Abstract
We have been developing a reinforcement learning technique called Bayesian-discrimination-function-based Reinforcement Learning (BRL). This is recognized as an effective technique for autonomous specialization, which is a new concept for cooperative multi-robot systems. Homogeneous robots with BRL learn behavior to be heterogeneous so as to play different roles for developing cooperative behavior. In this research, BRL is applied to arm-type autonomous robots, the task of which is lifting an object without tilting it. Then, we investigate the autonomous specialization process by means of observing behavior transitions about their role assignment.
