Home

Submitted version of the thesis - AIRWiki

1. Exported constants Module private variables extern u8 pImgBufferData extern u32 BufferSize Exported macro Private functions Exported functions void CalculateBallPosition u8 X u8 Y u8 Dpx double result 4 void CalculatePositionTop TPoint point s16 result 3 void CalculatePositionGround TPoint point s16 result 3 void Blob_Search void endif VISION_H_ Vision c B 1 Microprocessor Code 107 Standard include include 91x map h include utils h include definitions h include definitionsLib h include CamInt_x24 h include usb_lib h include usb_conf h include usb_prop h include usb_pwr h include usb config h include RGBCube h include blobsearch h include math h include Misc h include Vision h include Camera_Init h include Motion h Include of other module interface Local includes Private typedef Private define define PRINT UN Private macro Private variables u8 pImgBufferDat
2. LOK OK EE REGE AAA AAA AA AA AA AAA AA AAA AA AA AAA AA AAA AA AA AA AAA AA AAA AA AAA AA AA AAA AA AAA AAAAAAAAAA Function Name main Description Main program Input None Output None Return None EE EE E SEE SEE AA AAA AA AAA AA AA AAA AA AA AAA A AA GE EE KA AAA KA AAA OR ROR RA AAA RARA AAA AA AAA AA AA HAS int main void tifdef DEBUG debug endif Configure the system clocks SCU Configuration Configure the GPIO ports GPIO Configuration Configure x24 Camera x24 HwConfig 8000000 x24 SendPatches B 1 Microprocessor Code 87 Configure the USB USB_Configuration USB_Init LED ON LEDR MSDelay 200 LED OFF LEDR MSDelay 200 LED ON LEDR MSDelay 200 LED OFF LEDR MSDelay 200 LED ON LEDR MSDelay 200 LED OFF LEDR To initialize camera parameters Init To initialize motors configure pins and PWM values Init Motors To arrange camera head position available position SERVO_TOP SERVO MID SERVO BOTTOM Position Camera SERVO TOP To initialize bumpers and configure pins Init Bumpers MSDelay 2000 To define blob color amp geometry Init_BlobParameters the initial values bool search_direction TRUE u8 game_end 0 bool blob_found FALSE bool color_done FALSE u8 cam_position SERVO
3. include 91x map h include utils h include definitions h include definitionsLib h include usb lib nh include usb conf h include usb prop h include usb pwr h include usb config h B 1 Microprocessor Code 111 include math h include Misc h Include of other module interface headers 7 Locat includes SS SS SS SS SS SS SS Private typedef Je Private define ASA 2 HS Ss eS SS Private macro Private variables Private function prototypes void Triskar s16 destination s8 omega s16 vt void Direct_Motion s16 destination s8 angle s16 speed Private functions void Triskar s16 destination s8 omega s16 vt s16 frontal_speed destination 0 s16 lateral_speed destination 1 double alpha 0 s16 velocity 2 velocity 0 frontal_speed cos alpha lateral_speed sin alpha velocity 1 frontal_speed sin alpha lateral_speed cos alpha ifdef NO_DEBUG USB_WriteString CalculatingyMotiony n Print Int velocity 0 frontal Print_Int velocity 1 lateral endif u8 R_robot 250 816 v_F velocity 0 816 v_L velocity
4. Enable USB clock SCU_AHBPeriphClockConfig __USB ENABLE SCU_AHBPeriphReset __USB DISABLE SCU_AHBPeriphClockConfig __USB48M ENABLE Configure USB D PullUp pin GPIO_StructInit amp GPIO_InitStructure GPIO InitStructure GPIO Direction GPIO_PinOutput GPIO InitStructure GPIO Pin USB_Dp_PullUp_GPIOx_Pin GPIO InitStructure GPIO Type GPIO Type 0OpenCollector GPIO Type PushPull GPIO InitStructure GPIO Alternate GPIO_OutputAlti GPIO Init USB Dp PullUp GPIO GPIO_InitStructure USB Dp PullUp OFF MSDelay 100 USB Dp PullUp ON USB endPointConf VIC_Config USBLP ITLine VIC IRQ USB Priority VIC_ITCmd USBLP ITLine ENABLE 94 Appendix B Documentation of the programming LOK OK EG XE SE AAA AA AAA AA AAA AA AA AAA AA AA AAA AA AAA AA AA AAA AA AA AAA AA AA AAA AA AAA AA AAA AA AA AA AAA A Function Name GPIO_Configuration Description Configures the different GPIO ports Input None Dutput None Return None EE EE E ARA HOR AAA E EE EE E OE OE EG E SE IA AH A AA A KA EE OR KA AAA KA AAA AAA OE E EE OE AAA AAA AAA o oe oe HAS void GPIO Configuration void GPIO InitTypeDef GPIO_InitStructure Configure LEDR GPIO StructInit amp GPIO_InitStructure GPIO InitStructure GPIO Direction GPIO_PinOutput GPIO InitStructure GPIO Pin LEDR GPIOx Pin GPIO InitStructure GPIO Type GPIO Type PushPull GPIO InitStructure GPIO Al
5. set h3 Color 0 2 0 2 0 set hl LineStyle set h2 LineStyle set h3 LineStyle if v_F 0 line 0 0 0 vl Color 1 0 0 if v F lt 0 fill 0 pl 0 05 p1 0 05 vl v1 0 12 pl v1 0 12 pl r else fill 0 p1 0 05 p1 0 05 vl vl 0 12 pl vl 0 12 pl r end text 0 15 vl vl v_F end 130 Appendix B Documentation of the programming if v L 0 line 0 v2 0 0 Color 1 0 0 if v L 0 fill v2 v2 0 12xp1 v2 0 12 p1 0 p1 0 05 p1 0 05 7r else fill v2 v2 0 12xp1 v2 0 12 p1 0 p1 0 05 p1 0 05 7r end text v2 0 1 v2 v_L end if omega 0 if omega gt 0 theta linspace pi 3 pi 2 100 fill 0 12 pl 0 0 R robot 6 R_robot 6 p1 0 05 R_robot 6 pl 0 05 5 text 0 14x p1 R_robot 6 p1x0 05 lomega else theta linspace pi 2 4 pi 3 100 fill 0 12 xpl 0 0 R_ robot 6 R_robot 6 p1 0 05 R robot 6 pl 0 05 b text 0 12 pl R_robot 6 p1 0 05 omega end rho ones 1 100 R_robot 6 xr yr pol2cart theta rho line xr yr end ruota 1 if vt 1 0 line R_robot cosA R_robot cosA vt 1 sl sinA R_robot sinA R robo sinA vt 1 slxcosA Color 0 1 0 offsetl R robot cosA cvt 1l sl sinA R_robotxsinA vt 1 sl cosA if vt 1 slxsinA lt 0 M m_rot alpha else M m rot pi alpha end m a
6. 191 7146 0 80 3591 0 191 2730 61 2765 0 0 1 T_cr 3 4 4 0 0160 0 9987 0 0478 0 6271 0 6634 0 0252 0 7478 245 6112 0 7481 0 0437 0 6622 44 2694 C 3 4 1 0 0 0 0 0 0 0 0 0 0 0 0 H 3 3 0 0 0 0 0 0 0 0 0 InvH 3 3 1 0 0 0 0 0 t_ 4 3 OF 0 0 O Matrix_Mult3334 K T_cr C Matrix_Mult3443 C t H B 1 Microprocessor Code 109 Inverse H InvH double position 3 0 0 0 double image_point 3 point X point Y 1 vectorByMatrix3x1 InvH image point position result 0 result 1 result 2 s16 position 0 position 2 s16 position 1 position 2 s16 position 2 position 2 void CalculateBallPosition u8 X u8 Y u8 Dpx double result 4 1 double image_point 3 X Y 1 double inv_K 3 3 4 0 005216086284917 0 0 419160051539105 0 0 005228129700905 0 320361489617536 0 O 1 XY double position 3 0 0 0 vectorByMatrix3x1 inv_K image point position double norm norm sqrt position 0 position 0 position 1 position 0 position 1 position 2 int X_axis int Y_axis double double double double fstar Pc 4 position 1 position 2 position 2 position 0 norm position 1 norm position 2 norm 300
7. Exported macro Private functions Exported functions void Init_Motors void void Calculate_Order s16 speed u8 size void Position_Camera u8 position void Set_Directions u8 Motori u8 Motor2 u8 Motor3 void Stop_Motor u8 Motor void Bumpers u8 bumper 6 void Set_Speed s16 speed u8 mod void Stop Servo void void Turn u8 direction endif MOTOR_INIT_H_ Motor Init c Standard include include 91x map h include utils h include definitions h include definitionsLib h include CamInt_x24 h include usb_lib h include usb_conf h include usb_prop h include usb_pwr h include usb config h include math h include Motor_Init h Include of other module interface headers Local includes Private typedef Private define Private macro Private variables u8 control hit u8 servo position 0
8. complex Wheel alignment is critical in this drive system if the wheels are not parallel the robot will not translate in a straight line Figure 2 6 shows MRVA a robot with this drive mechanism Figure 2 6 MRVA robot with synchro drive mechanism The car type locomotion or Ackerman steering configuration is used in cars The limited maneuverability of Ackerman steering has an important advantage its directionality and steering geometry provide it with very good lateral stability in high speed turns The path planning is much more difficult Note that the difficulty of planning the system is relative to the environment On a highway path planning is easy because the motion is mostly forward with no absolute movement in the direction for which there is no direct actuation However if the environment requires motion in the direction for which there is no direct actuation path planning is very hard Ackerman steering is characterized by a pair of driving wheels and a separate pair of steering wheels A car type drive is one of the simplest locomotion systems in which separate motors control translation and turning this is a big advantage compared to the differential drive system There is one condi tion the turning mechanism must be precisely controlled A small position error in the turning mechanism can cause large odometry errors This sim plicity in line motion is why this type of locomotion is popular for human driven vehicles
9. b 1 b 2 A 1 0 x 0 100 4 1 1 x 1 100 A 1 2 x 2 100 A 2 0 x 0 100 A 21 1 x 1 100 A 2 2 x 2 100 void vectorByMatrix4x1 double A 4 double 1 4 b 0 b 1 b 2 b 3 void Matrix int int int int int int for void Matrix_ int int int for void Matrix int int int double b 4 ATO 0 0 ACO 1 x 1 AC0 2 x 2 A O0 3 x 3 A 11 0 x 0 A 1 1 x 1 A 11 2 x 2 A 11 3 x 3 A 21 0 0 A 2 1 x 1 A 21 2 x 2 A 2 3 x 31 A 3 0 0 A 3 1 x 1 A 3 2 x 2 A 3 3 x 31 Mult double a1 4 double 82 4 double a3 1 41 1 i 0 j k a b C BOO i 0 i lt a itt for j 0 j lt b j for 0 k lt c k a3 i1 j a1 i x a2 x1 j Mult3334 double a1 3 double a2 4 double a3 4 1 i 0 0 k 0 i 0 i lt 3 itt for j 0 j lt 4 j for k 0 k lt 3 k a3 i jl a1 i k a2 x1 j Mult3443 double ai 4 double a2 3 double a3 3 1 i 0 j 0 k 0 116 Appendix B Documentation of the programming for i 0 i lt 3 1 for j 0 j lt 3 j for k 0 k lt 4 k a3 i j ai il k a2 x1 j void Inverse double A 3 float B 3 3 the float C 3 3
10. bB prctile B 25 75 maskR im 1 16 gt floor bR 1 amp im 1 16 lt ceil bR 2 122 Appendix B Documentation of the programming maskG im 2 16 gt floor bG 1 amp im 2 16 lt ceil bG 2 maskB im 3 16 gt floor bB 1 amp im 3 16 lt ceil bB 2 mask maskB amp maskR amp maskG subplot 1 3 1 hold on title mask_R hold on imshow maskR subplot 1 3 2 hold on title mask G imshow maskG subplot 1 3 3 hold on title mask B imshow maskB figure subplot 1 2 1 hold on title mask RGB imshow mask subplot 1 2 2 hold on title orginal picture imshow im B 3 Object s Position Calculator position_calculatorm 1 Ball Position Calculator This script does transformation for the 3D ball position to 2D robot B 3 Object s Position Calculator 123 Translation vector position of chesboard in camera reference system Tc ext 17 929188 0 849442 64 715079 Rotation matrix Rc ext 0 042070 0 998317 0 039914 0 614786 0 057357 0 786605 0 787571 0 008554 0 616165 transformation of camera to chess board T eche Rezext Tezext Ho eye 3 0 0 0 the position of the origin image point 81 62 1 K calibration matrix K 191 71462 0 80 35911 0 191 27299
11. u8 servo flag 0 u8 motori speed 100 Appendix B Documentation of the programming u8 motor2_speed u8 motor3_speed Private function prototypes void Init_Motors void void Calculate_Order s16 speed u8 size void Position_Camera u8 position void Set_Directions u8 Motori u8 Motor2 u8 Motor3 void Stop_Motor u8 Motor void Bumpers u8 bumper 6 void Set_Speed s16 speed u8 mod void Stop Servo void void Turn u8 direction Private functions void Set Speed si16 speed u8 mod u8 direction NUM MOTORS u8 k for k 0 k lt NUM MOTORS k 1 direction k speed k gt 0 1 O speed k Abs speed k for k 0 k lt NUM MOTORS k if speed k 0 1 Stop Motor k if speed k gt 2500 speed k 2500 for k 0 k lt NUM MOTORS k if speed k 0 speed k speed k 75 mod 90 Set_Directions direction 0 direction 1i direction 2 u8 speed 0 u8 speed 1 u8 speed 2 motori_speed motor2_speed motor3_speed ifdef NO_DEBUG Print_Int motori_speed mi applied Print Int motor2_speed m2 applied Print Int motor3 speed m3 applied B 1 Microprocessor Code endif void Stop_Motor u8 Motor 1 switch case 0 case 1 case 2 case 3 void Set_Directi if Moto
12. From that histograms we create the masks Figure 5 6 by finding the upper bounds and lower bounds for each color channel As a final step we create the total mask that is returning the target object s color boundaries 64 Chapter 5 Vision in a form that can be converted into the rules The result of the color selec tion script with the original picture is shown in Figure 5 7 mask RGB orginal picture a Mask to be Applied b Orginal Picture Figure 5 7 The mask applied to the orginal picture The boundaries found for the object are converted into rules that will add or subtract the color from the RGB Cube The rules are checked by color selection program provided with the ST software coming with the camera The script can be found in Appendix B 1 in Color Histogram Cal culator Using both the color selecting program and an offline Matlab script we have been able to define the color in a more precise way and improved the results for object detection But the results diverged as the lighting conditions change in the environment For example the laboratory where we have performed tests is exposed to direct sunlight in the morning while in the afternoon it is not We introduced different configurations for the same color to avoid the need for sampling of the previously defined color By controlling the color before each run manually and recalculating when necessary resulted a stable solution to the problem 5 3 Tracki
13. num2str omega urad s y_ref y_ref delta text R_robot 0 6 y_ref v n2 num2str vn 2 mm s y_ref y_ref delta text R_robot 0 6 y_ref v_ t3 num2str vt 3 umm s y_ref gt y_ref delta text R_robot 0 6 y_ref v_ n3 num2str vn 3 mm s y_ref y ref delta text R_robot 0 6 y_ref omega_1 num2str om 1 rad s y_ref gt y ref delta text R robot 0 6 y_ref omega_2 num2str om 2 rad s y ref y ref delta text R robot 0 6 y ref omega_3 num2str om 3 rad s axis equal hold off 134 Appendix B Documentation of the programming Appendix C User Manual This document gives information about the robot the settings that needs to be made before the starting a run We use two different programs in the robot The first one is the ST software that is used during the color selection and other demo features provided with the software The second one is the software which we implemented for the Thesis the game software In order to run the game we need also the ST color selection interface since it offers a good visual feedback on the selection process The ST program uses the QT Framework and installation is necessary to use the ST color selection interface C 1 Tool chain Software In order to implement the robot to flash the written software in to the mi croproces
14. the int i j double X 1 3 transpose of a matrix A adjunct matrix of transpose of a matrix A not adjunct of float x n 0 n is the determinant of A for i 0 j 0 j lt 3 j if j 2 n ACi j Ali 1110 ALi 21 11 else if j 1 n A illj Ali 11 j 1 Ali 2 0 else n A il j Ali 11 j 1 Ali 2 j 21 for i 2 j 0 j lt 3 j if j 2 n A i j Ali 1110 Ali 21 11 else if 1 n A ill j Ali 11 j 1 Ali 2103 else n A il j Ali 11 j 1 Ali 2 j 2 if n 0 x 1 0 n else for i 0 i lt 3 1 for j 0 j lt 3 j B i j A j i c 0 0 B 1 1 B 2 2 B 2 1 B 1 2 Cfo 1 1 B 1 0 B 2 2 B 2 0 B 1 2 c 0 2 B 1 0O B 2 1 B 2 0 B 1 1 c 1 0 1 B 1 Microprocessor Code 117 B 0 1 B 2 2 B 21 11 B 0 2 C 1 1 B 0 0 B 2 2 B 2 0 B 0 2 C 1 2 1 B 0 0 B 2 1 B 21 0 B O 11 Cc 2 0 B 0 1 B 11 2 BI11 1 B 0 2 c 2 1 1 B 0 0 B 1 2 B 1 0 B 0 2 C 2 2 B 0 0 B 1 1 BI11 0 3 03 1 for i 0 i lt 3 i for j 0 j lt 3 j 1 11 C i j x definitions h DK KR KR AKKKRA KKK KKKKKEEKEK C COPYRIGHT 2007 STMicroelectronics Akk A Ak A Ak Ak A Ak Ak Ak
15. b3 motor send endif else control hit else ifdef NO_DEBUG Print Int control hit hit endif Blob_Search ifdef NO_DEBUG Print Int BlobCount Lblobsuyfound n endif TPoint n_point u8 k s8 myBlob s16 maxArea maxArea 0 myBlob 1 among the available blob the blob with the biggest size is selected as the blob for k 0 k lt BlobCount k if Blobs k ColorSetID color 1 if maxArea lt Blobs k Area maxArea Blobs k Area myBlob k if BlobCount 0 f if myBlob lt 0 random search ifdef NO_DEBUG USB_WriteString Searching for a blob n endif in order to prevent continue turning to search for blob counter introduced if blob_found TRUE counter else if counter gt 3 blob_found FALSE search_direction changes the turning side to left to right or vice versa 90 Appendix B Documentation of the programming if search_direction TRUE Turn RIGHT else Turn LEFT blob_found FALSE else search direction is reversed search_direction search_direction controlling if the blob is with the correct color blob_found TRUE counter 0 n_point X Blobs myBlob Centroid X n_point Y Blobs myBlob OuterBox Bottom s16 position 3 for different camera positions different calibration
16. 893 877 862 847 833 820 806 794 781 769 757 746 735 725 714 704 694 685 676 667 658 649 641 633 625 617 610 602 595 588 581 575 568 562 555 549 543 538 532 526 521 515 510 505 500 495 490 485 481 476 472 467 463 459 454 450 446 442 439 435 431 427 424 420 417 413 410 406 403 400 397 394 391 388 385 382 379 376 373 370 368 365 362 360 357 355 352 350 347 345 342 340 338 336 333 u8 pImgBufferData u32 BufferSize Tx24_ImgJob Job 2 CompletedJob 0 u8 CurrentJob u8 CamState 0 O Uniitialized 1 Running 2 Paused TTimeStamp Tcami Tcam2 vs32 OldErr 0 vs32 Ierr 0 Nerr 0 Private function prototypes Private functions void Init_BlobParameters void Define blobsearch constraints BlobSearchSetOption 5 0 120 0 160 Define blobsearch geometry options correct tihs BlobSearchSetColParam i1 30 2000 50 13000 26 153 BlobSearchSetColParam 2 30 2000 50 30000 26 153 my green day 7 RGBCube AddCube 0 6 4 8 0 1 1 cloudly day ZA RGBCube AddCube 0 6 4 8 1 3 1 RGBCube_SubCube 3 5 3 5 1 3 1 end day RGBCube_AddCube 0 6 4 6 2 3 1 RGBCube_SubCube 3 5 2 4 1 3 1 RGBCube_SubCube 5 7 4 6 2 4 1 RGBCube_SubCube 3 5 4 6 2 4 1 my
17. After finishing these steps we should deselecting the Visual Rule Editor and Save settings In order to test the whole process we should select continuous and click 7 Get Blobs The selected color will be 140 Appendix C User Manual BlobSearch RGB 160x120 on YCbCr400 60x60 tion YCbCr400 C Image stream JPEG up to 540x480 Figure C 6 The main screen of the ST s software represented according to the parameters defined under the Visualization tabs Figure C 7 C 3 Main Software Game software The project should be imported to the Eclipse in the same way with the ST s software The color information and the blob geometry parameters are found under the source Camera_Init c The function that contains these C 3 Main Software Game software 141 Camera BlobSearch FD Corners p Color Classes Get Blobs Continuous Save to Flash Load from Flash Yisualization Blob Color Blob Geometry visual Rule Editor Save settings Class 1 Red C Green C Blue R G Min s C Custom R B Min Bhs C YCbCr Get from Visual Rule Editor Figure C 7 The second step at importing Launch Configurations information are placed in the Init BlobParameters function BlobSearchSet ColParam u8 IDCol u16 MinPer u16 MaxPer u16 MinArea u16 MaxArea u8 MinCirc u8 MaxCirc taking the parameters defined should be updated with the values found in the
18. For example when all four wheels spin forward the robot as a whole moves in a straight line forward However when one diagonal pair of wheels is spun in the same direction and the other diagonal pair is spun in the opposite direction the robot moves laterally These omnidirectional wheel arrangements are not minimal in terms of control motors Even with all the benefits few holonomic robots have been used by researchers because of the problems introduced by the complexity of the mechanical design and controllability In mobile robotics the terms omnidirectional holonomic and non holo nomic are often used a discussion of their use will be helpful Omnidirec tional simply means the ability to move in any direction Because of the planar nature of mobile robots the operational space they occupy contains 12 Chapter 2 State of the Art only three dimensions which are most commonly thought of as the x y global position of a point on the robot and the global orientation 0 of the robot A non holonomic mobile robot has the following properties e The robot configuration is described by more than three coordinates Three values are needed to describe the location and orientation of the robot while others are needed to describe the internal geometry e The robot has two DOF or three DOF with singularities A holonomic mobile robot has the following properties e The robot configuration is described by three coordinates The in
19. TOP u8 counter 0 u8 color 1 s16 speed 3 u8 replay_counter 0 while 1 reading the bumpers values u8 check bumpers 6 0 7 Bumpers check bumpers to blink led Led SMO controlling whether the first target acquired 88 Appendix B Documentation of the programming if color_done TRUE to check whether game end satified Stop_Motor 3 MSDelay 100 cam_position SERVO_TOP Position_Camera SERVO_TOP color color_done FALSE blob_found FALSE if color gt 2 if replay_counter lt 5 1 replay_counter color 1 else 1 Stop_Motor 3 while 1 else else of color_i_done if control hit 1 checking collision do something if check_bumpers 0 check_bumpers 4 1 1 check_bumpers 3 1 1 check_bumpers 5 go back speed 0 1000 speed 1 1000 speed 2 0 Set Speed speed NORMAL MSDelay 100 hit front else if check_bumpers 1 1 hit right go left speed 0 750 speed 1 750 speed 2 1200 Set_Speed speed NORMAL MSDelay 100 else if check_bumpers 2 1 hit left go right speed 0 750 speed 1 750 speed 2 1200 Set Speed speed NORMAL MSDelay 100 B 1 Microprocessor Code 89 ifdef NO_DEBUG Print Int speed 0 bi motor send Print Int speed 1 b2 motor send Print Int speed 2
20. else Motor stop 1 GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR1_DIRECTION_A Bit_RESET GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR1_DIRECTION_B Bit_RESET break stop 2 GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR2_DIRECTION_A Bit_RESET GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR2_DIRECTION_B Bit_RESET break stop 3 GPIO WriteBit MOTOR_DIRECTION_GPIO MOTORS DIRECTION A Bit RESET GPIO WriteBit MOTOR DIRECTION GPIO MOTORS DIRECTION B Bit RESET break stop all GPIO WriteBit MOTOR DIRECTION GPIO MOTOR1 DIRECTION A Bit RESET GPIO WriteBit MOTOR DIRECTION GPIO MOTOR1 DIRECTION B Bit RESET GPIO WriteBit MOTOR DIRECTION GPIO MOTOR2 DIRECTION A Bit RESET GPIO WriteBit MOTOR DIRECTION GPIO MOTOR2 DIRECTION B Bit RESET GPIO WriteBit MOTOR DIRECTION GPIO MOTORS DIRECTION A Bit RESET GPIO WriteBit MOTOR DIRECTION GPIO MOTORS DIRECTION B Bit RESET break ons u8 Motori u8 Motor2 u8 Motor3 1 r2 FORWARD 4 GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR2_DIRECTION_A Bit_SET GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR2_DIRECTION_B Bit_RESET GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR2_DIRECTION_A Bit_RESET 102 Appendix B Documentation of the programming GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR2_DIRECTION_B Bit_SET if Motori FORWARD GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR1_DIRECTION_A Bit SET GPIO WriteBit MOTOR DIRECTION GPI
21. mechanical construction and hardware architectures Chapter 4 is the de 1 4 Thesis Structure 3 tailed description of the control what is necessary to replicate the same work Chapter 5 is the vision system description in detail what has been done which approaches are used and the implementation Chapter 6 con cerns game design for the tests made and the evaluation of the game results Chapter 7 concludes the presentation Chapter 1 Introduction Chapter 2 State of the Art Advances in computer engineering artificial intelligence and high tech evolutions from electronics and mechanics have led to breakthroughs in robotic technology 23 Today autonomous mobile robots can track a per son s location provide contextually appropriate information and act in re sponse to spoken commands Robotics has been involved in human lives from industry domain to daily life applications such as home helper or recently entertainment robots The latter introduced a new aspect of robotics entertainment which is intended to make humans enjoy their lives from a various kind of view points quite different from industrial applications 17 Interaction with robot is thought of a relatively new field but the idea of building lifelike machines that entertain people has fascinated us for hun dreds of years since the first ancient mechanical automaton Up to our days there have been major improvements in the development of robots We
22. since when the ball gets further from the robot the diameter information is not reliable This results to calculate a wrong value of the ball position In this trials we realized also the effects of the white balance and exposure compensation for the color information The camera tries to automatically adjust the color information as the ball gets further from the light source Since the blob elimination is dependent on both the shape of the ball and the color of the blob the information changed by the camera results to definition of wrong blob information To clarify with an example the ball shape is checked by the circularity information calculated by the camera but the whole ball is not composed of the same color since the light is affecting the color so the center can be more bright or the corner more dark etc The camera with the automatic setting of the white balance and exposure compensation tends to change the color in order to have a better picture but by doing so the color temperature is changed So the blob search eliminates the non color points as well as the non shape objects and detects the ball in another shape such as an open rectangle This whole change also results to a change of the diameter of the ball and the distance information is not calculated correctly The diameter of the ball is a problematic issue due to the low cost optics Instead we decided to change our ball into a cylindric shape and assume that the object w
23. the script that is used to calculate the motor contributions according to a set point motor control behavior and the software imple mentation 4 1 Wheel configuration The configuration space of an omnidirectional mobile robot is a smooth 3 manifold and can then be locally embedded in Euclidean space R The robot has three degrees of freedom i e two dimension linear motion and one dimension rotational motion There are three universal wheels mounted along the edge of the robot chassis 120 apart from each other and each wheel has a set of rollers aligned with its rim as shown in Figure Be cause of its special mechanism the robot is able to simultaneously rotate and translate Therefore the path planning can be significantly simplified by directly defining the tracking task with the orientation and position er rors obtained by the visual feedback For a nonholonomic robot the robot s velocity state is modeled as the motion around its instantaneous center of rotation ICR As a 2D point in the robot frame the ICR position can be represented using two indepen dent parameters One can use either Cartesian or polar coordinates but singularities arise when the robot moves in a straight line the ICR thus lies at infinity Hence we used a hybrid approach that is defining the robot 40 Chapter 4 Control position both in Cartesian and polar coordinates Normally the position of the robot is represented by Cartesian coordi n
24. 1 double d_cosA 0 8660 double d_sinA 0 5000 u8 cosA d_cosA 100 18 sinA d_sinA 100 s16 v 3 v_F v_L omega R_robot 100 ifdef NO_DEBUG print velocity vector Print Int v 0 v0 Print Int v 1 vi1 Print Int 2 v2 endif 18 k 1 816 MF 3 3 4 cosA sinA 100 cosA sinA 100 0 100 100 112 Appendix B Documentation of the programming ifdef NO_DEBUG print MF for k 0 k lt 3 k 4 USB_WriteString n for 1 0 1 lt 3 1 Print Int MF k 1 MF 1 endif vectorByMatrix3x1_s16 MF v vt ifdef NO_DEBUG motor speeds found Print Int vt 0 vtO Print Int vt i1 vti Print Int vt 2 vt2 endif void Direct_Motion s16 destination s8 angle s16 speed ifdef endif ifdef endif 816 frontal speed destination 0 816 lateral speed destination 1 double alpha 0 s16 velocity 2 velocity 0 frontal speed cos alpha lateral_speed sin alpha velocity 1 frontal_speed sin alpha lateral_speed cos alpha NO_DEBUG USB_WriteString CalculatingyMotiony n Print Int velocity 0 frontal Print_Int velocity 1 lateral Triskar velocity angle speed NO_DEBUG Print_Int speed 0 speed0 Print_Int speed 1 speed1 Print_Int speed 2 speed2 Misc h B 1 Microprocessor Code 113 k Motion h Created on Jan 19 2011
25. One major problem is the blob search algorithm The algorithm used inside is unknown and this forced us to define color information very precisely since we cannot eliminate the non object blobs easily We can eliminate the problem only by using the area perimeter and the circularity information which in case the target is a ball is not enough for most cases to track the blob So this makes the color definition important The color is defined in RGB Cube using RGB444 format and it is not very clear to understand the correctness of the defined color from the color codes In order to ease this process the color selection is made by the ST software by selecting the desired color from the image or directly programming the rules if the color values are known Since the resolution is very low the color consists of shadows or existence of blob not belonging to the searched object made us to improve the color selection The current ST software is sufficient to visualize the selected color and test the blob search but color definition is difficult since in order to define the color the pixels should be selected one by one is not possible to select an area Moreover the rules which is the additions and subtrac tions of the color codes to RGB Cube found by the ST software are not in the compact form the same rule for a color can be added or subtracted more than one To improve the color selection we decided to use an of fline calculator that
26. PWM signal and vary the duty cycle to act as a throttle 10096 duty cycle full speed 096 duty cycle coast 5096 duty cycle half speed etc After some testing we optimized the percentage of the duty cycle in order achieve a better performance This optimization will be mentioned later in Control Chapter For the motor control we started by using the H Bridge motor control circuits provided by our sponsor The initial tests have been performed by implementing the correct PWM waves using these boards Later we real 36 Chapter 3 Mechanical Construction ized that the boards were configured to work at 8 V This forced us to make the decision of buying new batteries or new control circuits Evaluating the prices we ended up buying new control circuits that are rated for 5 V MC33887 motor driver integrated circuit is an easy solution to connect a brushed DC motor running from 5 to 28 V and drawing up to 5 A peak The board incorporates all the components of the typical application plus motor direction LEDs and a FET for reverse battery protection A micro controller or other control circuit is necessary to turn the H Bridge on and off The power connections are made on one end of the board and the con trol connections 5V logic are made on the other end The enable EN pin does not have a pull up resistor so it be must pulled to 5 V in order to wake the chip from sleep mode The fault status FS active low output pin may
27. Some robots are omnidirectional meaning that they can move at any time in any direction along the ground plane x y regardless of the ori entation of the robot around its vertical axis This level of maneuverabil ity requires omnidirectional wheels which present manufacturing challenges Omnidirectional movement is of great interest to complete maneuverability 2 1 Locomotion 11 Omnidirectional robots that are able to move in any direction x y 0 at any time are also holonomic There are two possible omnidirectional config urations The first omnidirectional wheel configuration has three omniwheels each actuated by one motor and they are placed in an equilateral triangle as de picted in Figure 2 7 This concept provides excellent maneuverability and is simple in design however it is limited to flat surfaces and small loads and it is quite difficult to find round wheels with high friction coefficients In general the ground clearance of robots with Swedish and spherical wheels is somewhat limited due to the mechanical constraints of constructing om nidirectional wheels Figure 2 7 Palm Pilot Robot with omniwheels The second omnidirectional wheel configuration has four omniwheel each driven by a separate motor By varying the direction of rotation and relative speeds of the four wheels the robot can be moved along any trajectory in the plane and even more impressively can simultaneously spin around its vertical axis
28. The resulting point is the distance of the robot to the target object This information is later used to calculate the motor contributions Among the mentioned approaches we started with the first configura tion for the ball The idea was to detect a ball no matter whether it is on the ground or up in the air We started by testing with a red ball since the color red blue green were initially defined for the camera The results were not sufficient to follow the ball for greater distances because of several effects One of them is related with the camera resolution Normally the camera can work with 160x120 resolution for the blob search This does not cause a problem if the ball is close to the camera but as the ball goes further from the camera the ball represented in the image is getting smaller and smaller Due to the low cost optics at the worst case the ball becomes 1 pixel in the picture Using only this pixel it is impossible to detect the ball Even though we can recognize the ball for some cases in most of the cases we cannot Due to the low cost optics it is possible to have several other 1 pixel size red blobs that are not our ball In order to detect the ball inside the blob search algorithm we need to eliminate the blobs that are not the ball we are searching 5 1 Camera Calibration 57 Later in order to solve this problem we came up increasing the size of the ball Again this trial did not resulted as we expected
29. also to break down the motors The pins 5 3 5 2 1 0 4 0 4 1 are for the tactile bumper sen sors The detailed description can be found in definitions h inside main software 144 Appendix D Datasheet 333IDOSI AZEEZE 2 NGQREND 2IZIZENE GIMINDND 82223222 22272222 22222222 te d su mm sr LE 5f si HER 5 i pa 23 33 Ba cs TT m oe DIETS SERENA TRECE mS ITE A ES Figure D 1 The schematics for the RVS Module board 145 Das 2d hmi 31 5 31 BosssaaoguadusES 5555555 55559093 Sio gt Figure D 2 The schematics for the STL Mainboard 140 Appendix D Datasheet o o PIO er a 59 gGPIlO4l 1 Figure D 3 The pin mappings of the board with corresponding GPIO in the microcon troller
30. and floating point operations using this approach 66 Chapter 5 Vision Chapter 6 Game The robot will be used as a testbed to develop robogames As a testbed a simple game is implemented using almost all the components available on the robot The game is used also to verify the correctness of the solutions implemented We created a simple game to test the robot Basically the robot has to go to several targets in a sequence by avoiding obstacles until the final target is acquired The game flow can be expressed in a better way with an algorithmic approach The class diagram is shown in Appendix A 1 and the flow dia gram in Appendix A 2 The microcontroller starts by controlling the game end status The game end status is composed of conditions First it checks whether the target is acquired If it is acquired then it also checks whether this is the last target and ends the game by staying at the current position If it is not the last target the algorithm steps from the game status check phase and continues the search Before performing any other operation the collision status is controlled If a collision is detected a proper command is sent to motors to get rid of the obstacle The next step is blob search and blob tracing The camera searches for the target at its vision side If no blob is found the robot performs a turn around its center of rotation until a blob is found or a collision detected Normally the collisi
31. be left disconnected if it is not needed to monitor the fault conditions of the motor driver if it is connected it must use an external pull up resistor to pull the line high IN1 and IN2 control the direction of the motor and D2 can be PWMed to control the motor s speed D2 is the not disabled line it disables the motor driver when it is driven low another way to think of it is that it enables the motor driver when driven high Whenever D1 or D2 disable the motor driver the FS pin will be driven low The feedback FB pin outputs a voltage proportional to the H Bridge high side current providing approximately 0 59 volts per amp of output current Voltage Divider and Voltage Regulator Circuit Batteries are never at a constant voltage For our case 7 2 V battery will be at around 8 4 V when fully charged and can drop to 5 V when drained In order to power microcontroller and especially sensors which are sen sitive to the input voltage and rated to 5 V we need a voltage regulator circuit to output always 5 V The design of the circuit that will be used in voltage regulation merged with the voltage divider circuit that will be used for battery charge monitor shown in Figure 3 16 To operate voltage divider circuit the following equation is used to de termine the appropriate resistor values Vi o un Ri Re a V is the input voltage Ri and Ro are the resistors and V is the output voltage 3 7 Hardware Architectur
32. calculates the histogram for the selected object s colors As a first step we started by disabling automatic white balance and au tomatic exposure control Before getting into the white balance the concept of color temperature needs to be introduced Color temperature is just a way of quantifying the color of light It is measured in degrees Kelvin K Normal daylight has a color temperature of around 6 500K Warmer light has a lower color temperature The warm light that occurs late in the after noon might have a color temperature of around 4 000K Cooler light has a higher color temperature The bluish light that sometimes occurs in twilight periods of the day might have a color temperature of about 7 500K So our concept of warm and cool light is tied directly to the color temperature The 5 2 Color Definition 59 warmer yellow the light the lower the color temperature the cooler the light blue the higher the color temperature Tungsten Bulb household variety Sunrise Sunset clear sky Daylight with Clear Sky sun overhead Color temperature is also important for cameras The camera manufac turers knew that the color of the light would affect the colors delivered by the camera Therefore they decided to deal with the problem by designing the cameras to automatically measure the light temperature and to make adjustments as the light changes color That is why we can shoot in the rel atively neutral light with camera
33. camera itself is connected to a servo that enables the camera head to move freely in 120 in the vertical axis as shown in Figure 3 9 and 3 10 This configuration gives us the flexibility to generate different types of games using the vision available in a wide angle and interchangeable height 3 5 Bumpers The last mechanical component is the collision detection mechanism to avoid obstacles in the environment There are lots of good solutions to this issue By using different types of sensors such as sonars photo resistors 30 Chapter 3 Mechanical Construction Figure 3 9 The position of the camera Figure 3 10 The real camera position IR sensors tactile bumpers etc Among them the simplest are the tactile bumpers A tactile bumper is probably one of the easiest way of letting a robot know if it s colliding with something Indeed they are implemented by electrical switches The simplest way to do this is to fix a micro switch to robot in a way so that when it collides the switch will be pushed in making an electrical connection Normally the switch will be held open by an inter nal spring Tactile bumpers are great for collision detection but the circuit 3 6 Batteries 31 itself also works fine for user buttons and switches as well There are many designs possible for bump switches often depending on the design goals of the robot itself But the circuit remains the same They usually implement a mechanical b
34. correcting dead reckoning errors in mobile robots 1994 H Everett and L Feng J Borenstein Navigating mobile robots Sys tems and techniques AK Peters Ltd Natick MA USA 1996 J L Jones and Anita M Flynn Mobile Robots Inspiration to Imple mentation Wellesley D Kragic Real time tracking meets online grasp planning Proceedings of IEEE International Conference on Robotics and Automation 2001 E Malis 2d 1 2 visual servoing IEEE Transaction on Robotics and Automation 1999 S Papert Mindstorms Children computers and powerful ideas New York Basic Books 1980 M Resnick A study of children and concurrent programming Interac tive Learning Environments vol 1 no 3 pp 153 170 1991 F Ribeiro I Moutinho P Silva C Fraga and N Pereira Three omni directional wheels control on a mobile robot 80 BIBLIOGRAPHY 39 41 42 43 P Boulanger F Blais S F El Hakim A mobile system for indoors 3 d mapping and positioning Optical 3 D Measurement Techniques IV Zurich Wichmann Press 1997 C E Guestrin R Sukthankar and M Paskin S Funiak Distributed localization of networked cameras Fifth Int l Conf on Information Processing in Sensor Networks 2006 S Thrun Probabilistic algorithms in robotics AI Magazine vol 21 no 4 pp 93 109 2000 Manuela M Veloso Entertainment robotics Commun ACM 45 50 63 March 2002 G Xiang Y H Liu K Li Y Shen Uncalibra
35. in order to go a specified position in the world It is possible to obtain three different movement model using the inverse kinematics model These are e Linear Movement e Rotation e Mixed Angular and Linear Movement Linear Movement 4 2 Matlab Script 43 This is case where the angular speed of robot should be zero and there should be a linear movement in the displacement which is the composition of Vr and Vy vectors The software simulator that we built calculates the motor contribution The user inputs three variables frontal speed lateral speed and angular speed and the program outputs each motor contribution Figure 4 2 shows the result of the simulator for the linear movement forward Via 200 n2 V 1501 si V ul A F gt 100r M T 50r S SS m dl vp 3500 m s V4 3206 0889 mm s v 350 mm s j v4 1446 8911 mm s 50L Op 0 rad s V 5 2856 0889 mm s Vio 2053 1089 mm s shank v V g 350 mm s 100 x Vg 73500 mm s o 130 3288 rad s 150 F w 116 1012 rad s 2 14 2276 rad s 200 F 250tb 1 Viag 300 200 100 0 100 200 300 Figure 4 2 A linear movement going slightly on the left on a line calculated by simulator In Figure 4 2 Vr represents the frontal speed is set to 3500 mm s Vr lateral speed to 350 mm s and W represents the angular velocity which is always 0 rad s for the linear movement The resulting motion is a linear movement towards the direction which is th
36. in the afternoon and then shoot the next day in the cool light of early morning and still probably get reasonable colors in both situations even though the color of the light was different Cameras correct for the change in light temperature using white balance With auto white balance the camera attempts to determine the color temperature of the light and automatically adjusts for that color tempera ture Auto white balance works reasonably well under the following condi tions e The application does not require absolute maximum color accuracy e There is not a preponderance of one color in the scene being pho tographed e Adjustments for the color temperature of the light As mentioned in the previous paragraph in auto white balance mode the camera does its best to determine the color of the light and make appropriate adjustments However the methodology that is used to do this requires that certain assumptions be made These assumptions do not always match perfectly the scene being photographed As a consequence the auto white balance option does not always yield perfect results Accordingly problems 60 Chapter 5 Vision may occur using auto white balance when the conditions listed above are violated Therefore auto white balance may not be a good choice if Absolute color accuracy is required There is a lot of one color in the scene The preponderance of one color can fool the auto white balance function into as
37. k k AK AK k K File Name definitions h Author AST Robotics group Date First Issued 11 May 2007 Version 1 0 Description generic definitions and pin assignments file for Dongle 2K IAAAAAAAIAAAAAAAAAAIAAAAAAAHAA CCRC CCR ICO ICR AAA AA RICO AAA IAA AAA AAA History 28 May 2007 Version 1 2 11 May 2007 Version 1 0 ARA AHORA ARA AHORA AAA IAA AAIAAARAAAAAARAAAAIAARAAAAAARAAAAAARAAARAAHAAA AAA AAA AAA AAA A THE PRESENT SOFTWARE WHICH IS FOR GUIDANCE ONLY AIMS AT PROVIDING CUSTOMERS WITH CODING INFORMATION REGARDING THEIR PRODUCTS IN ORDER FOR THEM TO SAVE TIME AS A RESULT STMICROELECTRONICS SHALL NOT BE HELD LIABLE FOR ANY DIRECT INDIRECT OR CONSEQUENTIAL DAMAGES WITH RESPECT TO ANY CLAIMS ARISING FROM THE CONTENT OF SUCH SOFTWARE AND OR THE USE MADE BY CUSTOMERS UF THE CODING INFORMATION CONTAINED HEREIN IN CONNECTION WITH THEIR PRODUCTS ARA AHORA AAA AHORA AAA IAA ARRANCA ROA ORAR AAA AAA AA AAA AAA KA A AAA KARA RA ARA Define to prevent recursive inclusion ifndef __DEFINITIONS_H define __DEFINITIONS_H lt UART2_RaD N A UART2_ReD UART2_TzD N A UART2_TzD I2C_SCL P3 4 SW I2C SCL I2C_SDA P3 5 SW I2C SDA CAM CE P3 6 GPIO_OUT 118 Appendix B Documentation of the programming CAM_PCLK P3 0 Ext_Dma_In CAM_VSYNC P3 2 EXT_INT_2 CAM_
38. needs of system modularity and using standard components as much as possible After modeling the chassis of the robot and having se lected the wheels the actuators have been chosen The hardware design has affected the choice of sensors used for estimat ing the state of the system and the design of a microcontroller based control logic The last part of the work consisted in carrying out experiments to es timate the position of the robot using the data from the sensors and to improve the performance of the control algorithm The various contributors affecting the performance of the robot behavior have been tested allowing to observe differences in performance and alternative solutions have been implemented to cope with limitations due to low cost of HW and low com putational power 1 3 Achievements We have been able to develop a robot that is able to follow successfully a predefined colored object thus implementing many interesting capabilities useful for robogames We have faced some limitations due to the low cost constraints The main challenges have been caused by the absence of motor encoders and low cost optics We have done some optimizations and tunings to overcome these limitations 1 4 Thesis Structure The rest of the thesis is structured as follows In chapter 2 we present the state of the art which is concentrated on similar applications the tech niques used and what has been done previously Chapter 3 is about the
39. order to move with agility in the home The space between the two plates is around 6 cm which allows us to mount sensors and any device that will be added in the future The total weight of the structure is approximately 1 8 kg including motors and batteries 3 2 Motors 23 The initial design of the chassis was a bit different from the final configu ration seen in Figure 3 1 Even though the shape of the components did not change the position and orientation are changed in the final configuration The motor holders initially were intended to be placed on the top of the bottom plexiglas layer At the time when this decision was taken we were not planning to place the mice boards but only to put the batteries and the motor control boards Later with the decision of placing the mice boards in this layer in order to get more space we decided to put the motors to their final position So this configuration increases the free space on the robot layers to put the components and also increases the robot height from the ground that will result to better navigation Another change has been made by placing the second plexiglas layer Initially we placed that layer using only three screws with each a height of 6 cm The idea was using minimum screws so that the final weight will be lighter and the plexiglas will be more resistant to damage Later when we placed the batteries motor controller boards and the camera with its holder the total wei
40. pi 2 alpha end m arrl M m arr offsetl offsetl offsetl v_txt1 M v txt offsetl fill m arr1 1 m arr1 2 g text v txt1 1 v txt1 2 v n2 end ruota 3 if vt 3 0 line 0 vt 3 s1 R_robot R_robot Color 0 1 0 offsetl vt 3 sl R robot if vt 3 sl lt 0 M m rot pi 2 else M m rot pi 2 end m_arrl M m arr offsetl offsetl offsetl v txtl Mxv_txt offsetl fill inicare 1 m arr1 2 g text v_txt1 1 v txt1 2 v t3 end if vn 3 gt 0 line 0 0 R robot R_robot vn 3 s1 Color 0 1 0 offsetl 0 R robot vn 3 s1 if vn 3 sl lt 0 M m rot pi else M m rot 0 end m_arrl M m arr offsetl offsetl offsetl v_txtl M v_txt offsetl fill m arr1 1 m arr1 2 g B 4 Motion Simulator 133 end text v_txt1 1 v txt1 2 v n3 end y ref 0 delta 0 095 R robot text R_robot 0 6 y_ref v 11 num2str vt 1 umm s text R _robot 0 8 y_ref v_ F num2str v F nm s y ref y ref delta text R_robot 0 6 y_ref v_ nl num2str vn 1 mm s text R robot 0 8 y ref v L1 num2str v L mm s y ref y ref delta text R_robot 0 6 y_ref v_ t2 num2str vt 2 umm s text R robot 0 8 y _ref omega_ R _
41. previous step with ST s software IDCol is the color class defined and it is be the same also with the corresponding RGBCube_AddCube u8 RMin u8 RMax u8 GMin u8 GMax u8 BMin u8 BMax u8 IDCol and RGBCube_SubCube u8 RMin u8 RMax u8 GMin u8 GMax u8 BMin u8 BMax u8 IDCol functions that defines the color coding The values that are tested with S T s software should be updated in RGBCube_AddCube and RGBCube_SubCube functions 142 Appendix C User Manual The robot is ready to play the game after building the code and pro gramming the microprocessor through program_USB that can be found in Eclipse from the Make view Figure C 8 856 2 x E r El amp td gt ES STLCam lard 3 dep i settings gt fdlib E Iar4ASM gt include 3 Lib source gt STDLib gt usblib program_PP program USB Figure C 8 The programming of the microprocessor is done through program USB The process can be controlled through the console Appendix D Datasheet The pin mappings of the board to the corresponding GPIO port of the mi crocontroller can be seen in the figure D 3 Mainly GPIO pins 1 6 1 4 1 3 represents the PWM wave outputs for the motorl motor2 motor3 in the same order The pins 0 6 0 2 0 4 and 0 7 0 3 0 5 are the direction pins for the motor control that is used to control which direction is forward which is reverse and
42. than 18 of the light falling on them The exposure system doesn t know that the scene should look bright so it calculates an exposure that produces a middle gray image that is too dark Subjects that are darker than middle gray such as black cloth reflect less than 18 of the light falling on them The exposure system calculates an exposure that makes the image middle gray and too light The contrast or difference in brightness between the subject and the background can fool an exposure system particularly if the subject occu pies a relatively small part of the scene compared to the background The brightness of the background is so predominant that the automatic exposure system adjusts the exposure to render the overall brightness as a middle gray If the main subject is lighter than the background it will be overexposed and too light If it s darker than the background it will be underexposed and too dark Depending on the arrangement of the lighting some subjects may be too contrasty with brightly lit highlights and deep shadows The range of brightness may exceed the range that can be captured by the camera In these cases adjustments should be made in the lights to balance out the light and to lower the contrast However deciding whether the highlight or shadow areas are most important for the final picture the exposure setting should be made appropriately The perfect exposure retains details in both the highlights and shadows
43. the ones used in videogames or eventually in the physical building of the tele operated robots as it happens with RoboWars 11 A third main stream concerns robots that have been developed by roboticists to autonomously play games e g Robocup Here the accent is on the ability to program the robots to be autonomous but little effort is spent in the eventual playful interaction with people often avoided as in most of the Robocup leagues The last main stream concerns robots that act as more or less like mobile pets In this case interaction is often limited to almost static positions not exploiting rich movement nor high autonomy the credibility of these toys to really engage healthy people such as kids is not high According to the AIRLab report 19 a new category of games where the players are involved in a challenging and highly interactive game ac tivity with autonomous robots called as Highly Interactive Competitive RoboGames HI CoRG is introduced The idea is to take the videogame players away from screen and console and to make them physically interact with a robot in their living environment In this context some heuristics from videogames adapted to be applied on this HI CoRG games In our thesis we focused on developing a robot for games that can be count to the HI CoRG category 2 4 Games and Interaction 17 We introduce now some of the robots developed in the past related to the human robot interaction
44. the front wheels rads in our case alpha pi 6 gt vF frontal speed setpoint mm s ts axis vL lateral speed setpoint mm s gt 0 if oriented to right y axis omega angular velocity setpoint rad s gt 0 if CCW plot draw a plot with all vectors 1 plot 0 don t plot this function returns vt vector of tangential wheel velocities mm s vn vector of sliding velocity of the wheels due to rollers mm s gt 0 if directed from the center to the extern omega w angular velocity of the wheels mm s gt 0 if CW looking at the wheel from the center of the robot You can call that function like vt omega w vn TriskarOdometry 250 24 6 pi 6 1000 1000 0 1 cosA cos alpha sinA sin alpha v v F v L omega R_robot MF cosA sinA 1 cosA sinA 1 0 1 i ML sinA cosA 0 sinA cosA 0 1 0 0 vt MFxv om gt vt R_wheel vn MLxv B 4 Motion Simulator 129 if plot 1 scalature kl max abs v F abs v L vl v_F k1 R_robot 2 v2 v_L k1 R_robot 2 pl R robot 2 m arr 0 0 04x p1 0 04xp1 0 1 pl 0 0 v txt 0 05xp1l 0 2xp1 sl R robot 2 max abs vt vn plot figure 1 hold on hl line 0 R_robot cosA 0 R robot sinA h2 line 0 R_robot cosA 0 R robot sinA h3 line 0 0 0 R robo i t set hl Color 0 2 0 2 0 2 2 2
45. this configuration so we decided to keep it 3 6 Batteries The robot s battery life without the need of recharging is crucial for the game The game play must continue for about 2 hours without any in terruption This brings the question of how to choose the correct battery LiPo batteries are suitable battery choice for our application over conven 32 Chapter 3 Mechanical Construction Figure 3 12 The bumper design Figure 3 13 Robot with foams springs and bumpers tional rechargeable battery types such as NiCad or NiMH for the following reasons e LiPo batteries are light in weight and can be made in almost any shape and size e LiPo batteries have large capacities meaning they hold lots of power in a small package e LiPo batteries have high discharge rates to power the most demanding electric motors 3 7 Hardware Architecture 33 In short LiPo provide high energy storage to weight ratios in an endless variety of shapes and sizes The calculation is made to find the correct battery The motors are consuming 250 mA at free run and 3300 mA for the stall current For the all three motors we should have the following battery lives Battery Capacity C urrent_Draw Battery_Life 2 2500mAh 750mA 6 6hours 2 2500m Ah 9900mA 0 5hours using the 2 batteries each having a capacity of 2500 mAh The battery life shows changes according to the current draw of the motors In case each motor is consuming
46. toolboz K 191 71462 0 80 35911 0 191 27299 61 27650 0 0 1 lis x y z the distance of the chess board from the robots center in mm x 550 B 3 Object s Position Calculator 125 y 150 z 0 T rch indicates the transformation robot to chess board T rch 1 0 0 x 0 1 0 y 0 0 1 7 I transformation are rewritten for clarification camera to world Tech T cch 0 0 0 1 world to camera Teche inv T cch1 world to robot T rch1 T rch 0 0 0 Ll robot to world T chr inv T rch1 T_re T_rch1 T_chc T cr inv T_rc Ter Tende ge ROBOT Gl a en ole ROBOT a sig aeree ROBOT 2 Serta oe es t_ 1 0 0 010 000 000 1 H KxT_cr t_ Hinv inv H point Hinv 72 119 1 result point point 3 result 3 Calibration for servo_top position at ground tranlastion m Tc ext 100 613176 95 664945 539 117313 14 120 Appendix B Documentation of the programming rotation matrix Rc ext 0 025851 0 998468 0 048927 0 364120 0 054986 0 929727 0 930993 0 006219 0 364984 transformation of camera to chess board T ch Reuext Tc ext K internal camera parameters from image toolbox K 191 71462 0 80 35911 0 191 27299 61 27650 0 0 1 x y z the distance of the chess board from the robots center in mm x 600 y 100 2 gt 0 T rch indicates the transformation robot to chess board Tach 1 0 0 x 0 1 0 y 0 0 1 7 IE transformation are rewrit
47. will review the literature for the robots that are related with our de sign We divided the review in subsections like Locomotion Motion Models Navigation and Interaction and Games 2 1 Locomotion There exists a great variety of possible ways to move a robot which makes the selection of a robot s approach to motion an important aspect of mobile robot design The most important of these are wheels tracks and legs 33 6 Chapter 2 State of the Art The wheel has been by far the most popular motion mechanism in mo bile robotics It can achieve very good efficiency with a relatively simple mechanical implementation and construction easiness On the other hand legs and tracks require complex mechanics more power and heavier hard ware for the same payload It is suitable to choose wheels for robot that is designed to work in home environment where it has to move mainly on a plain surface There are three major wheel classes They differ widely in their kinemat ics and therefore the choice of wheel type has a large effect on the overall kinematics of the mobile robot The choice of wheel types for a mobile robot is strongly linked to the choice of wheel arrangement or wheel geometry First of all there is the standard wheel as shown in Figure P 1 a The standard wheel has a roll axis parallel to the plane of the floor and can change orientation by rotating about an axis normal to the ground through the contact point The standa
48. 150 int Dreal 62 red ball w 6 2 cm diam 160 9969 fstar Dreal Dpx position 0 fstar Dreal Dpx position 1 fstar Dreal Dpx position 2 1 T_wr 4 4 4 1 0 O X axis 0 1 O Y axis 1 0 0 1 O 0 0 O 1 7 T_wc 4 4 0 0012 0 3587 0 9334 350 8669 1 0000 0 0039 0 0002 110 Appendix B Documentation of the programming 140 5637 0 0037 0 9334 0 3587 203 8752 0 0 0 1 0000 double inv_T 4 4 0 0 0 O Matrix_Mult T_wr T wc inv T vectorByMatrix4xi1 inv T Pc result Motion h Motion h Created on Jan 19 2011 Author Administrator ifndef MOTION_H_ define MOTION_H_ k k Includes S555 SSS 5556 Soo Sess SS DS SS SS SS DS Exported types Exported constants Modute Private Vvaruables 225 22 a SS Exported macro Private functions Exported functions 5 void Triskar s16 destination s8 omega s16 vt void Direct_Motion s16 destination s8 angle s16 speed endif MOTION_H_ Motion c Standard include
49. 250 mA in free run current will result 6 6 hours of batteries life On the other hand with the stall current it will be 0 5 hour battery life Since the motor will not always work in stall current or the free run current the choice of 2500 mA batteries Figure 3 14 seems to be enough to power the robot for at least 2 hours Figure 3 14 J tronik Battery Li Po Li POWER 2500 mA 2S1P 7 4V 20C 3 7 Hardware Architecture During the development of the robot we used several hardware pieces such as microprocessor camera motor control boards voltage regulator circuit voltage divider circuit Most of them were already developed systems and we did not focus on the production details of them We only created the voltage regulator and divider circuit which we used in order to power the boards and measure the battery level of charge 34 Chapter 3 Mechanical Construction The main component in our system is the the STL Main Board known also as STLCam The STL Main Board is a low cost vision system for ac quisition and real time processing of pictures consisting of a ST VS6724 Camera 2 Mpx a ST STR912FA Microcontroller ARM966 96MHz and 16MB of external RAM PSRAM BURST The schematics of the STL Main Board is shown in Appendix D 2 ST STR912FAZ44 Microcontroller The microcontroller main components are a 32 bit ARM966E S RISC processor core running at 96MHz a large 32bit SRAM 96KB and a high speed 544KB Flash memory The A
50. 45 and 90 of duty cycle Mapping 300 mm s to the 45 and 2500 mm s to 90 did not resulted as expected since reflecting speed changes is not possible In order to express most of the speeds we decided to tune the PWM sent to motors Main idea is to bound to speeds in a dynamically changeable manner We introduced a fuzzy like behavior that is based on the robot s 48 Chapter 4 Control distance to the target position Each distance is divided by a constant then multiplied with the multiplier defined as FAST NORMAL and CLOSE Then another constant which is the minimum value that can initiate the motors movement is added Hence the mapping that each value set by the PWM signal generates a motion is achieved The multipliers are defined according to the distance of the robot to the target object For the distances between 1200 mm and greater the corresponding multiplier is classified as FAST between 1200 mm and 750 mm the multiplier is NORMAL and for the distances between 750 mm and 300 mm it is CLOSE By defining such control mechanism as FAST NORMAL CLOSE we are able to create three different speed limits for the motors which are all mapped from 60 to 85 of the duty cycle where the motion is visible As an example case when the robot detects a target at 3 meters the motor speeds will be set with multiplier FAST until the distance to the target reaches 1 2 meters When reached to 1 2 meters the motor speed will be set with multip
51. 61 27650 0 0 1 T rch indicates the transformation robot to chess board x y z the distance of the chess board from the robots center in mm x 600 y 0 z 0 T rch 1 0 0 x 0 1 0 y 0 0 1 7 transformation are rewritten for clarification camera to world Tow T cch 0 0 01 world to camera T_wc inv T_cw world to robot T wr T rch 0 0 0 1 robot to world Trw inv T_wr 124 Appendix B Documentation of the programming position inv K image_point l sqrt x 2 y 2 l the distance of the origin point selected in the image to the robot center Dreal 200 diameter of the ball in mm Dpx 29 pirel counter of diameter in image the fx constant pixel unit mm fstar Dpx 1 Dreal the position should be normalized since position 3 is not 1 position position norm position Pc fstar Dreal Dpx position 1 inv T_cw xT_rw Pc result T wrxT wcxPc result ans gives the position of the ball in real world in mm ans 1 x ans 2 y ans 3 z coordinates in real world 2 Calibration for the object at Ground tranlastion matrix coming from image toolboz Tc ext 160 700069 27 498986 492 532634 rotation matrir coming from image toolboz Rc ext 0 009783 0 999883 0 011800 0 428525 0 014854 0 903408 0 903477 0 003781 0 428620 transformation of camera to chess board T cch Rc_ext Tc ext 7 K internal camera parameters from image
52. Author Administrator tifndef MOTION_H_ define MOTION_H_ Includes 22 92 2222222222222 SRS SSS Re SSeS 22 RAS SA Exported types r lt Exported constants Module private variables Exported macro Private functions Exported functions void Triskar s16 destination s8 omega s16 vt void Direct_Motion s16 destination s8 angle s16 speed endif MOTION_H_ Misc c Standard include include 91x map h include utils h include definitions h include definitionsLib h include CamInt_x24 h include usb_lib h include usb_conf h include usb_prop h include usb_pwr h include usb config h include RGBCube h include blobsearch h include math h Include of other module interface headers JE Local includes ss So 5 Ss sO S552 SS SS E Private typedef Private define 5 Private macro Private variables Private fun
53. CLK_IN P3 7 T1_PWM_OUT CAM BUS P9 ALL GPIO IN LED_R P6 0 GPIO_OUT 77 SW_1234 27 0123 6510 IN 77 USB_Dp P7 5 GPIO_OUT lt define RVS MB define STL_MB define NO_DEBUG define PresentationStringConst STLCam nV 0 41 nAST Robotics nSTMicroelectronics n define TMsg_EOF OxFO define TMsg BS OxF1 define TMsg BS EUF OxF2 define TMsg_MaxLen 128 define DEV_ADDR 0x32 define USB_Dp_PullUp_GPIO GPIOO define USB Dp PullUp GPIOx Pin GPIO Pin 1 define Cam724 IRQ Priority 0 Highest 15 Lowest define SERVO_TIM_ITPriority 0 define x24_DMA_ITPriority 1 define x24_VSYNC_INT_ITPriority 2 define x24_SYNC_TIM_IPriority 3 define USB_Priority 5 define MOTOR1_Priority 9 define MOTOR2_Priority 10 define MOTOR3_Priority 11 ifdef STL_MB define LEDR_GPIO GPI03 define LEDR_GPIOx_Pin GPIO_Pin_7 endif define DE_GPIO GPI04 define DE_GPIOx_Pin GPIO_Pin_3 define x24_VSYNC_GPIO GPIO7 define x24_VSYNC_GPIOx_Pin GPIO_Pin_7 define x24_VSYNC_INT_WIU_Line u32 WIU_Line31 B 1 Microprocessor Code 119 define x24_VSYNC_INT_WIU_Line_N 31 define x24 VSYNC INT ITx LINE EXTITS ITLine define MY GPIO GPI03 define MY GPIOx Pin GPIO Pin 2 define MY MY GPIO DR MY GPIOx Pin 2 define MY ON O MY 0xFF define MY OFF MY 0x00 define LEDR LEDR GPIO DR LEDR GPIOx Pin 2 define USB Dp
54. ClockConfig __GPI09 ENABLE GPIO DeInit GPIO9 Enable VIC clock SCU AHBPeriphClockConfig VIC ENABLE VIC DeInit Enable WIU clock SCU APBPeriphClockConfig WIU ENABLE WIU DeInit Enable WIU clock SCU_APBPeriphClockConfig __I2C0 ENABLE Enable DMA clock SCU_AHBPeriphClockConfig __DMA ENABLE DMA_DeInit Enable TIM0123 clock SCU_APBPeriphClockConfig __TIMO1 ENABLE B 1 Microprocessor Code 93 TIM_DeInit TIMO TIM DeInit TIM1 SCU_APBPeriphClockConfig __TIM23 ENABLE TIM DeInit TIM2 TIM DeInit TIM3 SCU TIMPresConfig SCU TIMO1 4800 10KHz SCU TIMExtCLKCmd SCU TIMO1 DISABLE Disable external pin SCU TIMPresConfig SCU TIM23 4800 10KHz SCU TIMExtCLKCmd SCU TIM23 DISABLE Disable external pin SCU_APBPeriphClockConfig __12C0 ENABLE I2C DeInit 1260 SCU_AHBPeriphClockConfig __FMI ENABLE SCU_AHBPeriphReset __FMI DISABLE LOK OK EE E SE AAA AA AAA AA AA AAA AA AAA AA AA AAA AA AAA AA KA AAA AAA AA AAA AA AAA AA AA AAA AA Eo oo oko eo ooo Function Name USB Configuration Description Configures the USB Input None Output None Return None EE EE ARA AR ARA E EE EE E OE OE EO AOR IA AAA AAA A KA A OR A KARA AAA OK OR AAA RA AAA RARA AAA RARA AAA AA HAS void USB Configuration void 1 GPIO InitTypeDef GPIO InitStructure USB clock MCLK 2 48MHz SCU USBCLKConfig SCU_USBCLK_MCLK2
55. DER FOR THEM TO SAVE TIME AS A RESULT STMICROELECTRONICS SHALL NOT BE HELD LIABLE FOR ANY DIRECT INDIRECT OR CONSEQUENTIAL DAMAGES WITH RESPECT TO ANY CLAIMS ARISING FROM THE CONTENT OF SUCH SOFTWARE AND OR THE USE MADE BY CUSTOMERS OF THE CODING INFORMATION CONTAINED HEREIN IN CONNECTION WITH THEIR PRODUCTS AAA AAA AA AO RO RO RO ROIO ARA AA AAA AA AAA AA RA AAA AA AAA AAA AAA ARA AAA AAA AA AAA RR RR AA AAA ARAS FE Includes S3 222 gt H 2 PP SA A Lue edem include 91x lib h include definitions h include CamInt_x24 h 86 Appendix B Documentation of the programming include utils h include math h include usb_lib h include usb_conf h include usb prop h include usb pwr h include usb config h include definitionsLib h include Misc h include Camera_Init h include Motor_Init h include blobsearch h Private typedef Private define Private macro Private variables extern u8 control_hit Private function prototypes void SCU_Configuration void void USB_Configuration void void GPIO_Configuration void Private functions
56. Direction GPIO_PinInput GPIO InitStructure GPIO Pin GPIO_Pin_0 GPIO_Pin_1 GPIO InitStructure GPIO Type GPIO Type PushPull GPIO Init GPIO4 amp GPIO InitStructure void Bumpers u8 bumper 6 control_hit 0 if GPIO ReadBit GPI05 BUMPER_FRONT_LEFT Bit SET bumper 0 1 if GPIO_ReadBit GPI04 BUMPER_FRONT_RIGHT Bit_SET bumper 4 1 if GPIO ReadBit GPIO1 BUMPER_LEFT Bit SET bumper 2 1 if GPIO ReadBit GPI05 BUMPER RIGHT Bit SET bumper 1 1 if GPIO ReadBit GPIO1 BUMPER BACK LEFT Bit SET bumper 3 1 if GPIO ReadBit GPIO4 BUMPER BACK RIGHT Bit SET bumper 5 1 u8 i for i 0 i lt 6 i if bumper i 1 control_hit 1 ifdef NO_DEBUG Print_Int i no Print_Int bumper i Bumper endif void Turn u8 direction 106 Appendix B Documentation of the programming s16 left_speed 3 550 550 550 s16 right_speed 3 550 550 550 if direction LEFT Set_Speed left_speed NORMAL else Set_Speed right_speed NORMAL MSDelay 100 Stop_Motor 3 MSDelay 100 Vision h Viston h Created on Jan 19 2011 Author Administrator ifndef VISION_H_ define VISION_H_ Includes include 91x_lib h include CamInt_x24 h Exported types
57. For the auto exposure system this is as difficult If there is even a little too much exposure the image is too light and details are lost in the highlights If there is too little exposure the image is too dark and details are lost in the shadows When confronted with any subject lighter or darker than middle gray exposure compensation is used to lighten or darken the photograph that the camera would otherwise produce To lighten a picture the exposure is increased This is useful for setups where the background is much lighter than the subject or when photograph ing very light objects such as white china on a white tablecloth To darken an image the exposure is decreased This is useful for setups where the 62 Chapter 5 Vision background is much darker than the subject or when photographing very dark objects such as black china on a black tablecloth From the parameters available in our camera we disabled the automatic exposure by setting DIRECT_MANUAL_MODE in which the exposure pa rameters are input directly and not calculated by the camera We don t want the camera automatically adjust the exposure since this will cause to lighten or darken the picture dynamically Instead by setting the exposure directly to a constant value we ensure that the picture will remain with the same enlightenment together with the target we are searching for After setting the white balance and exposure compensation we took sample images F
58. GH and update the desired motor speed of the output compare register 1 When counter reaches the defined value the motor speeds will be updated and the pins will be reset It will remain LOW until output compare register 2 value which is the full period is reached Using that algorithm we generate the PWM signal to run the motor For each motor we decided to use a separate timer since the timers are available to use and not needed by other applications It is also possible to use only one timer to generate the PWM s for all the motors using the interrupts and output compare registers but we decided not to use like this for the simplicity of the algorithm since this will need a better synchroniza tion and optimization to run efficiently The control of the servo that is changing the camera head pointing direc tion is also implemented using PWM signals For the servo PWM we don t have the problem of control that we faced with the motors So no tunings 50 Chapter 4 Control are implemented for this case and the PWM creation is implemented sim ply as follows The PWM which controls the servo position is implemented with the first timer at with the motor The period of PWM in first timer to control the motor is divided by a constant in order to achieve a smaller PWM period which is around 2ms for the servo The rest of the idea is the same The PWM is generated with the help of the output compare mode The servo position is initializ
59. IT_OC1 TIM_IT_0C2 ENABLE Enable 1141 Output Comparel interrupt TIM ITConfig TIM1 TIM_IT_TO TIM_IT_OC1 TIM_IT_0C2 ENABLE Enable 1143 Output Comparel interrupt TIM_ITConfig TIM3 TIM_IT_TO TIM IT OC1 TIM_IT_0C2 ENABLE VIC_Config TIMO_ITLine VIC IRQ 9 VIC ITCmd TIMO ITLine ENABLE VIC_Config TIM1_ITLine VIC IRQ 10 VIC ITCmd TIM1 ITLine ENABLE VIC_Config TIM3_ITLine VIC IRQ 11 VIC_ITCmd TIM3 ITLine ENABLE Start TIM_CounterCmd TIMO 1111 START TIM_CounterCmd 1111 1111 START TIM_CounterCmd 1113 1111 START u8 i s16 speed 3 for i 0 i lt NUM MOTORS i 1 speed i 0 Set_Speed speed CLOSE void Position_Camera u8 position 0 servo_position servo_flag position void Init Bumpers O 1 GPIO InitTypeDef GPIO_InitStructure GPIO_StructInit amp GPIO_InitStructure GPIO InitStructure GPIO Direction GPIO InitStructure GPIO Pin GPIO Pin 2 GPIO InitStructure GPIO Type GPIO Init GPIO5 amp GPIO InitStructure GPIO_StructInit amp GPIO_InitStructure GPIO_PinInput GPIO_Pin_3 GPIO_Type_PushPull B 1 Microprocessor Code 105 GPIO InitStructure GPIO Direction GPIO_PinInput GPIO InitStructure GPIO Pin GPIO Pin 0 GPIO Pin 7 GPIO InitStructure GPIO Type GPIO Type PushPull GPIO Init GPIO1 amp GPIO InitStructure GPIO_StructInit amp GPIO_InitStructure GPIO InitStructure GPIO
60. Mode x24 InitStruct DesiredFrameRate Num Framerate 250 x24 InitStruct DesiredFrameRate Den 250 x24 InitStruct ExtClockFregMhz Num 8 x24 InitStruct ExtClockFregMhz Den 1 x24 InitStruct JPEGClockDerate JpegDerat x24 InitStruct SyncEnabled O0 x24 InitStruct ImageWidth Width x24 InitStruct ImageHeight Height x24 InitStruct ImageFormat ImgFormat c24 ImageFormat RGB 565 24 Imagel x24 InitStruct JPEGSqueezeValue JpegCompr use 100 800x600 150 1024 x24 InitStruct SysClock HalfSysClock 2 0 35MHz Range 35 270 MHz 320 240010Hz RGB 565 x24 InitStruct MemBlocksNum 2 x24 InitStruct ImgBlocksNum 15 x24 InitStruct ImgBlockSize BufferSize x24 InitStruct pDataMemory pImgBufferData x24 InitStruct ExposureCompensation 5 x24 biExposureCompensation x24 Init x24 InitStruct Job 0 pNext amp Job 1 Job 1 pNext amp Job 0 CurrentJob 0 x24 GrabFrameStart amp Job 0 Motor Init h Motor_Init h Created on Jan 19 2011 Author Administrator ifndef MOTOR_INIT_H_ define MOTOR_INIT_H_ B 1 Microprocessor Code 99 FE 663 lt lt lt lt lt lt lt lt Sos Se Se eS eS SSS pS Se ea ese E Exported types Z Exported constants Module private variables
61. O MOTOR1 DIRECTION B Bit RESET else 1 GPIO WriteBit MOTOR DIRECTION GPIO MOTOR1 DIRECTION A Bit RESET GPIO WriteBit MOTOR DIRECTION GPIO MOTOR1 DIRECTION B Bit SET if Motor3 FORWARD GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR3_DIRECTION_A Bit_SET GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR3_DIRECTION_B Bit_RESET else GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR3_DIRECTION_A Bit_RESET GPIO_WriteBit MOTOR_DIRECTION_GPIO MOTOR3_DIRECTION_B Bit_SET void Calculate_Order s16 speed u8 size u8 pass 1 u8 sorted 0 u8 i while sorted amp amp pass lt size 1 sorted 1 for i 0 i lt size pass i if speed i gt speed i 1 1 int temp speedlil speed i speed i 1 speed i 1 temp sorted 0 pass void Stop_Servo 1 servo_flag STOP_SERVO B 1 Microprocessor Code 103 void Init_Motors GPIO_InitTypeDef GPIO_InitStructure GPIO_StructInit amp GPIO_InitStructure GPIO InitStructure GPIO Direction GPIO_PinOutput GPIO InitStructure GPIO Pin MOTOR1_Pin MOTOR2 Pin MOTOR3 Pin GPIO InitStructure GPIO Type GPIO Type PushPull GPIO InitStructure GPIO Alternate GPIO_OutputAlti GPIO Init MOTOR GPIO amp GPIO InitStructure Config Servo PIN GPIO_StructInit amp GPIO_InitStructure GPIO InitStructure GPIO Direction GPIO_PinOutput GPIO InitStructure GPIO Pin SERV
62. O Pin GPIO InitStructure GPIO Type GPIO Type PushPull GPIO InitStructure GPIO Alternate GPIO_OutputAlti GPIO Init SERVO GPIO amp GPIO InitStructure GPIO_StructInit amp GPIO_InitStructure GPIO InitStructure GPIO Direction GPIO_PinOutput GPIO InitStructure GPIO Pin MOTOR1_DIRECTION_A MOTOR1 DIRECTION B MOTOR2 DIRECTION A MOTOR2 DIRECTION B MOTORS DIRECTION A MOTORS DIRECTION B GPIO InitStructure GPIO Type GPIO Type PushPull GPIO InitStructure GPIO Alternate GPIO_OutputAlti GPIO Init MOTOR DIRECTION GPIO amp GPIO InitStructure TIM InitTypeDef TIM InitStructure TIM StructInit amp TIM InitStructure TIM InitStructure TIM Mode 111 0CM CHANNEL 12 for both use 12 TIM InitStructure TIM OC1 Modes TIM TIMING TIM InitStructure TIM 0C2 Modes TIM TIMING TIM InitStructure TIM Clock Source TIM CLK APB TIM InitStructure TIM Prescaler 47 127 TIM InitStructure TIM Pulse Length 1 50 TIM InitStructure TIM Pulse Length 2 200 TIM InitStructure TIM Pulse Level 1 TIM HIGH TIM InitStructure TIM Pulse Level 2 TIM HIGH TIM InitStructure TIM Period Level TIM_LOW HHA HAHAHAHAHA AH TIM_Init TIMO amp TIM_InitStructure TIM Init TIM1 amp TIM InitStructure TIM Init TIM3 amp TIM_InitStructure 104 Appendix B Documentation of the programming Enable TIMO Output Comparel interrupt TIM_ITConfig TIMO TIM_IT_TO TIM_
63. PER_BACK_LEFT GPIO_Pin_7 1 7 define BUMPER_BACK_RIGHT GPIO Pin 1 4 1 define BUMPER RIGHT GPIO Pin 3 5 5 define BUMPER LEFT GPIO_Pin_0 1 0 define LEFT 1 define RIGHT define CLOSE 8 define NORMAL 4 define FAR 2 endif kxkkkkkkkkkkkkkkkkkxk C COPYRIGHT 2007 STMicroelectronics END OF FILE B 2 Color Histogram Calculator color_calculator m data structure to keep images R G Bis the sample images should be read 1 by 1 im imread table3 png fprintf done t Z the part containing the object should be selected sub imerop im 1 3 imshow sub subR double sub 1 subG double sub 2 subB double sub 3 B 2 Color Histogram Calculator 121 R R subR G G subG B B subB save load matlab After taking all samples the histogram drawn automatically 6 Il QW cc e IH S subplot 1 3 1 hist double R 16 title R hold on subplot 1 3 2 hist double G 16 title G hold on subplot 1 3 3 hist double 16 title B hold on bound finding amp the mask and the object shown to see the result of Ythe color coding im imread table3 png the bounds for each color channel the values that should be pass the microprocessor for color coding bR prctile R 25 12 5 75 12 5 bG prctile G 10 90
64. POLITECNICO DI MILANO Corso di Laurea in Ingegneria Informatica Dipartimento di Elettronica e Informazione BE SO Kes oo CE gt iy DEVELOPMENT OF A LOW COST ROBOT FOR ROBOGAMES AI amp R Lab Laboratorio di Intelligenza Artificiale e Robotica del Politecnico di Milano Coordinator Prof Andrea Bonarini Collaborator Martino Migliavacca MS Thesis ANIL KOYUNCU matricola 737109 Academic Year 2010 2011 POLITECNICO DI MILANO Corso di Laurea in Ingegneria Informatica Dipartimento di Elettronica e Informazione DEVELOPMENT OF A LOW COST ROBOT FOR ROBOGAMES AI amp R Lab Laboratorio di Intelligenza Artificiale e Robotica del Politecnico di Milano Mi ecc E Anil Koyuncu Prof Andrea Bonarini Academic Year 2010 2011 To my family Sommario L obiettivo del progetto lo sviluppo di un robot adatto ad implementare giochi robotici altamente interattivi in un ambiente domestico Il robot verr utilizzato per sviluppare ulteriori giochi robotici nel laboratorio AIR Lab Il progetto stato anche utilizzato come esperimento per la linea di ricerca relativa alla robotica a basso costo in cui i requisiti dell applicazione e il costo costituiscono le specifiche principali per il progetto del robot stato sviluppato l intero sistema dal progetto del robot alla realizzazione del telaio delle componenti meccaniche e elettroniche utilizzate per il controllo del robot e l acquisi
65. PullUp USB Dp PullUp GPIO DR USB Dp PullUp GPIOx Pin lt lt 2 define DE DE_GPIO gt DR DE_GPIOx_Pin lt lt 2 define LED_ON Led Led 0xFF define LED_OFF Led Led 0x00 define LED TOGGLE Led Led Led define DE_ON De De 0xFF define DE_OFF De De 0x00 define IS BTN PRESSED Btn Btn 0 define WAIT BTN Btn while IS BTN PRESSED Btn while IS BTN PRESSED Btn define USB Dp PullUp OFF USB Dp PullUp OxFF define USB Dp PullUp ON USB_Dp_PullUp 0x00 define motor direction define FORWARD 1 define REVERSE 0 define servo position define SERVO_TOP 0 define SERVO MID 1 define SERVO BOTTOM 2 define STOP SERVO 1 the servo pins on micro define SERVO GPIO GPIO1 define SERVO_Pin GPIO Pin 1 motor pins define NUM_MOTORS 3 define MOTOR_GPIO GPIO1 define MOTOR1_Pin GPIO_Pin_6 define MOTOR2_Pin GPIO_Pin_4 define MOTOR3_Pin GPIO_Pin_3 define MOTOR_DIRECTION_GPIO X for the first input which is INA define MOTOR1_DIRECTION_A GPIO_Pin_4 define MOTOR1 DIRECTION B GPIO Pin 5 define MOTOR2 DIRECTION A GPIO_Pin_2 define MOTOR2_DIRECTION_B GPIO_Pin_3 define MOTOR3_DIRECTION_A GPIO_Pin_6 120 Appendix B Documentation of the programming define MOTOR3_DIRECTION_B GPIO_Pin_7 bumper pins define BUMPER1 GPIO5 define BUMPER2 GPIOO define BUMPER3 GPI04 define BUMPER_FRONT_LEFT GPIO_Pin_2 5 2 define BUMPER_FRONT_RIGHT GPIO_Pin_0 4 0 define BUM
66. RM966E S core can perform single cycle DSP instructions good for speech recognition audio and embedded vision algorithms ST VS6724 Camera Module The VS6724 is a UXGA resolution CMOS imaging device designed for low power systems particularly mobile phone and PDA applications Man ufactured using ST 0 184 CMOS Imaging process it integrates a high sensitivity pixel array digital image processor and camera control functions The device contains an embedded video processor and delivers fully color processed images at up to 30 fps UXGA JPEG or up to 30 fps SVGA YCbCr 4 2 2 The video data is output over an 8 bit parallel bus in JPEG 4 2 2 or 4 2 0 RGB YCbCr or Bayer formats and the device is controlled via an I2C interface The VS6724 camera module uses ST s second generation SmOP2 packaging technology the sensor lens and passives are assembled tested and focused in a fully automated process allowing high volume and low cost production The VS6724 also includes a wide range of image enhancement functions designed to ensure high image quality these include automatic exposure control automatic white balance lens shading compensation de fect correction algorithms interpolation Bayer to RGB conversion color space conversion sharpening gamma correction flicker cancellation NoRA noise reduction algorithm intelligent image scaling special effects MC33887 Motor Driver Carrier All electric motors need some sort of controller
67. The motor controller may different features and complexity depending on the task that the mo tors will have to perform 3 7 Hardware Architecture 35 The simplest case is a switch to connect a motor to a power source such as in small appliances or power tools The switch may be manually operated or may be a relay or conductor connected to some form of sensor to auto matically start and stop the motor The switch may have several positions to select different connections of the motor This may allow reduced voltage starting of the motor reversing control or selection of multiple speeds Over load and over current protection may be omitted in very small motor con trollers which rely on the supplying circuit to have over current protection 51 53 Vin 52 54 Figure 3 15 Structure of an H bridge highlighted in red The DC motors cannot be controlled directly from the output pins of the microcontroller We need the circuit so called motor controller motor driver or an H Bridge The term H Bridge is derived from the typical graphical representation of such a circuit An H Bridge Figure 3 15 is built with four switches solid state or mechanical When the switches S1 and 4 are closed and S2 and 83 are open a positive voltage will be ap plied across the motor By opening S1 and S4 switches and closing S2 and S3 switches this voltage is reversed allowing reverse operation of the motor To drive motors we used a
68. a u32 BufferSize double BlobSample 10 2 u8 index 0 extern Tx24_ImgJob Job 2 extern CurrentJob extern u8 control hit Private function prototypes see header Private functions void Blob_Search void if Job CurrentJob State MY_ON x24_StateAcquisitionEnd Image is acquired CompletedJob amp Job CurrentJob BlobSearch u16 CompletedJob gt pFirstBlock gt pData MY OFF CurrentJob 1 void CalculatePositionTop TPoint point s16 result 3 1 Homography for servo top 108 Appendix B Documentation of the programming double K 3 3 191 7146 0 80 3591 0 191 2730 61 2765 0 0 1 T_cr 3 4 1 0 0755 0 9948 0 0683 8 6194 0 1867 0 0531 0 9810 271 8015 0 9795 0 0868 0 1817 72 3125 1 C 3 4 0 0 0 0 0 0 0 0 H 3 3 0 0 07 0 0 07 0 0 0 InvH 3 3 0 0 0 0 0 t_ 4 3 OF 0 0 O Matrix_Mult3334 K T_cr C Matrix_Mult3443 C t H Inverse H InvH double position 3 0 0 O X double image point 3 point X point Y 1 vectorByMatrix3x1 InvH image point position result 0 s16 position 0 position 21 result 1 s16 position 1 position 21 result 2 s16 position 2 position 21 void CalculatePositionGround TPoint point s16 result 3 1 Homography for servo_top double K 3 3
69. a mobile robot by including the parame ters characterizing the non systematic error with the state to be estimated While the majority of prior research has focused on formulating the pose es timation problem in the Cartesian space Aidala and Hammel 29 among others have also explored the use of modified polar coordinates to solve the relative bearing only tracking problem Funiak et al 40 propose an over parameterized version of the polar parameterization for the problem of target tracking with unknown camera locations Djugash et al fur ther extend this parameterization to deal with range only measurements and multimodal distributions and further extend this parameterization to improve the accuracy of estimating the uncertainty in the motion rather than the measurement 14 Chapter 2 State of the Art 2 3 Navigation A navigation environment is in general dynamic Navigation of autonomous mobile robots in an unknown and unpredictable environment is a challenging task compared to the path planning in a regular and static terrain because it exhibits a number of distinctive features Environments can be classified as known environments when the motion can be planned beforehand or partially known environments when there are uncertainties that call for a certain type of on line planning for the trajectories When the robot nav igates from original configuration to goal configuration through unknown environment without any prior descri
70. a motion model is to capture the relation ship between a control input to the robot and a change in the robot s pose Good models will capture not only systematic errors such as a tendency of the robot to drift left or right when directed to move forward but will also capture the stochastic nature of the motion The same control inputs will almost never produce the same results and the effects of robot actions are therefore best described as distributions 41 Borenstein et al cover a variety of drive models including differential drive the Ackerman drive and synchro drive Previous work in robot motion models have included work in automatic acquisition of motion models for mobile robots Borenstein and Feng describe a method for calibrating odometry to account for systematic er rors Roy and Thrun propose a method which is more amenable to the problems of localization and SLAM They treat the systematic errors in turning and movement as independent and compute these errors for each time step by comparing the odometric readings with the pose estimate given by a localization method Alternately instead of merely learning two simple parameters for the motion model Eliazat and Parr 15 seek to use a more general model which incorporates interdependence between motion terms including the influence of turns on lateral movement and vice versa Martinelli et al propose a method to estimate both systematic and non systematic odometry error of
71. al The design and perception of humanoid robot heads In in Proceedings of the DIS Conference pages 321 326 ACM Press 2002 Joseph Djugash Sanjiv Singh and Benjamin Grocholsky Modeling mobile robot motion with polar representations In Proceedings of the 2009 IEEE RSJ international conference on Intelligent robots and sys tems IROS 09 pages 2096 2101 Piscataway NJ USA 2009 IEEE Press BIBLIOGRAPHY 79 25 26 27 gt 30 31 32 33 34 35 36 37 38 Han et al A new landmark based visual servoing with stereo camera for door opening ICCAS2002 2002 Frizera Vassallo et al et al Visual navigation combining visual servo ing and appearance based methods SIRS 98 Int Syrup on Intelligent Robotic Systems 1998 Shirai et al et al Wiimedia motion analysis methods and applications using a consumer video game controller Proceedings of the 2007 ACM SIGGRAPH Symposium on Video games 2007 G D Hager A tutorial on visual servo control IEEE Transaction on Robotics and Automation vol 12 no 5 pp 651 670 1996 V Aidala and S Hammel Utilization of modified polar coordinates for bearings only tracking IEEE Transactions on Automatic Control vol 28 no 3 pp 283 294 1983 R Holmberg Design and Development of Powered Castor Holonomic Mobile Robots PhD thesis Stanford University 2000 J Borenstein Umbmark a method for measuring comparing and
72. and games Kismet Figure 2 8 a is a robot made in the late 1990s at Massachusetts Institute of Technology with auditory visual and expressive systems in tended to participate in human social interaction and to demonstrate sim ulated human emotion and appearance This project focuses not on robot robot interactions but rather on the construction of robots that engage in meaningful social exchanges with humans By doing so it is possible to have a socially sophisticated human assist the robot in acquiring more sophisticated communication skills and helping it to learn the meaning of these social exchanges A Furby Figure 2 8 b was a popular electronic robotic toy resembling a hamster owl like creature which went through in 1998 Furbies were the first successful attempt to produce and sell a domestically aimed robot A newly purchased Furby starts out speaking entirely Furbish the unique lan guage that all Furbies use but are programmed to speak less Furbish as they gradually start using English English is learned automatically and no matter what culture they are nurtured in they learn English In 2005 new Furbies were released with voice recognition and more complex facial movements and many other changes and improvements AIBO Artificial Intelligence Robot Figure was one of several types of robotic pets designed and manufactured by Sony There have been several different models since their introduction on May 11 1999 AIBO
73. as been ported to the microcontroller The rest of the implementation not mentioned above focuses on pin assignments in Appendix D 3 definitions of the constants creation of the header files 72 Chapter 6 Game Chapter 7 Conclusions and Future Work After the tests performed proper modifications have been implemented to solve the detected problems discussed previously At last we have been able to obtain a working robot satisfying most of the constraints defined at the beginning The robot dimensions are more or less same with the designed dimensions Even though we don t have a proper method to measure the robot s maximum speed the maximum speed achieved by the robot is fast enough as expected The obstacle avoiding is implemented with bumpers and supported with foams and springs The camera head is placed on a servo can be moved up and down The fully charged batteries provide enough power to move the robot for at least 2 hours It is difficult to es timate the total price of the robot The camera board that is used is not available in the market and the total cost of the components in camera board might be misleading to determine the cost of the camera board Moreover for the other components the prices show differences as the order quantity changes In conclusion the final price might match the cost constraint 250 euro The robot is fully working autonomously playing the designed game without any interrup
74. ater than 3 or blob found FALSE we perform a turn around the center and set the blob_found FALSE That mechanism en abled us to detect a target that was not detected even if it is on side Even if the target is on side and it is not detected for 3 times we perform the turn To process blob in myBlob parameter the first step is controlling the camera position Depending to the camera position the position of the target in real world is calculated The motor contributions are calculated using Iriskar function According to the calculated position of the target the following cases is executed If the camera is set to SERVO TOP and the distance of the robot to the target is e Greater than 1200 mm The motor speeds from Triskar are set with multiplier FAST e Between 1200 mm and 700 mm The motor speeds from Triskar are set with multiplier NORMAL e Less than 700 mm The camera head position is changed to SERVO MID 71 If the camera is set to SERVO_MID the position of the target is cal culated with CalculatePositionGround function The motor contributions are calculated using Triskar function and the motor speeds from Triskar are set with multiplier CLOSE If the distance of the robot to target is less than 300 mm the target is marked as acquired by setting color done TRUE The detection of target acquired and game over is done by controlling color_done on every loop When the color_
75. ates which is a point in X Y plane By polar coordinates the position is described by an angle and a distance to the origin Instead of repre senting robot position with a single coordinate the hybrid approach is used as follows The robot pose is defined as X7 Y7 a where Xy and Y are linear positions of the robot in the world Let a denote the angle between the robot axis and the vector that connects the center of the robot and the target object The transformation of the coordinates into polar coordinates with its origin at goal position p y Az Ay and a 6 atan2 Ax Ay 4 1 Then the calculated angle a is passed as the parameter to the simulator in order to test the motor contributions calculated for the motion Later the behavior tested with simulator is implemented for microcontroller with Triskar function in Appendix A 1 and the angle a calculated after ac quiring the target is passed to Triskar function to calculate the motor contributions on board to reach the target The inverse kinematics model is simple It was considered that the rep resentative coordinates of the robot were located in its center Each wheel is placed in such orientation that its axis of rotation points towards the center of the robot and there is an angle of 120 between the wheels The velocity vector generated by each wheel is represented on Figure 1 1 by an arrow and their direction relative to the Y coordinate or robot front directio
76. be actively powered to spin along any direc tion There have not been many attempts to build a mobile robot with ball wheels because of the difficulties in confining and powering a sphere One mechanism for implementing this spherical design imitates the first com puter mouse providing actively powered rollers that rest against the top surface of the sphere and impart rotational force Figure 2 3 Spherical Wheels The wheel type and wheel configuration are of tremendous importance 8 Chapter 2 State of the Art they form an inseparable relation and they influence three fundamental char acteristics of a maneuverability controllability and stability In general there is an inverse correlation between controllability and maneuverability The number of wheels is the first decision Two three and four wheels are the most commonly used each one with different advantages and disad vantages The two wheels drive has very simple control but reduced ma neuverability The three wheels drive has simple control and steering but limited traction The four wheels drive has more complex mechanics and control but higher traction 98 The differential drive is a two wheeled drive system with independent actuators for each wheel The motion vector of the robot is the sum of the independent wheel motions The drive wheels are usually placed on each side of the robot A non driven wheel often a castor wheel forms a tripod like support struc
77. c which is produced by merging two transwheels where the rollers are covering all the wheel which will enable the movement in any direction eas ily and more consisting model by reducing the possible power transmission loss that can be occur when merging the two wheels by hand a Omniwheel b Transwheel c Double Transwheel Figure 3 6 Wheels In order reach the maximum speed of 1 m sec we should have the following equation speed circumref erence rps speed diameter x pi x rps As it can be seen from the equation the speed is also related with the rota tion per second rps of the wheels which is determined by the motor So the dimension choice of the wheels are made keeping the rps in mind The rpm necessary to turn our wheels with the maximum speed of 1 meter sec ond the is calculated as follows 1000mm second diameter pi rps 1000mm second 49 2mm pi rps rps S 6 4 28 Chapter 3 Mechanical Construction rpm 388 As a result of the calculation using the omniwheel with outer diameter of 49 2 m we will need a motor that can run around 388 rpm to reach the maximum speed Figure 3 7 The wheel holder bar that is making the transmission between motors and wheels The transmission between motors and wheels is achieved by the bar which is lathed from an aluminum bar Figure 3 7 The bar is placed inside the wheel and locked with a key using the key ways in the wheel and the bar For
78. camera in order to calculate the camera matrix Figure 5 1 By taking the pictures of the checker board makes it possible to use the information of its dimensions in order to calibrate the camera by that the unknown parameters of the camera can be calculated such as focal length principal point skew radial distortion etc Figure 5 1 The images used in the camera calibration In order to calculate the homography matrix which makes the transfor mation of the points from camera coordinate to robot coordinate we wrote a script that makes the calculation for us Having the camera calibration matrix from the toolbox the homography matrix H is calculated by the script The whole camera calibration and the projection between the world points and camera as explained in details as follows The matrix K is called the camera calibration matrix E x pa K f y Dy where f and f represent the focal length of the camera in terms of pixel dimensions in the x and y direction respectively Similarly p and py are 5 1 Camera Calibration 53 the principal points in terms of pixel dimension ge Knox a where Xcam as X Y Z 1 to emphasize that the camera is assumed to be located at the origin of a Euclidean coordinate system with the principal axis of the camera point straight down the z axis and the point Xcam is expressed in the camera coordinate system In general points in the space will be expressed in terms of a differ
79. ction prototypes 114 Appendix B Documentation of the programming void Led_SM void void Print_Int int a u8 str void vectorByMatrix3x1 double A 3 double x 3 double b 3 void vectorByMatrix4x1 double A 4 double x 4 double b 4 void Matrix Mult double a1 4 double 82 4 double a3 4 void Matrix_Mult3334 double 81 3 double a2 4 double a3 1 41 void Matrix_Mult3443 double 81 4 double 2 3 double a3 3 void Inverse double A 3 double X 1 31 void vectorByMatrix3x1_s16 s16 1 3 s16 x 3 s16 b 3 Private functions void Led_SM void static int Count 0 Count if Count lt 50000 1 LED_ONC LEDR else LED_OFF LEDR if Count gt 60000 Count 0 void Print_Int int a u8 str u8 myStr 10 USB WriteString str USB_WriteString Num2String a myStr USB_WriteString myStr USB_WriteString r n void vectorByMatrix3x1 double A 3 double x 3 double b 3 1 b 0 A O O x 0 A OJ 1 x 1 ALO 2 x 2 b 1 A 1 0 x 0 A 11 1 x 1 A 11 2 x 2 b 2 A 2 0 x 0 A 21 1 x 1 A 21 2 x 2 void vectorByMatrix3x1_s16 s16 A 3 s16 x 3 s16 b 3 b 0 A O O x 0 100 A O 1 x 1 100 A O 2 x 2 100 B 1 Microprocessor Code 115
80. ded to cover problems due to the lack of encoders to control the position of the motors 43 PWM Control Motors are controlled by applying a PWM signal to the H Bridges In order to control velocity we did not use any encoders so we don t have feedback about the motor rotation or the displacement achieved with the result of the applied PWM To have an odometry information without using encoders we decided to review and replicate the work done at AIRLab about using the mice odometry to get feedback from the motors On the previously work a set of mice were reconfigured to obtain the position information and calculations were made from these readings to find the odometry The working configuration was able get a feedback when the distance between the mouse sensors and lens was about 1mm and the distance between lens and ground was also 1 mm but this distance is too small for our robot In order to increase the distance to 3 cm which is the distance between the ground and the chassis of the robot we experimented changing the lens of the mice The mice work like a camera and the image is captured by the sensor inside it through a lens placed in front of it We tested lenses with different focal lengths and different orientations in order to achieve the 3 cm con figuration During the experiments we obtained different performance on different grounds and different lighting conditions We tried several illumi nation alternatives such as las
81. done is set as TRUE the camera head position is set to SERVO TOP color_done and blob_found are flagged as FALSE and color is incremented by 1 Since the current configuration is working with 2 colors the case where color gt 2 is con trolled for the game over case For that case we introduced a variable as reply counter which sets the number of replies rounds to end the game In the code this variable is set as 5 that made the game to find and go to the targets 5 times in the same color order defined The work flow in the main loop is shown in Appendix The class diagram that is showing the relation between the modules is reported in Appendix A We increased the modularity of the pro gram by separating the code to improve reuse breaking a large system into modules makes the system easier to understand By understanding the be haviors contained within a module and the dependencies that exist between modules it s easier to identify and assess the ramification of change We used the naming convention _Init for initialization classes functions The algorithm are implemented inside the Motion and Vision classes The mo tion algorithm is implemented by porting the software simulator previously written to the microcontroller The vision algorithms are also implemented in a similar manner The camera calibration calculating the position of the object was previously implemented in a Matlab script Later the script h
82. e Output None Return None ak o ok AE AE AE AE AAA IAN AIAR ooo AE AE AE AE oko AE AE AE AE ook AE AE AE A ook RIOR AE AE AE AE AE AE AEA RIA oko ooo oe ok Y void SCU_Configuration void Initialize PLL SCU_MCLKSourceConfig SCU_MCLK_OSC SCU_FMICLKDivisorConfig SCU_FMICLK_Div1 FMI_Config FMI_READ_WAIT_STATE_2 FMI_WRITE_WAIT_STATE_O FMI PWD ENABLE FMI LVD ENABLE FMI FREQ HIGH 92 Appendix B Documentation of the programming SCU_RCLKDivisorConfig SCU_RCLK_Divi SCU_HCLKDivisorConfig SCU_HCLK_Divi SCU_PCLKDivisorConfig SCU_PCLK_Div2 SCU BRCLKDivisorConfig SCU BRCLK Divi SCU PLLFactorsConfig 192 25 2 PLL 96 MHz SCU PLLCmd ENABLE PLL Enabled SCU MCLKSourceConfig SCU MCLK PLL MCLK PLL SCU_PFQBCCmd ENABLE Enable GPIO 0 1 2 3 4 5 6 7 8 9 Clocks SCU_APBPeriphClockConfig __GPI00 ENABLE GPIO DeInit GPIOO SCU APBPeriphClockConfig GPIO1 ENABLE GPIO DeInit GPIO1 SCU_APBPeriphClockConfig __GPI02 ENABLE GPIO DeInit GPIO2 SCU APBPeriphClockConfig GPIO03 ENABLE GPIO DeInit GPIO3 SCU APBPeriphClockConfig GPIO04 ENABLE GPIO DeInit GPIO4 SCU APBPeriphClockConfig GPIO5 ENABLE GPIO DeInit GPIO5 SCU_APBPeriphClockConfig __GPI06 ENABLE GPIO DeInit GPIO6 SCU APBPeriphClockConfig GPIO7 ENABLE GPIO DeInit GPIO7 SCU_APBPeriphClockConfig __GPI08 ENABLE GPIO DeInit GPIO8 SCU_APBPeriph
83. e 37 With the appropriate selection of resistor R and Ra based on the above information V will be suitable for the analog port on microcontroller The divider is used to input to the microcontroller a signal proportional to the voltage provided by the battery so to check its charge Note that a fully charged battery can often be up to 20 more of its rated value and a fully discharged battery 20 below its rated value For example a 7 2 V battery fully charged can be 8 4 V and fully discharged 5 V The voltage divider circuit allows to read the battery level from the microcontroller pins directly that will be used in order to monitor battery charging level changes Figure 3 16 The schematics of voltage divider and voltage regulator circuit From the Figure 3 16 the pin Vout is used in order to monitor the bat tery life which is placed on the left side of schematics The divided voltage is read from the analog input of the microcontroller The voltage regulator circuit which is placed on the right part of the schematics is used to power the STLCam board with 5 V through the USB port It can be also used in order to make the connection with the PC to transmit some data which we use for debug purposes 38 Chapter 3 Mechanical Construction Chapter 4 Control The motion mechanism is inspired from the ones that have been introduced In this Chapter we will explain the wheel configuration model move ment types
84. e and seek In 2004 RoBallet advanced these constructionist activities further blending elements of projected vir 16 Chapter 2 State of the Art tual environments with sensor systems that reacted to children dancing in a mediated physical environment The realm of toys and robotic pets has also seen the development of a wide array of interactive technologies e g Furby Aibo Tamagotchi and more recently Microsoft s Barney 9 which has been integrated with T V based video content Interactive robotic en vironments for education are now being extended to on line environments such as CMU s educational Mars rover 8 and becoming popular through robotics challenges such as FIRST Robotics Competition 3 BattleBots 1 and Robot World Cup soccer tournaments such as Robocup 42 The games related with robots so called robogames are categorized into four branches according to AIRLab report 19 One is the videogames where robot characters are simulated Soccer Simulation League in RoboCup Soccer is an example of this kind of games The Simulation League focuses on artificial intelligence and team strategy Independently moving software players agents play soccer on a virtual field inside a computer This pro vides a context to the game but also allows to escape all the limitations of physical robots Another one is the tele operated physical robots where the player is mainly in the manipulation of remote controllers similar to
85. e composition of Vr and Vr The motor contributions and their angular velocities are also shown in Figure 4 2 44 Chapter 4 Control Rotation The second case is the rotation For that particular configuration we only considered the movement caused by rotation without any displacement In the software simulator the frontal speed and lateral speed are set to zero and the angular speed is calculated from the robot s pose a using 4 1 An example outcome of the simulator is shown in Figure 4 3 The robot is mov ing around its center with an angular speed of 0 099669 rad s without any displacement in X Y plane 150 100 50 100 150 200 250 0 mm s VE 0 mm s Og 0 099669 rad s di 24 9172m 9 0 mm s tl 2 24 9172 mm s 0 mm s 3 24 9172 mm s 0 mm s O 1 0129 rad s 0 1 0129 rad s OL 1 0129 rad s t t lt lt lt lt lt lt 5 5 200 100 0 100 200 300 Figure 4 3 The angular movement calculated by simulator All the theory for the calculation of the angular movement is same as in the linear movement case only values for some parameters change The angular velocity of the robot body wg is calculated with a in the hybrid approach we mentioned in the beginning of the section 4 2 Matlab Script 45 Mixed Angular and Linear Movement This is the case where the movement of the robot is composed of both the a
86. ed at the same time with the motors then the position is controlled calling the Position _Camera function that can take SERVO_TOP SERVO_MID SERVO_BOTTOM as the different posi tions for the camera servo Indeed this camera positions are used during the vision inspection to reach the target and calculate the motor contribution according the distance of the target object 3 unique camera head position are implemented using the visual feedback from the ST software to find the best configuration in terms of depth of field and field of view If the target is more far than 750 mm the camera position is set to SERVO_TOP which can see the environment from 750 mm to robot up to 4 5 meters If the target is closer than 750 mm to robot the camera head position is lowered to SERVO_MID in order to provide a vision from robot boundary to 750 mm The last position is implemented for testing reasons and also provides a ready configuration to set the camera head in any position Chapter 5 Vision The vision system is focused on the implementation of the camera for object detection and calculating the position of this object according to the robot To do so the relation between different coordinate systems which are image plane coordinate system robot coordinate system and real world coordinate system should be defined In order to go to the target it is necessary to know where the object is placed Using the blob search algorithm we ac quire the p
87. ent Euclidean coordinate frame known as the world coordinate frame The two frames are related via a rotation and a translation The relation between the two frames can be represented as follows x KRII C X where X is now in a world coordinate frame This is the general map ping given by a pinhole camera In Figure 5 2 A general pinhole camera P KRJI Cl has a 9 degrees of freedom 3 for K 3 for R 3 for C The parameters contained in K are called internal camera parameters The parameters of R and C which relate the camera orientation and position to a world coordinate system are called the external parameters In a compact form the camera matrix is P E RI where t RC The camera matrix which consists of internal and external camera pa rameters is calculated by the camera calibration toolbox automatically Having the internal and external camera parameters from the toolbox we compute the homography H Homography is an invertible transformation from the real projective plane to the projective plane We developed two different approaches to determine the position of a target according to the robot coordinates In the first case the target ob ject is a ball and we are using the information coming from the diameter of the ball In this case the transformation should be a 3D to 2D since the ball can be represented in world coordinates system with the X Y Z and in the camera coordinates with two values X Y In the sec
88. er led and table lamp We were able to find the correct illumination using the led with a wave length readable from the sensor And for the lens we used a configuration with a focal length of 10 mm leaving the distance between the lens and the sensor at 15 mm and lens to ground at 30 mm Instead of using encoders that each cost around 50 euro the possible configuration costs around 5 euro but the work is not finished yet and we decided to include this on the robot later Since there is no information on the motor speed and displacement of the robot the control process mainly depends on vision and it is not very precise There are situations where we cannot determine whether the motor is moving When the robot is stuck in a place and the sensors do not detect 4 3 PWM Control 47 a collision such as entrance of the wheels to a gap in the floor since we don t have a feedback on the motors we cannot determine whether the motors are turning The motors are controlled by PWM generated from the microcontroller To drive motors we used a PWM signal and varied the duty cycle to act as a throttle 100 duty cycle full speed 0 duty cycle coast 50 duty cycle half speed etc After some testing we realized that the percentage of the duty cycle and the speed are not changing with the same ratio e g 50 of the duty cycle does not correspond the half speed Also there is a minimum percentage of the duty cycle which the wheels start tu
89. evel of my thesis and also thank to Martino Davide Luigi Simone and other member of AIRLAB for their support and help to resolve my problems Thanks to all of my friends who supported and helped my during the university studies and for thesis Arif Burak Ugur Semra Guzide and Harun and to my house mate Giulia and the others I forgot to mention Last but not least thanks to my beloved family for their endless love and continuous support Anneme babama ve ablama Contents ommario I III HI m e e 8 eo H lt 1 Introductio 1 Ll Goal RR Rss 1 1 2 Context Motivations eA 1 1 3 _ Achievements s 2 14 Thesis Structure 0 000008 vee 2 2 State of the Art 5 lt 5 22 Motion Models o aa a a 13 23 Navigation 14 2 4 Games and Interaction 15 3 Mechanical Constructio 21 3 1 Chassis i es so ze tomm oo RR modo ROS RR E Sox UR 21 4 23 LIL 26 2 28 2444 20 3 0 Batterieg sa a mie ee EE we RE 31 3 4 Hardware Architecture 33 39 4 1 Wheel configuration 39 4 2 Matlab Script 42 43 PWM Contro 46 5 1 Camera Calibratio 5 2 Color Definitio gt Q D 7 Conclusions and Future Wo
90. eveloped visual control method for the end effector of the robot with two un calibrated cam eras estimating the distances based on cross ratio invariance 2 4 Games and Interaction Advances in the technological medium of video games have recently included the deployment of physical activity based controller technologies such as the Wii 27 and vision based controller systems such as Intel s Me2Cam 13 The rapid deployment of millions of iRobot Roomba home robots and the great popularity of robotic play systems such as LEGO Mindstorms and NXT 5 now present an opportunity to extend the realm of video game even further into physical environments through the direct integration of human robot interaction techniques and architectures with video game ex periences Over the past thirty to forty years a synergistic evolution of robotic and video game like programming environments such as Turtle Logo 36 has occurred At the MIT Media Lab these platforms have been advanced through the constructionist pedagogies research and collaborations of Sey mour Papert Marvin Minsky Mitch Resnick and their colleagues leading to Logo 7 Star Logo 37 programmable Crickets and Scratch 6 and Lego MindStorms 37 In 2000 Kids Room demonstrated that an immersive educational gaming environment with projected objects and characters in physical spaces e g on the floor or walls could involve children in highly interactive games such as hid
91. f a dock station and implementation of the autonomous charging behavior The circuit to measure the battery level is already op erational The software to check the battery level and then go to the dock station must be implemented The second addition is the mice boards that we are going to use as odometers Experiments have been made to modify the previous work done at AIRLab using the mice as an odometry device We have not been able to design the working prototype and the control circuit that calculates the position information from the mice yet Previously the position informa tion was calculated with a PIC and a PC The new design will work on the ARM microprocessor and the interface between the camera board and mice should be designed and implemented The third addition is the self adaptation of the robot to battery charge level Using the output of battery monitoring circuit the battery level can be measured These measurements are going to used in the implementation to detect the battery level and change the speed multipliers to keep the robot in the same speed even if the battery level is going low Also the robot is going to adapt itself to make the decision of going to the docking station for reaching when a low battery status detected Implementation of the future developments will result to have a better control in motion mechanism and introduce more autonomous robot behav 75 ior which will be useful in the research
92. fficult to determine the color with the ST color picker software Instead we decided to find the color histogram that is the distribution of the selected color properly in the RGB channels separately To do so we took several samples from the camera in different distances in different illuminations to find a proper average The calculated results are later tested with the ST color picker software since it offers a good interface to see whether the color is selected correct or not in a visual way The third step is related to the estimation of the objects distance to the robot As mentioned before the diameter of the ball is a problematic issue due to the low cost optics hence we decided to use an object with cylindric shape and the second configuration for the transformation This distance is easier to measure independently from the shape of the object and will always change as the object s position changes in the world During the implementation to the microcontroller to increase the perfor mance we tried to avoid matrix and floating point operations The matrices that can be defined statically are defined statically To avoid floating point operations the floating points are converted into integer then multiplied by a constant most of the times 1000 The operations are made in integer and the result is returned as floating point by dividing with the constant We ensure that the microcontroller will not spend too much time in matrix
93. g of the motor since aluminum has great energy absorbing characteristics 12 The connection can be seen clearly in Figure 3 5 The component is attached to the plexiglas from the L shaped part using a single screw This gives us the flexibility to dis attach the component easily and to change the orientations of them if needed Figure 3 5 The connection between the motors chassis and wheels 3 3 Wheels The target environment of our robot is a standard home The character istic properties of this environment that are important for our work are as follows Mainly the environment is formed by planes surfaces such as par quet tile carpet etc In order to move freely to any direction on these surfaces and reach the predefined speed constraint we selected the wheels and a proper configuration for them The decision to choose omnidirectional wheel was motivated but there are lots of different omnidirectional wheels available on the market Among them we made a selection considering the 3 3 Wheels 27 target surface maximum payload weights of the wheels and price The first selected wheel was the omniwheel shown in Figure The wheel consists of three small rollers which may affect the turning since the cover age is not good enough Also the wheel itself is heavier than the one with the transwheel shown in Figure A single transwheel is 0 5 oz lighter than an omniwheel Another model is the double transwheel seen in Figure 3 6
94. ght was too much to be handled by the three screws And additionally we placed 6 more screws with the same height as before These screws allowed us to divide the total weight on the plate equally on all the screws and also enabled us to install springs and foams to implement bumpers that protect the robot from any damage that could be caused by hits 3 2 Motors The actuator is one of the key components in the robot Among the possible actuation we decided to go with DC motors Servo motors are not powerful enough to reach the maximum speed Due to noise and control circuitry requirements servos are less efficient than uncontroled DC motors The control circuitry typically drains 5 8mA on idle Secondly noise can draw more than triple current during a holding position not moving and almost double current during rotation Noise is often a major source of servo inef ficiency and therefore they should be avoided Brushless motors are more power efficient have a significantly reduced electrical noise and last much longer However they also have several disadvantages such as higher price and the requirement for a special brushless motor driver Since they are running at high speed we need to gear them down This would also add 24 Chapter 3 Mechanical Construction some extra cost Also the Electronic Speed Controllers ESC are costly and most of them do not support multiple run motors The minimum price for the motor is about 20 dollar
95. h wheel The calculation of the independent motor contributions is made using the following formulas Vu Vi Mp V Viz 42 Chapter 4 Control cosA sinA 1 VE where Mp cosA sinA 1 V Vr 0 1 1 omega gt Rrobot A is the angle of the front wheels which is 30 for our robot Indeed the values 0 866 0 5 0 866 0 5 0 1 in the Figure 4 1 coming from the projection of cosine and sine of the motor vectors in the X Y plane which are later used statically in the microcontroller for the calculation of the mo tor contributions Vp represents the frontal speed Vz lateral speed and omega Robot represents the angular velocity of the body The vector of tangential wheel speed is represented as Vi1 Vt2 Vt3 and vector of sliding velocity of the wheels due to rollers is represented as Vai Vn2 Vn3 The vectors can have a positive or negative direction which represents the direction in which the motor has to move forward or back wards respectively The desired speeds set as frontal speed Vr and lateral speed Vr are projected through the motor axes in order to find motor con tributions The angular velocity of the wheels is found by dividing the V of the desired wheel to the radius of the wheel It can be formulated as follows WI Va wal Vt LR wheel W3 Via 4 2 Matlab Script In order to test the behavior we implemented a software simulator in Matlab that is calculating the motor contributions
96. he image based visual control method does not need to reconstruct in 3D space but the image Jacobian matrix needs to be estimated The controller design is difficult And the singular problem in image Jacobian matrix limits its application 28 Hybrid control method attempts to give a good solution through the combination of position and image based visual control methods It controls the pose with position based method and the position with image based method For example Malis et al provided 2 4 Games and Interaction 15 a 2 5 D visual control method Deguchi et al 22 proposed a decoupling method of translation and rotation Camera calibration is a tedious task and pre calibration cameras used in visual control methods limit a lot the flexibility of the system Therefore many researchers pursue the visual control methods with self calibrated or un calibrated cameras Kragic et al gave an example to self calibrate a camera with the image and the CAD model of the object in their visual control system Many researchers proposed various visual control methods with un calibrated cameras which belong to image based visual control methods The camera parameters are not estimated individually but combined into the image Jacobian matrix For instance Shen et al 43 limited the working space of the end effector on a plane vertical to the optical axis of the camera to eliminate the camera parameters in the image Jacobian matrix Xu et al d
97. he low level modules are the assembly code that is coming with the ST tool chain which already defines the registers memory mappings all the communica tions with the chips We did not concentrate on the assembly code all the written software implements the high level part The development mainly focused on the generation of PWM using timers initialization of components algorithms and utilities Like all the microcon troller programs the core runs in the main loop Before entering the main loop we initialize all the components first Initialization is made step by step as follows e Configure the system clocks e Configure the GPIO ports e Configure x24 Camera e Configure the USB e Initialize camera parameters Initialize motors Set initial camera position 69 e Initialize sensors e Initialize the blob search and color parameters In order to initialize the camera the RGB_Cube which contains the color information and BlobSearch are initialized We also set the automatic white balance to off and exposure control to direct manual mode The appropriate parameters for the image format frame rate sensor mode and clock are set for the blob search at this point The second step is initialization of the motors The corresponding pins in the microcontroller for the motors are set to control the direction of motors and PWM generation The PWM generation is made by using the timers were available in the microcontroller Output c
98. igure in order to define selected object s color using the script Figure 5 4 Samples are taken from different lighting conditions and different distance to the target object After acquiring the samples the next step is creation of the histogram to find the distribution of the RGB values in the target object s color The histogram created is shown in Figure 5 5 An image histogram is a type of histogram that acts as a graphical repre sentation of the color distribution in a digital image It plots the number of pixels for each color value By looking at the histogram for a specific image 5 2 Color Definition 63 9000 T T T T 6000 T T T T T 7000 8000 6000 5000 7000 5000 F 6000 40004 4000 5000 3000 4000 3000 H 3000 2000 2000 2000 1000 1000 1000 Figure 5 5 The histogram formed from the samples taken for each color channel a viewer will be able to judge the entire color distribution at a glance For this we wrote a script in Matlab We took as many samples as we could from the environment where the object is placed in different lighting conditions From the sample images captured the part with the object is cropped area with the target color in order to find the color distribution of the pixels forming the object mask G a Mask R b Mask G c Mask B Figure 5 6 The mask for each channel by setting the upper and lower bounds
99. ill always be on the ground This reduces the mapping to a 2 D to 2 D mapping problem since Z the height will be always zero Instead of calculating the diameter of the ball we determine the distance be tween the end of the image and the intersection point of the object with the ground This distance is easier to measure independently from the shape of the object and will always change as the object s position changes in the world The mentioned mapping also makes easier the blob search since it is not very important to detect the object clearly as a cylindric If we can recognize some part of the object we can ensure that the object is more or less close to the calculated point This also gave us the flexibility for the color information since we do not have the obligation to detect all the object we can define a more generic color code that is working for most of the cases We used the second configuration in both camera positions Even the internal camera parameters are always same in our camera the external parameters are effected with a translation or rotation Indeed when the camera head position is changed it results with the usage of a H matrix for each transformation 58 Chapter 5 Vision The whole algorithm and script is found in Object s Position section in Appendix B 2 5 2 Color Definition The camera we are currently using has a UXGA resolution CMOS and it is set to 160x120 resolution for the blob search algorithm
100. ional conference on Machine learning 2004 N Tomatis A Tapus and R Siegwart A Martinelli Simultaneous localization and odometry calibration for mobile robot IEEE RSJ International Conference on Intelligent Robots and Systems 2003 M Asada R D Andrea A Birk H Kitano and Manuela Veloso Robotics in edutainment In Proceedings of the IEEE International Conference on Robotics and Automation 2000 ICRA 00 volume 1 pages 795 800 April 2000 Bobick et al The kidsroom Communications of the ACM Vol 43 No 3 60 61 2000 A Bonarini Designing highly interactive competitive robogames AIR Lab report 2010 David Cavallo Arnan Sipitakiat Anindita Basu Shaundra Bryant Larissa Welti santos John Maloney Siyu Chen Erik Asmussen Cyn thia Solomon and Edith Ackermann C and ackermann e roballet exploring learning through expression in the arts through construct ing in a technologically immersive environment In in Proceedings of International Conference on the Learning Sciences 2004 2004 M Tan and Y Shen D Xu A new simple visual control method based on cross ratio invariance EEE International Conference on Mechatronics Automation 2005 K Deguchi Optimal motion control for image based visual servoing by decoupling translation and rotation Proc Int Conf Intelligent Robots and Systems 1998 Carl F Disalvo Francine Gemperle Jodi Forlizzi and Sara Kiesler All robots are not created equ
101. is able to walk see its environment via camera and recognize spoken com mands in Spanish and English AIBO robotic pets are considered to be autonomous robots since they are able to learn and mature based on ex ternal stimuli from their owner their environment and from other AIBOs The AIBO has seen use as an inexpensive platform for artificial intelligence research because it integrates a computer vision system and articulators in a package vastly cheaper than conventional research robots The RoboCup autonomous soccer competition had a RoboCup Four Legged Robot Soc cer League in which numerous institutions from around the world would participated Competitors would program a team of AIBO robots to play games of autonomous robot soccer against other competing teams The developments in Robocup lead to improvements in the mobile robots 18 Chapter 2 State of the Art ni AIBO d Spykee e Rovio Figure 2 8 Robots developed for games and interactions The domestically aimed robots become popular in the market One of them is Spykee Figure 2 8 d which is a robotic toy made by Meccano in 2008 It contains a USB webcam microphone and speakers Controlled by com puter locally or over the internet the owner can move the robot to various locations within range of the local router take pictures and video listen to surroundings with the on board microphone and play sounds music or vari ous built in recordings R
102. is usually comes from the encoders A round shaped differential drive configuration is shown in Figure 2 4 In a tricycle vehicle Figure 2 5 there are two fixed wheels mounted 2 1 Locomotion 9 Figure 2 4 Differential drive configuration with two drive wheels and a castor wheel on a rear axle and a steerable wheel in front The fixed wheels are driven by a single motor which controls their traction while the steerable wheel is driven by another motor which changes its orientation acting then as a steering device Alternatively the two rear wheels may be passive and the front wheel may provide traction as well as steering Figure 2 5 Tri cycle drive combined steering and driving Another three wheel configuration is the synchro drive The synchro drive system is a two motor drive configuration where one motor rotates all wheels together to produce motion and the other motor turns all wheels to change direction Using separate motors for translation and wheel rotation guarantees straight line translation when the rotation is not actuated This mechanical guarantee of straight line motion is a big advantage over the differential drive method where two motors must be dynamically controlled to produce straight line motion Arbitrary motion paths can be done by actuating both motors simultaneously The mechanism which permits all wheels to be driven by one motor and turned by another motor is fairly 10 Chapter 2 State of the Art
103. lier NORMAL until the detected target distance is less than 700 millimeters And finally the motors speeds will be set with multiplier CLOSE from 700 mm until the target is reached This mechanism allows to achieve more precision in the motion since the mapping guarantees that each motor speed set by PWM will generate a reasonable motion Moreover instead of trying to map all the speeds uniquely indeed we are trying to map only one third of the speeds from the predefined speed ranges to 60 85 of the PWM value in the duty cycle Even though the robot goes to target with an unexpected path the path is corrected as the robot comes closer since the slow motor speeds are more precise and easy to achieve Even though the solution is not working perfectly the robot is able to go to the target successfully almost every time The travel time and number of movements in order to reach the target are not optimal but this result should be regarded as normal since we are not using any encoder to detect the movement in the motors The need for an optimization of the PWM signal is also related to the result of the miscalculation of the required torque We selected motors that do not have enough torque power for our robot This brought the problem of the arrangement of the PWM signal in order to run the motors Since the motors have less torque than we expected almost 50 of the PWM is wasted First trials to optimize this inefficiency have been made by cha
104. line to develop more interesting robogames 76 Chapter 7 Conclusions and Future Work Bibliography 10 11 12 13 Battle bots http www battlebots com Camera calibration toolbox for matlab http www vision caltech edu bouguetj calib_doc index html First robotics competition www usfirst org uploadedfiles who impact brandeis st udies 05fll underserved summary pdf Item http www item24 com en Lego group http mindstorms lego com en us default aspx Lifelong kindergarten http llk media mit edu projects php Logo foundation http el media mit edu logo foundation The mars autonomy project http www frc ri cmu edu projects mars Microsoft actimates interactive barney to in teract with barney amp friends on pbs http www microsoft com presspass press 1997 sept97 mspb spr mspx Plexiglass physical properties http www rplastics com phprofplac html Robowars http www robowars org Working with aluminum http www search autoparts com searchautoparts article articledetail jsp id 162604 Intel playTMme2cam computer video camera http www intel com support intelplay me2cam Intel Corpo ration 2008 14 irobot roomba vacuum cleaning robots iRobot Corp 2008 TT 78 BIBLIOGRAPHY 15 18 19 20 24 ea A Eliazar and R Parr Learning probabilistic motion models for mobile robots In Proceedings of the twenty first internat
105. m 2 ANIL 781AA302A1 w Batteries y Computer 9 Disk drives E Display adapters 2 DVD CD ROM drives IDE ATA ATAPI controllers Keyboards LibUSB Win32 Devices 3 Mice and other pointing devices Monitors E9 Network adapters 4 Ports COM amp LPT 4 Printer Port LPT1 JE COM Port COM4 System devices Universal Serial Bus controllers 1 8 9 E E RET ET E Figure C 5 The drivers can be validated from Device Manager To test the color found with color calculator the following steps should be followed From the tabs at the top BlobSearch should be selected In side BlobSearch tab color class number should be set accordingly with the number of the color we want to control Figure C 7 Under the Blob Color tab Visual Rule Editor should be checked to control the color The color code defined with inside the main software for the thesis will be explained in details in a few steps could be tested at that tab by adding the color codes to the Rules List that can be found on the lower right bottom of the software When the color information is ready it should be captured by clicking the Get from Visual Rule Editor in the Custom menu The next step is setting the Blob Geometry that defines the size of the color we are searching for
106. n are 30 150 and 270 respectively For this type of configuration the total platform displacement is achieved by summing up all the three vectors contributions given by gt PP Fr lt Fi F3 First of all some definitions need to be considered Figure represents the diagram of the mobile robot platform with the three wheels It was assumed that the front of the robot is in positive Y direction and the posi tive side to counter clockwise The three wheels coupled to the motors are mounted at angle position 60 60 and 180 degrees respectively It is 4 1 Wheel configuration 41 sons MY Wheel 1 A 0 866 0 5 Wheel 2 X axis Wheel 3 Refers to the turning direction Y axis Figure 4 1 The wheel position and robot orientation important to remember that the wheel driving direction is perpendicular to the motor axis therefore 90 degrees more The line of movement for each wheel when driven by the motor and ignoring sliding forces is represented in Figure 4 1 by solid black arrows The arrow indicates positive direction contribution The total platform displacement is the sum of three vector components one per motor and is represented as a vector applied to the platform body center In order to find out the three independent motor contributions the com position of the vectors represented by red arrows is projected on axes rep resenting the line of movement of eac
107. next step is converting the direction from camera to the ball d pan 55 5 1 Camera Calibration n gt TH X Axis Camera Plane E World Coordinate System gt X Axis ES PR AY 7 1 W Robot Coordinate System Figure 5 3 Camera robot world space into the direction vector and its normalized vector d norm in order to find the direction from camera plane to the ball u gt d yan Kls 1 gt dbal norm gt de bail E From the direction d norm and the line Inew the point P pan that is 3D homogeneous coordinates of the points with respect to camera c n ind 1 56 Chapter 5 Vision To calculate Pn which is 3D homogeneous coordinates of the point in world the transformation matrix camera to robot T is used R Ro pC P va Tc Po pan The second configuration is for the object on the ground This case is simpler since it requires a 2 D to 2 D transformation The transformation between camera and robot is used again and the calculation is the same as the previous configuration The homography is defined for this case as 100 Ri 0 00 27 The third row of the matrix is all zeros since the we assume that the H KxTg object is at ground with Z 0 hence this reduces the mapping from 3D to 2D To calculate the object position Pobject we will use the inverse homography on the image point Pimage 1 Pobject H x P image Pobject is then normalized
108. ng The final working configuration of the object detection is made by solving the problems described previously We can explain these in three titles First one is finding the correct color information We start by disabling the automatic white balance and automatic exposure compensation This allows us to have the color as it is seen directly from the camera by not adjusting 5 3 Tracking 65 any color values It means if the object is exposed to light the parts that are exposed to light are not considered as white but a tone of target color similarly with the previous configuration the dark pixel such as the parts in the shadow are considered as black but with the new configuration they are also considered as a tone of the target color Second step is calculation of the color information in a better way Pre viously we calculated the information by ST color picker software With this software we select the color region we want to assume as the blob color The problem of this approach is that the blob may not consists of a single rule RGB color code Some parts of the object are darker some parts are lighter Also with the distance the color that is composing the object is changing Additionally the colors that are not considered as red blue green represented again as rules RGB color codes and it is not possible to understand by just looking at the color code whether it is the correct color coding or not These makes it very di
109. ng ing the frequency of the PWM to a higher or a lower frequency but these trials did not give good results Later we found the proper configuration 4 3 PWM Control 49 mentioned previously by limiting the minimum and maximum PWM values of the motors and by introducing the fuzzy like control to introduce differ ent limits for the different distances to target Even though the lack of the encoders reduced the performance we were able to produce an acceptable configuration Another step is initialization of the motors The corresponding pins in the microcontroller for the motors are set to control the direction of motors and PWM generation The PWM generation is made by using the timers available in the microcontroller Output compare mode of the timers is used in order to create the PWM needed to run the motors Output compare function is used to control the output waveform and indicates when a period of time has elapsed When a match is found between the output compare register and the counter the output compare function e Assigns a value to pins e Sets a flag in the status register e Generates an interrupt Using the described output comparison we created the desired PWM wave to run the motors At first the timer clock is set to 8 mHZ Then output compare register 1 is set for the desired motor speed as some per centage of the full period of the cycle Output compare register 2 is set to full cycle that will set the pin to HI
110. ngular and linear movement In other words the robot goes to the spec ified target while it is also rotating The calculation made by the software simulator uses all the parameters frontal speed lateral speed and angular movement The calculation of the frontal and lateral speed is found by sub tracting the desired position of the robot from the initial position Figure Ei 7 Ve 12 200r n2 150 3 AF 100 M P d 50 E s 0 oe di kc Y oL VE 3500 mm s i Vai 8181 1718 mm s V 350 mm s Vaj 1446 8911 mm s 50l 4 0 099669 rad s V 5 2881 0061 mm s Vio 2053 1089 mm s 100 L v Vig 374 9172 mm s 7 du v 3500 mm s Q 129 3159 rad s 150 n 117 1141 rad s 15 2405 rad s 200r 250 dg 1 1 300 200 100 0 100 200 300 Figure 4 4 The mixed angular and linear movement calculated by simulator In Figure 4 4 the resulting motion is not only a linear movement towards the direction which is the composition of Vp and Vr but also an angular movement in clockwise direction with an angular speed of 0 099669 rad s All these three different configurations are used in the implementation of the robot The whole script is found in Motion Simulator in Appendix 46 Chapter 4 Control B 3 The most common used one is the mixed angular and linear movement Almost every path was initially implemented with this movement approach Others have been used where optimization is nee
111. obot is focused on construction of the robot chassis motor holders motor and wheel connections camera holder the foam covering the robot batteries and hardware architectures 3 1 Chassis The main principles for the construction of the chassis are coming from similar projects from the past which are the simplicity of assembly and disassembly the ease of access to the interior and the possibility of adding 22 Chapter 3 Mechanical Construction and modifying elements in the future We decided to use some design con straints revising these according to our goals The design is started with the choice of the chassis made of plexiglas One advantage of using plexiglas it is 4396 lighter than aluminum 10 An other advantage that is affecting our choice is the electrical resistance of the plexiglas that will isolate any accidental short circuit One of the major problems with plexiglas is the difficult processing of the material However it has to be processed only once hence this is negligible Figure 3 1 The design of robot using Google SketchUp The preliminary design has been created with Google SketchUp allow ing to define the dimensions of the robot and the arrangement of various elements shown in Figure 3 I This model has been used to obtain a descrip tion of the dynamics of the system The robot is 125 mm in diameter wide and 40 cm in height meeting the specification to be contained in a footprint on the ground in
112. obot laugh laser guns etc through the speaker Spykee has a WiFi connectivity to let him access the Internet using both ad hoc and infrastructure modes Similar to Spykee with different kinematics models and more improve ments Rovio M Figure 2 8 e is the groundbreaking new Wi Fi enabled mobile webcam that views and interacts with its environment through video and audio streaming According to our goals we investigated the previously made robots since we thought we could benefit from the techniques used Furbies have expres sions but they don t move While Rovio has omnidirectional Spykee has tracks for the motion but they lack entertainment AIBO has legs and a 2 4 Games and Interaction 19 lot of motors but these brings more cost and high complexity for the devel opment 20 Chapter 2 State of the Art Chapter 3 Mechanical Construction We started our design from these specifications of the robot e a dimension of about 25 cm of radius 20 cm height e a speed of about 1 m sec e sensors to avoid obstacles e a camera that can be moved up and down e power enough to move and transmit for at least 2 hours without recharging e the robot should cost no more than 250 euro In the development process we faced some problems due to the limita tions from the specifications Main causes of these problems are related with low cost that is coming with our design constraints The mechanical construction of the r
113. ompare mode of the timers used in order to create the PWM needed to run the motors Output compare function is used to control the output waveform and indicate when a period of time has elapsed When a match is found between the output compare register and the counter the output compare function e Assigns a value to pins e Sets a flag in the status register e Generates an interrupt Using the described output comparison we created the desired PWM wave to run the motors and camera servo The sensors which are the bumpers for the current state of the robot are initialized by setting the corresponding pins in the microcontroller As a final step in the initialization the blob search and color parame ters are set The blob search parameters which are grid size top bottom left and right borders of the image are defined Similarly the blob search geometry options are defined which are the color id minimum maximum area of the blob minimum maximum circularity of the blob and minimum maximum perimeter of the blob For each color the blob search geometry options should be defined separately with the proper coding Lastly the color information should be defined for each color we are searching The color information is calculated either by the ST color picker 70 Chapter 6 Game software or from the Matlab script taking samples from the environment in order to calculate the histogram to find the correct color coding The color val
114. on should not be controlled for the turn around the center of rotation but as we discussed previously we cannot guarantee the correctness of the movement since we lack encoders for the motors When a blob is found the first step is checking the color information of blob since the difference between targets is made by the color information If the cor 68 Chapter 6 Game rect target is found we calculate the distance of the object from the robot According to the distance found the motor contributions are calculated and the robot starts going to the target We introduced two different camera head positions At the start the camera is always set at the normal position which can see the environment from 750 mm from the robot up to 4 5 meters According to the result of the distance calculated at the previous step the camera head position maybe be lowered in order to detect and go to the target in the next step The target is set as acquired when the distance between the robot and the target is below a certain threshold The whole algorithm we can be seen in Appendix A A 2 in a clear way The software that is running the robot is working on board on the mi crocontroller The software is divided into sub modules in order to ease the development of the process In this section we will give the details of the software introducing the sub modules supporting with the diagrams The software is composed of low level and high level modules T
115. ond case we assume that the target object is on the ground and we are using the information coming from the intersection point of the object with the ground This 54 Chapter 5 Vision C Camera centre p Principal point x Point on image M Point in world coordinates f Focal length Figure 5 2 Perspective projection in a pinhole camera means the object can be represented by a X Y Z position in the world co ordinate system with Z 0 This reduces the transformation into a 2D 2D transformation and needs a new H matrix We need to calculate the transformation matrix between the robot and camera TE coordinates The transformation matrix between world frame and camera T coordinates is known from the external camera parameters The world frame to robot frame transformation matrix T 7 is calculated by the target object position and robot center shown in Figure 5 3 TR is derived as follows T Tg TY For the ball localization case the calculation is made for the pixel diam eter Dpx of the ball at known distance l and the real diameter Dpan For the camera we introduced a parameter f to indicate the dimension of a unit pixel which is a statical parameter of the camera Using this information the distance of the ball lnew can be calculated by counting the pixels of the diameter of the ball Dy and by multiplying this by f l x D X Dhan law El Dral px gt The
116. orange B 1 Microprocessor Code 97 RGBCube_AddCube 6 8 3 5 1 3 2 RGBCube_AddCube 5 7 1 3 0 1 2 night RGBCube_AddCube 0 6 4 8 0 2 1 RGBCube_SubCube 3 5 3 5 1 3 1 night void Init_Cam void 18 ImgBufferData 2 BUFFERSIZE_DEMO_BS 18 RGBCubeData 4096 BufferSize BUFFERSIZE_DEMO_BS pImgBufferData ImgBufferData pRGBCubeData RGBCubeData RGBCube_Clear BlobSearchInit white balance amp exposure compensation settings x24_WriteReg8 0 1380 0 wb x24 WriteReg8 0x1080 2 ae x24 WriteRegi6 0x1095 1000 x24 WriteRegi6 0x109d 240 x24 WriteRegF900 0 1081 1 0 u8 ImgFormat x24 ImageFormat RGB 565 116 Width 160 ui6 Height 120 u8 Framerate 25 u8 SensorMode x24 SensorMode SVGA u8 HalfSysClock 80 150 u8 JpegDerat 10 u8 JpegCompr CameraSetParam ImgFormat Width Height Framerate SensorMode HalfSysClock JpegCompr JpegDerat void GetCompletedJob void if Job CurrentJob State x24 StateAcquisitionEnd CompletedJob amp Job CurrentJob CurrentJob 1 void CameraSetParam u8 ImgFormat u16 Width u16 Height 98 Appendix B Documentation of the programming u8 Framerate u8 SensorMode u8 HalfSysClock u8 JpegCompr u8 JpegDerat Tx24 InitTypeDef x24 InitStruct Initialize the x24 module x24 StructInit amp x24 InitStruct x24 InitStruct SensorMode Sensor
117. osition of the object in the pixel coordinates We need a mapping between those three coordinates to command the movement There are sev eral approaches to determine the object position according to the robot as mentioned in Chapter 2 Among those we will use the one for the calibrated cameras The rest of Chapter describes camera calibration the script that calcu lates the transformation color definition and the script to define the color 5 1 Camera Calibration The initial step for the algorithm is to find the camera calibration The Camera Calibration Toolbox 2 is used to calculate the camera calibration matrix that is necessary to find the projection between the world and image points represented by homogeneous vectors Normally the object is placed on the real world coordinate systems and its position can be measured from the origin In order to find the distance of the camera plane to the object we need a transformation that maps the 3D points for the object that is assumed to place in ground the point is a 2D point in world coordinates since height is 0 all the times of the objects in the 3D world coordinates to 52 Chapter 5 Vision the 2D points in the image plane coordinates To find the transformation matrix we use the camera calibration matrix calculated from Camera Cal ibration Toolbox 2 In general the Camera Calibration Toolbox works as follows The sample pictures of the checker board it taken by the
118. ption of the environment it obtains workspace information locally while it is moving and a path must be incre mentally computed as the newer parts of the environment are explored 26 Autonomous navigation is associated to the capability of capturing infor mation from the surrounding environment through sensors such as vision distance or proximity sensors Even though the fact that distance sensors such as ultrasonic and laser sensors are the most commonly used ones vision sensors are becoming widely applied because of its ever growing capability to capture information at low cost Visual control methods fall into three categories such as position based image based and hybrid 28 The position based visual control method re constructs the object in 3D space from 2D image space and then computes the errors in Cartesian space For example Han et al presented a posi tion based control method to open a door with a mobile manipulator which calculated the errors between the end effector and the doorknob in Carte sian space using special rectangle marks attached on the end effector and doorknob As Hager pointed out the position based control method has the disadvantage of low precision in positioning and control To improve the precision El Hakim et al proposed a visual positioning method with 8 cameras in which the positioning accuracy was increased through iteration It has high positioning accuracy but poor performance in real time T
119. rMode u8 HalfSysClock u8 JpegCompr u8 JpegDerat void Init_BlobParameters void endif CAMERA_INIT_H_ Camera_Init c Standard include include 91x map h include utils h include definitions h include definitionsLib h include CamInt_x24 h include usb_lib h include usb_conf h include usb_prop h include usb_pwr h include usb config h include RGBCube h include blobsearch h include math h include Camera Init h Include of other module interface headers Local includes Private typedef define BUFFERSIZE DEMO BS 160 120 2 Private define define TimeDiv 250000000 Private macro Private variables const u16 SyncOutPeriods 150 9999 8332 7142 6249 5555 4999 4545 4166 3846 3571 3333 3125 2941 2777 2631 2500 2381 2272 2174 2083 90 Appendix B Documentation of the programming 2000 1923 1852 1786 1724 1666 1613 1562 1515 1470 1428 1389 1351 1316 1282 1250 1219 1190 1163 1136 1111 1087 1064 1042 1020 1000 980 961 943 926 909
120. rd wheel has two DOF The caster offset standard wheel also know as the castor wheel shown in Figure 2 1 b has a rotational link with a vertical steer axis skew to the roll axis The key difference between the fixed wheel and the castor wheel is that the fixed wheel can accomplish a steering motion with no side effects as the center of rotation passes through the contact patch with the ground whereas the castor wheel rotates around an offset axis causing a force to be imparted to the robot chassis during steering 30 a Standard Wheel b Castor Wheel Figure 2 1 The standard wheel and castor wheel The second type of wheel is the omnidirectional wheel Figure 2 2 The omnidirectional wheel has three DOF and functions as a normal wheel but 2 1 Locomotion 7 provides low resistance along the direction perpendicular to the roller di rection as well The small rollers attached around the circumference of the wheel are passive and the wheel s primary axis serves as the only actively powered joint The key advantage of this design is that although the wheel rotation is powered only along one principal axis the wheel can kinemati cally move with very little friction along many possible trajectories not just forward and backward Figure 2 2 Omniwheels The third type of wheel is the ball or spherical wheel in Figure It has also three DOF The spherical wheel is a truly omnidirectional wheel often designed so that it may
121. rk Bibliography A Documentation of the project logic B Documentation of the programming B 2 Color Histogram Calculato B 3_ Object s Position Calculato B 4 Motion Simulato C 2 Setting up the environment Qt SDK Opensource C 3 Main Software Game software D Datashee VIII B 1 Microprocessor Codd C 1 Toolchain Software 51 51 58 64 67 73 76 81 85 85 120 122 127 135 135 138 140 143 List of Figures 2 1 The standard wheel and castor wheel 6 2 2 Omniwheeld 7 2 3 Spherical Wheels 7 Csi ia 9 lt 10 2 7 Palm Pilot Robot with omniwheeld 11 2 8 Robots developed for games and interactions 18 3 1 The design of robot using Google SketchUp 3 2 The optimal torque speed curve 24 44 25 m 26 26 lt 27 TE 28 2 20 4 30 442 30 uM ae 31 32 xor edades 32 MCI 33 CD 41 a RITO 43 M 44 T 45 IX Chapter 1 Introduction 1 1 Goals The aim of this thesis is to develop an autonomous robot implementing the main functionalities needed for low cost but interesting robogames There are some predefined constraints for the robot One of the most important property is being the lowest cost possible We limit the maximum cost to 250 euro In order
122. rning and a maximum percentage of the duty cycle at which the speed remains as full speed after that The minimum PWM value to start the motion is 45 of the duty cycle and the speed set to full speed after exceeding the 90 of the duty cycle which is the maximum PWM that can be applied Since there is a non linear increase between the ratio of rotation of the motors and applied PWM rather than a linear one the speed changes are not distinguishable In other words the range of PWM signal that enables movement is very narrow and the difference between the minimum speed and maximum speed is not very apparent The script that is calculating the motor contributions according to the target position is not taking into account the motor s maximum or mini mum speed The script can return a motor contribution with 4000 mm s or 3 mm s which is cannot be performed physically by the motors After some experiments to limit the script that calculates the motor contributions we decided to bound the PWM signal value to be applied to the motors The maximum PWM value defined for a motor is 2 5 meters second and the minimum PWM value is around 300 millimeters second Using both the limits coming the physical characteristic of motor and the PWM rotation limitations we implemented a map that transforms the motor contribution calculated by the script into a PWM signal To do so all the speeds between 2500 mm s and 300 mm s should be mapped to PWM signals between
123. rrl M m arr offsetl offsetl offsetl v txtl gt Mx v_txt offsetl B 4 Motion Simulator 131 fill Cmzare marte 245 g text v txt1 1 v txt1 2 v t1 end if vn 1 0 Ter robot cosA R_robot cosA vn 1 s1 cosA R _robot sinA R_robo sinA vn 1 s1 sinA Color 0 1 0 offset1 R_robot cosA vn 1 sl cosA R_robot sinA vn 1 sl sinA if vn 1 sl cosA lt 0 M m_rot pi 2 alpha else M m rot pi 2 alpha end m arrl M m arr offsetl offsetl offsetl v txtl Mrv_txt offsetl fill m arr1 1 m arr1 2 g text v txt1 1 v txt1 2 v n1 end ruota 2 if vt 2 0 line R_robot cosA R_robot cosA vt 2 s1 sinA R robot sinA R ro sinA vt 2 s1 cosA Color 0 1 0 offsetl R _robot cosA vt 2 slxsinA R_robot sinA vt 2 s1lx cosA if vt 2 sl sinA lt 0 M m rot pi alpha else M m rot alpha end m arrl M m arr offsetl offsetl offsetl v txtl Mrv_txt offsetl fill m arr1 1 m arr1 2 g text v txt1 1 v txt1 2 v t2 end if vn 2 gt 0 line R_robot cosA R_robot cosA vn 2 s1 cosA R robot sinA R ro sinA vn 2 sl sinA Color 0 1 0 132 Appendix B Documentation of the programming offsetl R_robot xcosA vn 2 slxcosA R_robot sinA vn 2 sl sinA if vn 2 sl sinA lt 0 M m rot pi 2 alpha else M m rot
124. s and for the ESC is around 30 dollars The minimum expected price for motors and controller will be a least 90 dollars if we can run the 3 motors on a single controller Also there should be an extra cost to gear them down We made some calculations to find the most suitable DC motor for our system In order to effectively design with DC motors it is necessary to understand their characteristic curves For every motor there is a specific torque speed curve and power curve The graph in Figure 3 2 shows a torque speed curve of a typical DC motor D C Motor Torque Speed Curve Stall Torque ts No Load Speed wn x e Rotational Speed Figure 3 2 The optimal torque speed curve Note that torque is inversely proportional to the speed of the output shaft In other words there is a trade off between how much torque a motor delivers and how fast the output shaft spins Motor characteristics are frequently given as two points e The stall torque Ts represents the point on the graph at which the torque is a maximum but the shaft is not rotating e The no load speed wn is the maximum output speed of the motor when no torque is applied to the output shaft The linear model of a DC motor torque speed curve is a very good ap proximation The torque speed curves shown below in Figure are calcu lated curves for our motor which is Pololu 25 1 Metal Gearmotor 20Dx44L mm 3 2 Motors 25 Torq
125. s and motion mechanism is used if cam_position SERVO_TOP CalculatePositionTop n_point position ifdef NO_DEBUG Print_Int n_point X X in cam coordinates Print Int n point Y Y injcam coordinates Print_Int position 0 X jin real coordinates Print Int position 1 Y jin real coordinates endif double alpha 0 alpha atan2 position 1 position 0 s8 omega_modifid alpha 100 Triskar position omega_modifid speed if Abs position 0 gt 1200 speed 2 speed 2 2 Set_Speed speed FAR else if Abs position 0 gt 700 B 1 Microprocessor Code 91 Set_Speed speed NORMAL else cam_position SERVO_MID Position_Camera SERVO_MID ifdef NO_DEBUG Print_Int speed 0 1 motor send Print Int speed 1 2 motor send Print Int speed 2 3 motor send endif servo_top else servo mid CalculatePositionGround n point position double alpha 0 alpha atan2 position 1 position 0 s8 omega_modifid alpha 100 Triskar position omega_modifid speed speed 2 speed 2 2 Set Speed speed CLOSE if n point Y gt 115 color_done TRUE 3 blob else of control hit LOK OK EK E SE AAA AA AAA AA AA AAA AA AAA AA AA AAA AA AAA AA AA AA AAA AA AAA AA AAA AA AA AAA AA AAAA AAA AAA AAA Function Name SCU_Configuration Description Configures the system clocks Input Non
126. sor we use the ST ARM Toolchain Unzip the archive STLCam XXXz Redistrib zip in the Eclipse workspace directory Now you need to import the project From Eclipse choose File Import General Existing project into workspace browse for the STLCam directory and then click Finish shown in Figure After the import completed and the compilation ended successfully we need to import the Launch Configurations to the tool chain From Eclipse choose File Import Run Debug Launch Configurations and browse for the Launch Configurations 003 and then click Finish T he step can be seen from the Figures below Before programming the microprocessor the drivers for the J TAG pro 136 Appendix C User Manual Import Import Projects Select a directory to search for existing Eclipse projects Import Select Import launch configurations from the local file system Figure C 2 The import screen for Launch Configurations C 1 Tool chain Software 137 ll Import Launch Configurations mE ni xj Import Launch Configurations AS Import launch configurations from the local file system Y dl From Directory C Documents and Settings Administrator Desktop test Launch Configurations 003 Browse Launch Configurations 003 STM32F107xx DEBUG launch B STM32RET6_DEBUG launch E sTR9 DEBUG launch B str91x launch Hae Overwrite existing launch config
127. suming that the light has a lot of that color in it This can result in an incorrect white balance and a color cast No adjustments made for the temperature of the light wanted In certain cases the color of the light is what makes the photograph A sunset is an example Without the rich warm colors of the light a sunset just isn t a sunset Auto white balance may attempt to make adjustments to correct for the warm color of the sunset light This would produce an image with less saturated colors or colors that were different than what has been seen The camera we are using supports different settings for the balance These are OFF No White balance all gains will be unity in this mode AUTOMATIC Automatic mode relative step is computed here MANUAL_RGB User manual mode gains are applied manually DAYLIGHT_PRESET DAYLIGHT and all the modes below fixed value of gains are applied here TUNGSTEN_PRESET FLUORESCENT_PRESET HORIZON_PRESET MANUAL _color_TEMP FLASHGUN_PRESET Among the possible options we set the white balance OFF Because in the application we require absolute color accuracy and we don t want the camera to change the temperature of the color since we want to detect the target with a constant color parameter defined using the samples we took 5 2 Color Definition 61 Another automatic adjustment can be done on exposure Subjects lighter than middle gray such as a white china plate reflect more
128. ted visual servoing of pla nar robots Proceedings of IEEE International Conference on Robotics and Automation 2002 Appendix A Documentation of the project logic Documentation of the logical design which is documenting the logical de sign of the system and the design of SW This appendix shows the logical architecture implemented 82 Appendix A Documentation of the project logic control hit unsigned char assa S CRI DR Sa GPIO Configuration X24 HwConfigl 24 SendPatches control hit unsigned char servo position unsigned char gt 0 servo flag unsigned char 0 USB Configuration LIS8 Init Init Cam Init Motors Position Camera Init_Bumpers Init_BlobParameters pimgBufferData unsigned char BufferSize unsigned long Completedlob unspecified 0 Init_Cam Init_BlobParameters GetCompletedJob Position_Camera CameraSetParam Set_Directions uses Stop Motori Bumpers Set_Speed Turnf 2 C uit gt c 3 pimgBufferData unsigned char BufferSize unsigned long job 2 Led SM Currentlob main c Print Int Blob_Search vectorByMatrix3x1 CalculatePositionTop vectarByMatrixdx1 CalculatePositionGround Matrix Multi CalculateBallPosition Matrix Mult3334 Matrix Mult3443 vectorByMatrix3x1 s16 Figure A 1 The class diagram of
129. ten for clarification camera to world T cch1 T cch 0 0 0 1 world to camera T che inv T cch1 world to robot T rch1 T rch 0 0 0 1 robot to world T chr inv T rchl T_rc T_rchl T_chc T_cr inv T_rc Tier Tera xa ROBOT i pese E ROBOT 3 T cr 2 BOBOT S 2 ren T Gr Sa c 0 0 1 0 B 4 Motion Simulator 127 000 0 0 E H KxT_cr t_ Hinv inv H point Hinv 19 99 1 result point point 3 result B 4 Motion Simulator robot_distance m The distance should be expressed in terms of frontal lateral cartesian coordinate angle to the object polar coordinate X 3500 Y 350 deltaX X deltaY Y p sqrt deltaX deltaX deltaY deltaY angle atan2 Y X alpha 0 frontal speed gt 3500 350 lateral speed vf frontal speed cos alpha lateral speed sin alpha new x vl frontal_speed sin alpha lateral_speed cos alpha new y The triskar function called like this vt omega_w vn TriskarOdometry 250 24 6 pi 6 vf vl angle 1 Triskar m 128 Appendix B Documentation of the programming function vt om vn TriskarOdometry R robot R_wheel alpha v_F v L omega plot R robot robot radius by mean of distance of the wheels from the center of the robot mm 25 cm in our case 250 mm R_wheel wheel radius mm 24 6000 mm yaricap alpha angle of
130. ter nal geometry does not appear in the kinematic equations of the robot so it can be ignored e The robot has three DOF without singularities e The robot can instantly develop a force in an arbitrary combination of directions x y 0 e The robot can instantly accelerate in an arbitrary combination of di rections x y 0 Non holonomic robots are most common because of their simple design and ease of control By their nature non holonomic mobile robots have fewer degrees of freedom than holonomic mobile robots These few actuated degrees of freedom in non holonomic mobile robots are often either inde pendently controllable or mechanically decoupled further simplifying the low level control of the robot Since they have fewer degrees of freedom there are certain motions they cannot perform This creates difficult prob lems for motion planning and implementation of reactive behaviors However holonomic offer full mobility with the same number of degrees of freedom as the environment This makes path planning easier because there are no constraints that need to be integrated Implementing reactive be haviors is easy because there are no constraints which limit the directions in which the robot can accelerate 2 2 Motion Models 13 2 2 Motion Models In the field of robotics the topic of robot motion has been studied in depth in the past Robot motion models play an important role in modern robotic algorithms The main goal of
131. ternate GPIO_OutputAlti GPIO Init LEDR GPIO amp GPIO InitStructure Configure MY GPIOU GPIO_StructInit amp GPIO_InitStructure GPIO InitStructure GPIO Direction GPIO_PinOutput GPIO InitStructure GPIO Pin MY GPIOx Pin GPIO InitStructure GPIO Type GPIO Type PushPull GPIO InitStructure GPIO Alternate GPIO_OutputAlti GPIO Init MY GPIO amp GPIO_InitStructure kkkkkkkkkkkkkkkkkkxk C COPYRIGHT 2007 STMicroelectronics END OF FILE Camera_Init h k Camera_Init h Created on Jan 19 2011 Author Administrator ifndef CAMERA_INIT_H_ tdefine CAMERA_INIT_H_ Inctud s 22332 SS DD O SS SS SS SR tinclude 91x_lib h tinclude CamInt_x24 h Exported types B 1 Microprocessor Code 95 Exported constants Module private variables extern u8 pImgBufferData extern u32 BufferSize extern u8 CamState O Uniitialized 1 Running 2 Paused extern Tx24 ImgJob CompletedJob Jk Exported macro Private functions Exported functions void Init Cam void void GetCompletedJob void void CameraSetParam u8 ImgFormat u16 Width u16 Height u8 Framerate u8 Senso
132. the application may fix this problem Figure C 4 The error that can be caused if Qt libraries not found To solve the problem place the executable file under Qt libraries folder From the Windows Device Manager under Ports COM amp LPT find the STM Virtual COM Port and check port number In the example it is COMA Figure C 5 Use this information in the software In the fol lowing order we first Open the port Then Start Camera with the BlobSearch option Then Get Image to check the camera started cor rectly Continuous should be selected if we can to see the image as a video stream Figure Since there is no mechanism to save the pictures we capture the pictures by taking a snapshot in the windows The important point in capturing the images is to set Zoom parameters to 100 in order not to change the resolution of 160x120 The images should be cut also at the orginal res olution 160x120 from the snapshot images to achieve the correct values The matlab script color calculator that can be found at Appendix B 1 is used to calculate the color histogram of the demanded object It is advised to take as many samples as you can with different distances to robot and with different illumination to have better histogram C 2 Setting up the environment Qt SDK Opensource 139 Tr xKX YX i4 File Action View Help e mfa 2em mim
133. the most used classes 83 Initialize parameters search directions TRUE replay counter color done FALSE blob foundsFALSE cam position SERVO TOP Irc return to begining can m a begining f return to begining Check color done FALSE reply_counter color 1 FALSE LC FALSE i Turn the robot around its center ot ratation t 4rF TRUE fn fee SVEZIA color_done s 0 cam_position SERVO_MID Figure A 2 The flow diagram of the game algorithm 84 Appendix A Documentation of the project logic Appendix B Documentation of the programming The microprocessor code the scripts and other helper tools that are used during the implementation are included here B 1 Microprocessor Code main c DK KEK RK KKK KKK KKKKKEK C COPYRIGHT 2007 STMicroelectronics x kkk kxkkkxkkkxkk k File Name main c Author AST Robotics group Date First Issued 11 May 2007 Version 1 0 Description Main program body AAA AAA AA AO RO RO RO ROIO RARA AAA AAA AAA AAA AAA AAA AAA OR ARR AAA AAA AAA AAA ARR RA AAA History 28 May 2007 Version 1 2 11 May 2007 Version 1 0 AAA AAA AAA RO RO ROIO NOAA AA AAA AA AAA AA RA AAA AAA AA AAA AAA RA AAA AAA AAA AA ROR ROR eee AAA THE PRESENT SOFTWARE WHICH IS FOR GUIDANCE ONLY AIMS AT PROVIDING CUSTOMERS WITH CODING INFORMATION REGARDING THEIR PRODUCTS IN OR
134. the wheel configuration we preserved the popular three wheeled con figuration Figure 3 8 The control is simple the maneuverability is enough to satisfy the design specifications The details of this configuration will be mentioned in Control Chapter 3 4 Camera The camera positioning is tricky We needed a holder that should be light in the weight but also provide enough height and width to enable vision from the boundary of the robot at ground to the people face in the environment The initial tests have been made by introducing a camera holder using parts from ITEM 4 These parts are useful during the tests since they are easily configurable for different orientations and easy to assemble But the parts are too heavy and we decided to use an aluminum bar for the final configu ration The movement of the camera is done by the servo placed at the top of the aluminum bar this gave us the flexibility to have different camera 3 5 Bumpers 29 Figure 3 8 Three wheel configuration positions that will be useful to develop different games The camera is placed in a position on top of a mounted on aluminum bar that allows us to have the best depth of field by increasing the field of view The idea is to detect objects and visualize the environment be tween the boundary of the robot to all the way in the ground and up to 2 meters in height which allows also to see faces of people or the objects not on the ground Mechanically the
135. tion The game is finding the predefined colored ob jects and going to them in the defined order We tested the game with two objects colored in orange and green More colors and objects can be introduced easily following the steps indicated in Appendix C The color detection must be made with the color calculator script and must be tested with the ST color picker software In order to have stable performance it 74 Chapter 7 Conclusions and Future Work is advised to check the colors before each run since the illumination of the environment can change the color interpretation We defined two different configurations for each color one for the sunny day the other for the cloudy day both are working at our testing environment In order to solve the problems related to vision color selection script a custom white balance and an auto exposure is used to improve the be havior of the system affected by low cost vision The problems arisen from the motor control is solved by limiting the minimum and maximum PWM values of the motors and by introducing the fuzzy like control to introduce different limits for the different distances to target The use of these tun ings and optimizations enabled us to select low cost components which are performing enough well to develop a robot framework that will be used as a basis to implementing different robogames As for the future developments we plan two additions The first one it is the development o
136. to satisfy this constraint we focused on producing and reusing some of the components or choosing the components that barely work Another constraint is the target environment of the robot which is home environment The size and the weight of the robot have been chosen in a way that it can move easily in home environments The choice of the kinematics and the wheels are made according to these needs 1 2 Context Motivations Finding solutions for building robots capable of moving in home environ ment and to cooperate with people is a subject of much study prevalent in recent years many companies are investing in this area with the conviction that in the near future robotics will represent an interesting markets The aim of this work has been to design and implement a home based mobile robot able to move at home and interacting with users involved in games The presented work concerns the entire development process from building a model of the system and the design of the chassis mechanical components and electronics needed for implementation robot control and the acquisition of sensory data up to the development of behaviors 2 Chapter 1 Introduction The preliminary phase of the project has been to study the kinematics problem which led to the creation of a model system to analyze the rela tionship between applied forces and motion of the robot The work proceeded with the design of mechanical parts using solutions to meet the
137. ture for the body of the robot Unfortunately castors can cause problems if the robot reverses its direction The castor wheel must turn half a circle and the offset swivel can impart an undesired motion vec tor to the robot This may result in to a translation heading error Straight line motion is accomplished by turning the drive wheels at the same rate in the same direction In place rotation is done by turning the drive wheels at the same rate in the opposite direction Arbitrary motion paths can be im plemented by dynamically modifying the angular velocity and or direction of the drive wheels The benefits of this wheel configuration is its simplicity A differential drive system needs only two motors one for each drive wheel Often the wheel is directly connected to the motor with internal gear reduction Despite its simplicity the controllability is rather difficult especially to make a differential drive robot move in a straight line Since the drive wheels are independent if they are not turning at exactly the same rate the robot will veer to one side Making the drive motors turn at the same rate is a challenge due to slight differences in the motors friction differences in the drive trains and friction differences in the wheel ground interface To ensure that the robot is traveling in a straight line it may be necessary to adjust the motor speed very often It is also very important to have accurate information on wheel position Th
138. ue and Power vs Speed 250 Rotational Speed RPM Figure 3 3 The calculated curve for the Pololu 25 1 Metal Gearmotor 20Dx44L mm Due to the linear inverse relationship between torque and speed the maximum power occurs at the point where w lus and 7 IT The maximum power output occuring at no load speed with 7 500rpm 52 38rad sec and the stall torque w 0 282N m is calculated as follows P T w 1 1 Pax 2 3 gen Pmax 26 190rad sec x 0 141Nm 3 692W Keeping in mind the battery life the power consumption the necessary torque and the maximum speed we selected the Pololu motors shown in Figure In the design of the robot we decided to use low cost components In that sense we focused on producing components or re using components that can be modified according to our demands The mechanical production of the components took some time both for the design and the construction process e g the connectors between motors and wheels are milled from an aluminum bar however this reduced the overall cost The connection of motors with robot is made by putting the motor inside an aluminum tube merging it with the U shaped plate Figure B 5 Using such a setup helps 26 Chapter 3 Mechanical Construction Figure 3 4 Pololu 25 1 Metal Gearmotor 20Dx44L mm Key specs at 6 V 500 RPM and 250 mA free run 20 oz in 1 5 kg cm and 3 3 A stall not only protecting the motor from the hit damage but also coolin
139. ues found from the Matlab script should be tested with the ST color picker software in order to check the correctness of the results After the initialization phase completed the main control loop starts Before entering the main control loop the parameters such as search_direction game_end color_done blob_found etc that are going to be used locally in side the main loop are set to initial values The main loop starts by reading the values of the sensor data Later the color_done status which is con trolling the target acquired condition Initially it is set as FALSE since the camera did not acquire any images Before capturing any image the control_hit is checked to detect and clear the hit Until no collision is detected the capturing of the images will not start After the capturing is complete the blobs found with the correct color if any are sorted according to the their area and the one with the maximum area is set as myBlob If no blob found in this step myBlob is set to 1 to indicate no blob is found within constraints The sign of the myBlob is controlled to check whether a blob is found or not If no blob is found blob_found or counter status is checked This part is implemented so that to decrease the number of turns for each time a blob is not found Having the counter less then 3 we increment the counter and do not perform any turns In the case the counter is gre
140. urations without warning D lt Back Next gt __Finish Cancel Figure C 3 The second step at importing Launch Configurations grammmer and the Virtual COM Port for the USB driver of the board provided in the CD should be installed The PC should be restarted if it is necessary After installing the drivers successfully the next step is flashing ST s software into the microprocessor You can program the microprocessor by clicking the program_USB from the Make view If the programming is finished successfully you should see Flash Pro gramming Finished from the Console 138 Appendix C User Manual C 2 Setting up the environment Qt SDK Open source Download the 2009 04 version of the Qt SDK from here ftp ftp qt nokia com qtsdk qt sdk win opensource 2009 04 exe Do not download the last version at the moment our software works only with Qt 4 5 x Run the executable that will install on your pc the Qt Framework 4 5 3 and the Qt Creator IDE Unzip the archive STRVS Serial Executable zip in a directory of your choice If you receive the following error Figure C 4 while trying to launch the STRVS Serial 046 exe place the executable file in the installed direc tory of the Qt Framework and launch from that location STRYS Serial 046 exe Unable To Locate Component X eo This application has failed to start because QtCore4 dll was not Found Re installing
141. utton to short the circuit pulling the signal line high or low An example is the micro switch with a lever attached to increase its range as shown in Figure 3 11 The cost is nothing if compared to the other solu tions such as photo resistors and sonars and the usage is pretty simple since the values can be read directly from the microcontroller pins without having any control circuits Major drawback is its limited range but we tried to improve the range using the foam and the external springs attached to the foam Since the robot is light in the weight and collision can be useful in development of games we decided to use tactile bumpers n Figure 3 11 Bumpers are mechanical buttons to short the circuit pulling the signal line high or low The collision detection for robot is made with bumpers which are placed on the plexiglas every 60 Figure 3 12 The coverage was not enough so the bumpers are covered with foams which are connected to the springs The springs are enabling the push back of the switches the foams are increasing the coverage of the collision detection and also enhance the safety both for the damage that could be caused by the robot and to the robot from en vironment Figure 3 13 After some tests we realized there are still dead points which the collision are not detected We decided to cut the foam into three placing the around the robot leaving the parts with the wheel open The best results are obtained using
142. zione dei dati forniti dai sensori ed stato implementato un semplice gioco per mostrare tutte le funzionalit disponibili I compo nenti utilizzati sono stati scelti in modo da costruire il robot con il minor costo possibile Sono infine state introdotte alcune ottimizzazini ed stata effettuata una accurata messa a punto per risolvere i problemi di impreci sioni nati dall utilizzo di componenti a basso costo Summary Aim of this project is the development of a robot suitable to implement highly interactive robogames in a home environment This will be used in the Robogames research line at AIRLAB to implement even more interesting games This is also an experiment in the new development line of research where user needs and costs are considered as a primary source for speci fication to guide robot development We have implemented a full system from building a model of the robot and the design of the chassis mechanical components and electronics needed for implementation robot control and the acquisition of sensory data up to the design of a simple game showing all the available functionalities The selection of the components made in a manner that will make it with the lowest cost possible Some optimizations and tuning have been introduced to solve the inaccuracy problem arisen due to the adaption of low cost components III Thanks I want to thank to Prof Bonarini for his guidance and support from the initial to the final l

Submitted version of the thesis - AIRWiki

Contents

Download Pdf Manuals

Related Search

Related Contents