Feedback control of nonlinear multiplicative systems using neural networks: an application to electrically stimulated muscle

The authors consider the control of a class of nonlinear systems using feedback linearization methods, in which the control maps are learned using an artificial neural net. The class of systems investigated involves the multiplication of separate subsystems. The motivation for considering the real-time control of such systems is the fact that they represent the input-output properties of electrically stimulated muscle. Such systems must be controlled as part of neural prostheses that are designed to restore function to individuals who are paralyzed.<>