Optimal semi-adaptive search with false targets

Abstract

Searchers frequently encounter the presence of false targets or clutter, which appears indistinguishable from the real target and must be identified in a second stage of the search. False targets can significantly impede search operations, such as underwater recovery and mine warfare, when contact investigation is costly. Current literature optimizes these searches by applying Bayesian updates to the prior distributions for the real and false targets, in what is called a semi-adaptive search. We take full advantage of intermediate search results, along with soft information about the target, to build up-to-date maximum likelihood estimates of the location of the real target and the distribution of the clutter. Using these estimates in place of the priors, we update and improve the allocation of search effort as the operation progresses. In a detailed simulation study, this new approach increases the probability of finding the target by up to 12% over the optimal semi-adaptive plan without such estimates. These gains are robust to variation in the false target density, time to identify false targets, and total search time available.