Tunnel-Guided Assist-as-Needed Control for Hand Pre-Shaping: Development and Validation in Healthy Participants

Abstract

Rehabilitating hand function after stroke is challenging, particularly for individuals with spasticity. Grasp preshaping—an anticipatory component of motor control—is often undertrained in conventional therapy. This work introduces a synergy-based assist-as-needed (AAN) framework for preshaping using the Opposable Thumb Hand Exoskeleton for Rehabilitation (OTHER Hand). Desired motion is expressed in a synergy-dependent representation rather than time, with Gaussian Mixture Models and Regression (GMM/GMR) generating smooth, task-specific reference trajectories. Virtual corrective and tangential forces derived from joint errors and velocities adapt motor output to encourage active effort. A pilot study with five healthy participants tested the controller during both unconstrained movement and a simplified impairment condition created via constant feedforward torque. Although not a velocity-dependent model of spasticity, this perturbation provided a controlled proxy for evaluating adaptive assistance. AAN increased user engagement by 57.6% relative to passive position control and improved efficiency—defined as the ratio of user effort to trajectory error— by 10.1%-22.5% higher than for either force feedback or position control across both constraint levels. These results indicate that synergy-based minimal assistance can enhance engagement and performance, supporting its potential for clinical translation.