Muscle model control of knee joint torque in a powered lower-body orthosis.

Abstract

Powered orthoses that employ virtual muscle models for specifying motor commands may better produce natural, human-like joint torque and motion outputs. Here, we design and implement a muscle controller based on the standard 3-element Hill model that outputs a joint torque signal prescribed to the electric motors of a powered orthosis. Real-time simulation of muscle dynamics in a MATLAB/Simulink environment was expressed through a microcontroller on the Parker Hannifin Indego orthosis. To initially demonstrate and examine the force-length and force-velocity relationships of the virtual muscles, we programmed the Indego to perform standard isokinetic and isometric rehabilitation exercises at the knee joint. A Biodex-2 isokinetic dynamometer was used to apply constraints and actively measure changes in knee joint angle, velocity, and torque. The observed torque rise time (∼126 ms) and profiles for torque-angle and torque-velocity for the muscle-model at the knee joint of the orthosis compared favorably to results reported for humans undergoing the same exercises. We further demonstrate potential feasibility for real-world implementation through measures of root mean square error (RMSE) and coefficient of determination (R2), which indicate good matching between the commanded and measured torque outputs. This pilot study represents the first step in developing and implementing control strategies for powered orthoses based on human muscle models.