Paper, 2025.04

A2SC: Adaptive Speed and Sensorless Control System for Motor Controllers based on Deep-RL

PID speed control and sensorless control in home-appliance motors rely on eight control gains that are tuned by hand and then fixed, which leaves them exposed to overshoot when conditions change or disturbances occur. A2SC connects a PMSM-based motor simulator to a reinforcement learning system through a trigger module and uses LSTM-TD3 to adjust the control gains adaptively during operation. The work was done with LG Electronics, and the errors were compared against the existing approach on Matsui-Sensorless and PMSM-Sensorless motor simulations. Joonho Seo is the second of four authors, and the paper appeared as a poster in the CPS track of ACM SAC 2025.

Where

ACM SAC 2025

My role

Second author

Status

published