Solving disagreements in a multi-agent system performing Situation Assessment

Situation assessment and decision making in monitoring and surveillance scenarios are evolving from centralized models to high-level, reasoning oriented, net-centric models, according to new information fusion paradigms proposed by recent research. In this paper, we propose the design of a multi-agent architecture for situation assessment, where situations are classified through agent collaboration, in order to provide human operators with a synthetic vision, that points out the elements of the scenario that require human intervention. Specifically, our approach provides for: i) high level classification of situations based on OWL ontology reasoning; ii) distributed assessment by a protocol which solves disagreements possibly arising among agents' conclusions. Experiments in a real maritime surveillance scenario show that the proposed approach approximates the results of a centralized architecture, while preserving independency of decision makers and dramatically reducing the amount of communication required.