Pebbles: User-Configurable Device Network for Robot
Contents
1. 12 2012 Bahl P Padmanabhan V N RADAR An In Building RF based User Location and Tracking System In The Conference on Computer Communications Annual Joint Con ference of the IEEE Computer and Communications Societies vol 2 pp 775 784 2000 Haeberlen A Flannery E Ladd A M Rudys A Wallach D S Kavraki L E Practi cal Robust Localization over Large Scale 802 11 Wireless Networks In Annual Interna tional Conference on Mobile Computing and Networking pp 70 84 2004 Nishida Y Aizawa H Hori T Hoffman N H Kanade T Kakikura M 3D Ultrason ic Tagging System for Observing Human Activity In IEEE RSJ International Conference on Intelligent Robots and Systems pp 785 791 2003 Chang H Choi J Kim M Experimental research of probabilistic localization of service robots using range image data and indoor GPS system In IEEE International Conference on Emerging Technologies and Factory Automation pp 1021 1027 2006 Niwa H Kodaka K Sakamoto Y Otake M Kawaguchi S Fujii K Kanemori Y Sugano S GPS based Indoor Positioning system with Multi Channel Pseudolite In IEEE International Conference on Robotics and Automation pp 905 910 2008 Gonzalez J Blanco J L Galindo C Ortiz de Galisteo A Fernaindez Madrigal J A Moreno F A Martinez J L Combination of UWB and GPS for indoor outdoor vehicle localization In IEEE International Symposium on Intelligent Signal2. Processing pp 1 6 2007 Ubisense http www ubisense net Ishiguro H Distributed Vision System A Perceptual Information Infrastructure for Robot Navigation In International Joint Conference on Artificial Intelligence pp 36 41 1997 1D 16 17 18 19 20 21 22 2a 24 23 26 2T Saito S Hiyama A Tanikawa T Hirose M Indoor Marker based Localization Using Coded Seamless Pattern for Interior Decoration In IEEE Virtual Reality Conference pp 67 74 2007 Nakazato Y Kanbara M Yokoya N Localization System for Large Indoor Environ ments Using Invisible Markers In ACM Symposium on Virtual Reality Software and Technology pp 295 296 2008 Miyama S Imai M Anzai Y Rescue Robot under Disaster Situation Position Acquisi tion with Omni directional Sensor In IEEE RSJ International Conference on Intelligent Robots and Systems vol 4 pp 3132 3137 2003 Pugh J Martinoli A Relative Localization and Communication Module for Small Scale Multi Robot Systems In IEEE International Conference on Robotics and Automation pp 188 193 2006 Pugh J Raemy X Favre C Falconi R Martinoli A A Fast Onboard Relative Posi tioning Module for Multirobot Systems IEEE Transactions on Mechatronics vol 14 no 2 pp 151 162 2009 Roberts J F Stirling T S Zufferey J C Floreano D 2 5D Infrared Range and Bearing System for Collective Robo
3. peb ble to indicate that the user should rearrange it Utilizing LED and voice feedback the user places pebbles at desired goals and their transit points If the user places pebbles as in Fig 4 for example the robot can go to all rooms except the top middle room Since only routes connecting two arbi trary neighbor pebbles are considered possible trails the user can explicitly keep the robot from entering the top middle room Fig 4 Sample Deployment The robot can go to all rooms except the top middle room In this case the user explicitly keeps the robot from entering the top middle room When the user changes the layout of the rooms or introduces a new object that ob structs a route of the robot the user has to rearrange the pebbles near the change Pebbles periodically transmit signals and automatically detect topology changes Therefore the user can reconfigure the network by adjusting pebbles confirming that a pebble communicates with its neighbors as it did initially This is useful for exam ple when the user places an obstacle such as a piece of luggage in the room tempo rarily and wants to specify a path that avoids it Locations are labeled by attaching physical labels on the buttons of the remote con trol Fig 5 After labeling the user can select the target location of the robot by the name of the label We consider three kinds of labeling static semi static and dynam ic targets Static targets are
4. Pebbles User Configurable Device Network for Robot Navigation Kentaro Ishii Haipeng Mi Lei Ma Natsuda Laokulrat Masahiko Inami and Takeo Igarashi The University of Tokyo Tokyo Japan 2 JST ERATO IGARASHI Design Interface Project Tokyo Japan 3 Keio University Yokohama Japan kenta sakamura lab org mi iii u tokyo ac jp malei satolab itc u tokyo ac jp natsudadlogos t u tokyo ac 7p inami inami info takeo acm org Abstract This study proposes devices suitable for use by non experts to design robot navigation routes The user places landmarks called pebbles on the floor to tell navigation routes to a robot Using infrared communication the pebbles automatically generate navigation routes The system is designed such that non expert users can understand the system status to configure the user s target en vironment without expert assistance During deployment the system provides LED and voice feedback The user can confirm that the devices are appropriate ly placed for the construction of a desired navigation network In addition be cause there is a device at each destination our method can name locations by associating a device ID with a particular name A user study showed that non expert users were able to understand device usage and construct robot naviga tion routes Keywords Robot Navigation Tangible User Interface Navigation Landmark Non Expert User Fig 1 Pebbles Usage The user place
5. ask the robot to reach a specific location However existing methods require experts to deploy devices and therefore the home or office user needs expert assistance to configure locations and labels We propose a method that allows non expert users to configure the environment for robot navigation employing infrared communication devices called pebbles The robot receives topology information from these devices and moves to locations by tracing infrared signals We extend previously demonstrated implementation 6 and discuss usage scenario and deployment assistance We also conduct a user study to see if non expert users can understand device usage Fig 1 illustrates the basic con cept of the proposed method The user only has to place pebbles at desired locations and the pebbles will automatically construct a network topology After deployment the robot can move to any pebble location in this network Since infrared signals trav el in a straight line if one pebble has unobstructed line of sight to another pebble the robot can be expected to travel between these two pebbles With our method destina tions are determined by pebbles and thus it is easy to label destinations by associating a pebble device ID with a name In our current implementation the user can easily do this by putting physical labels on buttons on a remote control for example kitchen or entrance Fig 5 The user can select a labeled button for the target loca
6. bserved that the robot reached the final goal in most navigation trials We confirmed that nav igation can be done when a pebble is near a highly reflective wall Fig 8 c In this case the angle estimation of the robot has slight errors but the robot can move in a similar direction and can try the next step of angle estimation We also confirmed that normal wallpaper covered walls do not cause this problem robot pebble gt infrared signal a b c Fig 8 Possible Limitation If there are highly reflective walls like a in the environment the robot may receive the same signal from all directions and be confused b On the other hand the robot can deal with the case that a pebble is near a highly reflective wall c 4 Sample Deployment We took a pebble system to a demonstration house and conducted a trial deployment and navigation in a relatively realistic environment We took a first time user from our group and asked her to deploy the pebble system The goal of the deployment was to get a robot navigate between a kitchen and a studio Fig 9 shows part of the de ployment process which is just placing each pebble so that neighboring pebbles are within line of sight A total of seven pebbles were placed in a kitchen a living room a passage and a studio Fig 9 Deployment Process We took a pebble system and a first time user to a demonstration house and asked her to design navigation routes in this en
7. dy as well as uncover further research challeng es Ideas presented in this paper may be elaborated by an additional user study of the system 7 Conclusion We proposed a tangible user interface for designing navigation routes for robots Our method is designed for non expert users and solves problems associated with existing methods such as the technical complexity of labeling invasion in unwanted areas and complications for reconfiguration Our user study showed that the proposed de vices are easy to use and provide users with flexibility where planning customized navigation routes in simulated home and office environments We believe that robots as autonomous mobility entities improve convenience in home and office environments by providing more functionality such as monitoring and delivery but it is difficult to realize completely autonomous robots In this paper we demonstrated that a little help from the user improves the functionality of the ro bot as well as the naturalness in specifying locations We also assert that besides improving performance and functionality improving usability and other human fac tors is an important topic when using robots with humans Acknowledgments We thank Akihiko Ishida and Greg Saul for their valuable sug gestion in the initial stage of this work and their help in implementing an earlier ver sion of the proposed system We also thank Youichi Kamiyama for his help in im plementing the propos
8. ecked the possible trails by listening to the voice feedback We conclude that LED feedback could be improved in future revisions Two participants gave negative feedback for the entire system One 31f argued it seems too costly to maintain In the current implementation pebbles operate on battery power and require battery replacement after a certain period of time Since most pebbles with the exception of dynamic targets are placed at fixed locations for a long time we plan to revise the hardware so that power may be supplied from an outlet The other 26f did not want to have to think where to place them to construct a connected network We however think that deploying devices appropriately should remain the user s responsibility A user can have all the advantages of using pebbles such as explicitly marking areas to avoid and labeling locations by the small work of ensuring a direct line of sight between each pebble which is difficult for the system to do automatically Some people may prefer an effort saving method but others may prefer a more customizable method such as the proposed method Our proposed method provides yet another way to realize environment configuration for robots Most of the participants 26f 28m 30f 31f 50f and 62f mentioned it would be better if a pebble was smaller In the current implementation infrared communica tion reaches to at least 5m and this may be more than needed for the ho
9. ed localization sys tem using coded seamless patterns 15 Similar indoor localization systems have been subsequently proposed using invisible markers installed on the ceiling 16 These absolute positioning techniques can be used to determine robot position but navigating in the environment also requires environmental knowledge in particular map information Typically registering such map information requires expert knowledge and is not easy to reorganize after registration On the other hand relative positioning by devices emitting beacon signals is an other method of robot navigation Similar to physical guides such as rails and lines the robot can trace signals to move among the devices These devices also work to understand environmental structure by detecting if the beacon signals reach others as expected There have been some work where a robotic system uses such beacon de vices to determine the environment 17 20 and we share the basic idea of this previ ous work However the previous work does not discuss if an average user can deploy such devices which is our focus here In contrast a robot places the devices 17 or robots themselves are the devices 18 20 The automatic mapping techniques represented by SLAM are widely used for ro bots and autonomous vehicles 1 Various sensors are used for observing an envi ronment such as sonar 1 21 laser 22 and vision 23 25 The robot constructs a map of the environment by e
10. ed system We also thank Thien Anh Tran and Gary Muialaret for their help in testing the proposed system and taking a video at the demonstration house References 10 11 12 13 14 Leonard J J Durrant Whyte H F Simultaneous Map Building and Localization for an Autonomous Mobile Robot In IEEE RSJ International Workshop on Intelligent Robots and Systems vol 3 pp 1442 1447 1991 Shiomi M Sakamoto D Kanda T Ishi C T Ishiguro H Hagita N A Semi autonomous Communication Robot A Field Trial at a Train Station In ACM IEEE Annual Conference on Human Robot Interaction pp 303 310 2008 Ishu K Takeoka Y Inami M Igarashi T Drag and Drop Interface for Registration Free Object Delivery In IEEE International Symposium on Robot and Human Interactive Communication pp 228 233 2010 Park S Hashimoto S Indoor localization for autonomous mobile robot based on passive RFID In IEEE International Conference on Robotics and Biomimetics pp 1856 1861 2009 Park S Hashimoto S Autonomous Mobile Robot Navigation Using Passive RFID in Indoor Environment IEEE Transactions on Industrial Electronics vol 56 no 7 pp 2366 2373 2009 Mi H Ishi K Ma L Laokulrat N Inami M Igarashi T Pebbles An Interactive Configuration Tool for Indoor Robot Navigation In Annual ACM Symposium on User Interface Software and Technology Demonstrations pp 11
11. for a specific location after constructing a map However the internal represen tation of the constructed map is typically a set of coordinates which may be too com plicated for non expert users to associate with a specific named location such as kitchen We call this name association labeling In addition in such a system the robot may enter into areas where the user does not want the robot to enter On the other hand there are several implementations that fall into the category of environmental configuration by humans Physical guides such as rails for trolley robots and lines for line following robots are the simplest ways to realize such envi ronmental navigation support A robot can reach target locations by tracing these environmental guides Most other techniques employ absolute positioning because odometer based positioning has the fundamental limitation of accumulating position ing errors Some directly measure a robot s position using environmental sensors 2 3 while some other employ robot readable devices or tags that work as landmarks that allow the robot to determine its position in the environment 4 5 For a robot to navigate in these environments the map of the entire environment is required and is typically provided by the person who manually calibrates and installs sensors devic es or tags Combining this manual registration of map information with manual label ing of locations enables a user to naturally
12. has a sound speaker for voice feedback and a roof rack for delivering objects Fig 7 Roof Rack iRobot Create Fig 7 Robot Platform The robot has the same hardware as a pebble for receiving infrared signals The robot also has a sound speaker for voice feedback and a roof rack for delivering objects The closer a neighbor pebble is placed the more receivers out of eight receive sig nals Using this characteristic the robot roughly estimates its proximity to the target pebble by monitoring the number of receivers receiving signals from the target peb ble The robot has a bumper switch to detect collision and will change its direction if it detects obstacles including pebbles Because the bumper switch is weaker than a pebble s static friction the robot does not push the pebble to a new location when the robot bumps into it In this way the robot in the current implementation can navigate past objects on the floor Note that obstacle avoidance is beyond the topic of this pa per and if the robot is equipped with some mechanisms to detect an obstacle before collision such as a laser range finder the robot could avoid obstacles without colli sion The robot may happen to go out of line of sight of the network of pebbles In such a case the robot moves around in the local area to search for infrared signals as well as searches for another path to the final goal If it finds the current target pebble the robot resumes the naviga
13. me and office environment We thus think that with careful performance considerations the device could be made smaller We think the participants would be satisfied if we eventually made them small enough to be attached to home or office furniture One participant 29m commented that he would like to call the robot using a mo bile phone This is possible with our current implementation When a pebble is near the user the user just tells the system the ID of the pebble using a mobile phone If the user is taking a pebble as a dynamic target as described in Section 3 2 any place could be the robot s goal As an alternative approach a mobile phone that comes with infrared transmitters could work as a pebble to call the robot 6 Discussion and Future Work The proposed method provides a simple way for manually configuring an environ ment for robots in particular describing the structure of the environment and naming specific locations This is in a manner of speaking the technique of Simultaneous Labeling and Mapping Our method is one option for users having preferences as described in this paper and we do not aim to completely replace existing methods with the proposed method If a user does not care about the drawbacks of existing methods for mapping the environment those techniques can be used to achieve more sophisticated navigation For example in the case where the user can receive daily expert assistance more sensors ca
14. n be installed to achieve finer robot localization In the case where the user does not mind the robot exploring the environment automatic mapping helps the robot to perform more efficient navigation Conversely labeling is fundamentally a manual process and not well fitted for autonomous mapping There fore the proposed technique can be used to add the labels to locations in the already mapped environment Although we proposed a simple way of mapping and labeling non expert but pro active users may prefer more flexible solutions Seeking out different possibilities and providing an adequate degree of flexibility for each user s preference is a continuing topic of interest for us One direction is to use text data for labeling Inputting text may be a little more complicated than attaching physical labels on buttons but it still does not require expert knowledge If the system stores the text of the location names they can be used to specify locations in various forms such as typing drop down menu and voice input We are now developing a smartphone interface for label regis tration based on text data Fig 12 Fig 12 Labeling by Text Data We are developing a smartphone interface to label locations Labels based on text data can be seamlessly used to specify locations by typing drop down menu or voice input Another direction is to conduct a longitudinal study that could possibly reveal more practical limitations than our user stu
15. ng the system Task 1 uses a room and a passage Fig 11 a The two target locations do not have a direct view between each other and thus the participant needs to place pebbles at transit points After Task 1 the experi menter asks about the system to determine the understanding of the participant and demonstrates full function of the system The experimenter also accepts questions from the participant The second task Task 2 is intended to test the pebble system in a more complicated environment with four different room spaces and a passage Fig 11 b KA oO O target location Meeting Room a Task 1 b Task 2 Fig 11 Experimental Configuration We use a room and a passage for Task 1 and four rooms and a passage for Task 2 In both studies the participant is asked to think aloud so that the experimenter can observe the participant s thoughts The experimenter places the robot in the environ ment and hands out nine pebbles that are switched off 5 2 Detailed Procedure In Task 1 the participant is instructed to read the user manual After the participant finishes reading the manual the participant is asked to configure the environment where the robot can navigate toward two target locations using pebbles The user manual mentions how to use LED and voice feedback but it does not instruct or en courage the participant to use the feedback After deployment the participant is asked t
16. o direct the robot to go to two targets using two different methods by the remote control and by a pebble button The participant is allowed to adjust the positions of pebbles in case the robot cannot arrive at the target If the participant has difficulty completing the task the experimenter regards it as the participant could not under stand the system and provides necessary help In Task 2 the participant is asked to design a network for the robot with four target locations The participant is allowed to revise their deployment After deployment the experimenter checks to see if the network is constructed so that the robot can go to all four of the locations After that the participant is asked to move the robot between some of the target locations given some scenario such as A wife in the kitchen wants to deliver snacks and coffee to her husband in the living room After Task 2 we conduct a semi structured interview with the participant to ask opinions about their experiences using the system and how the participant thinks the system should be improved 5 3 Participants Eight non expert participants five female and three male with ages ranging from 26 to 62 26 year old female denoted as 26f 28 year old male denoted as 28m 29m 29m 30f 31f 50f and 62f were invited to participate The participants received 3000 yen roughly 25 dollars for taking part in the user study 5 4 Results Task1 Six out of eight participan
17. ofa or dynamic like father 3 3 Specifying a Goal After the navigation routes have been appropriately designed the user can instruct the robot to move to any pebble within the network A basic way to specify a goal is by pressing a button on the remote control Fig 6 left The application services of the robot will request the user to specify a goal when required to proceed with the task The user can also specify a goal by pressing the button on the pebble Fig 6 right This is useful when the user want to call the robot or describe a series of target loca tions by traveling to the locations If the button of the pebble is pressed LEDs on the pebble blink more slowly than when transmitting and receiving the signals which is described Section 3 2 The robot will reset the button press state after the task is com pleted or the user can cancel the button press state by pressing the button again Fig 6 Two Ways to Specify a Goal The user can specify a goal of the robot by pressing a button on the remote control attached to the robot left The user can also specify a goal by pressing the button of the pebble right 3 4 Robot Navigation The robot navigates by tracing infrared signals from pebbles Therefore the robot has the same hardware as a pebble for receiving infrared signals Fig 7 Our prototype employs the robotic platform iRobot Create the developer s version of iRobot Room ba vacuum cleaner The robot
18. pants reported that it was not much more difficult to plan more complicat ed navigation trails than the ones given in Task 1 After the deployment eight partici pants tested a total of 21 navigation trails In 17 of the tested navigation trails the robot successfully navigated to the specified place without the participant s help In the other four navigation trails the participants first experienced unsuccessful naviga tions caused by loss of the target pebble location After adjusting the pebbles the robot completed the navigation in all four trials We observed the participants tendency to reposition the navigation landmarks several times to find more efficient routes for example when trying to reduce the number of pebble landmarks We observed a participant 28m who tried to extend the navigation network to an additional room that was not required in Task 2 These observations provide evidence to support our claims that a pebble system has the ad vantage of being reconfigurable which provides users more freedom when designing navigation routes Interview We conducted an interview with each participant All participants were asked their opinions about the feedback Overall we obtained positive comments regarding the feedback that assists deployment and seven out of eight participants gave a more positive evaluation of the voice feedback One participant 30f answered that she never watched the LED indicators instead she always ch
19. s pebbles at desired locations Pebbles automatically construct a network topology After deployment the robot can travel to any pebble in the net work 1 Introduction Robots can be used for various applications in both home and office environments For example a robot can travel to predetermined locations at a specific time to collect environmental data Robots can also deliver a physical object such as a cup of coffee to a colleague upon a user s request If a robot moves an electric fan closer to a per son the fan can work more energy efficiently However it is difficult for a robot to know how to reach a specific target location in an unknown environment Meanwhile regarding user interaction it is preferable to specify target locations using natural expressions such as names of the locations In this study we propose a method that allows a non expert end user to provide navigation information to a robot and to natu rally specify locations While there are many ways to characterize existing methods to configure an envi ronment for robot navigation we divide the existing methods into two approaches automatic mapping by robots and environmental configuration by humans Automatic mapping by robots represented by simultaneous localization and mapping SLAM 1 is an effort saving method from the user s point of view The robot acquires map information by automatically exploring the environment and the robot can plan navi gation
20. sponsi bility The system provides LED and voice feedback to help the user to ensure pebble connectivity The user can see LEDs on the top of each pebble each of which indicates its con nection status to another pebble in a specific direction Fig 3 When transmitting and receiving infrared signals the indicator LEDs of corresponding channels synchro nously blink to indicate adjacent relationships The user can see if a pebble is visible to its neighbor pebbles In this way the user knows if two pebbles are placed too far apart or the path directly between them is blocked by some obstacles The user can then correct the problem Fig 3 LED Feedback When transmitting and receiving infrared signals the indicator LEDs of corresponding channels synchronously blink to indicate adjacent relationships to the user The LEDs of the transmitting pebble left indicate the transmission of the signal to all directions The LEDs of the receiving pebble right indicate the reception of the signal in the direction of the transmitting pebble In addition to the LED feedback the system also provides voice feedback The ro bot enumerates the IDs of pebbles in the network for example I can locate number one number three and number five The user then knows which pebbles are not visible to the robot and can identify which connection is not working as expected If the robot does not receive any signals from pebbles it says I cannot see any
21. tery set All components are packed in a plastic cylinder case 10 cm in diameter and 10 cm in height Fig 2 Each pebble has a unique ID that distinguishes it within the network A button labeled with this ID sits on the top for user input Indicator ARD cis os ares a Nfrared gt _ Receivers ransmitters B tton Fig 2 Pebble Hardware A pebble has 24 infrared transmitters 8 receivers 8 indicator LEDs and a button The infrared transmitters and receivers are arranged to cover all directions The indicator LEDs are used to provide primitive feedback to the user The button is used to receive user input The infrared communication unit has eight infrared transceiver modules each one contains a red LED indicator infrared receiver and three infrared signal transmitters The modules are placed 45 degrees apart so that all eight modules achieve 360 degree coverage Infrared signals are modulated using a 38 kHz carrier a frequency typically used for remote controlled appliances to prevent errors caused by noise The indica tor LEDs on the top provide primitive feedback 3 2 Deploying Pebbles The robot plans its navigation based on topological information from the pebbles and moves toward a goal according to pebble signals Therefore each pebble must com municate with at least one other pebble and no pebble should be completely isolated To arrange for each pebble to communicate with its neighbors is the user s re
22. the places whose name will not be changed after initial installation such as kitchen entrance and meeting room Semi static targets are usually fixed locations but might be relocated occasionally such as dining ta ble sofa and the desk of Ms XXX Because the user might make a change to the layout of a room in home or office the location associated with the table sofa or desk may be changed after the installation The user can easily reconfigure the navi gation network to fit the new layout by relocating the associated pebble The user can also carry a pebble around to define dynamic targets such as father and Mr YYY For instance users could carry a pebble with them to continuously indicate their current locations When one user wants to send a robot to another user the send ing user can simply specify the goal by pressing the button associated with the receiv ing user Note that the sending user does not need to know where the receiving user is in this scenario Pebbles can deal with all three scenarios in the same manner In par ticular the ability for non experts to deal with semi static and dynamic targets is the feature that differentiates our method from existing methods Fig 5 Labeling by Physical Labels The user attaches physical labels on a remote control to name locations The labels of the targets can be static like kitchen and entrance semi static like s
23. tics In IEEE RSJ International Conference on Intelligent Ro bots and Systems pp 3659 3664 2009 Yap T N Shelton C R SLAM in Large Indoor Environments with Low Cost Noisy and Sparse Sonars In IEEE International Conference on Robotics and Automation pp 1395 1401 2009 Montemerlo M Thrun S Koller D Wegbreit B FastSLAM A Factored Solution to the Simultaneous Localization and Mapping Problem In National Conference on Artifi cial Intelligence pp 593 598 2002 Davison A J Murray D W Mobile Robot Localisation Using Active Vision In Euro pean Conference on Computer Vision vol 2 pp 809 825 1998 Hwang S Y Song J B Monocular Vision Based SLAM in Indoor Environment Using Corner Lamp and Door Features From Upward Looking Camera IEEE Transactions on Industrial Electronics vol 58 no 10 pp 4804 4812 2011 Pirker K Ruther M Bischof H Schweighofer G Mayer H An Omnidirectional Time of Flight Camera and its Application to Indoor SLAM International Conference on Control Automation Robotics and Vision pp 988 993 2010 Ishii H Ullmer B Tangible Bits Towards Seamless Interfaces between People Bits and Atoms In ACM SIGCHI Conference on Human Factors in Computer Systems pp 234 241 1997 Dijkstra E W A Note on Two Problems in Connexion with Graphs Numerische Mathe matik vol 1 no 1 pp 269 271 1959
24. tion If it finds a new path to the final goal the robot starts a new navigation If it cannot find the way it stops navigation and says Can you help me I m lost Based on the constructed network topology we apply Dijkstra s algorithm 27 with a uniform edge cost for finding the shortest path All pebbles directly visible to the robot are tested as a starting point of a path After finding the shortest path the robot moves to each pebble in the path one by one For example if the shortest path is 7 3 2 5 the robot first moves toward pebble No 7 After the robot reaches pebble No 7 the robot then moves toward pebble No 3 and the robot repeats this process to finally arrive at pebble No 5 One fundamental limitation with the current implementation is confusion caused by highly reflective surfaces such as metal coated walls Fig 8 a We observed that the robot receives infrared signals from all directions if the robot is in a corner made of these walls and there is a pebble nearby Fig 8 b In such a case the robot does not understand the direction of the target pebble and cannot move forward using a tracing strategy We implement a short random walk in the local area when receiving the target pebble s signals from all directions until some of channels do not receive the signals Although this problem can happen in an environment with highly reflec tive walls it is the case that the robot happens to move into a corner and we o
25. tion In addition our method explicitly prevents the robot entering into the area where the user does not want the robot to do because the robot moves by tracing the constructed network trails Although our method requires the user to place each pebble so that it is within line of sight of neighboring pebbles each pebble can provide LED feedback and the robot can provide voice feedback to confirm connection status None of these configuration processes require expert knowledge Our contribution is presenting a user configurable method for robot navigation ra ther than the system s algorithms communication technology or network construc tion LED and voice feedback helps the user to construct an appropriate network and we describe the interaction design between the user and the devices 2 Related Work Several positioning techniques have been proposed for indoor environment where GPS technology provides only limited performance Radio based fingerprinting 7 is one of the trends in indoor positioning in most cases the system does not require major changes to infrastructure because of the recent widespread presence of wireless LAN 8 Some other approaches calculate positions by installing sensors in the envi ronment 9 13 including a commercial system 13 Another popular approach is to utilize cameras fixed in the environment 14 Other approaches use passive tags to calculate position Saito et al implemented an indoor marker bas
26. ts correctly deployed pebbles at appropriate places to generate navigation trails and successfully navigated the robot to the specified tar gets One participant 28m did not understand that the pebbles connected as a net work He attempted to move the robot step by step using only one pebble He placed one pebble at a transit point to move the robot and then took the pebble to the next transit point Another participant 26f did not notice the visibility restriction and the pebble network was disconnected between the inside and outside of the meeting room We observed a participant 29m who tried to use pebbles in an unexpected man ner He wanted to call the robot so he placed a pebble on the table and pressed the button as if he was pressing a service bell However because elevated communication is not supported by the current implementation the pebbles on the floor did not re ceive the user request and the robot failed to come Another participant 31f attempt ed to place pebbles in inconspicuous places e g under chairs or at corners She ar gued that navigation landmarks should not disturb users daily activities and they should not be placed at the center of the corridors Task2 We confirmed using the voice feedback that all participants completed an appropriate landmark deployment without the experimenter s help During the de ployment all participants used the voice feedback feature to confirm navigation trails The partici
27. vironment After deployment we asked her to try using the robot We gave the scenario A wife in the kitchen wants to deliver a glass of juice to her husband in the studio Fig 10 shows stills from the video record of the navigation procedure She called the robot and directed it to deliver a glass of juice to the studio using functions derived by the pebble system The robot successfully navigated and delivered a glass of juice indi cating that the pebble system can provide environmental knowledge for robot naviga tion even in an unfamiliar environment a b c 2 h i Fig 10 Navigation Procedure The user in a kitchen called a robot by tapping a pebble button a The user poured juice into a glass while the robot was moving toward the pebble b c The user directed the robot to travel to a studio using a remote control d The robot autonomously traveled out of the kitchen e f through a living room g i a passage J l and the studio m The other user received the glass of juice n o In this way users were able to pass physical things between remote locations by employing the robot 5 User Study 5 1 Goals and Methods We designed a user study to test if the pebble system is understandable and usable for end users to configure an environment The user study involves two tasks The first task Task 1 is to investigate the clarity of our system The participants are given only a user guide describi
28. xploration and the robot can use the constructed map to navigate toward a specific location in the map However a typical internal representa tion of the map is a set of coordinates and it is not trivial for a user to specify a desti nation or label a location within it In general this method requires another represen tation or visualization of the constructed map so that the user may specify a point in the map This paper deals with the issue of providing a user understandable represen tation as described in the next paragraph In the proposed method tangible user interface 26 is a key concept for providing a user understandable representation of the environmental map Placing a tangible device at each destination bridges the specific location in the user s physical world and the specific node in the system s internal topology map In our prototype system the internal representation of the environmental map is a network topology but the user does not need to know about it The user only has to specify and label a device at desired locations in the physical world and the system determines the steps to reach the devices from the network topology 3 Pebbles 3 1 Device Hardware A pebble is a self contained device that is capable of communicating with other peb bles Pebbles construct a network topology to generate navigation trails for the robot Each pebble consists of a microcontroller unit infrared communication unit and bat
Download Pdf Manuals
Related Search