Model-based reinforcement learning for on-line feedback-Nash equilibrium solution of N-player nonzero-sum differential games

Abstract

This paper presents a concurrent learning-based actor-critic-identifier architecture to obtain an approximate feedback-Nash equilibrium solution to a deterministic, continuous-time, and infinite-horizon N-player nonzero-sum differential game on-line, without requiring persistence of excitation (PE), for non-linear control-affine systems. Convergence of the developed control policies to neighborhoods of the feedback-Nash equilibrium policies is established under a sufficient rank condition. Simulation results are presented to demonstrate the performance of the developed technique.