Using Large-scale Multi-agent Simulation to Study Methods for the Detection, Tracking and Prediction of Threats

Clayton Morrison
University of Southern California

New tools are now available to study complex social
organizations. At the Center for Research on Expected Events (CRUE,
USC ISI), we use large-scale multi-agent simulation as a testbed to
study methods for detecting, tracking, and predicting present and
emerging threats to complex organizations. I will present two
simulators we have developed for this work. The simulators stand in
contrast to each other in the emphasis they place on population
features, and their use as analysis and decision-making tools. The
Hats Simulator emphasizes very large populations of simple agents
interacting over time, a very small portion of which plan, organize
and carry out malicious behaviors. All behavior (malicious or benign)
is structured around group cooperation to achieve goals. Hats has
proven to be a useful tool for analyzing mathematical and statistical
approaches to threat prediction and detection. From a different
perspective, the Assymetric Threat Assessment Tool (ATAT) models more
complex organizational structures in a spatial/geographical domain,
but on a smaller population scale than Hats. ATAT models the timing
and location of events based on feeback from prior event history and
the psychological, military, economic, social, infrastructure and
information (PMESII) factors of local populations. ATAT is a
prototype testbed in which to explore possible action policies in
complex and unstable populations.

Audio (MP3 File, Podcast Ready) Presentation (PDF File)

Back to Crime Hot Spots: Behavioral, Computational and Mathematical Models