State estimation using adaptive linear combiner and multilayer neural network

Abstract

The state estimation problem using artificial neural networks is considered. Stochastic systems are analyzed. The neural networks used are the adaptive linear combiner (ALC) and a multilayer network. An approach to train the network based on several Kalman filter solutions whose average is used as the desired output is developed. The performance of the training algorithms gives state estimates when measurement are presented. Examples are given for cases of high and low signal-to-noise ratio to illustrate the proposed approach.<>