Estimation of coordinates of a radio radiation source at direction finding in the process of motion based on dynamic neural networks

Abstract

The paper considers an algorithm for estimating the coordinates of radio radiation sources from measurements in the process of motion of a single radio direction finder. The algorithm is devised as a set of recurrent neural networks, with delays of inputs, a feedback, and a prediction model. Analysis of characteristics of coordinate estimation accuracy is performed — as compared with the statistical adaptive algorithm of filtering and with regard to the potential boundary of the error of optimal estimates.