From RoboLab to Aibo - Computer and Information Science
Contents
1. 1 1 2 E E 2 2k 2k ok 2k ok DRO OC GO AK k kkk k k walk forward for a specified distance in mm FRA A AI I A A a 1 2k 21 1 3 3 2k 2k 24 2 2 ok 2k 2k 2k 2k ok ak ak Forward pass a distance ARG distance PLAY ACTION WALK 0 distance RETURN PRB OO GAR kk kk k wait for a specified amount of time FRA A IO AAI I A A a 1 2k 1 24 3 3 2k k 2k 2K 2k kkk Wait_For_Time ARG time WAIT time RETURN DROBO OOO GO A a kkk k plays a specified sound FRA A ooo I A a Ik I 2K 3 2 2k 24 kkk kkk Play_Sound ARG sound PLAY MWCID sound RETURN GOO AK A ak kk kk plays a specified led pattern FRA A oo A I A A a 2k 2k 1 24 3 3 2k 2k 24 2 6 ok 2k 2k 2K kkk Display_LED ARG pattern PLAY MWCID pattern RETURN R CCODE Fig 5 Example This program will test whether the head sensor is pressed or not If the head sensor is released it will play sound number 3 otherwise if the head sensor is pressed it will display LED pattern number 2 The Aibo will then move forward 100mm and wait half a second This process will repeat while the distance sensor reads a value greater than 100mm 4 Summary We have presented the design and implementation of our prototype framework for providing a simple graphical behavior based interface to the Sony Aibo robot Built into the popular RoboLab graphical programming environme2. 4 Sklar E Parsons 5 RoboCupJunior a vehicle for enhancing technical literacy In Proceedings of the AAAI 02 Mobile Robot Workshop 2002 Papert S Situating constructionism Constructionism 1991 6 Slavin R When and why does cooperative learning increase achievement theoret ical and empirical perspectives In Hertz Lazarowitz R Miller N eds Interac tion in cooperative groups The theoretical anatomy of group learning Cambridge University Press 1992 145 173 7 Sklar E Eguchi A Johnson J RoboCupJunior Learning with Educational Robotics In Proceedings of RoboCup 2002 Robot Soccer World Cup VI Springer Verlag 2002 238 253 8 Goldman R Eguchi A Sklar E Using educational robotics to engage inner city students with technology In Kafai Y Sandoval W Enyedy N Nixon A S Herrera F eds Proceedings of the Sixth International Conference of the Learning Sciences ICLS 2004 214 221 9 Sklar E Eguchi RoboCupJunior Four Year Later In Proceedings of RoboCup 2004 Robot Soccer World Cup VIII Springer 2004 10 Sklar E Parsons S Stone P RoboCup in Higher Education A preliminary report In Proceedings of RoboCup 2003 Robot Soccer World Cup VII Springer Verlag 2003 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Tufts University RoboLab http www ceeo tufts edu robolabat
3. RoboLab As part of this process we analyzed the main differences between the Aibo platform and the LEGO Mindstorms platform to ensure that our be haviors are suitable for both The two platforms are physically quite different not only in terms of processor configuration but also in regard to the types of sensors and effectors provided The LEGO Mindstorms is typically constructed as a wheeled robot as depicted in Figure 2b though legged structures can be built The kit comes with two touch sensors and a light sensor and has the ability to support numerous other LEGO and commercial sensors The Aibo a four legged robot comes with a variety of built in sensors including multiple touch sensors distance sensors and a camera and it cannot support any other sensors without being dismantled With the capabilities of both platforms in mind we defined a set of proto type behaviors and control structures suitable for our behavior based palette in RoboLab illustrated in Figure 3 RoboLab is implemented on top of Na tional Instruments LabVIEW 24 Individual icons and full programs are saved as VIs virtual instruments Each icon or program can be seen as imitating an actual instrument 25 We used RoboLab s built in feature to create subVIs VI modules or subroutines in order to construct our customized behaviors These behaviors act as macros for sets of lower level RoboLab commands Al though the ability to expand R
4. Yacc greatly simplify compiler writing or translating between two programming language representations Lex generates the C code for build ing a lexical analyzer or lexer Any given lexer takes an arbitrary input stream and divides it into a sequence of tokens based on a set of regular expression patterns The Lex specification refers to the set of regular expressions that Lex matches against the input A deterministic finite state automaton generated by Lex performs the recognition of the expressions Lex allows for ambiguous spec ifications and will always choose the longest match at each input point Each time one of the patterns is matched the lexer invokes user specified C code to perform some action with the matched token Yacc is responsible for generat ing C code for a syntax analyzer or a parser Yacc uses the grammar rules to recognize syntactically valid inputs and to create a syntax tree from the corre sponding lexer tokens A syntax tree imposes a hierarchical structure on tokens by taking into account elements such as operator precedence and associativity When one of the rules has been recognized then the user provided code for this an action is invoked Yacc generates a bottom up parser based on shift reduce parsing When there is a conflict Yacc has a set of default actions For a shift reduce conflict Yacc will shift For reduce reduce conflicts Yacc will use the earlier rule in the specification In our case
5. Yacc reads the grammar description and the token declara tions from lasm2aibo y and generates a parser function yyparse in the file y tab c Running Yacc with the d option causes Yacc to generate definitions for the tokens in the file y tab h Lex generates a lexical analyzer function yylex in the file lex yy c by reading the pattern descriptions from lasm2aibo 1 and including the header file y tab h The lexer and parser are compiled and linked together to form the executable lasm2aibo From the main function in lasm2aibo the yyparse function is called which in turn calls the 1 function to obtain each token Our Yacc parser is generated from the lasm2aibo grammar specification Through a series of grammar productions the parser attempts to group se quences of tokens into syntactically correct statements Associated with each production is a set of semantic actions Given that many of the LASM com mand sequences for varying groups of behaviors are the same often behavior identification becomes part of the semantic analysis For instance all the mo tion behaviors including forward backward right and left produce the same LASM command output Furthermore there is no direct way to distinguish be tween activating an LED or a motor because they are both viewed as output devices Only by examining the values of some of the parameters and by mak ing certain assumptions can they be differentiated Having to perform specific
6. From RoboLab to Aibo A Behavior based Interface for Educational Robotics Rachel Goldman Azhar and Elizabeth Sklar 1 Google Inc 1440 Broadway New York NY 10018 USA rjg google com 2 Dept of Computer Science Graduate Center City University of New York 365 5th Avenue New York NY 10016 USA mazhar gc cuny edu 3 Dept of Computer and Information Science Brooklyn College City University of New York 2900 Bedford Avenue Brooklyn NY 11210 USA sklar sci brooklyn cuny edu Abstract This paper describes a framework designed to broaden the entry level for the use of sophisticated robots as educational platforms The goal is to create a low entry high ceiling programming environment that through a graphical behavior based interface allows inexperienced users to author control programs for the Sony Aibo four legged robot To accomplish this end we have extended the popular RoboLab application which is a simple icon based programming environment originally de signed to interface with the LEGO Mindstorms robot Our extension is in the form of a set of behavior icons that users select within RoboLab which are then converted to low level commands that can be executed directly on the Aibo Here we present the underlying technical aspects of our system and demonstrate its feasibility for use in a classroom 1 Introduction Many teachers have an interest in introducing robots into their classrooms for teaching a variet
7. behavior recognition during the semantic analysis is one of the limitations in volved in using LASM as our source program Ideally each behavior should be represented by a unique syntax allowing for quicker and cleaner parsing When a behavior is recognized the semantic actions involve calling the cor responding behavior function These pre defined behavior functions are used to generate the equivalent behavior in R CODE and help flag the R CODE sub routines that need to be included in the R CODE source file The final output of the translator includes the file R CODE R the R CODE program that matches the original RoboLab program and a behaviors txt file used for debugging purposes to ensure that behaviors sensors and control structures are appropri ately classified Running the lasm2aibo executable file with a LASM text file as input generates the file R CODE R This file should be placed on a R CODE ready memory stick in the OPEN R APP PC AMS directory 28 To execute the program insert the memory into Aibo turn on the power and watch Aibo come to life Throughout the process there are multiple places where errors could occur this includes lexical syntactic and semantic errors The lasm2aibo translator attempts to handle errors gracefully If the error is fatal a blank R CODE R file will be generated however if the error is non fatal then the translator will continue to process the source file In both cases a descriptive error mes
8. ceeo cessed January 16 2006 LEGO Mindstorms robotics invention kit http www legomindstorms com accessed February 1 2006 Goldman R From RoboLab to Aibo Capturing Agent Behavior Master s thesis Department of Computer Science Columbia University 2005 Chu K H Goldman R Sklar E Roboxap an agent based educational robotics simulator In Agent based Systems for Human Learning Workshop at AAMAS 2005 2005 Azhar M Q Goldman R Sklar E An agent oriented behavior based interface framework for educationa robotics In Agent Based Systems for Human Learning ABSHL Workshop at Autonomous Agents and MultiAgent Systems AAMAS 2006 2006 Sony AIBO http www us aibo com accessed January 16 2006 OPEN R SDE http openr aibo com accessed January 16 2006 Serra F Baillie J C Aibo Programming Using OPEN R SDK Tutorial http www cs 1th se DAT125 docs tutorial_OPENR_ENSTA1 0 pdf 2003 R CODE SDK http openr aibo com openr eng no_perm faq_rcode php4 accessed January 16 2006 YART Yet Another R CODE Tool http www aibohack com rcode yart htm accessed January 16 2006 Touretzky D S Tira Thompson E J Tekkotsu A framework for AIBO cogni tive robotics In Proceedings of the Twentieth National Conference on Artificial Intelligence AAAI 05 Menlo Park CA AAAI Press 2005 Baille J C URBI Towards A Universal Robotic Body Interface In Proceedings of the IEEE RSJ Internati
9. dvanced programming concepts and features to add power flexibility and complexity to programs Our experiences working with classroom teachers and young students have raised several issues that have motivated us to pursue the development of a behavior based interface which abstracts away the low level motor and sensor commands that often confuse inexperienced programmers or deter techno phobic students 13 Our longterm goal is to create a standard middle ground that can act as a sort of magic black box for current and future robotic platforms following several design criteria ease of use programmers only have to deal with high level icons novices will not get discouraged with low level text based syntax disappearing boundaries programmers are able to test and run the same behaviors on multiple agent platforms running the same RoboLab program on a LEGO robot and on an Aibo will help students understand about abstraction and behavior based control interoperability a standard behavior language is used for multiple platforms students do not need to learn different languages in order to use a variety of robot platforms or our simulator currently under development 14 15 and flexibility students from a wide range of backgrounds and teachers with a broad range of goals can use the system effectively accommodating different levels curricular needs academic subjects and physical environments for instructi
10. e and its lack of compilation however programmers have less control than with other lower level languages With only a few lines of R CODE users can program the Aibo to perform complex behaviors such as dancing kicking or walking Because R CODE does not require compilation it can be written in a plain text file on any operating system and saved directly on a memory stick that has the R CODE virtual machine pre loaded on it commands can be viewed as OPEN R macros where the degree of precision and control depends solely on the underlying OPEN R code Although R CODE is powered by OPEN R functions the developer does not have access to the underlying OPEN R subroutines R CODE is best suited for performing actions and various behavior sequences YART 20 is an R CODE front end developed by a hobbyist It provides a text based GUI with simple drag and drop functionality and produces files of R CODE commands This tool can also be used to generate customized behaviors via pre existing YART compatible Aibo personalities YART is an easy place to start programming simple behavior patterns However unlike RoboLab YART is a text based interface so the objects that the user drags with the mouse are bits of text whereas in RoboLab the user drags graphical icons Tekkotsu 21 developed at Carnegie Mellon University by Touretzky et al is an application development framework for intelligent robots that is compatible with the OPEN R f
11. he new state must be satisfied The low level functionality of the robot can be controlled using the OPEN R API 17 With the API programmers can experiment with image processing sensory feedback information and robot motion in order to develop original sets of behaviors for the Aibo However the OPEN R API can be quite hard to use for novice and even intermediate programmers As a result several interfaces and abstraction languages have been developed to sit on top of OPEN R in an attempt to hide the low level complexity from the inexperienced end user These include R CODE 19 YART Yet Another R CODE Tool 20 Tekkotsu 21 and URBI Universal Robotic Body Interface 22 23 4 C objects and OPEN R objects are not the same and should not be confused with each other 18 These languages interfaces vary in power and intricacy and each has its own goals and features as described briefly below OPEN R is the basis upon which R CODE Tekkotsu and URBI are built YART goes one step further abstract ing R CODE one level by providing a basic graphical user interface GUI to create and manipulate R CODE programs The R CODE SDK provides a set of tools that allow users to program the Aibo using the R CODE scripting language offering higher level commands than traditional programming languages such C and even OPEN R The benefits of R CODE being a scripting language are its simplicity to both learn and us
12. irections for future work 2 Background OPEN R is the programming interface designed for the Aibo provided for free via download from Sony s web site 17 OPEN R gives programmers the ability to develop software to control the low level hardware of the robot Some of Aibo ERS 7M2 B b LEGO Mindstorms Fig 2 Robot platforms the features of OPEN R include modularized hardware modularized software and networking support Because the hardware is modularized each module is connected by a high speed serial bus and can be exchanged for a new module Each software module in OPEN R is either a data object or a distributed object and is implemented in C OPEN R programs are built as a collection of concurrently running OPEN R objects which have the ability to communicate with each other via message passing Due to the modularity of the software individual objects can be easily replaced and each object is separately loaded from a Sony Memory Stick Furthermore OPEN R supports Wireless LAN and TCP IP network protocol Sony s formalism for describing the working cycle of OPEN R objects resem bles layered finite state automata 18 Each OPEN R object can have numerous states and must include an IDLE state At any given point an object can be in only one state and objects move from state to state using transitions In order to switch states an event must activate a transition to a new state and the pre condition of t
13. lt output from RoboLab is LASM LEGO Assembly Language We use this default saving the LASM commands in an output file whereas normally users send the LASM commands directly to the LEGO robot via the communication tower Once the LASM file is saved we invoke our trans lation program called lasm2aibo that converts the LASM commands into Aibo commands After examining the Aibo languages and interfaces discussed in sec tion 2 we decided to use R CODE to implement Aibo commands in our system because R CODE is easy to use and does not require compilation and because R CODE offers a more natural mapping to RoboLab With R CODE numer ous behaviors can be prototyped quickly and tested efficiently Both Tekkotsu and URBI are better suited for applications that require complex solutions and greater computational power A key challenge in designing lasm2aibo was to determine which LASM com mand s and what parameters are generated for each RoboLab icon Our be havior icons are macros comprised of multiple low level built in RoboLab icons which complicates the translation process as detailed below Taking a file of LASM commands as input our translater recognizes tokens that match relevant LASM commands numbers white space and delineators commas and new line characters and compiles or translates these into R CODE sequences We designed and implemented our translator using the UNIX tools Lex 26 and Yacc 27 Both Lex and
14. nt and connected to the Aibo via a translater lasm2aibo this framework represents the proof of concept for a longterm project aimed at bringing educational robotics into a broad range of classrooms The successful implementation of the direct translation from RoboLab to Aibo demonstrates the feasibility and viability of the process A user study is planned for Summer 2006 Although our direct translation is appropriate for small scale solutions cur rent work on this project is exploring a more abstract generalized framework for linking RoboLab or other graphical programming interfaces to a variety of robot platforms 15 Recent press releases have revealed that the LEGO Mind storms will be succeeded in August 2006 by a more sophisticated platform called NXT 29 and Sony has announced that production of Aibo halted in March 2006 30 Given the changing face of consumer robotics a flexible framework such as ours will help classrooms ease transitions from one robot platform to another by providing teachers and students with a familiar interface and an in tuitive behavior based methodology for programming no matter what hardware lies beneath References 1 Piaget J To Understand Is To Invent The Viking Press Inc New York 1972 2 Papert S Mindstorms Children Computers and Powerful Ideas BasicBooks 1980 3 Resnick M Technologies for lifelong kindergarten Educational Technology Re search and Development 46 4 1998
15. ntion System robot 12 RoboLab runs on a Mac Win dows or Unix computer The environment is highly visual and provides a good first experience with procedural programming concepts Entities such as motors and sensors are represented as rectangular icons on the screen and users drag and drop them with the mouse onto a canvas to create code The icons are strung together using wires and all programs are downloaded from RoboLab onto the LEGO robot via a communication tower connected to the computer s USB or serial port that transmits the program to the robot using an infra red signal A simple RoboLab program is illustrated in Figure 1 z sample vi Block Diagram ot Application forty Fig 1 A basic RoboLab program If the robot has wheels and motors attached to two of its ports labeled A and C this program will make the robot go forward for 2 seconds and then stop RoboLab s graphical programming environment tiers the levels of program ming which allow a novice to produce results quickly and acquire skills without having to read a sophisticated text or complicated application manual Ranging from Pilot to Inventor levels RoboLab is broad enough to ensure the suc cess of both beginners and advanced users The initial Pilot levels are completely graphical so even users who cannot read can be successful The more advanced Inventor levels incorporate a
16. oboLab s current set of icons is a remarkably powerful feature we are constrained by the necessity to use a set of pre defined icons as the underlying basis for each new icon For example as illustrated in Figure 4 our forward behavior icon is in reality a macro for the set of icons motor A forward motor C forward and wait for some amount of time The behaviors icons shown in the palette to the left can be used just like any other RoboLab icon As with basic RoboLab function icons wiring to gether a sequence of behavior icons creates a well formed RoboLab program The meanings of the first eight icons are from top left to right loop while back sensors are pressed loop while back sensors are not pressed branch according to state of back sensor move backwards begin behavior loop while distance sensor is less than or equal to parameter loop while distance sensor is greater than parameter branch according to state of dis tance sensor See 13 for a complete and detailed description of the behavior icons Fig 3 Behavior Icon Palette Macro Expansion Begin Distance fe ene Fig 4 Forward behavior icon and its underlying set of basic RoboLab icons The first step in our approach is for a user to write a program in RoboLab and save the contents of the program to a file in a manner that will preserve its functionality while allowing translation to Aibo commands to occur outside of RoboLab The defau
17. on Because of RoboLab s popularity in the classroom and its icon based pro gramming style it is well suited as the front end for our interface We have three underlying educational goals each supporting different pedagogical needs First we want to produce a low entry high ceiling interface to the Sony Aibo robot 16 pictured in Figure 2a that will encourage non traditional computer science students to learn programming allowing us to capitalize on the cuteness factor associated with Aibo while still providing students with a serious first adventure in programming Second we want advanced students to gain an appreciation of the modu larity of both a robot s controlling interface and its underlying hardware Just as Java and JavaScript are platform independent providing a robot programming environment that is platform independent will give students hands on experience with code abstraction witnessing the same code execut ing on both the LEGO robot and the Aibo and other platforms Third we want to create a motivating hands on environment with which to introduce more advanced students to behavior based concepts This paper is organized as follows We begin by reviewing several different programming interfaces designed for the Sony Aibo Then we describe some technical details about RoboLab and explain how our behavior based program mming interface operates Finally we end with a brief summary and mention d
18. onal Conference on Humanoid Robots Santa Monica CA USA 2004 Baille J C URBI Towards a Universal Robotic Low Level Programming Lan guage In Proceedings of the IEEE RSJ International Conference on Intelligent Robots and Systems Edmonton Canada 2005 National Instruments LabVIEW http www ni com labview accessed Jan uary 16 2006 LabVIEW User Manual http www ni com pdf manuals 320999e pdf ac cessed January 16 2006 Lesk M E Schmidt E Lex A Lexical Analysis Generator Bell Laboratories Murray Hill NJ 1975 Johnson S C Yacc Yet Another Compiler Compiler Bell Laboratories Murray Hill NJ 1975 T llez R R CODE SDK Tutorial v1 2 2004 LEGO Whats NXT LEGO Group Unveils LEGO MIND STORMS Robotics Toolset at Consumer Electronics Show http www lego com eng info default asp page pressdetail amp contentid 17278 amp countrycode 2057kyearcode karchive falsekbhcp 1 January 4 2006 Duffy J What happened to the Robot Age BBC News Magazine January 27 2006
19. ramework 17 It is based on an object oriented and event passing architecture that utilizes some of the key features of C like templates and inheritance Although Tekkotsu was originally created for the Sony Aibo the current version can be compiled for multiple operating systems Tekkotsu simplifies robotic application development by supplying basic visual processing forward and inverse kinematics solvers remote monitoring teleoperation tools and wireless networking support Tekkotsu handles low level tasks so the devel oper can focus on higher level programming It provides primitives for sensory processing smooth control effectors and event based communication Some of the higher level features include a hierarchical state machine formalism used for control flow management and an automatically maintained world map Addi tionally Tekkotsu includes various housekeeping and utility functions and tools to monitor various aspects of the robot s state The Universal Robotics Body Interface URBI 22 23 developed at Ecole Nationale Sup rieure de Techniques Avanc es ENSTA is an attempt to provide a standard way to control the low level aspects of robots while providing the high level capabilities of traditional programming languages URBI is based on a client server architecture where the server is running on the robot and is typically accessed by the client via TCP IP The client can be virtually any system or any other kind of compu
20. sage and the line number where the error occurred is displayed on the console via standard output 3 1 Example A detailed example follows First the RoboLab behavior based program shown in Figure 5a is constructed The resulting LASM is contained in Figure 5b Second our lasm2aibo module is executed taking the LASM file as input and creating an R CODE equivalent illustrated in Figure 5c This gets written to a Sony memory stick which also contains the R CODE virtual environment and is ready to be executed on the Aibo a RoboLab code delt 0 task 0 sent 0 senm 0 sent 1 senm 1 1 0 11002 chkl 2 100 1 9 0 Labe11003 chkl 2 1 0 9 1 1 11004 plays 2 wait 2 100 wait 2 0 jmpl 11005 11004 pwr 2 2 2 dir 0 2 out 2 2 wait 2 0 Label1005 pwr 1 2 100 dir 2 1 out pwr dir out wait 2 0 wait 2 50 ping jmpl 11002 Labe11003 endt 23 4 0 31 ne PPD BE 2 4 2 2 plays 5 b LASM code Start PLAY ACTION STAND WAIT WHILE 100 lt Distance IF 1 gt Head_ON THEN PUSH 2471 CALL Play_Sound 1 PUSH 1000 CALL Wait_For_Time 1 WAIT ELSE PUSH 2260 CALL Display_LED 1 WAIT ENDIF CLR SENSORS PUSH 100 CALL Forward 1 WAIT PUSH 500 CALL Wait_For_Time 1 WEND CLR SENSORS EXIT 2 2 E K K K k k kK K kk k k k kK kK kk k k K kK 3K K kK ka ok K aK kK K BEHAVIOR FUNCTIONS
21. ter thereby adding flexibility to URBI The URBI language is a high level scripting language capable of controlling the joints and accessing the sensors camera speakers or other hardware on the robot URBI has been primarily applied to entertainment robots because they tend to provide the most interesting interfaces and capabilities Each of these interfaces have informed our approach We selected RoboLab as the front end for several reasons RoboLab is already quite widely used in classrooms worldwide Teachers and students are comfortable with the interface and will not feel like they need to learn yet another programming environment in order to expand their use of robot platforms in the classroom In addition schools do not need to purchase another software package our methodology is a free extension to RoboLab Finally as detailed below RoboLab is constructed on top of a framework designed to support a wide range of hardware devices thus the concept of expanding its use to interface with a range of robot devices is a natural fit 3 Our Approach This section describes the approach to our implementation outlining the steps required for progressing from writing programs in RoboLab to generating code that is executed on the Aibo Given that one of our objectives with this work is to develop a behavior based programming interface for controlling Aibo we have designed a set of generic low level behaviors that can be graphically represented in
22. y of subjects other than specifically robotics from traditional technical topics such as programming and mechanical engineering to other areas such as mathematics and physical science where robots are used to demonstrate the concepts being taught It has long been recognized that hands on methods have a powerful impact on student learning and retention 1 3 The growing field of educational robotics the use of robots as a vehicle for teaching sub jects other than specifically robotics 4 employed that approach using the constructionist 5 process of designing building programming and debugging robots as well as collaboration and teamwork as powerful means of enlivening education 2 6 7 Even very young children have been successfully engaged in hands on learning experiences that expose them to some of the basic principles of science and engineering widening their horizons and preparing them for life in a highly automated technically challenging world For the past several years we have been designing and helping to implement educational robotics curriculum in inner city primary and middle school class rooms after school programs and summer schools undergraduate introductory programming courses and international robotic competitions 8 9 7 10 To sup port these curricula we have employed RoboLab 11 a widely used graphical programming environment developed at Tufts University for operation on the LEGO Mindstorms Inve
Download Pdf Manuals
Related Search