RL-Based TD3 Tuning of FOPI Controller for BLDC Motor Control

Abstract

This paper investigates the speed control of a Brushless DC (BLDC) motor using a fractional-order proportional-integral-derivative (FOPID) controller. Although fractional-order controllers offer additional tuning degrees of freedom compared with conventional PID controllers, determining suitable controller parameters remains a nontrivial task. To address this issue, a reinforcement learning approach based on the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm is considered. The algorithm is used to tune the FOPID controller parameters through interactions with the motor drive model. The obtained controller is evaluated under different operating conditions and compared with a conventional PID controller. The simulation results show lower tracking errors and improved transient responses for the TD3-tuned FOPID controller. These results indicate that reinforcement learning can be used as an effective tool for tuning fractional-order controllers in BLDC motor drive applications.