Model Free Difference Feedback Control of Stochastic Systems

This paper presents a model-free reinforcement learning (RL) algorithm for the output-difference feedback control (ODFC) of stochastic systems with measurement and process noise. A policy iteration algorithm is presented using a quadratic output-difference reward function to learn the optimal ODFC from noisy output difference measurements. It is proved that the proposed method trains stable dynamic control laws that approach the analytical optimal solution, even in the presence of process and measurement noise. Simulation results are presented to demonstrate the effectiveness of the proposed scheme.