A computational efficient Bayesian approach for active sonar localization under uncertainty in sound speed profile

Active sonar localization is challenging due in part to boundary interactions, relatively small aperture constraints, and the short coherence time associated with mobile bodies and dynamic environments. To address these challenges, a computational Bayesian approach is employed for joint inference of the wavevectors, characterizing the scattered field associated with the angle/Doppler spread arrivals. The resulting posterior is mapped to the posterior of the scattering body's location and speed under an uncertain sound speed profile (SSP) using a second-order variational Bayesian method. The approach employs a multivariate Gaussian model of the SSP with a lower-dimensional subspace representation of SSP uncertainty. The mobile scattering body’s range, depth and speed joint posterior is constructed using a computationally efficient solution based on the Laplace approximation and marginalization over uncertainty in sound speed. A case study using SSPs from the Mediterranean Sea is presented to lend credence to the approach. [Work supported by NUWC ILIR and ONR.]