Reinforcement Learning-Based Robust Adaptive Tracking Control for Multi-wheeled Mobile Robots Synchronization with Optimality

This paper proposes a new method based on reinforcement learning to design robust adaptive tracking control laws with optimality for multi-wheeled mobile robots synchronization in communication graph without requiring knowledge of drift tracking terms in node dynamics. Wheeled mobile robots are controlled by integrated kinematic and dynamic laws. Actor critic structures in the control scheme for every node is proposed such as only single NN is used to reduce computational cost and storage resources, but parameters of critic and actors are updated synchronously. Novel tuning laws for the NNs are designed not only to learn online adaptive solutions of cooperative Hamilton-Jacobi-Isaacs (HJI) equation on purpose of approximating optimal cooperative tracking performance index functions and robust direct adaptive tracking control inputs as well as worst case disturbances but also to guarantee closed-loop stability in real-time. The convergence and stability of the overall system are proven by Lyapunov techniques. The simulation results on multi-wheeled mobile robots systems verify the effectiveness of the proposed controller.