Fatigue‐Driven Adaptive Assistance for a Semi‐Active Upper‐Limb Occupational Exoskeleton

Upper‐limb exoskeletons reduce overuse injuries in overhead work, but effectiveness depends on aligning assistance with biomechanics and user’s physical state. Most controllers rely on joint kinematics and task phase estimation, rarely accounting for fatigue from demanding tasks such as handling heavy handheld tools. This study presents fatigue‐driven assistance algorithm for a semi‐active upper‐limb occupational exoskeleton (OE). Fatigue is estimated in real time via electromyography (EMG) and heart rate (HR); median EMG frequency and averaged HR feed a linear regression controller that updates assistance according to physiological demand. Three conditions are evaluated: no exoskeleton, fixed assistance, and fatigue‐driven adaptive assistance, using muscle activation, HR, respiratory quotient (RQ), work duration, and perceived exertion as metrics. Adaptive assistance increases task duration by 41.2% ( p = 6.9 × 10 −4 ) versus no exoskeleton and reduces EMG activity in the anterior deltoid (AD) (26.5%, p = 0.002), posterior deltoid (23%, p = 0.02), upper trapezius (UT) (39%, p = 0.01), and medial deltoid (27%, p = 0.06). Fixed assistance yields smaller, nonsignificant duration increase of 9.9% ( p = 0.75), reducing AD by 16% ( p = 0.06) and UT by 27% ( p = 0.05). Both conditions reduce HR and RQ. Overall, results support fatigue‐driven control strategies for OEs.