Task-based methods for evaluating electrically stimulated antagonist muscle controllers

Abstract
Single-joint motor neural prosthesis control algorithms were tested in a novel animal model. The model consisted of a human subject who provided joystick inputs to a controller. The controller output determined the stimulus activation levels of two antagonist muscles which manipulated the ankle joint of an intact, anesthetized cat. Using visual feedback, the subject manipulated the system to perform positioning tasks which simulated normal activity of an intact limb. Three controllers were evaluated, open-loop reciprocal control, P-D closed-loop reciprocal control, and open-loop cocontraction control. The results demonstrated that in the presence of visual feedback, open-loop cocontraction control compared favorably in performance to a P-D closed loop controller. This has a practical value for the implementation of clinical neural prostheses since it suggests that in some cases, feedback transducers may not be required for fine control.