Distributed Battle Management Techniques

Many technical issues need to be addressed before a distributed BM/C2 architecture can be designed and implemented. Specific issues include: what should each BM/C2 node know in terms of models, sensor data, and overall system goals? What algorithm isused by each node? How do the BM/C2 nodes coordinate their processing in order to avoid conflicts and achieve high performance? What happens if there is latency in communication? In this Phase I effort, Information Extraction and Transport, Inc. (IET)proposes to support the Ballistic Missile Defense-Decision Architecture (BMD-DA) efforts to build a generic BM/C2 decision architecture by developing algorithms that allow multiple cooperating BM/C2 nodes (agents) to coordinate their decisions to achievenear-optimal performance in a distributed BM/C2 architecture. These algorithms are designed to estimate and communicate the impact of evidence on object classification using learned Bayesian network factoring of posterior joint likelihood distributions.We will also develop algorithms that estimate the performance of learned Bayesian network factorings these posterior joint likelihoods. The result will be an experimental harness where we will demonstrate the feasibility and utility of using learnedBayesian network factorings of posterior joint likelihoods to support efficient distributed situation assessment and dissemination. Information Extraction and Transport, Inc. (IET) anticipates that benefits from Phase I of the proposed SBIR effortinclude: 1) a methodology for learning a set of Bayesian networks that represent the posterior joint likelihood over a set of variables that are shared by BMDS elements; 2) measures for the impact of evidence on a BMDS' system model and for estimating theresources and bandwidth requirements given a particular Bayesian network for the posterior joint likelihood; 3) a strategy for choosing the best learned network structure given the evidence, available resources, and impact; 4) an approach for updating thesystem model at a remote BMDS element given the selected Bayesian network for the posterior joint likelihood; and 5) an experimental harness to support the evaluation of this approach. The technology proposed has significant commercial application. Themarket for battlespace sensor control and processing software will be significant within the DOD and Intelligence Community. Critical mission areas that could be supported include force protection, facility monitoring, and mobile target tracking.Commercial sensor markets could benefit as well including applications using microsatellites, terrestrial and extraterrestrial robotics, and unmanned vehicles (e.g., UAVs and UUVs).