Bayesian Optimization with Gaussian Processes for Robust Localization

Abstract

We present a sample-efficient Bayesian optimization (BO) method to estimate underwater source localization robust to unknown tilt in a vertical line array. Rather than conducting exhaustive search of parameter space to estimate localization and tilt, BO uses a Gaussian process (GP) surrogate model of the Bartlett power objective function to guide sampling of the parameter space. Samples are suggested using a heuristic acquisition function that uses the GP to balance exploitation and exploration of parameter space. Using experimental data, we show that BO obtains better localization estimates than conventional grid search and quasi-random sampling strategies, and that robustness to array tilt comes with little additional computational cost.