Kalman Filtering with Position-Dependent Noise Modeling for Active Sonar

The Kalman Filter (KF) is commonly applied in problems that require tracking of target like sonar applications for Anti-submarine Warfare (ASW) or harbor surveillance. The conventional KF approach to implementing a 2D tracking problem assumes a nonlinear measurement equation that provides a measurement of the target’s range and bearing, corrupted by additive white Gaussian noise (AWGN). These nonlinearities and the noise make accurate positioning of the target challenging. In this paper, it’s proposed another KF procedure for estimating the target’s state, i.e., the target’s position and speed, where the measurement equation directly gives the target’s position in Cartesian coordinates combined with a position-dependent noise for an active sonar platform. This noise is modeled and used to derive the measurement noise covariance matrix of KF algorithm. The performance of the proposed algorithm was compared with that of classic filtering techniques, and the proposed method achieved a similar relative gain (measured in terms of Relative Mean Squared Error (RMSE)) compared to the best baseline results (Extended KF (EKF) and Unscented KF (UKF)), however its processing time is almost five times smaller than the second- fastest accurate filter (EKF), due to its simplicity.