Assist-as-needed Hip Exoskeleton Control for Gait Asymmetry Correction via Human-in-the-loop Optimization

TL;DR

This work addresses gait asymmetry in post-stroke rehabilitation by introducing an assist-as-needed hip exoskeleton controlled via human-in-the-loop optimization. An adaptive oscillator-based system detects gait phase and symmetry errors in real time, while an error-driven, subject-specific torque strategy provides targeted assistance without prescribing a fixed trajectory. A Bayesian optimization loop tunes two key gains to balance symmetry correction with patient participation, confirmed by healthy-subject experiments with simulated hemiplegia that show near-normal time trajectories and increased participation. The findings support AAN exoskeletons as a viable approach to personalize rehabilitation and actively engage patients, with implications for future bilateral designs and clinical validation.

Abstract

Gait asymmetry is a significant clinical characteristic of hemiplegic gait that most stroke survivors suffer, leading to limited mobility and long-term negative impacts on their quality of life. Although a variety of exoskeleton controls have been developed for robot-assisted gait rehabilitation, little attention has been paid to correcting the gait asymmetry of stroke patients following the assist-as-need (AAN) principle, and it is still challenging to properly share control between the exoskeleton and stroke patients with partial motor control. In view of this, this article proposes an AAN hip exoskeleton control with human-in-the-loop optimization to correct gait asymmetry in stroke patients. To realize the AAN concept, an objective function was designed for real-time evaluation of the subject's gait performance and active participation, which considers the variability of natural human movement and guides the online tuning of control parameters on a subject-specific basis. In this way, patients were stimulated to contribute as much as possible to movement, thus maximizing the efficiency and outcomes of post-stroke gait rehabilitation. Finally, an experimental study was conducted to verify the feasibility and effectiveness of the proposed AAN control on healthy subjects with artificial gait impairment. For the first time, the common hypothesis that AAN controls can improve human active participation was validated from the biomechanics viewpoint.

Figures (8)

• Figure 1: Illustration of patient-robot shared control and assist-as-needed concept during gait rehabilitation with exoskeletons.
• Figure 2: Assist-as-needed hip exoskeleton control for gait asymmetry correction, mainly consisting of the AO-based gait phase extraction and asymmetry detection, the high-level control for assisting as needed via human-in-the-loop optimization, and the low-level control for force tracking.
• Figure 3: Assistive torque $t_{d}$ as a function of the gait phase $\varphi^{imp}$.
• Figure 4: Activation function $\upsilon$ for the symmetry errors, such as $e_{fle}^{\theta}$, $e_{fle}^{\varphi}$.
• Figure 5: (a) Experimental protocol. (b) Feasible parameter region $A$ and five predefined sets of control parameters for the initialization of BO. (c) Overview of experimental environment and setup. (d)-(e) Experimental scenarios of different testing sessions. Specifically, the height of elastic ropes that are used for creating artificial gait impairment can be adjusted to accommodate the height of subjects.