Improved Rehabilitation Robot Trajectory Regeneration by Learning from the Healthy Ankle Demonstration

Abstract

The prevalence of ankle injuries in daily life has prompted the widespread application of rehabilitation robots. One of the important factors affecting robot-assisted ankle rehabilitation is the training trajectory which is usually regenerated from ankle movements. The traditional trajectory regeneration method is not suitable for the clinically recommended periodic ankle movements. In this paper, an improved robot trajectory regeneration method based on the individual characteristics is proposed to provide training reference trajectory for rehabilitation robots. This method extracts sample characteristics from the demonstration of the healthy ankle and reconstructs the sample space. Based on Learning from Demonstration (LfD) technology, the reference trajectory is regenerated for the rehabilitation of the injured ankle. The analysis of statistics and the regeneration of spatial features are performed to prove that this proposed method can regenerate the rehabilitation reference trajectory by learning from the healthy ankle demonstration.