Simulation is widely used for military training and experimentation. In many cases the local populace is a vital part of the virtual world; whether it is the interaction they have with trainees or the added realism they provide through their presence. However, creating a scenario with a realistic population is a laborious and time consuming process, and therefore has to be omitted or simplified quite often.

Although the behavior of the locals, their daily pattern-of-life, is inextricably linked with the features and infrastructure of the physical terrain in which the population lives, they are currently still being treated as disjunctive assets in most computer generated forces tools. Therefore, we propose an integration of physical and human terrain in the generation process that has the potential to i) reduce the effort needed for an instructor to create and maintain scenarios, and ii) improve the realism of the simulation.

The objective of this research is automatic generation of a local population and their typical behavior, based on general rules for daily life, resulting in, e.g., people leaving in the morning to go to work in the fields or factories, children leaving for school. These generated behavior patterns abide by local demographics and physical terrain characteristics, such as the locations of residential areas and other infrastructural elements.

This paper describes a novel, instructor-centric approach to scenario creation, which considers human and physical terrain data as a combined source for the generation of population behavior. The authors discuss the design of the human terrain data model, and its incorporation into an existing physical terrain generation process.