Lego Mindstorms Robots as a Platform for Teaching Reinforcement Learning

摘要

This paper examines the suitability of LegoMindstorms robotic kits as a platform for teaching theconcepts of reinforcement learning. The reinforcementlearning algorithm Sarsa was implemented on board anautonomous Mindstorms robot, and applied to twolearning tasks. The reasons behind the differing resultsobtained on these two tasks are discussed, and severalissues related to the suitability of Mindstorms as aplatform for teaching or experimentation withreinforcement learning are identified.