Lloyd Greenwald and Donovan Artz
While robot platforms have played a role in artificial intelligence and robotics education for over 30 years, the cost and size of these platforms have limited their reach. Recently, low-cost robot platforms have emerged, extending hands-on educational benefits to a diverse audience. In prior work, we present and discuss the construction and implementation of a course based around a series of detailed lab exercises using these platforms to tackle basic problems in computer science, artificial intelligence, robotics, and engineering. In that work we discuss the overall educational lessons and curricular themes that can be accomplished with these platforms. We observe that in that course, as in many similar courses, the extensive time spent on low-level engineering and computer science leaves little time for artificial intelligence education. In this paper we focus on the use of these platforms to achieve artificial intelligence education goals, assuming as pre-requisites basic engineering and computer science lessons. We first discuss the tradeoffs an educator must face when deciding to employ low-cost robots in artificial intelligence education, using localization as an example exercise. We then provide step-by-step instructions for using a Handy Board-based mobile robot kit to teach neural networks. We then extend this lesson to teaching Bayesian networks. These example exercises demonstrate that low-cost platforms have matured sufficiently to become a standard tool for teaching artificial intelligence and robotics to advanced undergraduate and beginning graduate students.