Reinforcement learning-based development of time-optimal current trajectories for permanent magnet synchronous motor drives under voltage and current constraints

This paper presents a control algorithm to enhance torque dynamics of a permanent magnet synchronous motor (PMSM) within voltage and current limits by development of time-optimal current vector trajectories utilizing reinforcement learning. During the high-speed operation, torque dynamics is degraded due to voltage limit of motor drive system and high back electromotive force (emf) voltage. This paper proposes trajectory optimization methods to improve torque dynamics of PMSM drives. The reinforcement learning algorithms, Q-learning and Deep Deterministic Policy Gradient (DDPG), are applied to trajectory optimization, and the characteristics and effects of each algorithm are analyzed.