Reinforcement Learning for Adaptive Optimal Stationary Control of Linear Stochastic Systems

This article studies the adaptive optimal stationary control of continuous-time linear stochastic systems with both additive and multiplicative noises, using reinforcement learning techniques. Based on policy iteration, a novel off-policy algorithm, named optimistic least-squares-based is proposed, which able to find iteratively near-optimal policies problem directly from input/state data without explicitly identifying any system matrices, starting an initial admissible policy. The solutions given by proposed iteration are proved converge small neighborhood solution probability one, under mild conditions. application algorithm triple inverted pendulum example validates its feasibility effectiveness.