Cross-Network Association, Prediction, and Learning

A fundamental challenge in multi-domain ISR is associating observations across domains. Despite recent advances in network-based association, the amount of human analytical effort currently required to associate information between virtual domains (e.g. the same event being described in different publicly available sources of information) with sensor information from the physical domain (e.g. EO/IR images) makes it infeasible to perform large-scale cross-domain associations of observations. In this effort, we will develop new inferential methods for aggregate data association across virtual and physical domains and prediction of network structure and correlated activities and events across these domains. The research will also address challenges inherent in gathering information in virtual vice physical domains, including balancing additional network exploration in the virtual domain with tasking of sensors to obtain new observations in the physical domain.