Personalized Forward Prediction Simulation Improves the Efficiency of Human-in-the-Loop Optimization

“Human-in-the-loop” (HIL) optimization can be used to obtain appropriate exoskeleton assistance modes to improve walking economy. Given that traditional HIL optimization based on metabolic energy consumption may lead to subject discomfort and high time costs, this study focuses on the application of personalized forward prediction simulation in exoskeleton assistance optimization. We first designed a variable speed experiment to collect gait information from subjects to obtain personalized forward prediction simulation optimization results. Then, we conducted two types of HIL optimization experiments under two initial parameters: simulation optimization values and empirical values. Finally, two sets of validation experiments and one set of experiments in multi-speed mode were conducted. The results show that choosing the simulation-optimized initial values can effectively accelerate the optimization process and improve the efficiency of HIL optimization. Moreover, the simulation-optimized moments show better auxiliary performance than the empirical moments under different speed conditions. Our results provide scientific evidence and technical support for the development and application of personalized, high-performance assistive devices for a wider range of subjects in the future.