Enhancing Disturbance Robustness through TD3 based Adaptive Bandwidth Control

Taeyeon Kim; Minhyuk An; Changoh Woo; Jae Suk Lee

doi:10.1109/SCEMS67400.2025.11267380

Enhancing Disturbance Robustness through TD3 based Adaptive Bandwidth Control

Taeyeon Kim
, Minhyuk An
, Changoh Woo
, Jae Suk Lee

Department of Electrical Engineering

Jeonbuk National University

Research output: Contribution to conference › Conference paper › peer-review

Abstract

In this paper, a control method is proposed in which the current controller for a Permanent Magnet Synchronous Motor (PMSM) is replaced by a reinforcement learning (RL) agent. The system is designed to output a constant bandwidth regardless of the changing parameters, overcoming the issue where the bandwidth typically fluctuates with changes in current. The algorithm applied for RL is the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, which is composed of two critics and one actor. By using TD3, it is possible to prevent overestimation in the critic's value function, which can lead to selecting sub-optimal actions instead of optimal ones. In this paper, a control method is proposed in which the current controller for a PMSM is replaced by a reinforcement learning agent.

Original language	English
Title of host publication	2025 IEEE 8th Student Conference on Electric Machines and Systems, SCEMS 2025
Publisher	Institute of Electrical and Electronics Engineers
ISBN (Electronic)	9798331565640
DOIs	https://doi.org/10.1109/SCEMS67400.2025.11267380
State	Published - 2025
Event	8th IEEE Student Conference on Electric Machines and Systems, SCEMS 2025 - Pusan, Korea, Republic of Duration: 2025.11.20 → 2025.11.22

Publication series

Name	IEEE Student Conference on Electric Machines and Systems (SCEMS)
ISSN (Electronic)	2771-7577

Conference

Conference	8th IEEE Student Conference on Electric Machines and Systems, SCEMS 2025
Country/Territory	Korea, Republic of
City	Pusan
Period	25.11.20 → 25.11.22

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 7 Affordable and Clean Energy

Keywords

Bandwidth
Current control
DDPG
PMSM
TD3

Access to Document

10.1109/SCEMS67400.2025.11267380

Cite this

Kim, T., An, M., Woo, C., & Lee, J. S. (2025). Enhancing Disturbance Robustness through TD3 based Adaptive Bandwidth Control. In 2025 IEEE 8th Student Conference on Electric Machines and Systems, SCEMS 2025 (IEEE Student Conference on Electric Machines and Systems (SCEMS)). Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/SCEMS67400.2025.11267380

@inproceedings{c772c07700244bbfb7ea8dd8d3b3c57d,

title = "Enhancing Disturbance Robustness through TD3 based Adaptive Bandwidth Control",

abstract = "In this paper, a control method is proposed in which the current controller for a Permanent Magnet Synchronous Motor (PMSM) is replaced by a reinforcement learning (RL) agent. The system is designed to output a constant bandwidth regardless of the changing parameters, overcoming the issue where the bandwidth typically fluctuates with changes in current. The algorithm applied for RL is the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, which is composed of two critics and one actor. By using TD3, it is possible to prevent overestimation in the critic's value function, which can lead to selecting sub-optimal actions instead of optimal ones. In this paper, a control method is proposed in which the current controller for a PMSM is replaced by a reinforcement learning agent.",

keywords = "Bandwidth, Current control, DDPG, PMSM, TD3",

author = "Taeyeon Kim and Minhyuk An and Changoh Woo and Lee, \{Jae Suk\}",

note = "Publisher Copyright: 2025 IEEE.; 8th IEEE Student Conference on Electric Machines and Systems, SCEMS 2025 ; Conference date: 20-11-2025 Through 22-11-2025",

year = "2025",

doi = "10.1109/SCEMS67400.2025.11267380",

language = "English",

series = "IEEE Student Conference on Electric Machines and Systems (SCEMS)",

publisher = "Institute of Electrical and Electronics Engineers",

booktitle = "2025 IEEE 8th Student Conference on Electric Machines and Systems, SCEMS 2025",

}

Kim, T, An, M, Woo, C & Lee, JS 2025, Enhancing Disturbance Robustness through TD3 based Adaptive Bandwidth Control. in 2025 IEEE 8th Student Conference on Electric Machines and Systems, SCEMS 2025. IEEE Student Conference on Electric Machines and Systems (SCEMS), Institute of Electrical and Electronics Engineers, 8th IEEE Student Conference on Electric Machines and Systems, SCEMS 2025, Pusan, Korea, Republic of, 25.11.20. https://doi.org/10.1109/SCEMS67400.2025.11267380

Enhancing Disturbance Robustness through TD3 based Adaptive Bandwidth Control. / Kim, Taeyeon; An, Minhyuk; Woo, Changoh et al.
2025 IEEE 8th Student Conference on Electric Machines and Systems, SCEMS 2025. Institute of Electrical and Electronics Engineers, 2025. (IEEE Student Conference on Electric Machines and Systems (SCEMS)).

Research output: Contribution to conference › Conference paper › peer-review

TY - GEN

T1 - Enhancing Disturbance Robustness through TD3 based Adaptive Bandwidth Control

AU - Kim, Taeyeon

AU - An, Minhyuk

AU - Woo, Changoh

AU - Lee, Jae Suk

PY - 2025

Y1 - 2025

N2 - In this paper, a control method is proposed in which the current controller for a Permanent Magnet Synchronous Motor (PMSM) is replaced by a reinforcement learning (RL) agent. The system is designed to output a constant bandwidth regardless of the changing parameters, overcoming the issue where the bandwidth typically fluctuates with changes in current. The algorithm applied for RL is the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, which is composed of two critics and one actor. By using TD3, it is possible to prevent overestimation in the critic's value function, which can lead to selecting sub-optimal actions instead of optimal ones. In this paper, a control method is proposed in which the current controller for a PMSM is replaced by a reinforcement learning agent.

AB - In this paper, a control method is proposed in which the current controller for a Permanent Magnet Synchronous Motor (PMSM) is replaced by a reinforcement learning (RL) agent. The system is designed to output a constant bandwidth regardless of the changing parameters, overcoming the issue where the bandwidth typically fluctuates with changes in current. The algorithm applied for RL is the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, which is composed of two critics and one actor. By using TD3, it is possible to prevent overestimation in the critic's value function, which can lead to selecting sub-optimal actions instead of optimal ones. In this paper, a control method is proposed in which the current controller for a PMSM is replaced by a reinforcement learning agent.

KW - Bandwidth

KW - Current control

KW - DDPG

KW - PMSM

KW - TD3

UR - https://www.scopus.com/pages/publications/105030133718

U2 - 10.1109/SCEMS67400.2025.11267380

DO - 10.1109/SCEMS67400.2025.11267380

M3 - Conference paper

AN - SCOPUS:105030133718

T3 - IEEE Student Conference on Electric Machines and Systems (SCEMS)

BT - 2025 IEEE 8th Student Conference on Electric Machines and Systems, SCEMS 2025

PB - Institute of Electrical and Electronics Engineers

T2 - 8th IEEE Student Conference on Electric Machines and Systems, SCEMS 2025

Y2 - 20 November 2025 through 22 November 2025

ER -

Enhancing Disturbance Robustness through TD3 based Adaptive Bandwidth Control

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this