EMG-based Human-in-the-loop Optimization of Ankle Plantar-flexion Assistance with a Soft Exoskeleton

Abstract

Individualized control strategies were considered as an alternative approach for wearable robotic devices to increase the capability of walking economy. Human-in-the-loop (HIL) optimization can identify appropriate assistive control parameters for specific users based on real-time measurement of human physiological objectives. Recent studies on HIL optimization use metabolic cost as the physiological objective to adjust the control parameters. However, the long measurement times and low signal-to-noise ratio presented significant challenges for the wearable experiments. We presented an EMG-based HIL optimization method with a soft exoskeleton to rapidly and automatically generate customized control parameters for ankle plantar-flexion assistance. Experimental results demonstrated that this method can identify nearly optimal control parameters within a certain parameter space for specific subject in 7.5 minutes. When comparing the assistive condition that exoskeleton provided assistance with optimized control parameters to the assistive conditions that the exoskeleton provided assistance with initial control parameters and the exoskeleton unpowered, the soleus muscle activity was reduced by 11.5% (p < 0.001) and 17.6% (p < 0.001), respectively. Our results demonstrated that the EMG-based HIL optimization can be a promising approach to rapidly improve the assistive performance for exoskeletons.