Reinforcement Learning Control of Hydraulic Servo System Based on TD3 Algorithm

This paper aims at the characteristics of nonlinear, time-varying and parameter coupling in a hydraulic servo system. An intelligent control method is designed that uses self-learning without model or prior knowledge, order to achieve certain effects. The quantity can be obtained current moment through continuous iteration strategy–value network, online self-tuning parameters realized. Taking system as experimental object, twin delayed deep deterministic (TD3) policy gradient was used reinforce learning Additionally, setting compared using (DDPG) linear–quadratic–Gaussian (LQG) based on linear quadratic Gaussian objective function. To compile reinforcement algorithm deploy it test platform controller for testing, we Speedgoat prototype target machine build fast platform. MATLAB/Coder compute unified device architecture (CUDA) were generate an S-function. results show that, with other tuning methods, proposed effectively optimize improve dynamic response when tracking signals.