Constant-Force Control of a Lead-Screw-Driven Polishing Actuator Using Multi-Scenario PSO Tuning

Abstract

This study addresses contact-force fluctuations in robotic polishing caused by local surface protrusions, variations in contact parameters, and force-sensor noise. A reduced-order force-actuator model is established by combining a closed inner speed loop, lead-screw transmission, and unilateral contact dynamics. A multi-scenario particle swarm optimization (PSO) objective is then used to tune a PID force controller and compare it with fixed-gain PID and ANFIS baselines under matched numerical conditions. To address optimizer-setting dependence and derivative-link noise sensitivity, additional MATLAB batch simulations are conducted for PSO hyperparameter sensitivity and derivative-filter/noise matching. The sensitivity results show that representative PSO settings produce similar validation performance, with mean peak deviations ranging from 0.8871 N to 0.9042 N and worst-case peak deviations from 1.7302 N to 1.7401 $\mathbf{N}$. The filter study shows that an excessively small derivative-filter time constant amplifies sensor noise and control effort, whereas 2-10 ms provides a practical compromise depending on noise intensity. The obtained results are simulation-level evidence based on the reduced-order model and require future calibration using measured force curves from a physical polishing platform.