A Relationship Model Between Optimized Exoskeleton Assistance and Gait Conditions Improves Multi-Gait Human-in-the-Loop Optimization Performance

Exoskeletons have been shown be able to reduce metabolic cost of human walking. Determining the suitable assistance is challenging due to individual variability in response and the need for tailored assistance across different gait conditions. Human-in-the-loop (HIL) optimization has been proposed to address this issue, but current implementations often suffer from prolonged optimization cycles. In this study, we establish a model capturing the relationship between optimized assistance parameters and gait conditions (speed and slope) through a series of HIL optimization experiments spanning various gait conditions. The validation results showed that the desired assistance torque calculated by the model closely aligns with the assistance torque obtained through HIL optimization, and the calculated assistance reduced metabolic cost by 11.95% (p<0.001) and RMS soleus activity by 22.28% (p=0.049) compared to the case without assistance. The optimized assistance using model values for initialization after two generations significantly reduced metabolic cost by 12.1% (p<0.001) and RMS soleus activity by 24.8% (p=0.033), and produced larger benefits than using the empirical values after four generations, with a 50% increase in efficiency. Results suggested that the relationship model can help to improve multi-gait exoskeleton assistance customization efficiency and effectiveness both by optimization parameter initialization and direct parameter assertion. These advancements expand the applicability of HIL optimization and improve the effectiveness of exoskeleton assistance.