Iterative Information Fusion using a Reasoner for Objects with Uninformative Belief Values

We describe an approach for iterative information fusion using a context-dependent Reasoner called Pequliar. The system basically consists of a query processor with fusion capability and a Reasoner with learning capability. The query processor first performs a query to produce some initial results. If the initial results are uninformative, then the Reasoner guided by the user creates a more elaborate query by means of some rule and returns the query to the query processor that executes it and returns a more informative answer. Rules may be initially specified by the user and subsequently learned by the Reasoner. Examples of iterative queries are drawn from multi-sensor information fusion applications.