A robot laboratory for teaching artificial intelligence

There is a growing consensus among computer science faculty thatit is quite difficult to teach the introductory course onArtificial Intelligence well [4, 6]. In part this is because AIlacks a unified methodology, overlaps with many other disciplines,and involves a wide range of skills from very applied to quiteformal. In the funded project described here we have addressedthese problems by ” Offering a unifying theme that draws together the disparatetopics of AI; ” Focusing the course syllabus on the role AI plays in the corecomputer science curriculum; and ” Motivating the students to learn by using concrete, hands-onlaboratory exercises. Our approach is to conceive of topics in AI as robotics tasks.In the laboratory, students build their own robots and program themto accomplish the tasks. By constructing a physical entity inconjunction with the code to control it, students have a uniqueopportunity to directly tackle many central issues of computerscience including the interaction between hardware and software,space complexity in terms of the memory limitations of the robot’scontroller, and time complexity in terms of the speed of therobot’s action decisions. More importantly, the robot themeprovides a strong incentive towards learning because students wantto see their inventions succeed. This robot-centered approach is an extension of theagent-centered approach adopted by Russell and Norvig in theirrecent text book [11]. Taking the agent perspective, the problem ofAI is seen as describing and building agents that receiveperceptions as input and then output appropriate actions based onthem. As a result the study of AI centers around how best toimplement this mapping from perceptions to actions. The robotperspective takes this approach one step further; rather thanstudying software agents in a simulated environment, we embedphysical agents in the real world. This adds a dimension ofcomplexity as well as excitement to the AI course. The complexityhas to do with additional demands of learning robot buildingtechniques but can be overcome by the introduction of kits that areeasy to assemble. Additionally, they are lightweight, inexpensiveto maintain, programmable through the standard interfaces providedon most computers, and yet, offer sufficient extensibility tocreate and experiment with a wide range of agent behaviors. At thesame time, using robots also leads the students to an importantconclusion about scalability: the real world is very different froma simulated world, which has been a long standing criticism of manywell-known AI techniques. We proposed a plan to develop identical robot buildinglaboratories at both Bryn Mawr and Swarthmore Colleges that wouldallow us to integrate the construction of robots into ourintroductory AI courses. Furthermore, we hoped that theselaboratories would encourage our undergraduate students to pursuehonors theses and research projects dealing with the building ofphysical agents.