A Bayesian Network Model for Tracking in Support of Discrimination

The objective of the Ballistic Missile Defense System (BMDS) is todetect, identify, and track targets using sensor observations, andengage the lethal objects using interceptors.To achieve the desired level of effectiveness, the system mustdistinguish between the lethal and non-lethal objects.Object identification is performed by a discrimination systemthat relies on a tracking system to associate measurement data to eachobject being tracked.In missile defense, the presence of many closely spaced objectsresulting from debris and countermeasures makes the associationprocess very difficult.If the data association is not perfect and the resulting uncertaintyin the association is not known, then discrimination systemperformance falters.The Phase I project identified the measures of uncertainty relevant tothe discrimination process and developed an initialBayesian Network Probabilistic Tracker (BNPT) to assess thefeasibility of extracting these measures of uncertainty from aMultiple Frame Assignment (MFA) Tracker or aMultiple Hypothesis Tracker (MHT).The proposed Phase II project will develop a full-featured BNPT,integrate the BNPT with an MFA Tracking System, and develop acomprehensive querying interface.Such a BNPT/MFA system can fully support theProject Hercules Decision Architectureand represents the next generation tracking system.