Agent based modeling (ABM) actually refers to two different types of modeling.
In one case, the agents (typically represented by dots) have a small set of rules of behavior (such as move toward dots of one color and move away from dots of another color). A single model run over a few hundred model time units may take only a fraction of a second. Hundreds of thousands of runs are used to analyze the emergent behaviors of the dots with respect to the rule settings. When a behavior is exhibited that seems to match some behavior of animals (e.g., flocking of birds or schooling of fish), these rule settings are regarded as generating the behavior. With enough experience, small models can be built that may replicate some complex situation (e.g., the actions of children playing the "capture the flag" game, or perhaps a simple military operation).
In the second case, the agents have much more complex sets of behaviors that are intentionally designed to employ the logic of the real world agents they are meant to represent. A single model run may represent the same few hundred model time units or multiple thousand model time units and may take several minutes of computer time. The number of replications possible in this case drops into the realm of tens or possibly hundreds of runs.
Select one of the links below.
|2001||Defense Modeling and Simulation Office (DMSO)||Specify the architecture and create an OOTW Toolbox.|
|2012||Air Force Research Laboratory||Support the GLENS project development.|
Return to Website Entrance.
Solving Complex Operational and Organizational Problems
Dr. Dean S. Hartley III, Principal