9361524 Stubberud The artificial neural network (ANN) has been used with varying degrees of success in the feed back loop to provide the necessary control for nonlinear or uncertain systems. Recently, research has been done to train a neuro-controller, as well as a neuro-observer, using a relatively new extended Kalman filter (EKF) training paradigm. However, both functions have not been implemented with ANNs, which train on-line. ORINCON will investigate the capability of implementing a neuro-controller and neuro-observer on-line using the EKF paradigm. All systems encounter parameter variations over time from their nominal design that the control must adapt to or be robust enough to handle. This research will allow the entire feedback loop to be adjusted for parameter changes. ***
