Level-Ground and Stair Adaptation for Hip Exoskeletons Based on Continuous Locomotion Mode Perception.

Abstract

Hip exoskeleton can provide assistance to users to augment movements in different scenarios. The assistive control for hip exoskeleton involves the interactions among exoskeleton, user, and environment, which depends on the environment perception (to predict locomotion) to design control strategy combined with gait mode and so on. Current exoskeleton control still needs to be improved in adaptation to continuous locomotion mode and different users. To address this problem, we have employed a learning-free (i.e., non-data-driven) environment perception method to improve hip exoskeleton adaptive control toward continuous locomotion mode. The adaptive control experiments were conducted on level ground and stairs on 7 subjects. The prediction accuracy for steady locomotion mode was more than 95% for each subject (ranged from 95.7% to 99.7%). The prediction accuracy for each locomotion mode transition ranged from 87.5% to 100%, and the transition timing could be detected before the end of transition period. Compared with learning-based (data-driven) approaches, our method achieves better performances in adaptive control for hip exoskeleton and shows some generalization for subjects.