Decentralized multiplatform data fusion
Hugh F Durrant-Whyte, Rob H. Deaves, Phil Greenway · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1998
The decentralized data fusion paradigm is the more general case of centralized fusion. Although centralized fusion is currently well understood and applied widely, a decentralized approach offers many (potential) advantages. These include significant improvements in modularity, fault tolerance, and scalability. In particular, decentralized techniques directly support the development of 'plug and play' systems, and can provide valuable insights into the relationship between system level performance and that of sensor sub-systems. There can also be a provable equivalence to optimal centralized algorithms. This paper describes the design of a preliminary, comprehensive demonstration of the concept of fully decentralized, modular sensing technology. The demonstrator will allow for multiple sensor platforms each equipped with one or more range-only, bearings-only, or range and bearing sensors providing limited view, asynchronous measurements of the target environment. This work builds on the latest developments in the information-based representations of the data fusion process, as these make explicit the value (or utility) of each data fusion decision.