Adaptive user-specified control of movements with functional neuromuscular stimulation

Functional neuromuscular stimulation (FNS) systems can provide a person with thoracic level spinal cord injury the ability to stand, with limited maneuverability. The authors aim to develop a system that enhances maneuvering ability by providing the user with control of stimulation to adjust limb position, but nonlinear system properties make it difficult for the user to control limb position. In this work, the authors have implemented an adaptive neural network controller to linearize the system input/output properties as seen by the user. They have tested their controller on a model of a swinging lower leg that is controlled via stimulation. The results of a series of simulations indicate that the adaptive controller can successfully linearize input/output properties for a wide range of model parameters and also that the controller performs well over a wide range of values of learning rate and network size