Reinforcement Learning-Based Position Tracking Control for Proportional Directional Control Valve-Based Electro-Hydraulic System

This paper presents a reinforcement learning-based discrete-time control strategy for nonlinear electro-hydraulic actuator systems equipped with a proportional directional control valve. The proposed method incorporates a Lyapunov-based stability framework, ensuring bounded tracking errors and neural network weights despite modeling uncertainties and external disturbances. A discrete-time dynamic model of the electro-hydraulic actuator system is derived, and unmeasured system states are estimated using an extended Kalman filter. The controller is validated through both MATLAB simulations and real-time hardware experiments involving sensors, a hydraulic cylinder, a proportional directional control valve, and a data acquisition system. The obtained results demonstrate accurate trajectory tracking in the presence of modeling uncertainties, disturbances, and real-time adaptability of neural network weights, highlighting the effectiveness of the proposed controller for industrial electro-hydraulic systems.