Cooperative behavior acquisition mechanism for a multi-robot system based on reinforcement learning in continuous space

摘要

This paper describes an approach to controlling an autonomous multi-robot system. One of the most important issues for this type of system is how to design an on-line autonomous behavior acquisition mechanism which is capable of developing each robot's role in an embedded environment. Our approach is applying reinforcement learning that uses Bayesian discrimination method for segmenting the continuous state and action spaces simultaneously. In addition to this, neural networks are provided for predicting the other robots' moves at the next time step in order to support the learning in a dynamic environment that originates from the other learning robots. The output signals are utilized as the sensory information for the reinforcement learning to increase the stability of the learning problem. A homogeneous multi-robot system is built for evaluation.