Distributed classification in a multi-source environment
Tod M. Schuck, J.B. Hunter · 2003
Current methods of determining the classification of objecls (such as aircrafi) across the battlespace in military systems are by design, stovepiped, plafform-centric designs. Classification data and information are generally either sent at a very low level (e.g. every contributing response) or worse, as a decision such as FRIENDor F-16 without metrics lo quanrifi the level of assurance of the decision. The former method potentially overwhelms a large network when many nodes are simirltaneously reporting and receiving information. The latter method leaves each receiving node with no information to help resolve diflerences between its local decisions, and those coming from other nodes. In this paper, a method to share information that allows for distributed decisions across multiple nodes is given. It is based on a classi/ication vector that conveys probabilistic and evidential object informarion, as well as measurements of information content and confusion.