An Abductive Framework for Level One Information Fusion

This article argues for, and describes some of the advantages of, construing level one information fusion, as a task of abductive inference or inference to the best explanation. Such an approach enables certain benefits, such as, an expectation-based critique of hypotheses, and an elegant system for revising old beliefs, which may gainfully be exploited. It also introduces several relevant dimensions to reasoning based on the explanatory relations between hypotheses and data, that are closed to traditional approaches. The design principles of a software system, Smart-ASAS, that attempts to solve the level one fusion task of entity tracking and re-identification, are described, along with an example that illustrates its capabilities