Numerical Optimization of Search for a Moving Target

Abstract : The problem of computing an optimal search plan for a moving target is addressed. The majority of the report describes numerical techniques which have been developed to compute optimal search plans for the very broad class of problems in which the target's motion can be modeled by a discrete-time stochastic process and the detection function is exponential. A very efficient algorithm is given to find optimal search plans when the target's motion is modeled by a mixture of discrete time and space Markov processes. A second algorithm is presented to solve the variation of this problem that one encounters when the search effort at each time period is restricted to be uniform over an arbitrary rectangular region. The latter is intended to approximate the problem of choosing a sequence of sonobuoy fields to maximize the probability of detecting a submarine. Examples show that one can often find rectangular plans that are almost as effective as the optimal plan. In addition to the above, an algorithm is presented to find optimal plans for arbitrary discrete time target motion processes which can be modeled by Monte Carlo simulation. All the algorithms have been programmed in FORTRAN and run on a Prime 400 minicomputer. Examples of optimal plans calculated by these algorithms are presented.