Robust matched-field processing: A subspace approach

This dissertation is concerned with the realities of applying matched-field processing in realistic scenarios. Matched-field processing (MFP) is a model-based source localization method which is a function of the ocean environmental parameters. When the values of the environmental parameters are imprecisely known, the performance of MFP can be severely degraded. These uncertainties may arise from imprecise measurement of the environmental parameters or from random environmental effects such as surface and internal wave motion. The environmental parameters may also vary from the source to the receiver. These effects prevent the use of a single propagation model. In this dissertation, MFP methods which are robust to environmental uncertainties are derived.;Two algorithms are developed to address the problem of MFP in an uncertain environment. The first is suitable for scenarios where a small number of environmental parameters are uncertain. The Replica-Subspace Weighted Projections algorithm (RSWP) is derived as a computationally-efficient approximation to the maximum-likelihood estimator. It estimates the source location parameters jointly with the uncertain environmental parameters. The second algorithm, called the multiple uncertainty-RSWP algorithm (MU-RSWP), is suitable for use when a large number of environmental parameters are uncertain. It is based on a computationally-efficient approximation to the maximum a posteriori (MAP) estimator. Both of these methods also have interpretations in terms of vector spaces. Results are shown for both algorithms using simulated and experimental data. It is also shown that the MU-RSWP algorithm can be applied directly for robust MFP when using a short array.