Maximum likelihood estimation for long-range target tracking using passive sonar measurements

A Newton-type method is used to solve the target motion analysis (TMA) problem with respect to bearing and frequency measurements from a passive sonar system. In many long-range sonar situations the TMA problem is ill conditioned and suffers from a small signal-to-noise ratio. Although Kalman filters have been investigated extensively it is known that maximum likelihood (ML) estimation is superior in these cases. The main reason for the good performance of the ML method is that the underlying numerical optimization problem deals with the ill conditioning of the problem. This work illustrates how the conditioning depends on the geometry of the tracks and the signal-to-noise ratio. Monte Carlo simulations with respect to the measurement noise show the influence on the ML estimation performance for three specific cases concerning multileg situations and bottom bounce measurements