Q-learning Optimized Active Disturbance Rejection Control for Fast Tool Servo Systems

Aiming at the issue of insufficient control accuracy caused by time-varying cutting force disturbance in ultra-precision machining process of fast tool servo (FTS), an active disturbance rejection control (ADRC) algorithm based on Q-learning is proposed. Firstly, the structure of the linear active disturbance rejection controller (LADRC) is designed based on the second-order nonlinear dynamic model of FTS. Subsequently, within the active disturbance rejection control framework, the Q-learning algorithm is employed to rapidly and adaptively tune the controller parameters. Finally, the effectiveness of the proposed method is verified by simulation. The simulation results show that compared with the traditional LADRC and sliding mode control (SMC), the proposed method (Q-LADRC) achieves a maximum tracking error of 0.01891 micrometers for a sine wave trajectory with 5-micrometer amplitude and 100 Hz frequency, which is about 26.1 % of LADRC and 43.4% of SMC. The root-mean-square error (RMSE) of 0.01083 micrometers is obtained, which is about 22.0% of LADRC and 36.7% of SMC. Under the step disturbance of 1-micrometer amplitude, the anti-interference recovery time of Q-LADRC is 16.7% and 93.8% shorter than that of LADRC and SMC, respectively. This method overcomes the reliance on manual experience in traditional parameter tuning, significantly enhancing the system's robustness against uncertain disturbances and its trajectory tracking accuracy.