Situation assessment via Bayesian belief networks

Abstract

We present here an approach to battlefield situation assessment based on a level 2 fusion processing of incoming information via probabilistic Bayesian Belief Network technology. A belief network (BN) can be thought of as a graphical program script representing causal relationships among various battlefield concepts represented as nodes to which observed significant events are posted as evidence. In our approach, each BN can be constructed in real-time from a library of smaller component-like BNs to assess a specific high-level situation or address mission-specific high-level intelligence requirements. Furthermore, by distributing components of a BN across a set of networked computers, we enhance inferencing efficiency and allow computation at various levels of abstraction suitable for military hierarchical organizations. We demonstrate them effectiveness of our approach by modeling the situation assessment tasks in the context of a battlefield scenario and implementing on our in-house software engine BNet 2000.