Optimization algorithm for passive source localization

This paper proposes and evaluates by numerical simulation a hybrid algorithm that combines global and local optimizations for the problem of localizing an unknown source. Coherent observations of the signal at distributed sensors result in a highly multimodal function. The algorithm is designed to find the global peak of this function efficiently and accurately. The algorithm starts as a deterministic global optimization algorithm known as DIRECT (DIviding RECTangles). With DIRECT, a lower bound is found to the unknown function over increasingly smaller intervals. When the search interval reaches the scale of the carrier wavelength, the global search switches to a local search. It is shown that this approach speeds up convergence by a factor of two compared to DIRECT, and by many orders of magnitude compared to an exhaustive grid search.