Control of Soft Wearable Robots for Elbow Assistance and Rehabilitation: A Simplified Solution Using Disturbance Observer and Nonsmooth Feedback

• Published in: IEEE transactions on neural systems and rehabilitation engineering Vol. 33; pp. 2988 - 2999
• Main Authors: Wang, Jiajin, Xu, Baoguo, Wang, Xin, Lai, Jianwei, Wei, Xiangshan, Zhao, Zishuo, Lu, Ye, Shen, Ying, Wang, Hongxing, Song, Aiguo
• Format: Journal Article
• Language: English
• Published: United States IEEE 2025
• Subjects: Actuators; Adult; Algorithms; Analytical models; Bending; Biomechanical Phenomena; composite controller; Computer Simulation; disturbance observer; Elbow; Elbow Joint; Equipment Design; Exoskeleton Device; Feedback; Feedforward systems; Humans; Male; Medical diagnostic imaging; Nonlinear Dynamics; Pneumatic systems; rehabilitation robotics; Robot sensing systems; Robotics - instrumentation; Soft robotics; Stroke Rehabilitation - instrumentation; Wearable Electronic Devices; Wearable robotics; Wearable robots
• ISSN: 1534-4320; 1558-0210; 1558-0210
• DOI: 10.1109/TNSRE.2025.3594750

Soft wearable robots present a promising approach for elbow assistance and rehabilitation. However, most existing devices rely on open-loop control strategies, which renders individual customization cumbersome and incapable of adapting to dynamic interactions. The challenges in implementing closed-loop control arise from the complex nonlinearities of soft robots and the unpredictable external disturbances encountered during human-robot interactions. To achieve closed-loop control of elbow soft wearable robots, this paper proposes a disturbance observer (DO)-based nonsmooth feedback (NSF) method. Specifically, a tailored DO is designed to enhance feedforward compensation by analyzing the unique characteristics of the disturbances encountered in practical systems. A nonrecursive NSF is employed to suppress residual disturbances and nonlinearities, with the finite-time stability of the closed-loop system rigorously guaranteed. The proposed method balances efficacy and simplicity by leveraging the concise models of both the system and disturbances to enhance performance while avoiding intricate modeling. Moreover, its nonrecursive design results in a straightforward control law and facilitates implementation. Extensive comparison and ablation experiments validate the superiority of the proposed method over existing approaches. Human trials involving 8 healthy subjects and 7 stroke patients demonstrate that our method enhances task performance, reduces muscle strain during elbow assistance scenarios, and significantly improves elbow motor function in rehabilitation training.