Multi-robot concurrent learning of fuzzy rules for cooperation

In this paper, a fuzzy logic based reinforcement learning algorithm is proposed for multi-robot concurrent learning of cooperative behaviors. In contrast to traditional reinforcement learning that can only learn discrete and finite behaviors, the proposed fuzzy reinforcement learning controller can generate continuous and infinite behaviors by learning the optimal fuzzy control rules. Furthermore, a distributed learning coordination algorithm is proposed to eliminate the interference among learning robots. The strategy is to control the learning speed according to the progress of learning. The fuzzy learning controller is applied to multi-robot tracking of multiple moving targets. Simulation results demonstrate the efficacy of the learning controller.