To prepare soldiers for missions in densely populated environments, simulation-based training needs to immerse them in a similarly populated training environment. One challenge in simulating such an operationally realistic experience is re-creating the same patterns of life that the real population exhibits in response to alternative courses of action. The complexity of such behaviors and of their interdependency with military operations have motivated a need for AI methods that can generate realistically dynamic patterns of life. Unfortunately, most existing behavior-generation approaches rely on manual scripting of patterns of life, with insufficient flexibility to support the “free thinking” that real-world civilian populations exhibit on a daily basis. We have instead applied a multiagent social-simulation framework, PsychSim, for autonomous behavior generation for individuals and groups, across the range of socio-cultural backgrounds relevant to a simulated operating environment. PsychSim provides reusable mechanisms for the cross-cultural decision-making that forms the basis for the patterns of life implemented in this work (e.g., an individual may choose a route to work that avoids the site of a recent firefight, a crowd of civilians may decide to cheer or protest a blue-force unit’s actions). We use decision-theoretic agents to choose the behavior they think best advances their goals. Unlike typical agent-based social simulation, the agents’ behavior will not be determined by manually authored rules. Instead, the agents will form perceptions of their current situation, their current options, and the relative desirability of those options in terms of their expected outcomes. The agents will thus be sufficiently free-thinking to respond in a robust way to whatever situation they find themselves in, regardless of what path the human behaviors or exogenous events have taken the scenario. We illustrate how this underlying foundation can support a variety of relevant patterns of life taken from operationally relevant scenarios.
Keywords
AGENT-BASED SIMULATION,AI,PATTERN OF LIFE,SYNTHETIC ENVIRONMENT
Additional Keywords