A versatile knee exoskeleton mitigates quadriceps fatigue in lifting, lowering, and carrying tasks

The quadriceps are particularly susceptible to fatigue during repetitive lifting-lowering and carrying (LLC), affecting worker performance, posture, and ultimately lower-back injury risk. Although robotic exoskeletons have been developed and optimized for specific use cases like lifting-lowering, their controllers lack the versatility or customizability to target critical muscles across many fatiguing tasks. Here we present a task-adaptive knee exoskeleton controller that automatically modulates virtual springs, dampers, and gravity and inertia compensation to assist squatting, level walking, and ramp and stairs ascent/descent. Unlike end-to-end neural networks, the controller is composed of predictable, bounded components with interpretable parameters that are amenable to both data-driven optimization for biomimetic assistance and subsequent application-specific tuning, for example, maximizing quadriceps assistance over multi-terrain LLC. When deployed on a backdrivable knee exoskeleton, the assistance torques holistically reduced quadriceps effort across multi-terrain LLC tasks (significantly, except for level walking) in 10 human users without user-specific calibration. The exoskeleton also significantly improved fatigue-induced deficits in time-based performance and posture during repetitive lifting-lowering. Finally, the system facilitated seamless task transitions and garnered high effectiveness ratings post-fatigue over a multi-terrain circuit. These findings indicate this versatile control framework can target critical muscles across multiple tasks, specifically mitigating quadriceps fatigue and its deleterious effects.

Methods
Data collected from 10 able-bodied participants performing non-fatiguing (S1) and fatiguing (S2) lifting-lowering-carrying (LLC) tasks with and without a bilateral knee exoskeleton. dataset_S1 consists of electromyography, kinematics, torque, and footsensor data from non-fatiguing LLC tasks. dataset_S2 consists time, posture, and perceptual measurements from fatiguing squat lifting-lowering and carying tasks.