Bayesian Algorithm-Based Force Profiles Optimization of Hip-Assistive Soft Exosuits Under Variable Walking Speeds

Abstract

Relevant research highlights humans’ capacity to continuously adapt their walking speed to minimize metabolic energy consumption during overground free walking. Past studies have shown that soft exosuits assisting in hip flexion and extension can reduce metabolic costs and regulate gait parameters during human locomotion. This emphasizes the need to fine-tune hip exosuit parameters to align with walking speed, thereby enhancing metabolic efficiency. This study aims to optimize assistive force parameters of hip exosuits across different walking speeds, providing insights for optimizing force profiles in outdoor walking. We employed a human-in-the-loop approach with Bayesian optimization to determine optimal force profiles for hip assistance. Six subjects performed treadmill walking at four fixed speeds (0.84, 1.16, 1.48, and 1.8 m/s), optimizing control parameters for each speed and establishing a Bayesian experience (BXE) linking walking speed to optimal parameters. Furthermore, we developed a real-time force optimization controller based on the BXE for adjusting the force parameters of assistance. Outdoor walking experiments with the same subjects showed that BXE-optimized profiles significantly reduced metabolic costs compared to fixed profiles. This study underscores the importance of optimizing assistive forces for varying walking speeds in humans.