Spatial animat agents can be used to construct sophisticated spatially rich macroscopic models to study complex and emergent phenomena based only on localised microscopic control parameters. We have developed a predator-prey based animat model that we have successfully used to explore collective behaviours including herding, battlefront formation; segregation and population control. We describe our model architecture and how such agent based models can support large numbers (around one million) of animat agents for multiple generations of coexistence. In studying collective phenomena using compute r models it is important to develop quantifiable metrics and measurement apparatus in tandem with the model itself. We discuss some of the macroscopic metrics and statistical measurement approaches we have used to relate localised animat parameters to the emergent patterns of behaviour identified in our system. We show how these quantifiable approaches could be applied more generally to other agent-based models for decision support applications.
Keywords: spatial animat; complexity; emergence; quantifiable metric; decision support.
Full Document Text: PDF version.