Adaptive real-time threat assessment under uncertainty and conflict

This paper describes an adaptive fusion human-machine system supporting an analyst in recognizing threat. The system is designed in the framework of Transferable Belief Model for processing complex unreliable and uncertain data streams coming from multiple sources to improve threat recognition and detect new 'unknown' threat. The focus of the paper is on the latter: designing a method of detection of possible 'unknown' threat. Such problem arises in the open world environment, for example, in multisensor automatic target recognition systems, or in the situation assessment problem requiring selection of, sometime unknown, hypotheses about the state of the environment. The method of detecting possible unknown threat is based on the notion of conflict, which happens in the uncertain environment when multiple sources of information disagree. The results of a case study designed for sequential decision making for threat recognition in the littoral environment are also presented.