Higher Level Fusion For Catastrophic Events

Abstract:

The core purpose of higher level fusion (situation and threat assessment) is to infer and approximate the characteristics and critical events of the environment in relation to specific goals, capabilities and policies of the decision makers. The higher level fusion processes utilize fused data about objects of interest, dynamic databases, maps, and expert knowledge, and opinion for context processing. The result of higher level fusion is a coherent composite picture of the current and predicted situation, which provides human experts with essential information to help them understand and control the situation, and act effectively to mitigate its impact. Situation and threat assessment processing has to be adaptive to resource and time constraints, new and uncertain environments, and reactive to uncertain and unreliable heterogeneous inputs.

The presentation will discuss major challenges, specific requirements, and approaches to designing higher level fusion processes as applied to the problem of crisis management.

The higher level fusion processing described in the presentation exploits synergy between cognitive work analysis and ontological analysis of the specific domain, developed within the framework of a formal ontology. The combination of cognitive work analysis and ontology provides a formally structured and computationally tractable domain representation capturing the basic structures of relevant objective reality and users’ domain knowledge and requirements. This domain representation further serves as a basis for generating domain specific situational hypotheses and high-level reasoning about these hypotheses. The dynamic situational picture is built by analyzing spatial and temporal relations of the situational entities and entity aggregations at different levels of granularity, and their dynamics provided within the overall situational context. Special attention is paid to "inference for best explanation" aimed at discovery of the underlying causes of observed situational entities and their behavior. Belief Based Argumentation system, a reasoning framework considered, represents a generalization of Probabilistic Argumentation System. It allows for allocating rational belief in hypotheses about the environment by utilizing given knowledge to find and combine arguments in favor of and against them.

The presented methodology also includes multi-step inter-level and intra-processing information exchange comprising a quality control and a belief update steps.