A Realistic Model Reference Computed Torque Control Strategy for Human Lower Limb Exoskeletons

TL;DR

This paper introduces a robust, computationally efficient model reference computed torque control strategy for human lower limb exoskeletons, enhancing trajectory accuracy and handling uncertainties for neurorehabilitation.

Contribution

A novel control approach that accounts for parametric uncertainties and reduces computational load in exoskeleton trajectory control.

Findings

High accuracy in trajectory tracking demonstrated

Robustness confirmed under parametric uncertainties

Effective handling of model inaccuracies in experiments

Abstract

Exoskeleton robots have become a promising tool in neurorehabilitation, offering effective physical therapy and recovery monitoring. The success of these therapies relies on precise motion control systems. Although computed torque control based on inverse dynamics provides a robust theoretical foundation, its practical application in rehabilitation is limited by its sensitivity to model accuracy, making it less effective when dealing with unpredictable payloads. To overcome these limitations, this study introduces a novel model reference computed torque controller that accounts for parametric uncertainties while optimizing computational efficiency. A dynamic model of a seven-degree-of-freedom human lower limb exoskeleton is developed, incorporating a realistic joint friction model to accurately reflect the physical behavior of the robot. To reduce computational demands, the control system is split into two loops: a slower loop that predicts joint torque requirements based on input trajectories and robot dynamics, and a faster PID loop that corrects trajectory tracking errors. Coriolis and centrifugal forces are excluded from the model due to their minimal impact on system dynamics relative to their computational cost. Experimental results show high accuracy in trajectory tracking, and statistical analyses confirm the controller's robustness and effectiveness in handling parametric uncertainties. This approach presents a promising advancement for improving the stability and performance of exoskeleton-based neurorehabilitation.