Reinforcement learning-based control of a soft robot equipped with a electromagnetic actuator

Soft robots have attracted significant attention in recent years due to their high compliance, inherent safety, and bio-inspired designs, enabling a wide range of applications in industrial, research, and medical domains. Actuation with external magnetic fields provides a precise and contactless control mechanism. However, the nonlinear behavior and complex dynamics of these systems present substantial challenges for conventional control approaches. In this study, a novel reinforcement learning–based control framework is proposed for the motion control of a magnetically actuated soft robot using the proximal policy optimization algorithm. The soft robot is modeled as a continuous flexible rod with a permanent magnet attached to its tip, and the dynamic behavior of the system is numerically simulated. Simulation results demonstrate that the proposed reinforcement learning–based controller can accurately force the robot tip toward predefined targets while ensuring stable, robust, and reliable performance.

کلیدواژه ها:

Soft robot ، Electromagnetic actuator ، Reinforcement learning ، Proximal policy optimization algorithm