Densely Populated Urban Environments (DPUE)
The densely populated urban environment theme explores the modeling of populations of characters using techniques from artificial intelligence, distributed systems, and the commercial game industry. The goal of the DPUE effort is to develop a simulation framework capable of representing and simulating the complex dynamics and interactions of a densely populated urban location. The DPUE effort will support simulation of large populations (tens of thousands) of goal-directed characters acting within a significant geographic regions (square miles). Unlike other military modeling and simulation systems, DPUE will focus on simulation elements necessary to support non-combat military operations. A GFY06 study is developing requirements for the DPUE system and identifying existing technologies (such as ICT's Integrating Architecture) that can help quickly satisfy these requirements.
The primary research challenge in the DPUE effort is modeling large populations of goal-directed characters at multiple levels of detail. Included in this challenge is the need to be able to dynamically shift level of detail based on the actions of the user(s) and the needs of the system. The GFY06 study will result in a technical design for a densely populated urban environment simulation framework. The GFY06 study will also identify the tradeoffs and constraints that must be considered when defining the requirements for any single environment simulation (such as the ELECT urbanSIM game). For example, the requirements for the amount of detail in the physical environment and authoring time are closely related. A large environment (a square mile or more) with lots of detail including building interiors and the usual urban clutter will require a great deal of authoring time (man months or more). If new environments must be authored quickly (i.e., in hours or days, as for mission rehearsal), then the level of detail and fidelity in the environment will need to be limited.
As another example, the requirements for total population size, level of AI reasoning for each member of the population, and necessary processing power are related. A large population of characters, each driven by a sophisticated AI behavior generation system, will necessarily require a great deal of processing power and thus not run as a single machine application. If the ability to run on a single machine is a strict requirement (perhaps for portability), then either the size of the population needs to be reduced, or the sophistication of the AI system needs to be scaled back. The DPUE effort seeks to identify these tradeoffs and constraints and explore what is possible with current portable commodity hardware. In selected key areas, the DPUE effort explores extending current technologies to help address the current restrictions. For example, the DPUE effort is exploring a variable level of detail approach to AI which can dynamically adjust the level of sophistication of the AI system controlling a character based on the needs of the user and application.
Tags: artificial, environment, intelligence, modeling, population, simulation, urban
Goals
The intended result of the DPUE effort is a simulation framework. This is not a specific simulation system, but rather an overarching organizational structure and a collection of component technologies. The goal is to be able to create a broad range of specific simulation systems by plugging different component technologies into the DPUE framework based on the target requirements. For example, in the area of social simulation, candidate approaches might include PsychSim, the CAB approach, ICT's Virtual Humans or University of Pennsylvania's Performance Moderator Function Server (PMFServ) (Cornwell et al., 2003). A specific simulation system, built on the DPUE framework, might include any one (or combine more than one) of these social simulation systems.
In GFY07 and GFY08, the DPUE requirements analysis and simulation framework will support the ELECT urbanSIM game design and prototyping effort as well as the development of the ELECT urbanSIM evaluation-ready prototype (which will occur primarily in GFY08). As described in the GFY07 ELECT proposal, the urbanSIM game seeks to place the trainee in the role of a U.S. Army officer whose area of interest is a large town that includes a population of mixed cultures (40,000-50,000 people), significant infrastructure (water, power, roads, bridges), a civic government, a local police force and a number of local and foreign insurgent factions involved in a multi-sided power struggle. The DPUE requirements analysis will play a key role in the design of the ELECT urbanSIM game to ensure that the requirements established by the design team are achievable.
The GFY07/GFY08 DPUE effort will consist of four sub-tasks. The first sub-task is the development of the DPUE simulation framework based on the technical design developed by the GFY06 study. The second sub-task, and primary research thrust, will be an investigation of a variable level of detail AI system targeted at supporting very large populations, each appearing to the user to be driven by a sophisticated, goal-directed behavior generation system. The third sub-task will be the support of the ELECT urbanSIM game design and prototyping effort. The final subtask will be an investigation of how the population-level AI models can be distributed across a multi-machine cluster to increase both the size and fidelity of the population's behavior.