Passive tracking of a target based on Supervisory adaptive EKF and CKF

Abstract

This study proposed a neural network structure for passive tracking of a target by an observer. Since the passive tracking measurement equation is nonlinear, the extended Kalman filter (EKF) and cubature Kalman filter (CKF) methods were implemented. Due to the nonlinear nature of the passive tracking measurement equation, the conventional extended and cubature Kalman filters are not good candidates where bearing-only target tracking is a standard and traditional passive tracking method. The effectiveness of Kalman filters dramatically depended on measurement noise covariance (R). Since R is challenging to be determined and changed with environmental variations, an on-line adaptive filter is proposed. The adaptive structure is founded on the double-layer perceptron neural network, where its weights were updated by the steepest descent method to tune covariance matrix R. In the numerical simulations, it is assumed that tracking of a target was carried out in an underwater environment by sonar measurement. In this paper, in addition to the proposed method, the neural network extended Kalman filter (NNEKF), neural network cubature Kalman filter (NNCKF), Sage Husa adaptive cubature Kalman filter (SHCKF) are used to track the target. The simulation results show the effectiveness of the proposed method.