Ballbot Simulation System: Modeling, Verification, and Gym Environment Development

This paper introduces the development of a sophisticated simulation model for the ball-balancing robot (ballbot), implemented as a Gymnasium (Gym) environment in Python. The main purpose of this environment is to facilitate the application of Reinforcement Learning (RL) techniques to effectively control the ballbot system. Initially, a standard ballbot model is created using Solidworks, and a Unified Robot Description Format (URDF) file is generated to precisely capture the dynamics of the ballbot. To validate the model, simulations are performed within MATLAB/Simulink, comparing the ballbot URDF model against its nonlinear mathematical model with a simple PID controller. The simulation results demonstrate that the URDF model accurately represents the ballbot dynamics, exhibiting a comparable response to the mathematical model. Subsequently, the high-fidelity URDF model is integrated into the Gym environment using the Pybullet simulator. The ongoing objective of this research is to utilize the developed ballbot environment for RL-based control design.