A Comprehensive Review of Power-Assisted Exoskeletons: From Perception to Control Strategies

Abstract

Power-assisted exoskeletons represent a highly promising technology that can enhance human strength and improve productivity, with broad applications across military, industrial, construction, mining, and daily life domains. To provide natural and compliant assistance to the human body, an exoskeleton must accurately understand human motion intentions and rapidly make control decisions to drive joints to actively adapt to human movement. Over the past two decades, research on human motion intention sensing, understanding, and collaborative control strategies has been extensive. However, due to the complex, highly nonlinear interactions among humans, machines, and the environment—including physiological and random signal interference in perception, uncertain external forces, and sudden intention changes—no major breakthroughs have been achieved in the collaborative control of power-assisted exoskeletons. The existing review literature on this topic is relatively outdated. As a research contribution, this paper presents a comprehensive and systematic review from a critical perspective, classifying representative studies and related technologies across three key areas: the latest environmental-EMG-EEG-based human-robot interaction sensing and localization techniques, machine learning-based human motion intention recognition and prediction, and intelligent control strategies aimed at human-robot collaboration. From a macro perspective, the paper discussed existing challenges and unresolved issues across these three domains, while proposing forward-looking recommendations and potential research directions.