ARCHITECTURE OF A DEEP NEURAL NETWORK IN THE PROBLEM OF PREDICTING THE COORDINATES OF A MOVING TARGET

Download article

Authors

V.K. KEMAYKIN, Cand. Sc., A.S. POLYGAEV, stud.

Abstract

The use of a deep recurrent neural network in the problem of predicting the coordinates of a moving target under conditions of measurement inaccuracy is considered. The training of the neural network is carried out on a preliminary sample of coordinates of a limited length, errors in measuring the coordinates are modeled, the nature of the movement of the target during network testing differs from the model used in training the neural network and underlying the Kalman filter.

Keywords

artificial neural network, long short-term memory neural network, Kalman filter, time series forecasting, measurement filtering, extrapolation.