Integral Reinforcement Learning-Based Optimal Control for Nonzero-Sum Games of Multi-Player Input-Constrained Nonlinear Systems

Abstract

This paper investigates an integral reinforcement learning (IRL)-based optimal control scheme to solve nonzero-sum games of multi-player input-constrained nonlinear systems with unknown drift dynamics. The IRL method is introduced to obviate the identification procedure of the unknown drift dynamics. In order to achieve Nash equilibrium for each player, the simplified optimal control policy for each player is obtained by solving the coupled Hamilton-Jacobi equation with the critic neural network only. Thus, the computational resource is saved and the control structure is easy to realize. The closed-loop system is guaranteed to be uniformly ultimately bounded based on the Lyapunov stability analysis. Simulation results illustrate the effectiveness of the developed control scheme.