Sensor Management.
Abstract
This grant is suppofting the development an approach to sensor management based on information theoretic metrics. These methods can be used to optimize data collection with agile multimode sensors such as phased array radar systems, electro-optical imaging systems and passive radio frequency systems to improve detection, tracking and classification. For FY96 an information theoretic metric was extended to a more general type of probability density referred to as joint multitarget densities for use when the targets are dynamic and their number is unknown. In this application there is correlation between target-number uncertainty target-location uncertainty which must be accounted for in the statistics model and in the sensor management metric. Joint multitarget conditional probabilities are used to solve this problem. Target dynamics are modeled as Markov processes. The expected discrimination gain for sampling any cell is computed and the sensor is directed at the maximizing cell for each sample. In comparison to direct sampling, optimizing the discrimination increases the probability of detecting and localizing the targets. These results are summarized in an invited conference paper. An expanded version of this paper is now in preparation for submission to a refereed journal.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 30, 1996
- Accession Number
- ADA311311
Entities
People
- Avner Friedman
- Keith Kastella
Organizations
- University of Minnesota