The Formation Control of a team of Manipulators Based on Reinforcement Learning and Sliding Mode Control

In this article, sliding mode control (SMC) and reinforcement learning (RL) approaches is adopted to develop the optimized leader-follower formation control for a group of manipulators. The SMC architecture composed of communication interaction among the robots is designed to improve the robustness of multi-robot systems. To achieve optimal formation control performance, identifier-actor-critic mechanism of the reinforcement learning approach is proposed. This proposed method has two main advantages: i) the RL updating laws to train the actor and critic networks more simpler; ii) this method can smoothly steer multiple agent system to achieve the leader-follower formation. The effectiveness of this control method is demonstrated by theory and simulation.