Design and Model-Free Reinforcement Learning Based Control of a Modular Self-Balancing Robotic System
Published in IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), 2024
Authors
Mishek Musa, & Uche Wejinya
Abstract
The concept of utilizing a modular version of the two-wheeled self-balancing robot is motivated by the need for modern robotic systems to be scalable, making it possible to build robots of various sizes and capabilities by adding or removing modules, redundant in the case of system failure, and reconfigurable and customizable for industries with unique requirements, such as manufacturing, healthcare, and agriculture. With this change comes the added complexity of dealing with a high degree of parametric uncertainty due to variations in parameters like mass, friction, and center of gravity. This necessitates the design of a control system capable of overcoming these issues and so model-free reinforcement learning (RL) is investigated for this purpose.
Recommended citation: Musa, M., & Wejinya, U. (2024, July). Design and Model-Free Reinforcement Learning Based Control of a Modular Self-Balancing Robotic System. In 2024 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM)
Download Paper