 25 Right Ratio 0 25 Turn 0 OldTurn 0 timel 0 9 inside HallsRead time2 0 9 ActDist 7 ActDeg 0 timel 0 91 inside HallsRead time2 0 91 ActDist 7 ActDeg 0 timel 0 92 inside HallsRead time2 0 92 ActDist 7 ActDeg 0
146. n Marcos Texas 2000 M S Mechanical Engineering Texas A amp M University 2007
147. n pulses per wheel revolution as discussed in earlier chapters Table 7 5 Basic measurement variables Name SI Unit Formula Description Equivalence LHall z 0 997 cm Not Applicable Accumulation of left Hall sensor pulses RHall z 0 997 cm Not Applicable Accumulation of right Hall sensor pulses ActDist 0 997 cm LHall RHall Average of the distance ws traveled by both Hall sensors ActDeg 1 063 LHall RHall Difference between both Hall sensors 107 Both actual distance ActDist and actual degree ActDeg are conveniently close to the standard measurement of a single degree of arc and a single centimeter of distance respectively For general purposes ActDist is treated as though it is equal to 1 cm and ActDeg is treated as though it is equal to 1 This result was not intentional and is only a product of the actual physical dimensions of construction It should be noted that from this point on any reference in this thesis to any variable will be in terms of the ad hoc units described in Tables 7 5 and 7 6 7 4 2 SP controller error signal computation The error signal used by the SP controller is called offset error OffError As seen in Figure 7 7 the offset error is defined as the distance between the robot and the hypothetical straight path plus the intended offset reference OffsetRef value Note that this distance is in the Y axis direction and each theta 6 is equal t
148. n using a 12 V power supply was significantly less abrupt than when using the 24 V system It therefore seemed obvious that a 12 V system would be easier to control than the original 24 V power supply Since the new motor controllers could handle a range of PWM output voltages there was no reason to keep the 24 V configuration A new 12 V re charger was acquired and the appropriate cabling was used to facilitate easy recharge of the robot A basic diagram of the new power supply system is show below in Figure 3 7 29 g MC Left DPDT Switch Laptop Doo lr OO Power 15A L_ Supply fuse 7 2 ea MC Power oa f So Switch Me Rigt gt 120 VAC to Lo OO 16 VDC ii Len 477 Plug converter arn oo E Drive Main power 77 Motor j f connector 10 A fuse 12 VDC 120 2 ea Right Jones VAC converter Drive Plug ad Motor 2 SP Jones 80A Plug 80A fuse fuse AN ic I AN E Battery 1 Battery 2 12 VDC Charger with 120 VAC Supply Figure 3 7 System diagram of the 12 V supply 3 3 2 System hardware overview In Section 3 2 an overview of what the robot could do w
149. nd rotational errors is discussed in Chapter VIL The first step of the data conversion program will set a equal to zero take the Xj and Y image values and map them into the real X and Y values in Figure 6 12 The radius Rr and angle Og are computed as an intermediate step in the mapping transformation The angular error a is measured through the data extraction module as part of the 87 image evaluation and is known at this stage in the procedure As seen in Figure 6 12 the actual distance between the physical origin just below the camera and the optical mark is independent of the rotation angle This equality is indicated by setting R equal to Ro The three equations below explain how the transformation of the error data 01 Ri a is mapped into the non zero a error data 62 Ro a Q 0 6 5 X R xcos 6 6 Y R xsin 0 6 7 At this point in the data conversion module the three error values X2 Y2 and a are known and are then passed to the error data encoding program see Figure 6 5 The next sub section will discuss the procedure used to encode these three error values 6 3 8 Error data encoding The second program of the data conversion module is responsible for encoding the real vector data into a single 6 digit number This subsection will present how this program was developed Figure 6 13 illustrates how the encoding scheme works The center cross represents the desired stopping position of the robo
150. ng half of Chapter VII explained the LA and SP controllers The development history of both controllers was included in order to give some insight into the use and structure of the control algorithm Experiments were run to test the effectiveness of both controllers described in this chapter In the case of the SP controller an experiment was conducted that induced a disturbance torque and the effect was measured The amount of the torque was much greater than would occur during normal operation In the case of the LA controller measured courses with right and left angled turns were set up and data was collected This data shows that the robot is able to maintain a consistent rate of rotation during turns which allow for accurate path planning 126 CHAPTER VIII CONCLUSIONS AND SUMMARY This is the second master s level research project to use an electrically powered wheelchair as a base for an autonomous robot The first research project using this hardware was conducted by Homji whose thesis was titled Intelligent Pothole Repair Vehicle IPRV 2 Homji originated the idea of using a modified wheelchair to demonstrate the concept of an autonomous road repair vehicle that would be used to fill potholes When he finished his research the author of this paper took responsibility of the IPRV hardware and began his own research project The selection of a novel research topic that relied upon the existing hardware proved to be a difficu
151. ngle Dim HallCntRight As Byte Dim HallCntLeft As Byte Dim RRip As Integer Dim LRip As Integer Dim PrePulseR As Boolean Dim PrePulseL As Boolean Dim DegRef As Single output of control law w input OSE Dim ActDist As Single Dim OldDist As Single Dim DelDist As Single This is the delta distance Dim DegError As Single Both and differance in reference from actual Dim DegSign As Integer value either or 1 Dim OffError As Single OffsetX OffsetRef Dim absOffError As Single use for the control equation must have absolute value Dim OffsetX As Single summed fractional distance both and Dim ActDeg As Single fractional in degrees both and Dim DelX As Single incremental X error Dim OldTurn As Integer Compares with Turn to indicate a change Dim MotInc As Single define as Single due to equation Dim Turn As Integer a 1 indicates a left turn and a 1 indicates a right turn Dim RTurn As Single Dim LTurn As Single increments each trip through Turn sub routines Dim SpecTurn As Single output of Spectrum Block Dim DegRefFlg As Boolean change sign of RefDeg if OSE is negative Dim FirstLoop As Boolean identifies 1st loop of new Case Blue Green i P CONTROLLER VARIABLES Dim TurnDelta As Integer values 1 gt 6 add to Turn Variables this affects turn rate Dim OldTurnDelta As Integer stores the previous value for write to MotCont Dim PControl As Boolean Flags for P
152. nsing system several changes in the design were implemented and then tested The errors were gradually reduced as the entire system 62 was improved The addition of the NAND ICs to this circuit helps to create a more robust and trouble free position sensing system 5 4 Summary of Electronics System Chapter V discussed the 5 V power supply and the position sensing system that is installed on the IFB By the end of this chapter all of the functions of the IFB have been described The reason for the 4 bit counter IC is presented along with the circuit diagram of the position sensing system What remains to be discussed is the input and output connection to the IFB The DIO connection to the IFB is discussed in Chapter VII as part of the ROS 63 CHAPTER VI OPTICAL CORRECTION SYSTEM 6 1 Introduction The optical correction system OCS is designed to reduce the positional error that might occur by the end of each dead reckoning path Subsection 3 2 2 presented the first discussion of the optical method used to correct for positional errors As stated earlier there are four basic tasks necessary to complete the optical correction part of the program e Capture image e Send image e Process image Return correction data Figure 6 1 is a functional diagram that implies the OCS is outside of the ROS In actuality the OCS exists on both the client and server computers From the point of view of the ROS the optical system is external
153. nstalled on the robot s server computer This utility allows the user to define the number of bits in each channel and define if the channel is to be used as an input or an output One limitation of this PCMDIO device is that each channel must be defined within each port In other words the channel cannot exist on two ports As an example the single bit Channel 5 is placed between Channel 4 and Channel 6 General A D D A Digital 1 0 Configuration i xi Digital 1 0 Digital Input Output Selectable Ports 94 Select Page fo m Port 0 Bit 5 4 2 30 ir fin fin fi fi fr fr fn a PPP Pp pe Port 1 Bit U eA S Ao ee 1 ou ou ou ou ou ou aPFrER REP EE Port 2 Bel ae 46 25 436 ee HED gt gt gt Se af PPP ee be PCMDIO 24 channel digital 1 0 Figure 7 3 PCMDRIVE configuration utility Table 7 1 lists the channels as they are organized for use in the PML robot The channel numbers in the table can be compared to the channel numbers in Figure 7 3 for consistency Table 7 1 PCMDIO channel allocation Logical Bits Input Function Channel Channel Output 0 4 Input Left Hall sensor 1 4 Input Right Hall sensor 2 1 Output Counter chip 74LS191 clear 3 2 Input Counter max min and ripple 4 4 Output Right motor control speed value 5 1 Output Latch 54LS175 for MC 7 speed 6 4 Output Left MC 7 speed value 7 1 Output Currently not used
154. ntegrates all of the smallest measurement errors and computes the sum to get the total error or OffsetY value In this manner it functions in a very similar way to an integral controller One more step is needed before the offset error signal is determined The OffsetRef variable is used by the SP controller as a way to determine the perpendicular distance that the robot will track along a line that is 109 parallel to the 0 centerline When the OffsetRef value is a positive the robot will track to the right of the hypothetical 0 line A negative OffsetRef value will cause a parallel track on the left side of this center line The objective of the controller is to drive both the OffError and the ActDeg values to zero When this is true the robot will be on the correct path determined by the offset reference value and have the correct angular orientation determined by the PDB 7 4 3 One to one map between OffsetRef and RefAng The previous subsection gave a description of how the OffError variable was computed The SP control algorithm is responsible for reducing the path separation distance while keeping the robot on the correct heading Figure 7 8 presents a second order polynomial used as a one to one map that relates the OffError to the reference angle RefAng variable The RefAng variable describes the response to the magnitude of OffError It should be noted that the RefAng will always have an opposite sign than the OffEr
155. nts and Settings All Users Documents robot_lap primary txt For Output As 1 Call Initialize2 starts the motion Indx 90 reset for subsequent Case Green runs use for troubleshooting Indx 10 is the row location of Ist call line 100 on Excel While bInRun True Loaded in the Initialization If NewLine True Then only do when newline Call ReadTable If bInRun False Then GoTo 1000 Call HallsRead do every loop DoEvents If ActDist gt OldDist Then do with new distance info Call HallsCompute End If If Turn lt gt OldTurn Then only do when new command Call MotCont End If DoEvents 1000 Wend ends the start while loop Unload Me When end is reached in Excel table and dist End Sub This is the end of the Click Event for the CmdStart button XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX 142 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Private Sub cmdStop Click blnRun False End Sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX 1 Reads Excel table 2 Loads Variables 4 Increments Index Set conditions for the next read XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub ReadTable TrigCol 2 column 2 is the distance trigger value from the table TrigFwdCell R amp CStr Indx 10 amp C amp CStr TrigCol TrigFwdCell is a String leave no spaces between the string elements With TxtDist This is the format for reading a cell in an Excel
156. nufacturing environment 5 It would be very difficult to add a visual processing capability to the PML robot which would match the most advanced type of vision processing schemes found in the most recent generation of manufacturing robots However a basic understanding of the more complex systems would set a high end goal for any future development For example one high level area of development would be the use of a perceptual control manifold PCM that would combine the dynamics and system model of the robot with the sensor data in a mathematical space 6 2 2 Developmental Challenges To achieve any of the goals listed in the previous section it is necessary to develop key abilities This section focuses on areas of development in both computation and hardware that are needed to accomplish the goals above Any of the challenges presented below could be a potential topic for research with the PML robot 10 2 2 1 Hardware based front end image pre processor One of the strategies developed to deal with a complex video stream is the use of a hardware based front end image pre processor Pears et al propose that a field programmable gate array FPGA or a special purpose digital signal processor DSP be used as a high bandwidth raw video feed low level feature extractor to be implemented as a hardware layer with parallel processing capability 7 Because real time image processing requires a great deal of computational capacity
157. o the ActDeg value at the moment of measurement Direction of Travel 68 gt ARDEN DelDist Centerline 0 Position Start 5 T DelY4 OffsetRef zh 7 Si DelDist Direction DelY2 Babit Offset Error Angel o 8 0 DelYs gt Actual path re Y Axis Position and X Axis Direction Direction of Travel Y Axis Figure 7 7 Error signal used by SP controller 108 In order to compute the OffError signal several new variables need to be determined These variables are found in Table 7 6 As can be seen from Figure 7 7 and Table 7 6 the offset error is an accumulation or sum of the DelY offset values The equation used to determine the OffError signal is seen below OffError gt DelY OffsetRef i n Table 7 6 SP controller variable definition 7 1 Name SI Formula Description Equivalence DelDist z 0 498 cm ActDist OldDist Typically equal to 1 2 ActDist unit DelY 0 1104cm DelDistxsin ActDeg Smallest unit of distanc e in Y axis from 0 path OffsetY 0 1104cm DelY Sum of error from 0 i n path Offset z 0 1104 cm Not Applicable PDB input determines Ref parallel path offset OffError 0 1104 cm OffsetY Offsetref SP control error signal DegError 1 063 RefAng ActDeg Defined as difference between desired reference angle and actual heading angle In effect the SP controller i
158. of Rg Figure 6 9 demonstrates that this second assumption is also true Each of the five lines on this figure are also derived from Table 6 1 data The purpose of this figure is to relate the real radius position from the grid to the measured 0 from the image The X axis of Figure 6 9 displays the known radius positions on the grid Each line of radius from the grid has eight intersections at 7 5 cm 3 in intervals therefore there are eight points on each line in this figure The Y axis of this figure displays the 0 values measured from the image As an example the point on the graph that relates to the R18N20 intersection can be found at the leftmost point of the 20_Theta line in the figure As was true in 81 the previous case the proof of decoupling is seen by the horizontal lines in Figure 6 9 As the real radius is varied the angle of the image 0 remains constant Theta Independent of Radius 15 00 A Soe 4 amp 77 amp A A 10 00 T 5 00 me 8 s 8 2 D 7 0 00 i o H amp g 5 00 Real Theta e E E e E OR eee degrees 20 Theta oO 4000 10 Theta 0 Theta a ee ee ee a 10 Theta A 20 Th 15 00 0 eta 15 18 21 24 36 39 42 45 27 30 33 Real Radius inches Figure 6 9 Graphical proof of second condition for decoupling 6 3 6 Development of the one to one map equation Before either of t
159. of run Case yellow use to run the Camera Call RunCamera runs camera when yellow is sent 140 Case blue use to input correction and run correction frmServer BackColor vbBlue PControl False enable so CorrectCompute will do entire program Call Initialize2 load variables CaseBlueFlg True disable EMO reset in Initialize2 CorCarryDeg 0 necessary to make ActDeg 0 so correction is true DirAngle 0 in order to force the ActDeg 0 upon new Case Blue CorrectIndex 1 reset each time into Case Blue flagDB2 1 the case blue flag While blnRun True DoEvents Print 1 Print 1 inside While loop Blue bInRun blInRun Call HallsRead finds ActDist and ActDeg If ActDist gt OldDist Then Call CorrectCompute End If DoEvents If Turn lt gt OldTurn Then Call MotCont End If the datasend is operating Wend blnRun True End Select xxxxxxXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXK error Print 1 Tab 12 end of dataArrival block strData strData strData blue this is a debug line use to discover if End Sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX 141 Private Sub cmdStart_Click Use this command button to run individual Case Green runs It should not close the form or call the camera Case Yellow Close 1 only keep the latest form when using the manual start during course design Open C Docume
160. oint in the development of the data conversion program a has been set equal to zero A discussion of the procedure used to incorporate a non zero a is presented below Figure 6 12 shows a real space representation of the robot s actual position at the end of its dead reckoning run in relation to its desired stopping position In this case the robot has stopped to the right of the centerline and closer than the desired 61 cm 24 in With the robot at the indicated position two direction angles D and D2 are considered in the evaluation of a non zero a On the right side of Figure 6 12 there are two screen images that represent the webcam image taken at the two different angles 86 Optical Mark Screen Image 1 Position D1 Desired Stop Screen Image 2 Position is 24 inches from center of Optical Mark Actual Stop Position X Real Position D2 Y Real Figure 6 12 Strategy used to convert non zero error data The strategy used by the data conversion module to correct for rotational errors has two basic parts The first part is to find the orthogonal distances X and Y while assuming that a is actually equal to zero The next part is to transform this pair of numbers using the actual alpha into X and Y2 The three error values X2 Y2 and a are then encoded and passed to the Server side optical correction module An explanation of how the robot uses this data to correct its positional a
161. oller During early experiments the SP controller worked well at following a straight path in the intended direction The controller was able to follow the OffsetRef parallel line as designed Although the robot was able to travel in a straight direction with the appropriate heading the robot body had a tendency to oscillate by 2 as it traveled This oscillation resembled in appearance what a car would do on a wet street under acceleration For purposes of definition this behavior is called a fishtail motion The solution to this undesirable motion was to supplement the SP controller with a small angle control algorithm 111 The source of this problem was investigated and found to be caused by an over response by the motor control system at angles near the limit of measurement It should be noted that there are fundamental limitations in the hardware that make it difficult to eliminate this fishtail behavior Because of the low resolution of the Hall sensors the minimum angular measurement is limited to an arc of 1 It is impossible to control to a level of accuracy beyond the minimum resolution of the primary measurement input The fishtail behavior became noticeable or visible when the oscillation reached 2 The original SP controller that used OffError as its error signal was not able to control for an angular error of less than 2 The goal of the supplemental small angle controller is to reduce the oscillation to 1
162. on the motor rotor disk The selection of three magnets instead of four was done in order to make sure that there was as little overlap of the magnetic flux as possible 6 0 5 0 4 0 3 0 Volts 2 0 0 0 Time ms Figure 5 4 Signal output from right Hall sensor 5 3 5 Use of the NAND IC in the position sensor circuit The circuit diagram in Figure 5 3 shows that the output of both Hall sensors is first input into a NAND before the signal is sent to the counter IC Figure 5 5 shows the same 61 signal as in Figure 5 4 after it has been conditioned by the NAND The three effects of conditioning the signal are listed below followed by their explanation Invert the duty cycle e Adda Schmidt trigger to the Hall sensor output for a threshold control e The output signal has noticeably less noise 6 0 5 0 4 0 3 0 Volts 2 0 0 0 Time ms Figure 5 5 Signal output from right NAND The inversion of the duty cycle may be considered to be a cosmetic change however a duty cycle of less than 50 is considered to be desirable The Hall sensor comes with its own embedded Schmidt trigger which allows for the output to be strictly on or off With the addition of the NAND Schmidt there is another level of threshold filtering in an attempt to insure that the input to the counter ICs will be uncorrupted During the development of the position se
163. onless units i 99 99 50 24 00 Encoded Error Number 782445 Figure 6 13 Error data encoding scheme The center of the error window is the desired stopping position and represents an error of zero in both the X and Y directions This center position has an X and Y dimensionless value that is equal to 50 A value greater than 50 is interpreted as a positive error and a value less than 50 indicates a negative error 89 The 6 digit number is grouped into three 2 digit values The first group represents the X error the second group represents the Y error and the third group represents the angular error The sign of the error value is defined in terms of the correction distance or angle For example in this case the robot stopped short of the mark and is required to go an extra distance in the course correction step Distance will then be added to the X distance A 2 digit number group representing the X error has been computed in Figure 6 13 The X error has a value of 5 6 inches which is divided by 0 2 in The quotient is added to 50 because the addition of distance is needed in order to correct for the error in position The encoded value for the X error as shown in this figure is 78 and constitutes the first two numbers in the 6 digit code The same algorithm is applied to both the Y error and angular error the resulting two number pairs are then appended to the X error This process yields the 6 digit encoded error number 6 3 9 Us
164. ossible Two research papers by Sharma and Sutanto 6 and Chaumette and Hutchinson 11 presented methods of using a visual sensor to control an industrial robot with the goal of intercepting an object in a 3 D space The first paper discusses the development of a perceptual control manifold PCM that is an expansion of the configuration space to include a set of visual parameters or image features When the control loop is iteratively solved the video signal is seen as an input to the system The PCM space would take a large amount of computational requirement and special knowledge in order to implement 12 Chaumette and Hutchinson described a visual servo control method from two points of view An Image Based Visual Servo control IBVS is compared to a Position Based Visual Servo control PBVS using stability analysis and performance criteria This paper discussed the mathematical development of both models with the performance goal of a visual intercept of an object in 3 D space 11 Both of these control algorithms use a visual feature set that is included in the process control model 2 2 4 The development of an inexpensive secondary sensor system After the Hall sensor based position and direction control is complete the next step would logically be the addition of a secondary sensor system The second system should have several potential applications such as the ability to avoid obstacles or the ability to map the local environ
165. ot would swing wide during the high angle turns During this stage in the controller development it was assumed that when a turn was implemented at a set base speed the turn radius would always be the same Because the actual response of the MC 7s was not ideally consistent the robot would make inconsistent turns During high angle turns the heading rotation was not consistent Because of this inconsistent rotation the robot would exit its high angle turns with a different heading each time The solution was to devise a controller that would control the rotation rate during large angle turns 116 There is an important difference in the fundamental problem that each controller will face When the robot is operating under SP control a combination of right turns left turns and straight motion commands will be used to maintain the planned path The bottom chart of Figure 7 11 shows how the SP controller uses a combination of turns and straight motion during its normal operation The LA controller is different than the SP controller due to the fact that during a large angle turn the LA controller will only use a constant turn If for example the robot is making a 90 turn to the right the motor control system will only command a right turn Therefore the method used to control the rate of rotation must be able to vary the difference in speed between the right and left sides 7 5 1 Design of LA controller During normal operation when t
166. pboard GetData Exit Sub error End Sub RR 8 8 E 2S FAS E E E K 2S E E E 2S 2S 2S E E E RR FS 2S 2S 2S E E k E 9S 2S E 2S 2S 2S 2g E k k k 2S E 2S k k k k k k k k k k Procedure waitTime Purpose Waits for a time specified in milliseconds Inputs MilliSec wait time in milliseconds E E E E E 28 28 RR E E E E E E E E E RR E E E E E E E E E E E E 2S 2S E E E E E E E E E E E 2S 2G 2 2 2 2 2S 2S 2K OK EK Public Sub waitTime ByVal MilliSec As Long Dim oldTime As Long oldTime Timer 1000 Do DoEvents Loop Until Timer 1000 oldTime gt MilliSec End Sub 180 181 APPENDIX C EXAMPLE OF OPERATIONAL DATA FILE TYPE 1 This appendix shows the result of a primary type operational data file as described in Table 7 3 Data from this file has been truncated only the first second of SP control is shown The next data to be shown after this skip in time begins at 4 s and demonstrates the Primary file output during LA control Inside sub Zero Inside Sub Neutral data arrived green CorCarryDeg 0 inside green ActDeg 0 DirAngle 0 Turn 0 OldTurn 0 Inside sub Zero Distance 52 FwdSpeed 7 DirAngle 0 OffSetRef 0 timel 0 07 inside HallsRead time2 0 07 ActDist 0 ActDeg 0 Start HallsCompute 10 OffsetX 0 OffError 0 DegRef 0 DegError 0 SpecTurn 0 LTum 0 RTum 0 MotlInc 1 Left Ratio 0 Right Ratio 0 Tum 0 OldTurn 10 End HallsCompute MOTCONT Inside MotCont Turn 0 OldTurn 1
167. ping algorithm The first program of the data conversion module is responsible for mapping a vector in the image space into a vector in the real space This subsection will discuss how the mapping program works and the way in which it was developed A first set of experiments indicated that the best way to represent the image space would be to use a polar coordinate system A polar coordinate grid was constructed and placed on the floor in front of the robot and its webcam Figure 6 6 is a webcam image of one of the grids used to produce the data needed to develop the mapping algorithm The point just below the camera lens was marked by the plumb bob and served as the origin for the both the robot and the camera The grid is composed of concentric arcs that are three inches apart and drawn on white craft paper The closest arc to the camera has a radius of 46 cm 18 in and the furthest arc is 99 cm 39 in from the origin The orange centerline is defined as having an angular measurement theta 0 of 0 The black radial lines fan out from both sides of the centerline every 10 74 Figure 6 6 Webcam image of polar grid Each intersection on the image has a designator used to define its unique position The first part of the designator is the radius value and is indicated as R The second part is the angular measurement and is indicated as either P for positive or N for negative with the centerline indicated as PO For examp
168. r The design of the cross is printed in Figure3 2 A webcam was attached to the front of the robot and is angled in a manner such that a mark that is exactly 61 cm 24 in in front of the robot will be at the center of the image This optical target was placed on the floor 61 cm in front of the intended ending position of the dead reckoning path If the robot misses its intended final position the image of the cross in the picture will not be located in the center As seen in Figure 3 3 when the robot is in position A left and short of desired position the image on the screen will look like the one in Figure 3 4 right and long of center 21 Position A at camera Robot s forward motion direction NS Direct line of sight Y real error is Desired path ending position 24 Inches Figure 3 3 Top view of robot in relation to optical mark When the robot operates in the normal mode it listens for a wireless command from the user In transmission control protocol and Internet protocol TCP IP the computer that listens is called a Server and the computer that requests is called the Client The robot is therefore defined as the server and the desktop user s computer is defined as the client The second step of the optical correction function is to send the image to the client computer for processing The web camera is connected to the laptop through its USB port and is controlled by
169. r a run of 20 feet will cause a positional accuracy error greater than four inches The conclusion is that the accuracy of the Hall sensor count is critical to the operation of the PML robot as defined in this research The desired goal for accuracy on the position count is zero errors in the pulse count over a standard run length In order to determine the minimum cycle speed required several facts need to be known e Maximum wheel rotation speed see Table 4 3 e Number of pulses per wheel rotation see Subsection 4 2 2 Nyquist s rule for sampling a pulse signal 17 Table 4 3 indicates that the maximum wheel rotation found during testing is 46 5 RPM Also in Chapter IV it was noted that the gear reduction ratio was 32 1 With each rotation of the motor rotor there will be three pulses from the magnets Combining these facts means that there will be 96 pulses for each rotation of the wheel Equation 5 1 calculates the maximum expected frequency of the Hall sensor pulses when the robot is moving at its fastest velocity It should be noted that when the robot is traveling at its fastest speed there will be two completely independent pulse signals that operate near this rate Lae 6 pulses x 1 min 4 pulses 46 5 9 74 min rev S S 5 1 58 A frequency at this rate corresponds to a pulse signal length of 13 ms According to Nyquist s theory the sampling rate should be at least twice the frequency of the underlying si
170. r user commands The visual interface used by a VB 6 program is called a form Figure 7 5 shows the client form as seen by the operator of the robot After the basic connection has been established between the client and the server computers the operator clicks the CONNECT button on the client form This will prompt the client and server VB 6 programs to connect using the existing wireless network Once program connection has been established the operator has 98 control of the robot through the client interface color coded buttons Table 7 2 lists the buttons and their command purpose Table 7 2 Command button description Button Name Command Purpose CONNECT Connects the VB 6 client and server programs UNLOAD Shuts down the client form GREEN Start Begins the programmed dead reckoning path YELLOW Takes an image when robot is at rest Camera RED Stop User commanded emergency stop of robot motion BLUE Send_ After user image evaluation cues program to send data Data and begin course correction path Because the robot may not be in the line of sight of the operator the color bar serves as a visual cue of the robot s status The color bar indicates that the robot is performing the actions as described by the command buttons When the color bar is white it indicates that the course correction path has completed and it is waiting on the operator to initiate the next dead reckoning prog
171. rammed path The second responsibility of the wireless is the offloading of the image processing to a remote computer At the current level of development the user is responsible for evaluating the position and rotation of the optical mark in the image Chapter VI describes this method and the algorithm used to extract the position from the image When the user has taken the three measurements and input them into the Excel spreadsheet the next step is to click the Send Data button on the client form The data is collected from the spreadsheet sent to the ROS and the server program then begins normal operation 99 7 2 3 Operational data The next peripheral to be discussed is the operational data in the form of three text files that are created every time the robot operates Table 7 3 lists the index number the name and the purpose of the three files These files are extremely useful when troubleshooting problems with the robot s motion They make a record of the complicated interactions that occur during each run The files are written to the directory that is shared between the client and the server This allows the operator to access them remotely during the intervals of a run when the robot is not moving One limitation in the recording of the operational data is that each time the program runs it will overwrite the previous files Therefore if data is to be saved it must be copied prior to the next run when the existing files will
172. related to cost would remain constant into the future 1 Even now four decades later this trend seems to continue It is reasonable to assume that as processor speeds increase the capabilities of robots will increase as well It is instructive to review the earliest application of robots that was successful Automotive manufacturing was the first industry to profit from the use of robots by incorporating them into manufacturing plants Typically these non mobile robots performed repetitive pick and place tasks and were successful because of the repetitious nature of their movements Several decades ago many learned people in the field of robotics had expectations that artificial intelligence would rapidly develop new capabilities similar to human thinking Their predictions for the most part have not come true In spite of the constant development of processors the question has to be asked why have the abilities of robots developed at such a slow rate Consider an example of a simple task for a human that is very difficult for a robot to reproduce A young child of around five years is able to play catch with a ball over a distance of a couple of meters as long as the ball is not thrown with too much speed or is too erratic Each throw will have a slightly different path velocity and time of flight Each time the child will catch the ball in a different place and with a different orientation of the arms and physical body This thes
173. ribes several constants and two functions used to drive the webcam Much of the explanation for this module can be found in Chapter VI This code can be found in Appendix A 2 3 PCMMain bas This code module is user defined and has two basic purposes Its first function is to assist during the boot up sequence It also functions as an error trap for the DIO device The IPRV also had a code module with this same name For the PML its responsibilities have been significantly reduced This code can be found in Appendix A 3 4 PCMData bas This module was provided to me from Homji s thesis Originally this code was provided by the DIO manufacturer and is available as a support item download at www superlogics com This code is not included in this appendix 5 PCMDrvFunc bas This module was also provided by Homji s thesis and has support code used to operate the DIO device This code can also be found at www superlogics com This code is not included in this appendix 134 6 PCMUserFunc bas This module was provided by Homji s thesis and contains higher level functions used to access the PCMDIO device It contains the segments of code used to read and write from the DIO This code is not included in this appendix A 1 frmMC_cam3 This was copied from frmMC_cam 2 on 07 08 07 for the purpose of adding a major change to the code This change will use rotational speed as an input control to vary the turn speed This is
174. rip through set values and ICs Erase DegHistory put zeros in the array for subsequent turns TurnDelta 2 because the Ist 4 computes will be slow PTime timel 0 2 set initial time so PRATIO BLOCK will access 1st time incPC 0 reset 0 for array start set initial turn direction this will only be done once should be independent of any prior condition state INITIATE THE TURN If Sgn DegError gt 0 Then do a left turn Turn 1 turn oldturn diff will call MotCont OldTurn 1 Else do aright turn Turn 1 OldTurn 1 End If 152 Recompute the Turn Variables by calling MotCont Call MotCont this will also initiate a 1st turn w new Turn direction PRINT OUTS Print 1 Inside Initial block of P Control Print 1 TurnDelta TurnDelta Print 1 Right R Right R Right L Right L Print 1 Left R Left R Left L Left L Print 1 End If if PControlInitial true block ee ee ee eee ee PRATIO BLOCK san codes EE ee this block is accessed at a minimum of 1 10 second depending on when the control passes to HallsCompute If time gt PTime 0 1 Then determine the reading index value incPC2 incPC 1 the index in front of the current one is the oldest unless below conditio
175. rnDelta 3 End HallsCompute timel 4 83 inside HallsRead time2 4 83 ActDist 61 5 ActDeg 79 Start HallsCompute PRatio Block incPC 4 incPC2 0 Test PTime 4 83 DegHistory incPC 79 PRatio 5 DegHistory incPC2 84 pratio MOTCONT Inside MotCont Turn 1 OldTurn 1 Right R 7 Right L 2 Left R 2 Left L 7 end motcont TurnDelta 5 End HallsCompute timel 4 85 inside HallsRead time2 4 85 ActDist 62 ActDeg 78 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 86 inside HallsRead time2 4 86 ActDist 62 ActDeg 78 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 87 inside HallsRead time2 4 87 ActDist 63 ActDeg 78 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 88 inside HallsRead time2 4 88 ActDist 63 ActDeg 78 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 88 inside HallsRead time2 4 88 ActDist 63 ActDeg 78 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 89 inside HallsRead time2 4 89 ActDist 63 ActDeg 78 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 89 inside HallsRead time2 4 89 ActDist 63 5 ActDeg 77 202 203 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 9 inside HallsRead time2 4 9 ActDist 63 5 ActDeg 77 Start HallsCompute TurnDelta 5 End HallsCompute timel 4 91 inside HallsRead time2 4
176. rnDelta variable is changed the angular speed of either a left or right turn will also be changed It should also be understood that the LA Ratio is a measurement of the angular change in heading Therefore all that remains in completing the LA controller is to relate these two variables Figure 7 12 shows the development of the one to one mapping relationship between the LA Ratio and the TurnDelta variable The points on the graph indicate the preferred but not essential response of this map Their exact selection was based upon an analysis of 119 the desired response and experimental results This analysis had two primary objectives The objective of this functional map is the selection of a rate of rotation that was consistent and had a smooth appearance during actual testing When the second order polynomial in Figure 7 12 was loaded into the ROS the effective rotation rate was approximately 35 s The rotation rate has been verified by the data from repeated runs see Appendix C 1 Original Points 2 Order 6 a Equation Delta TurnDelta 0 0011 LA Ratio 0 1489 LA Ratio 7 1691 0 10 20 30 40 50 60 70 80 90 100 110 LA Ratio Figure 7 12 One to one map between LA ratio and TurnDelta When the robot is running and under LA control whenever the TurnDelta variable is changed the motor control program is called and will write a new DIO speed comman
177. rocessing software The primary development goal was the ability to maintain the accuracy and performance of the robot with inexpensive and low resolution hardware Combining the two abilities of dead reckoning and optical correction on a single platform will yield a robot with the ability to accurately travel any distance As shown in this thesis the additional capability of off loading its visual processing tasks to a remote computer allows the PML robot to be developed with less expensive hardware The majority of the literature research presented in this paper is in the area of visual processing Various methods used in industry to accomplish robotic mobility optical processing image enhancement and target interception have been presented This background material is important in understanding the complexity of this field of research and the potential application of the work conducted in this thesis The methods shown in this research can be extended to other small robotic vehicles with two separate drive wheels An empirical method based upon system identification was used to develop the motion controllers This research demonstrates a successful combination of a dead reckoning iv capability an optical correction method and a simplified controller methodology capable of accurate path following Implementation of this procedure could be extended to multiple and inexpensive robots used in a manufacturing setting To my wife and cat
178. roduce the lowest voltage at the Vtap point 41 5 V Resistors Q input Left Right 5kQ Re 6800 Re 6100 trim pot DIO 4 bit Run 2000 Run 2000 input R 3000 R2 3000 Re R 1500 R 1500 Ro 750 Ro 750 Viap Run Stop ser 3 bit block Figure 4 5 System diagram of n bit variable resistance blocks As indicated in Figure 4 2 the IFB voltage control system has two inputs and two outputs In order to run the robot in a straight path both wheels should have the same speed Sending the same 4 bit digital value to both inputs of the IFB voltage control system would hypothetically run the robot in a straight line However experiments indicated that even when the same voltage was applied to both MC 7s the speed of the two wheels was not identical In order to correct this problem a trim pot was installed on the right voltage circuit This allows the right side to be tuned so that at a standard speed both sides have approximately the same speed Implementing a feedback control system onto a stable hardware system is more desirable than implemented a control system onto an unstable system Note that the single trim pot is intended to be used only with one common speed since parallel speeds of both right and left are not precisely positioned due to inherent differences between the gearbox the motor and the MC 7s 42 The design of the IFB voltage control ha
179. ror in the final position This error will have both a positional and angular component Not shown in Figure 3 1 is an illustration of the fact that any error existing at the beginning of the path will have an impact on the final accuracy For example consider the case in which the starting orientation of the robot has a 1 error A simple calculation as in 3 1 will indicate that this small angular error at the beginning of the path will cause a 4 inch error by the end of a 20 foot path 20ftx 12inxsin 1 4 19 in 3 1 3 2 2 Optical correction overview The second sub function of the robot is the optical correction task Once the dead reckoning obstacle path is complete the robot will come to a complete stop Any positional and rotational errors need to be determined by the robot s webcam In order to accomplish this task four basic steps need to be completed e Capture image taking a picture of the waypoint mark e Send image transferring the bitmap image from the robot s computer to the remote server Process image image processing and data extraction e Return correction data encoding and decoding the correction data 20 Figure 3 2 Printed cross used as an optical target The first step in the optical correction function is the capture of the image of a stationary mark on the floor A standard size sheet of paper that measures 8Y2 by 11 inches with a printed image of a high contrast cross was taped to the floo
180. ror value For example if the robot has the correct heading but is offset some distance to the right of the intended path the RefAng variable will be negative and will define the angle needed to correct for the positive error Because the control response is in the form of a polynomial equation it is easy to adjust the robot s response to the input error by changing the equation The mapping function seen in the figure is only applied to the absolute value of OffError The output of the RefAng must be properly signed before it is used for further calculations in the SP controller This one to one map commands the robot to correct for large errors by allowing a sharper turn while commanding a lesser response for smaller offset error values The two heavy dashed lines on Figure 7 8 indicate the two cutoff values used by the mapping 110 equation The RefAng variable is limited to a maximum value of 20 When the OffError value is less than 5 the cutoff will set the RefAng equal to zero RefAng cut off values a ra ea ee ee eee ae ee ee Ser See ae e a 18 I 2nd order equation y 0 0024x 0 0287x 1 5863 ZK Points used to determine desired response I I l Offset Error cut off value l l l i OffSet Error 0 10 20 30 40 50 60 70 80 90 100 110 Figure 7 8 One to one map offset error to reference angle 7 4 4 Supplemental small angle oscillation contr
181. rrent 2 5 ms 129 e Implementation of path planning and collision avoidance artificial intelligence or other adaptive algorithms with an aim to fully automate the IPRV and eliminate the need for remote supervision The development of the PML robot addresses each of items but not in the specific manner indicated For example the use of a second processor is not needed if the purpose of the image stream is to serve as an intermittent guide for error correction The inclusion of an external clock to improve the sampling rate is not specifically needed as the counter chips act as a counting buffer and fulfill this need The implementation of path planning and automated mobility has been successfully implemented The design and modification of the PML robot was done with the intention of allowing for unspecified future work The use of a wheelchair as a physical base will allow for the addition of considerable weight to the frame Extra sensors and processors including an additional computer could be added in the future to expand the capability of the robot Future users could continue to use the Excel PDB as an input for path commands These users are able to interact with the existing ROS with an ease that will allow them to conduct their research without having to design a completely new operating system or motion controllers The research goal of building a robust robotic base has been successfully accomplished 1 2 3
182. rver_ConnectionRequest SendData when client calls server robot replies with echo check data 7 3 3 Modules The VB 6 program for the robot is technically called a project The project is a directory that includes the main program with subprograms five modules and a VB program file that describes the connections used for the project VB 6 programming is designed to use separate code modules written on their own forms When these modules are included in the project their contents are accessible to the main program A description of the responsibilities of each of the five modules is listed below CAM MODULE The purpose of this module is to define several public constants used by the main program and create several functions that control the messaging and aliasing functions used by the Windows OS A 2 PCMMAIN This module contains user defined variables and functions and is used to control the boot up sequence of the project Any errors that occur during the program operation will call the error trap in this module which is designed to un latch the error and allow for program shut down A 3 PCMDRVFUN This module was also written by SuperLogics and defines approximately 40 functions that all begin with PCM_ function in their name The purpose of these functions is the support and operation of the hardware device For example the first three functions in the list control the opening resetting and closing of the devic
183. s the resistors placed into cutout sections of a breadboard The purpose of this is to allow for easy removal and replacement so that speeds and intervals between speeds can be quickly changed Figure 4 6 is a photograph of the right IFB voltage control and shows the resistors placed in the breadboard section Figure 4 6 Image of right side resistor block 4 2 3 Voltage controller circuit design In order to implement the system diagram as seen in Figure 4 5 four resistors need to be selectively cut in or cut out by following the 4 bit digital input commands from the DIO In order to build this circuit three types of chips are used a Quad Bilateral Switch CD4066BC a Quad NAND Schmidt Trigger HCF4093B and a Quad D Flip Flop 54LS175 Each one of these circuits can control one of the MC 7 controllers 43 therefore it is necessary to have two voltage controller circuits in order to run the robot Figure 4 7 illustrates how a single logic input from the DIO is used to command whether the resistor Rx is cut in or cut out of the circuit This figure serves as an intermediate step in showing the connection between the previous system diagram and the subsequent complete circuit diagram Figure 4 7 shows the relationship between a single switched resistor of the system diagram drawing A and a modular circuit diagram drawing B which shows the logic connections between the four chips used
184. ser to input commands 96 Figure 7 4 Client and server wireless devices Figure 7 4 shows the two hardware devices used by the PML robot The specifications for each device are basically the same The designation and model number for the client wireless device is Netgear 54 Mbps Wireless USB 2 0 Adapter WG111 The designation and model number for the server wireless device is Netgear 54 Mbps Wireless PC Card WG511V2 Both of these devices provide reliable standards based 802 11g b 54 Mbps wireless local area network WLAN connectivity They both support wired equivalent privacy WEP encryption and work with Windows Me 2000 and XP operating systems 22 23 The connection between the client and the server was first established via an ad hoc network using the Windows XP operating system Once the basic wireless connection was established and file sharing is allowed the VB 6 client and server communicated by using the previously established network connection As shown in Figure 7 1 the 97 wireless connectivity was used to support two peripheral applications The first application is to support the user interface while the second allows the ROS to offload most of the image processing capability to the optical correction system OCS 0 x Color Bar Webcam Image White Of Optical Mark x Y Angle Excel Output X Error Y Error 6 Error Data Echo Server Encoded Data Echo Figure 7 5 Client form fo
185. set for case green FirstLoop True timeCarry 0 initial condition at start TumDelta 2 2 is the basic condition for I Control variations done in P Control set back to 2 at I Control re enable intStatus singleDigitalOutput gintlogicaldevice Count_CLR 1 sets the Z load to high voltage which is necessarary during the Initial sub for the 191 to be enabled If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus Call Zero resets the Hall counters and count variables OldTurn 0 _ avoids first call of ranging block for turn input Turn 0 Call Neutral writes neutral values to both L R and Fwd End Sub END Sub Initialize XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX SECOND INITIALIZE Use this for any variables that need to be reset or Conditions that need to be meet on entry into 2nd and later Case Green loops XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub Initialize2 timeInt Timer set initial time for motion use for data excel 165 EmoTime 0 resets at beginning of each Case Green EmoDist 0 so the 2nd Case Green run won t stop blnRun True put here for multiple runs green yellow green NewLine True _ turned off at end of Read Table Call Zero Set ActDist ActDeg 0 MotInc 0 since the increment is at the top FirstLoop True
186. severrerrerreese rense en ee een sern enes 52 Power supply configuration u u sssnereeerererrere reen reen reen nen ennen seerne en nee 55 Local measurement data from webcam image of grid 1 GW udu udscceerrrne 76 Evaluation of 20 line on polar grid eeeececeeceeceeeeeeseeeeeeseeeeneeeeaees 82 Average image radius compared to real radius eeeeeceeeeeeeeeeteeeeneeeees 84 PCMDIO channel allocation sccturss oct vast ees cedes bucket ces Nanaeeatdea cede 94 Command button description ceesceceeseecesececeeeeeceeeeeceeeeeceseeeeseeeenaeeees 98 Operational ata file TYPES sabctncsssesssedasecdtin aetecsected eo otia e a E eese 99 Ex miple of path databases a ii eene a AA EALES 101 Basic measurement variables u u u u seveeseereereereereenen sneen ennen serene ennen ee 103 SP controller variable definition u ssenverererrerrerreerse ren reense rense ener ener 108 Desisn OL ANU ae buer es arr tea e a a a a a E 117 CHAPTER I INTRODUCTION Computer operated robots have been around for more than 50 years Over that time there has been significant development in their capabilities The principal reason for their development has been linked to the increase in computing processing speeds It is widely known that the speed of processors have been increasing at a near constant rate for several decades Gordon Moore first observed this in 1965 when he predicted that the rate in the increase in processor speeds as
187. sheet LinkMode vbLinkNone LinkTopic ExcellC Program Files Microsoft Visual Studio VB98 vb_laptop MC_camera control xls Sheet1 LinkItem TrigFwdCell don t use paraenthesis around the String cell LinkMode vbLinkAutomatic End With Distance TxtDist Text Loads the Distance variable trigDist in Pulses TrigCol 3 column 3 is the forward speed from the table TrigSpdCell R amp CStr Indx 10 amp C amp CStr TrigCol TrigLRCell is a String With TxtFwd This is the format for reading a cell in an Excel sheet LinkMode vbLinkNone LinkTopic ExcellC Program Files Microsoft Visual Studio VB98 vb_laptop MC_camera control xls Sheet1 LinkItem TrigSpdCell don t use paraenthesis around the String cell LinkMode vbLinkAutomatic End With 143 FwdSpeed TxtFwd Text TrigCol 4 column 4 is the angle of direction in and TrigLRCell R amp CStr Indx 10 amp C amp CStr TrigCol TrigLRCell is a String With TxtDeg This is the format for reading a cell in an Excel sheet LinkMode vbLinkNone LinkTopic ExcellC Program Files Microsoft Visual Studio V B98 vb_laptop MC_camera control xls Sheet1 Linkltem TrigLRCell don t use paraenthesis around the String cell LinkMode vbLinkAutomatic End With DirAngle TxtDeg Text TrigCol 5 column 5 is the Offset value TrigLRCell R amp CStr Indx 10 amp C amp CStr TrigCol TrigLRCell is a S
188. t It has a power output range from 12 to 36 V and can accept three different types of control signal input The MC 7 controller will drive an electric motor in both the forward and reverse direction The design of the PML requires that both wheels will always travel in a single direction therefore this capability of the controller is disabled According to the manufacturer the MC 7 is able to output a continuous current of 35 A This is several times the maximum requirement of this system Experiments have shown that approximately 5 A is the maximum current used by the robot during normal operation As shown in Figure 3 7 each of the main power inputs for the MC 7 s has a 10 A fuse Figure 4 11 shows the modular diagram and the connections to one of the MC 7 Several points should be mentioned about the specific connections of the MC 7 and its current configuration 49 At the C7 position a 22 mu capacitor has one of its ends disconnected in order to eliminate the time based ramping function acceleration curve of the output PWM signal A connection wire has been placed between the T3 common voltage supply and one of the two direction command ports T4 or T5 The speed control signal is input to the T 13 pot wiper on the MC 7 The main power to the MC 7 is controlled by a kill switch When power is applied two indicator LEDs are lit The motor is powered by the T 11 Motor Negative and T 12 Motor Positive output connections
189. t at the end of its dead reckoning run The vector arrow represents the robot s actual stopping position and its angle at the end of its run The distance between these two positions represents the positional error In the case below the Y error is 5 2 inches to the right of the target and the X error is 5 6 inches short of the target A design limit was set on the size of the position error window The worst case ending position error should never be more than 25 cm 10 in in any direction from the intended target Therefore the position error window has a size of 51 cm 20 by 20 in 88 To the right and bottom of the Error Window the two measurement bars indicate how the X and Y errors are encoded Each of the 51 cm 20 in lengths is divided into 100 equal increments so the error data will have a 0 5 cm 0 2 in size resolution The reason for this is that the three error data numbers can then be encoded into one 6 digit number In the case of the angular measurements a negative angle is defined as a clockwise rotation and each degree is encoded as one integer Size of Position Error Window 20 by 20 00 a Angular difference from true vertical 0 0 C Sign convention Desired stopped for angles position Se 50 Example of X encoding stratagy XX 5 6 in 0 2in 50 78 X Axis K 78 x Y a Pa Y 5 2in Error Error Error Actual X 5 6in XX YY AA stopped i f Y Axis position i Dimensi
190. t returned to its normal path and direction after 3 8 s and 91 cm As can be seen in the figure the oscillation was limited to 1 from that point onward p ActDeg 10 Offset 45 Error 35 25 15 5 Turn 0 5 10 15 20 25 Commands T T T Right p moo 00 0 0 00000 00 000 o 0 DDO 0 000 0G M0 M0 00000 M00 000 G vo 0 000 wW Left 000 000M 00000 Hi 0000 000 0 80000 Aee a OD BO 000000 a0 GE aw o GOH 00000 a aq 0 5 10 15 20 25 Time s Figure 7 10 Results from CW torque experiment 115 Offset Error Turn Commands T T T Right o wwo MO W000 0 00 W000 0 00 wo oo CED GD 00 0 0 0000 0000 00 0 000 000 000 M000 00 Center omsssamsmoaosssoooxonnsoosssooosaosraosoosansm 010 0 ORDERED CINEREA CORR RNIINOONEECLOESROONRONE COREE Left 0 00 00 00 0 000 00 00000 0000000 000 0000 woow CRO GOOO 00 00000 0000000000 000 mom 000000000 000 00000 00 0 5 10 15 20 25 Time s Figure 7 11 Results from CCW torque experiment 7 5 LA Control Once the SP controller was functional the next stage in the PML development was to test all of the capabilities of the robot by running it on a designed course The course included two 90 turns one to the right and one to the left Several obstacles were placed on the course that required a series of turns to avoid After conducting many test runs it was discovered that on occasion the rob
191. tages will be produced that will drive the wheels the desired speeds In the 3 bit variable resistor speed block resistors are selected such that eight equal divisions of total resistance can be produced The strategy employed is to use a multiple of the lowest resistor Ro value Therefore R is twice the value of Ro and R gt is twice the value of R When these three resistors are added in series to the circuit the result is eight equally spaced values of resistance The R 3 resistor value is selected in order to produce a resistance that will result in a 1 5 V voltage that corresponds to the slowest speed of the robot When the R 3 resistor is cut out of the circuit the voltage will fall below the range needed to run the robot Therefore the R 3 resistor is used as a Go Stop command for the robot When the robot is in motion the Ro R and R gt resistors control for the eight speeds The R resistor is selected to solve 4 1 such that the voltages will be produced as shown in Figure 4 4 The exact resistor values for both right and left sides are tabulated in Figure 4 5 The circuit is designed so that a DIO value of zero for any of the n bit blocks will cut in its associated resistor For example a DIO zero value or the 4 bit value of 0 0 0 0 would have the maximum resistance and create the maximum voltage being available for the Vtap point A value of 15 or the 4 bit value of 1 1 1 1 would act like an electrical short and p
192. th computer screens The four colors used for this purpose are listed below In the VB6 code and as part of a case structure code module the sections of code are Case Green Responsible for the dead reckoning function Case Yellow Controls the webcam as part of the optical correction function Case Blue Responsible for the course correction function Case Red User activated emergency stop named appropriately by both Case and Color Excel Database ie Camera In Out_ The robot listens for any of the color commands from the client operator When the command arrives Data and Ovals indicate input or output from Motor Control Hall Sensors Control robot s OS to an external device O O 7 A Webcam L MC MC A IFB O lt Reads Excel database gets 4 numbers that define the path Measures distance traveled and orientation Path following with both Input and decode controllers stores run data correction number Path Data Use SP Control to correct position error Starts webcam focuses image Saves image to file shuts down RED User controlled emergency camera ES stop Figure 3 9 Operating system overview diagram the robot uses a case structure to branch to the appropriate section of 34 code In the next several paragraphs the function and basic design of the case structure co
193. the wheelchair user would indicate a general direction of travel and the chair would then be responsible for obstacle avoidance and basic path planning Because a robotic wheelchair should be able to interact with the user it is more properly a semi autonomous robot 3 A mixed capability could prove desirable For example the navigation system could rely on a map of the user s home and then shift to a system that did not have access to a map in other new locations 2 1 1 Household robots With the continuing development and reduction in cost of microprocessors the numbers of service robots have increased in recent years Because of these improvements robots have recently expanded into home use and are now being used for tasks previously done by people Several of these new applications could serve as a potential research topic for the PML robot For example obstacle avoidance is a necessary capability in most of these home use applications Investigations into this subject would present a rich topic for research 4 In general due to the newness of this field many challenges remain Two of the uses for home robots are discussed below e Lawn mowing In order to be efficient at this task the robot should have several abilities For example it should be able to estimate its position with accuracy Other necessary functions should be the ability to program for a specific location on the yard to be mowed If the lawn mower is electric powere
194. tring With TxtOff This is the format for reading a cell in an Excel sheet LinkMode vbLinkNone LinkTopic ExcellC Program Files Microsoft Visual Studio V B98 vb_laptop MC_camera control xls Sheet1 Linkltem TrigLRCell don t use paraenthesis around the String cell LinkMode vbLinkAutomatic End With OffsetRef TxtOff Text DETERMINE HIGH DEGREE TURN I CONTROL ENABLE DISABLE If FirstLoop True Then necessary in case a Case Green run ends w a 90 degree turn and then a new DirAngleOld DirAngle Case Green run starts w a 0 degree turn FirstLoop False Ist trip through Case Green run independent of index ff TEST FOR P CONTROL CONDITIONS If Abs DirAngleOld DirAngle gt 5 Then DirAngleOld DirAngle updates DirAngleOld so the next line won t go into P Control PControl True Stops I control in the HallsCompute sub 144 DegRef 0 don t really need this MotInc 1000 used to prevent a neutral write PControlInitial True use for Ist trip through P Controler Print 1 Print 1 Inside ReadTable set PControl PControl Print 1 End If nee nnn nnn nnnnn anna nanan PRINT OUT COMMANDS Print 1 NEWLINE i Print 1 Tab 12 Distance Distance FwdSpeed FwdSpeed Print 1
195. ulse onto the MC latch chips both RL and FWD XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Public Sub PulseMC intStatus singleDigitalOutput gintlogicaldevice MC CLK 1 to the Latch If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If waitTime 1 pulse of Latch clock width intStatus singleDigitalOutput gintlogicaldevice MC_CLK 0 If intStatus lt gt 0 Then Call pcmdioError gintlogicaldevice intStatus End If End Sub XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX Private Sub tcpServer ConnectionRequest ByVal requestID As Long If tepServer State lt gt sckClosed Then tcpServer Close End If Accept request and get request ID tcpServer Accept requestID lblRequest Caption requestID Call tcpServer SendData Connect End Sub 168 A 2 Camera bas Option Explicit Public hCap As Long Public Const WS CHILD As Long amp H40000000 Public Const WS VISIBLE As Long amp H10000000 Public Const WM_USER As Long amp H400 Public Const WM_CAP_START As Long WM_USER Public Const WM_CAP_DRIVER_CONNECT As Long WM_CAP_START 10 Public Const WM CAP DRIVER DISCONNECT As Long WM CAP START 11 Public Const WM_CAP_SET_PREVIEW As Long WM_CAP_ START 50 Public Const WM_CAP_SET_PREVIEWRATE As Long WM_CAP_START 52 Public Const WM CAP DLG VIDEOFORMAT As Long WM CAP START 41 Public Const WM_CAP_FILE_SAVEDIB As Long WM_CAP_START 25 Public Const WM_CLOSE a
196. unter 74LS191 g 25 2 a 8 i I er Left Wheel Figure 5 3 Modular and circuit diagram of position sensing system 5 3 3 Application of position buffer The benefit of adding a 4 bit buffer is that it allows the ROS cycle speed to be significantly and still not miss any position counts The specific ROS code that is used to interpret the 4 bit signal from the position sensing module is discussed in Chapter VII It should be noted that the position sensing system counts in a single digit hexadecimal number In other words there is no capacity for a second digit that is incremented every time the first digit goes from 15 to 0 The ROS code is responsible for keeping track of the ripple count of the second digit 5 3 4 Selection of magnets Figure 5 4 shows an output of the right Hall sensor as detected by an oscilloscope The motor was run at the DIO speed of six which turns the right wheel at 15 5 RPM The position sensor system counts the pulses as the magnets pass by the Hall sensor and will 60 count three complete pulses for each rotation of the motor rotor The voltage output of the Hall sensor goes to zero when the magnet is in proximity to the sensor and is approximately 5 V when the magnet is away from the sensor It should be noted that the pulse signal would appear to count the gaps between the magnets and not the magnets themselves The duty cycle of Figure 5 4 is approximately 60 and is produced with three magnets

ROGERS-THESIS - OAKTrust Home

Contents

Download Pdf Manuals

Related Search

Related Contents