Impedance Control of Lower–Limb Exoskeletons Using Bayesian Optimization for Adaptive Human–Robot Interaction

Abstract

Lower-limb rehabilitation exoskeletons require adaptive control strategies to accommodate patient-specific biomechanics and ensure safe and effective Human-Exoskeleton Interaction (HEI). Impedance control is widely adopted for this purpose; however, selecting appropriate impedance parameters remains challenging due to inter-subject variability and the dynamic nature of rehabilitation conditions. This paper presents impedance control of lower-limb rehabilitation exoskeletons using Bayesian Optimization (BO) for adaptive and synergistic interaction. The dynamics of the human and exoskeleton systems are derived using Euler-Lagrange mechanics, while their interaction dynamics is modeled as a simplified mass-spring-damper system. BO is employed to adapt the impedance parameters by minimizing an objective function designed to improve tracking performance, quantified by trajectory-tracking and interaction errors, while enhancing user participation, estimated through a defined index $\hat{\tau}_{h}$. Within the BO framework, Gaussian Process (GP) is used as the surrogate model to approximate the objective function, while Expected Improvement (EI) is used as the acquisition function to balance the exploitation of regions with low predicted cost and exploration of uncertain regions. Simulation results underscore the potential of BO as a practical and scalable approach to adaptive impedance tuning, enabling more synergistic HEI in lower-limb rehabilitation applications.