Information loss due to environmental variability and uncertainty in Bayesian localization of a narrowband source

Abstract

Performance degradation of Bayesian localization of a low-frequency narrowband acoustic source due to variability and imperfect knowledge of the acoustic environment is investigated in a computational study. The environmental variability is modeled as arising from water column fluctuations associated with a diffuse random linear internal wave field in a shallow-water ocean waveguide. The ambient noise spatial cross-spectrum is represented by a Kuperman-Ingenito model. For the case of complex Gaussian internal wave spectral amplitudes, a closed-form expression is derived for the conditional pdf, given source location, of the signal spectral values received on an acoustic array. Examples computed for a vertical receiver array quantify localization performance degradation as an increase in the entropy of the Bayesian source-location posterior. The effects of model bias, model spectral uncertainty, and medium variability as determined by the internal wave power spectrum are quantified separately and jointly. Potential extensions to more general models of medium variability are discussed. [Work supported by ONR.]