Variable Resolution Modeling for Distributed Battle Management

Missile defense battle management is an inherently distributed process, involving sensors, weapons, and targets that are distributed over a region spanning thousands of miles. While most agents need some sense of "distant events", they neither need nor can afford detailed, high-fidelity models of the entire situation. CleverSet's approach to distributed battle management is based in variable resolution modeling using Relational Influence Diagrams. This approach integrates recent advances in relational data modeling, Bayesian inference and data fusion, and distributed processing with limited bandwidth, in a scalable, open-platform, standards-based modeling environment to produce extensible and future-ready models for distributed battle management. These models provide a basis for estimation and control across a network of heterogeneous sensors and weapons connected by a potentially unreliable and constrained communication network. In this SBIR Phase I proposal, CleverSet proposes to develop representations and algorithms for distributed inference and control over a network of heterogeneous sensors and weapons connected by a communication network. The approach will be evaluated within CleverSet Modeler, CleverSet's relational Bayesian modeling environment using CleverSet's patent-pending synthetic variable language.