Integral Reinforcement Learning‐Based H∞$$ {H}_{\infty } $$ Optimal Tracking Control for Continuous‐Time Uncertain Systems With Application to Pulping Process

ABSTRACT This paper addresses the issue of optimal tracking control for continuous‐time uncertain systems, where a model‐data dual‐driven control algorithm is proposed. First, a model‐based robust controller is obtained by using the limited information on the uncertainty mainly to stabilize the system and obtain a certain tracking ability. Then, an integral reinforcement learning‐based optimal controller is developed to improve the tracking ability by learning the solution of the game algebraic Riccati equation online from the collected system operation data. This proposed dual‐driven approach can make full use of both the uncertain model and operation data to compensate for the model uncertainty. The stability of the discounted tracking error system is analyzed in detail. Finally, the proposed model‐data dual‐driven control approach is applied to the high‐consistency pulping process with model uncertainties and multiple disturbances. Numerical simulations verify the effectiveness of the proposed optimal control strategies.