Development of an Educational Platform for Reinforcement Learning Control of a Rotary Inverted Pendulum Based on LW-RCP

This study proposes a compact rotary inverted pendulum experimental platform designed for implementing reinforcement learning (RL) based controllers on physical systems. While RL demonstrates superior performance in simulated environments, applying it to real-world systems often leads to performance degradation and instability due to the sim-to-real gap and physical actuator constraints. In particular, when a stepper motor is employed as the actuator, its drive characteristics need to be considered when applying control commands generated by the learned policy to the physical system. This study treats the stepper motor as a high-pole-count synchronous motor and implements a closed-loop drive system using a field-oriented control (FOC)-based current control structure. Furthermore, a real-time control platform is established by integrating a lightweight rapid control prototyping(LW- RCP) environment with the rotary inverted pendulum system, enabling RL policies trained in simulations to be deployed directly onto the hardware. The proposed platform is designed with a desktop-sized, low-cost architecture to enhance accessibility in educational settings. The feasibility of applying RL-based controllers to actual systems is experimentally validated. Consequently, this research bridges the gap between theory and practice, providing an effective integrated environment for experiment-oriented control education and RL research.