Alain APPRIOU
Directeur de Recherche à l’ONERA (the french aerospace lab), Chatillon, France
Thursday 21st – 9h00-10h00
Belief Functions in Information Fusion Processes
The increased complexity of environments and operational needs requires information systems to process more and more disparate complementary sources, in order to provide a variety of information of higher level to different interactive components, whatever the organization they serve. Therefore problems arise mainly because of unsuitable, unreliable and heterogeneous data, conflicts in their interpretation, disparities in the frames of reference to consider, ambiguous associations, logic of combining, and decision principles to follow. A federative view of coherent operators that allow facing globally such a situation is provided in the framework of Belief Functions. The approach is based on a generic operator named “extension” that propagates knowledge from one frame of discernment to another, thanks to uncertain or imprecise relations that can be expressed between their elements. Beyond the benefit of its direct implementation, this operator provides a general formulation of different operators that constitute a complete, coherent and adaptive processing of multiple uncertain observations, from their modelling up to the required decision making. The particular conditions that lead to the traditional operators can be specified, and a few examples illustrate a suitable management of uncertainty processing thanks to the available tools.