Active Disturbance Rejection Control of Permanent Magnet Linear Synchronous Motor Based on TD3 Algorithm

This study investigates the controller design for permanent magnet synchronous linear motor (PMSLM) under variable load disturbances and internal parameter uncertainties, while applying deep reinforcement learning to optimize the controller parameters. We propose a TD3-ADRC strategy that integrates nonlinear active disturbance rejection control (ADRC) with the twin delayed deep deterministic policy gradient (TD3) algorithm. The nonlinear ADRC method, while effective, presents challenges due to its numerous parameters that are difficult to tune manually and exhibit coupling effects. The TD3 algorithm addresses these limitations by handling large-scale, highdimensional, and uncertain optimization problems, enabling effective multi-parameter optimization. Comparative simulation experiments demonstrate the performance of the proposed method over particle swarm optimization (PSO), genetic algorithm (GA) and differential evolution (DE), showing improved tracking accuracy and enhanced disturbance rejection capabilities.