Developing a Crowd Federate for Military Simulation

Crowds of non-combatants play a large and increasingly recognized role in modern military operations, and often create substantial difficulties for the combatant forces involved. U. S. military actions in Mogadishu, Bosnia, and Iraq exemplify the significant effects crowds may have on military operations. However, in spite of their potential significance, realistic models of crowds are essentially absent from current military simulations. For the scenarios considered likely in future conflicts the absence of crowds and of non-combatants in general would be a serious departure from realism. We are engaged in a two-phase research project aimed at developing a crowd modeling capability for military simulation. The first phase, recently completed, consisted of three parts: a requirements analysis to identify military simulation crowd modeling requirements, a literature survey to examine psychological research relevant to crowd modeling, and a design study to explore design issues in the implementation of a crowd simulation. In the second phase, now well underway, we are developing a crowd simulation, implemented as a distributed simulation federate, that will be interoperable with existing military simulations and will have a credible psychological basis for the crowd behavior it generates. In this paper we report on the crowd modeling project at an overview level. We first summarize the results of the first phase. Then the current status of the project is explained, including the process being employed to acquire direct information regarding the behavior of crowds, the design and implementation of the crowd federate, and the concept and development of historically accurate reference scenarios for use in validation and experimentation. Technical and modeling challenges (e.g., terrain correlation and quantitative psychological models) encountered so far will be identified and discussed. Finally, the future plans for the project are laid out, including two experiments planned to test the crowd federate; the first will examine the level of crowd behavior fidelity needed and the second will test the architectural reconfigurability of the crowd federate.