A Bayesian Network Model for Tracking in Support of Discrimination

"The success of hit-to-kill missile defense systems depends on correct discrimination of an entire collection of tracks. A number of organizations are now building Bayesian Network (BN) system models that assume near-perfect tracking knowledge, and theseBNs use single tracks and collections of tracks to define their inputs. An additional capability in support of this effort is the modeling of imperfections that occur in most realistic tracking systems, such as misassociaqtions, track termination andreinitiation, track spawning, track switching, and spurious and redundant tracks. On the other hand, Numerica is developing a Multi-Frame Assignment (MFA) tracker as a state-of-the-art tracking systems for BMD applications, but this tracking system doesnot perform target discrimination.Currently, the following issues need to be addressed: (1) inclusion of tracking knowledge in the system model, (2) reasoning using collections of tracks to understand the situation and to enhance the discrimination results, (3) incorporating results of thereasoning process to enhance tracking.To address these three issues, Numerica will first represent track imperfections in a BN. Second, Numerica will design an interface between the tracking system and the discrimination system such that the discrimination process starts, evolves, and reachesa conclusion. Numerica, Inc. has a very successful record of developing algorithms(specifically in the target trac