Multivariable adaptive control using artificial neural networks

The Extended Kalman Filter (EKF) is well known as a state estimation method for nonlinear systems. Recently this method has been used as a learning algorithm to estimate the parameters of a neural network used for identification of the process dynamics of a single input, single output system, and it was shown that this method offered superior capability over the conventional back-propagation algorithm (BP). In this paper we examine if the desirable characteristics that EKF provides over BP in identification are also true when this form of learning is used in the control of a multivariable dynamic model of a submarine vehicle.