Seamlessly Unified Reasoning Under Uncertainty for Missile Defense Architectures

Scientific Systems Company Inc. (SSCI) of Woburn MA and its subcontractor Lockheed Martin Tactical Systems (LMTS) of Eagan MN propose a radically new approach to expert systems theory that will greatly facilitate the effective design of BMD decision architectures. We will show that, as most commonly employed in the problem domain of interest---fusing information to obtain estimates of target states and estimates of uncertainty in those states---evidence-fusion using Dempster-Shafer, fuzzy logic, and rule-based can be rigorously subsumed within the Bayesian theory. Thus fusion of ambiguous measurements using Dempster's rule of combination '*' or Zadeh's conjunction operator '^' produce exactly the same answer as fusion using Bayes' rule. Rules---including second-order rules---can be processed in a recursive Bayes filter as just another kind of data. From a BMD decision architecture point of view, the following are immediate consequences: (1) the relationships between, and relative strengths and weaknesses of, the different expert systems approaches can be directly compared; (2) algorithm and architecture design can become more systematic and integrated; and (3) it is no longer necessary to "re-invent the wheel" when using fuzzy logic, Dempster-Shafer, and rule-based methods since we can draw from a huge "toolbox" of established probabilistic concepts and techniques. The approach can be rigorously extended to multisource-multitarget detection, tracking, and identification. The project team includes Dr. Ronald Mahler of Lockheed Martin. Lockheed Martin will provide both technical and commercialization support in the application of expert systems approaches to multisource-multitarget tracking.