Analysis of and heuristics for sensor configuration in a simple target localization problem

We investigate Bayesian methods and heuristics for management of a configurable sensor in a simple target localization problem. A target is located in one of M cells. A sensor, characterized by probabilities of correct detection and false alarm, repeatedly chooses a cell to interrogate; the resulting observations are used to update the posterior probability distribution of target location. Interrogations are repeated either a fixed number of times or until the probability of error drops below a pre-selected threshold. The Bayes optimal solution is exponentially complex, motivating the use of heuristics. Four heuristic rules are characterized using Monte Carlo simulation. Of these heuristics, choosing the most probable cell minimizes the number of observations, and the myopic Bayes optimal rule minimizes the probability of error.