Multi-Robot Concurrent Learning in Museum Problem

Multi-robot concurrent learning on how to cooperatively work through the interac-tion with the environment is one of the ultimate goals in robotics and artificial in-telligence research. In this paper, we introduce a distributed multi-robot learning algorithm that integrates reinforcement learning and neural networks (weighting network). By retrieving continuous environment state and implicit feedback (re-ward), the robots can generate appropriate behaviors without deliberative hard coding. We test the learning algorithm in the “museum ” problem, in which robots collaboratively track moving targets. Simulation results demonstrate the efficacy of our learning algorithms. Key words