Defence Applications of Agent-Based Information Fusion.

摘要

Information fusion and agent-based systems provide core capabilities in the NEC operational environment: uncertainty reduction and coordinated action. These are key enablers of situation assessment and effective resource allocation. In the military domain, uncertainty encompasses ambiguity and incompleteness, requiring flexible information fusion algorithms. Results and analysis are presented here for a Bayesian consensus algorithm applied to fault recovery in a sensor network, where some of the sensor fault types are unknown. The need for coordination raises architectural questions about where coordination is done and about its information exchange requirements. Two options are analysed: distributed coordination in which nodes establish shared situation awareness and run replicated central coordination algorithms, and decentralised agent-based coordination which is based on a market-based control representation. Decentralised coordination is shown to be more responsive than distributed coordination in sparsely connected networks with delays.