Lower-Limb Exoskeleton Trajectory and Structure Co-Optimization for Maximum Assistive Efficiency

Abstract

Improving assistance efficiency is paramount for booster exoskeleton design, requiring comprehensive consideration of the human-exoskeleton dynamics, with coupling mechanics playing a central role. This study proposes a purely simulation-based co-optimization framework that simultaneously considers both trajectory and structure parameters of the exoskeleton to maximize assistive efficiency under various coupling conditions. The framework integrates a two-layer optimization strategy for identifying coupling parameters and employs a co-optimization scheme based on the Neighborhood Field Optimization algorithm, utilizing surrogate models for inverse dynamic analysis. Four coupling models across five coupling positions and three surrogate models were examined. The nonlinear coupling model and Bayesian Neural Network surrogate achieved superior accuracy. Optimal exoskeleton segment lengths found to be shorter at varying degrees along the shank, but slightly longer at the thigh compared to corresponding human segment lengths. Furthermore, the optimized results significantly enhanced assistance efficiency, with the linear coupling model achieving torque reductions of 29.65% (knee) and 69.29% (hip). The most effective coupling positions were identified at 70% (shank) and 60% (thigh) of segment lengths. By simultaneously optimizing both the structure and trajectory parameters of the exoskeleton within the proposed framework, effective assistance to the wearer can be achieved. This study offers valuable insights and design strategies for the development and optimization of booster exoskeletons, contributing to enhanced human-exoskeleton coupling and performance.