DDF: An Evaluation of Covariance Intersection

DDF is an ONR-funded effort to evaluate the possible role of Covariance Intersection (CI) in future naval network-centric warfare (NCW) systems. Such systems, based on the concept of a plug-and-play information grid, are expected to yield key operational advantages, e.g., robustness and flexibility. In order to deliver these advantages, NCW systems require an effective way to control the data incest problem. This is a consequence of the proliferation of redundant data in general networks due to multiple circuits around point-topoint communication loops. Conventional data fusion methods, based on Kalman filters, require strict independence assumptions which cannot be satisfied in general network architectures. This paper proposes CI as the corner-stone of a rigorous data processing solution to the general decentralised data fusion problem. The proposal is based on the following observations: (1) CI requires no independence assumptions; (2) CI does not require the architectural constraints, which yield the NCW system benefits, to be relaxed; (3) CI fits naturally into a compelling mathematical foundation for NCW based on probabilistic information update; (4) CI can be integrated with Kalman methods to form hybrid data fusion algorithms for situations in which partial prior knowledge about the network topology can be assumed.