Balance Assist During Walking With a Lower-Limb Power-Assist Robot Using Multiple Virtual Tunnels

Abstract

Powered lower-limb exoskeletons hold promise for providing locomotion assistance to physically weak individuals, thereby enhancing their quality of life. However, ensuring the balance and safety of the human-robot system during dynamic walking remains a significant challenge for these robots, particularly under unexpected perturbations. This paper proposes a balance assist method that helps the wearer recover from forward or backward perturbations during walking, returning to a continuous gait without losing balance or requiring a recovery step, which is essential for wearers with limited physical capacity. In the proposed method, to detect gait instability early and reliably, four types of virtual tunnels across diverse biomechanical and physiological parameters are adopted simultaneously: center of pressure (CoP) progression, the sagittal center of mass (CoM) location, sagittal swing toe location, and electromyography (EMG) signals from relevant lower-limb muscles. In addition to employing multiple virtual tunnels for reliable walking instability detection, this paper introduces a balance assistive force generation method that mathematically decides the distribution of forces. The proposed method formulates the effect of balance assistive forces on each of the virtual tunnel parameters through a sensitivity matrix, and determines the appropriate force distribution using a least-squares-based framework that minimizes the difference between the detected deviations of virtual tunnel parameters and their predicted effects under applied balance assistive forces. This approach enables the robot to selectively correct only the abnormal parameters while minimizing interference with stable ones, thereby guiding the wearer’s motions back within their respective virtual tunnels and restoring a continuous, balanced gait. The effectiveness of the proposed method is evaluated through experimental evaluation with a lower-limb power-assist exoskeleton robot.