Locomotion Mode Identification and Gait Phase Estimation for Exoskeletons During Continuous Multilocomotion Tasks

Abstract

Gait phase estimation is important technology in controlling the exoskeleton robot to assist elderly walking. Several kinds of Gait estimation methods have been proposed, however, the previously proposed methods were mainly aiming at one kind of walking task, e.g., level ground walking. There are only a few studies aiming at continuous gait phase estimation during continuous multilocomotion tasks. In this article, we design a continuous gait phase estimator based on adaptive oscillator (AO) network. In order to overcome the problem that the traditional AO does not converge or converges slowly when the gait task is switching, a new structure of gait phase estimator, including a gait tasks classifier, an AO reset, a peak detector, and a model-based (MB) transition gait phase estimator is designed to improve the performance of AOs network. The switching unit is designed to reorganize the output gait phase. Considering the stabilization of the sensors in continuous multilocomotion tasks, the gait tasks classifier only utilizes the angle of hip joints. The results show that the constructed classifier has similar performance to other gait tasks classifiers and requires minimum sensing sources. The continuous gait phase estimation results during continuous multilocomotion tasks show that the proposed method has better performance than the traditional AO and the AO network with self-designed reset.