Recursive techniques for passive source location

The paper is concerned with the location of passive sources. Conceptually, this is viewed as a time delay estimation followed by a geometry determination. Signal, noise, and channel modeling questions affect the first block of the processor, i.e., the delay estimator. The second block is sensitive to the geometry description, namely the hypotheses on the dynamics, the array shape, and the relative observer/ /source configuration. Commonly used assumptions lead to decoupled effects which simplify the receiver structure. For deterministic array and source dynamics, a finite parameter description results. The receiver is designed via Maximum-Likelihood techniques. These do not encompass more general situations. To treat the problem of uncertain sensor location, or of stochastic dynamics, a different geometry description is considered. This description represents line arrays and motions as curves in space. Recalling simple facts from Differential Geometry, one is naturally led to describe the array geometry and/or the motion dynamics by a set of differential equations. This casts the passive positioning problem in the context of recursive Kalman-Bucy filtering. The problem of sensor uncertainty location and stochastic dynamics can then be dealt with, without having to consider Taylor series type arguments or other unnatural approximations.