The Design of Desired Collectives with Agent-based Simulation
Monday, November 3, 2003 - 2:10pm - 2:45pm
Akira Namatame (National Defense Academy)
Collective means any pair of a complex system of autonomous agents, together with a performance criterion by which we rank the behavior of the overall system. In examining collective, we shall draw heavily on the individual behavior. It might be argued that understanding how individuals behave is sufficient to describe collectives. In this presentation, I will take a different view. Although individual behavior is nested within important to understand, it is not sufficient to analyze emergent behaviors of collectives. These situations, in which an agent decision depends on the decisions of the others, are the ones that usually do not permit any simple summation or extrapolation to the aggregates. To make that connection we have to look at the micro-macro loop between agents and the collectivity. There is no presumption that a collection of interacting agents leads to collectively satisfactory results without any central authority. The system performance of interacting agents crucially depends on the type of interactions among agents as well as how they adapt to others. There are two closely related issues concerning collective, (1) the forward problem of how the fine-grained structure of the system underlying a collective determines its complex emergent behavior and therefore its performance, and (2) the inverse problem of how to design the structure of the system underlying a collective to induce optimal performance. I will discuss how agent based simulation contributes to answer these issues.