As smart manufacturing and automation technologies continue to advance rapidly, robotic arms have become indispensable in modern industry and technological applications. A research team led by Associate Professor Jim-Wei Wu from the Department of Electrical Engineering at National Central University (NCU) has successfully integrated reinforcement learning with classical and advanced control techniques to develop a highly stable and disturbance-resilient robotic arm. This innovation significantly enhances operational precision and, backed by solid theoretical foundations and research originality, has been published in the leading international journal IEEE Transactions on Cybernetics.
The proposed control method does not rely heavily on precise mathematical models. Even under environmental uncertainty or significant variations in system load, the system maintains robust control performance. This design greatly improves the adaptability and stability of robotic arms. In addition, the team introduced a novel disturbance observer design framework that overcomes the limitations of conventional approaches, which typically handle only limited types of disturbances. The new framework enables faster and more accurate responses to sudden or dynamic disturbances, further enhancing control precision and system stability while reducing dependence on complex sensing and computational resources.
Furthermore, the system incorporates an Actor–Critic artificial intelligence architecture, allowing it to simultaneously learn how to evaluate control performance and generate optimal control strategies. By integrating classical control principles as an initial foundation, the approach significantly reduces the need for extensive parameter tuning in practical applications. This result makes the system easier to design and enables it to reach stable operation more rapidly, laying a critical foundation for the future development of highly autonomous robotic systems.