Closing the Loop on Exoskeleton Motor Controllers: Benefits of Regression-Based Open-Loop Control

Lower-limb exoskeletons are widely researched to improve walking performance and mobility. Low-level sensor-less exoskeleton motor control is attractive for consumer applications due to reduced device complexity and cost, but complex and variable transmission system configurations make the development of effective open-loop motor controllers that are responsive to user input challenging. The objective of this study was to develop and validate an open-loop motor control framework resulting in similar or greater performance vs. closed-loop torque control. We used generalized linear regression to develop two open-loop controllers by modeling motor current during exoskeleton-assisted walking; a complex model used desired torque and estimated ankle angular velocity as inputs, while a simple model used desired torque alone. Five participants walked at 1.0-1.3 m/s on a treadmill with closed-loop and both open-loop controllers providing ankle exoskeleton assistance. Both open-loop current controllers had similar root-mean-squared torque tracking error (p = 0.23) compared to the closed-loop torque-feedback controller. Both open-loop controllers had improved relative average torque production (p < 0.001 complex, p = 0.022 simple), lower power consumption (p < 0.001 for both), and reduced operating noise (p = 0.002 complex, p < 0.001 simple) over the closed-loop controller. New control models developed for a different ankle exoskeleton configuration showed similar improvements (lower torque error, greater average and peak torque production, lower power consumption) over closed-loop control during over-ground walking. These results demonstrate that our framework can produce open-loop motor controllers that match closed-loop control performance during exoskeleton operation.