| Title |
Control System Design for Maglev Conveyor Systems with Time-Delay Via Reinforcement Learning |
| Authors |
노수영(Soo-Young Noh) ; 김창현(Chang-Hyun Kim) |
| DOI |
https://doi.org/10.5370/KIEE.2025.74.2.353 |
| Keywords |
Maglev Conveyor System; Reinforcement Learning; Deep Deterministic Policy Gradient; Time Delay Compensation; Noise and Disturbance Rejection |
| Abstract |
Magnetic levitation conveyor systems enable frictionless and contamination-free transport, making them suitable for sensitive environments such as semiconductor cleanrooms. However, their inherent nonlinearities, along with uncertainties such as noise, disturbances, and time delays, pose significant challenges to conventional controllers. To overcome these limitations, this study proposes a reinforcement learning based control strategy utilizing the deep deterministic policy gradient algorithm, selected for its effectiveness in handling continuous state-action spaces and learning stable control policies in complex nonlinear environments. The developed reinforcement learning based controller is designed to handle the system’s nonlinear dynamics and various uncertainties more effectively than traditional PID controller. Simulation results demonstrate that the proposed controller not only produces smoother transient responses and robust steady-state performance but also maintains stable operation under time delays. The proposed controller is applicable to magnetic levitation systems with various uncertainties. |