A Myography-Based Adaptive Gravity Compensation Strategy for Powered Upper-Limb Exoskeletons

The increasing adoption of upper-limb exoskeletons in industrial, domestic, and clinical settings calls for assistive strategies capable of adapting to human-in-the-loop requirements, including dynamic payload variations. This paper presents a myography-based adaptive gravity compensation (AGC) method that relies solely on biceps and triceps activity, without requiring large training datasets, subjectspecific anthropometry, or lengthy calibration. A theoretical analysis shows that the combined effects of acceleration and anthropometric uncertainties introduce inherent disturbances in payload estimation. To address this issue, two implementations of the proposed adaptive law are introduced: a continuous update strategy that compensates inertial effects, and a thresholdbased strategy that temporarily suspends adaptation when acceleration exceeds a predefined limit. Both approaches are validated through simulations and experiments on a 1-DoF elbow exoskeleton with fifteen healthy participants and compared against the only existing method that neglects acceleration effects. Results show that the threshold-based strategy yields the most accurate mass estimates, significantly reduces muscular effort, and provides the most favorable user experience.