Reconfigurable Computing for Missile C4I

The Missile Defense Agency is developing a Decision Architecture that utilizes a Bayesian Network (BN) model to integrate the information from sensors and provide estimates of the state of the threat. This operation must be carried out in real-time. The Bayesian networks used in the Decision Architecture are complex, owing to the large number of variables involved in describing a missile defense engagement and threat characteristics. The computational complexity of exact BN evaluation algorithms precludes them from being used in this context. Approximate BN evaluation algorithms suffer from problems since it is generally not possible to accurately constrain their errors since their performance is model and data dependent. Given the difficulties in identifying an algorithmic solution it is natural to investigate hardware capabilities that can support the real-time evaluation of complex Bayesian networks in missile defense applications. Dynamics Technology Inc. proposes to investigate the potential for utilizing reconfigurable hardware to support the efficient evaluation of Bayesian network models. Preliminary investigations indicated that Bayesian network evaluation algorithms exhibit characteristics that can be exploited by customized architectures. The Phase I effort will further refine this concept, determine its feasibility and lay the groundwork for developing a prototype system in Phase II.