Computational Bayesian inference for active sonar localization under uncertainty in sound speed profile

Area of Science:

• Ocean acoustics
• Signal processing
• Bayesian inference

Background:

• Active sonar localization faces challenges including small receive apertures, uncertain sound speed profiles (SSP), and limited coherence time.
• Accurate localization is critical for underwater object tracking and mapping.

Purpose of the Study:

• To develop a computational Bayesian approach for robust active sonar localization.
• To address challenges posed by uncertain sound speed profiles and limited acoustic data.
• To characterize the scattered acoustic field and infer target location and speed.

Main Methods:

• Inference on wavevectors from angle/Doppler spread arrivals to characterize the acoustic field.
• Mapping wavevector posterior density to scattering body location and speed using eigenray interpolation and marginalization.
• Modeling SSP uncertainty with a multivariate Gaussian and low-dimensional subspace modes.

Main Results:

• The proposed Bayesian method effectively handles small receive apertures, uncertain SSPs, and limited coherence time.
• Eigenray interpolation and marginalization successfully map wavevector information to target characteristics.
• Demonstrated efficacy in a case study using Mediterranean Sea sound speed profiles.

Conclusions:

• The computational Bayesian approach provides a robust solution for active sonar localization under challenging underwater conditions.
• Accurate localization is achievable even with significant uncertainties in the sound speed profile.
• The method offers a significant advancement in underwater acoustic sensing and target tracking.