BRECCIA: Unified Probabilistic Dynamic Geospatial Intelligence

Abstract

The study and analysis of geospatial data has progressed from simple Geographic Information Systems (GIS) (spatially organized layers of digital data with associated functional elements to broader geocomputation systems. Applied to Geospatial Intelligence analysis, this involves the combination of information expressed in logical form, computational form, as geographic information, and as sensor data. Each of these forms has its own way to describe uncertainty or error: e.g., frequency models, algorithmic truncation, floating point roundoff, Gaussian distributions, etc. We propose BRECCIA, a multi-agent Geospatial Intelligence analysis system, which receives information from humans (as logical sentences), simulations, and sensors, etc., where each piece of information has an associated uncertainty; BRECCIA then provides responses to user queries based on a probabilistic logic system which determines a coherent overall response to the query and the probability of that response, thus, ameliorating the human interaction with a complicated set of processes. In addition, BRECCIA attempts to identify concrete mechanisms (proposed actions) to acquire new data dynamically in order to reduce the uncertainty of the query response. The basis for this is a novel low-complexity approach to probabilistic logic analysis. Current knowledge-based GEOINT systems do not incorporate a broad notion of uncertainty quantification, although such a capability would allow decision makers to make more informed decisions, or to acquire more data before coming to conclusions. In addition, it would be better if system responses were provided with an explanation of how they were derived, as well as how the uncertainty was determined. In addition, intelligence, surveillance and reconnaissance support systems should generate dynamic path planning solutions which can dynamically include constraints on time, energy, or uncertainty reduction to inform the deployment of data measurement systems. The application studied here is Unmanned Aerial Vehicle (UAV) surveillance and reconnaissance in urban areas. Some work has been done in this general area. This research has led to two major research results: (1) the efficient combination of formal probabilistic logic methods with state-of-the-art physics-based uncertainty quantification methods, and (2) uncertainty driven active information data acquisition, demonstrated by UAV path planning, to optimize performance or to resolve contradictory information. In summary, BRECCIA is a dynamic, multi-agent uncertainty monitoring and reduction system; that is, uncertainty can arise due to a change in local conditions (e.g., the weather may make movement difficult) or new information may be available (e.g., obscurants in the air, new interesting sites). As a consequence, steps are taken to reduce the uncertainty of assessments, and precise reasons are offered to the user to explain uncertainty conditions and how they may be resolved.