Adaptive role development in a homogeneous connected robot group

A distributed autonomous approach to adaptive system design is investigated through a cooperative carrying problem (CCP), in which a homogeneous group of autonomous moving robots have to cooperate in carrying a load. A robot´s action generator is designed using two reinforcement learning (RL) units in order to cope with the dynamics in the CCP. The employed strategy is that each robot is learning its behavior as if it were in a static environment, by expanding the state space of the main RL unit using the prediction information of the nearest future that is generated by the supplementary RL unit. Computer simulations with two robots are conducted in order to illustrate the effectiveness of the proposed approach. The cooperative behaviors, which is adaptive role development, through autonomous task decomposition is observed in the achieved stable behavior