Abstract
This article is devoted to optimal adaptive control for a distributed parameter convection‐reaction system by reinforcement learning (RL) strategy. The goal is to develop an RL‐based control technique aiming to learn the solution of the involved matrix Riccati differential equation online. Initially, we present a policy iteration (PI) algorithm that solves the optimal control problem online without utilizing any knowledge about the system dynamics, unlike traditional control methods, which often rely on full system dynamics, making them impractical for distributed parameter systems. Furthermore, the convergence of the iterative algorithm is validated under the condition that the starting policy is stabilizing. Additionally, an improved version is introduced with the objective of making the RL algorithm implementable from a realistic standpoint. Numerical simulations are performed to show the effectiveness of the developed algorithm. The proposed method provides a reliable and implementable approach for adaptive control in convection‐reaction systems, extending RL techniques to distributed parameter systems.
Funding Information
  • Qatar University (CDIRCC‐2023‐112)