Efficient Employment of Imperfect Search Sensors in Complex Environments
Abstract
Defense planners must strive to develop and incorporate new, efficient procedures to allocate scarce resources in varied complex environments. We consider two discrete-time, discrete-space search effort allocation situations. Both involve the employment of an imperfect sensor, which is subject to both false-positive and false-negative errors. The area of interest, comprised of several disjoint area-cells, contains a single target of interest. In the first situation, the target moves according to a Markovian transition matrix, which is unknown to the sensor operator. The objective is to estimate the target's steady-state distribution, using only the sensor's detection signals and knowledge of its false-positive and false-negative rates. The second situation considers a stationary target, wherein the objective is to determine the area-cell occupied by the target, in the fewest expected number of investigations, to within certain operator-prescribed error tolerances. We develop an adaptive algorithm based on stochastic approximation for the first situation, and show that the resultant rate of error in determining target presence/absence in any area-cell converges to zero at the fastest possible rate. We propose a sequential elimination procedure for the second situation, which provides an efficient determination of target location and guarantees its error rate not to exceed the operator-prescribed tolerance.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2009
- Accession Number
- ADA510055
Entities
People
- Kurt E. Wilson
Organizations
- Naval Postgraduate School