Integrating Run-Time Incidents in a Large-Scale Simulated Urban Environment

The domain of Defense, Safety & Security offers interesting applications and research challenges for agent-based simulation of human behavior in urban environments. For example, virtual training of military personnel often requires complex urban scenarios to align with the current nature of operations. Previous research has focused on the simulation of a large population of civilians within an urban environment, introducing techniques such as pattern-of-life and activity generation. We have used this research as a basis for an approach that allows the injection of incidents(e.g., a crowd gathering, a shop lifting with a subsequent arrest) in a running, large-scale urban simulation. Such incidents can be used by an instructor during a training or exercise to teach a trainee to deal with unforeseen circumstances. The incidents are not explicitly scripted, but are generated by a planning algorithm within desired constraints such as what should happen, where,and when. We believe that this approach offers an intuitive and efficient way of specifying scenario-specific exceptions to daily behavior patterns in a representative urban setting. We describe an extendable prototype system aimed at demonstrating the approach. A necessary basis for the prototype system is the generation and simulation