Submarine tracking via fusing multiple measurements based on Gaussian sum mixture approximation

This paper presents a submarine tracking technique by fusing multiple direction of arrival (DOA), time difference of arrival (TDOA), and target frequency measurements via an extended Kalman filter bank (EKFB). These measurements are collected by two sonobuoys each containing a wideband sonar-sensor and a narrowband sonar-sensor. In the fusion process, the area that the target may be located in is divided into multiple subareas, and in each subarea, the conditional probabilities of DOA, TDOA and frequency measurements given the target state are approximated with a weighted Gaussian probability density function (PDF). The measurements´ conditional PDF´s are approximated by the summation of these weighted Gaussian PDF´s (Gaussian sum mixtures (GSM)). Using GSM to approximate the measurements´ conditional PDF´s, we obtain a smaller PDF approximation error than other methods (e.g., Taylor series linearization method), and achieve higher tracking accuracy. The tracking method is presented theoretically and then evaluated via simulation to compare it with an EKF. The comparison results confirm that the proposed method is more robust than an EKF.