Adaptive Optimal Prescribed Performance Solutions for Multi-player Systems with Tracking Error Constraints

This paper presented an RL-based robust controller design for the unknown servo system with disturbance. The system was identified with a three-layer NN. The robust controller is designed based on a zero-sum game theory, where the worst disturbance is studied to make robust control with certain anti-interference redundancy. A novel gradient law is developed to learn the observer NN parameter. Moreover, the identifier convergence and the stability of the robust controller are proved. Finally, the servo model is simulated with the proposed robust controller. The theorem and the conclusions indicate that the proposed RL-based robust controller can provide a good performance of the servo system.