Design and Adaptive Fuzzy Variable Admittance Control of A Wrist Rehabilitation Robot Based on A 3R Spherical Linkage
This paper presents a wrist rehabilitation robot based on a 3R spherical linkage and proposes a multimodal human–robot interaction (HRI) control method integrating surface electromyography (sEMG) and force sensing. The proposed robot provides three degrees of freedom (DoFs) of posture adjustment for wrist rehabilitation training and incorporates a 6-axis force sensor to establish an interactive rehabilitation platform. A variable admittance control strategy is further developed using fuzzy rules. In this strategy, sEMG signals are used to characterize muscle activation and active participation, while interaction force/torque information reflects the current mechanical interaction state. Based on these multimodal inputs, the admittance damping parameter is adjusted online. Experimental results show that the proposed method can identify different training states, including active exertion, abnormal spasticity, muscle weakness, and interaction discomfort. Compared with the fixed admittance controller, the physical-signal-based admittance controller, and the physiological-signal-based variable admittance controller, the proposed controller increases muscle activation by 5.88%, 4.79%, and 1.07% of the full-scale range, respectively, thereby demonstrating superior effectiveness in promoting active participation.