Neural network control of functional neuromuscular stimulation systems

 
 
 
 

Neural network control of functional neuromuscular stimulation systems

Show full item record


Title: Neural network control of functional neuromuscular stimulation systems
Author: Abbas, James Joseph
Description: A neural network control system has been designed for the purpose of controlling cyclic movements in Functional Neuromuscular Stimulation (FNS) systems. The design of the control system directly addresses three problems faced in the implementation of FNS control systems: customizing the control system parameters for a particular individual, adapting these parameters to account for changes in the musculoskeletal system, and resisting mechanical disturbances. The control system is implemented by a two-stage neural network that utilizes adaptive feedforward and feedback control techniques. The first stage of the neural network, the Pattern Generator, generates a cyclic pattern of activity. The design of this stage is based upon neural models of vertebrate motor control systems. The signals from the Pattern Generator are adaptively filtered by the second stage, the Pattern Shaper. A learning algorithm that accounts for system dynamics and input time delays was developed for use in adapting the Pattern Shaper filtering properties. Computer simulated models of electrically stimulated muscles acting on one- and two-segment skeletal systems were used to assess the potential utility of the neural network control system in FNS control. Results of the evaluation demonstrated that the control system can automatically c ustomize stimulation parameters, adapt them on-line, and resist mechanical disturbances. The control system was also demonstrated to be capable of controlling movements of multi-joint systems and of utilizing biarticular muscle effectively. The success of the control system in this evaluation indicates that it may provide significant improvements to existing FNS control system technology and suggests that the technique should be investigated further. These studies also indicate that this strategy may be appropriate for other applications in the control of dynamic, nonlinear systems with input time delays. The use of biologically motivated neural networks in the Pattern Generator provides this control system with unique features that are not readily available using existing control system techniques. The learning algorithm developed for use in the Pattern Shaper is particularly well-suited for use in engineering neural network control systems because it provides the ability to account for system dynamics and input time delays.
Permanent Link: http://rave.ohiolink.edu/etdc/view?acc_num=case1056554566
http://hdl.handle.net/2374.OX/16323
Date: 1992

Files in this item

Files Size Format View

There are no files associated with this item.

This item appears in the following Collection(s)

Show full item record