Distributed Object Discrimination for BMD

Object discrimination is one of the most important functions in ballistic missile defense since successful engagement and intercept of the warhead requires timely discrimination of lethal objects from decoys and other non-lethal objects. Multiple sensors exploiting diversity in phenomenology and viewing geometry can provide better discrimination information than a single sensor but exploiting this information requires a good fusion algorithm. Although centralized discrimination is theoretically optimal, distributed discrimination has advantages of lower communication bandwidth, robustness to failure, etc. The proposed research addresses key issues in distributed discrimination including choice of appropriate architecture, information to communicate among processing agents, optimal fusion algorithms, and communication strategies. It adopts the information graph model to analyze the dependence among processing agents so that information will not be double-counted. Object and sensor models represented by Bayesian networks are used to identify the minimum sufficient information that should be communicated and fused. An information-theoretic approach is used to decide when communication should take place. The Phase I effort will demonstrate the technical feasibility of distributed discrimination by integrating these components and comparing the resulting performance with that using centralized discrimination.