Autonomous Sensing and Deciding Framework Processor

Our objective is to develop an innovative cognitive knowledge-aided information processing framework to take very high rate intelligence data streams over wide areas and autonomously highlight Areas-of-Interest (AOIs) and targets for the image analyst without a priori knowledge of the area or location of the individual high interest targets. By using 100s of JADE based intelligent agents in Phase I, we successfully monitored AOIs, identified extended AOIs through link analysis, and detected incursions into restricted areas. In Phase II, we will extend the multi-agent system by designing, implementing, testing, and demonstrating a multi-agent autonomous sensing and deciding framework, including the development of knowledge bases, detecting anomalies between a real-time scene and the knowledge bases, mining the knowledge bases for new patterns and relationships, monitoring AOIs specified by the analyst, performing link analysis, and providing automated machine learning and ranking of anomalies within the multi-agent framework. Based on traffic pattern and pattern-of-life analysis, we derive knowledge bases and apply sound statistical analysis to detect changes between the current real-time scene and the derived knowledge bases. Through the use of Relevance Vector Machines, machine learning agents learn key characteristics that will be used to rank the importance of AOIs and individual targets.