On Non-intuitive Agent Dynamics in a General Complex System Simulation

Ted Carmichael and Mirsad Hadzikadic

Complex Systems Institute, College of Computing and Informatics, UNC Charlotte, Charlotte, NC.

TITLE: On Non-Intuitive Agent Dynamics in a General Complex System Simulation

ABSTRACT: We have implemented a simulation tool designed to serve as a general CAS (Complex Adaptive Systems) model, and – in previous work – have applied this model to both soft-tissue cancer and political dissent in a polity [1]. Here we focus on interesting behavior of the tool itself, rather than that which is specific to a particular domain.

We highlight two aspects of this system’s behavior that we found to be particularly surprising. First, and as shown in Figure 1, this simulation tool is able to replicate certain clusters of agents, and certain expected shapes for these clusters, such as that found in [2]. Our model can be thought of as a type of predator-prey system, with the patches as grass that grows and spreads, and the turtles as agents that “eat” the grass. However, unlike [2] and some other swarming models (such as found in [3]) the selforganizing behavior of the turtles in this model does not rely on communication between the turtles. At least not direct communication: the turtles cannot see or sense other turtles in any way; they can only affect each other indirectly, via their individual effects on the patches. Thus, direct interaction among swarming agents seems to be a sufficient, but not a necessary condition for swarming behavior.

Figure 1. Similarities of emergent swarm shapes between [2] and our CAS model.

The second aspect that we found interesting regards a number of examples of individual behavior for the agents that one would generally think of as being less effective, or less efficient. These may include agents that move more slowly across the environment, or requiring more “food” in order to reproduce. Yet in many instances such lower-efficiency behavior actually produces an overall benefit to the agent population as a whole, such as by increasing the total population of these agents, or reducing volatility in the population size. Thus, the species that these agents are a part of are stronger or more stable over all, even though the individuals are less so. We will highlight some specific examples of such phenomena and show how these affect the overall dynamics of the system.

References

1) Carmichael, T., Hadzikadic, M., Dréau, D., Whitmeyer, J.: “Characterizing Threshold Effects Across Diverse Phenomena,” in Advances in Information and Intelligent Systems, Ras, Z., Ribarsky, W., Eds. (Springer, New York, 2009).

2) Hawick KA, James HA, and Scogings CJ: “A Zoology of Emergent Patterns in a Predator-Prey Simulation model.” Computational Science Technical Note CSTN-0015, Massey University, March 2005.

3) Wilensky U: NetLogo, 1999. http://ccl.northwestern.edu/netlogo. Center for Connected Learning and Computer- Based Modeling. Northwestern University, Evanston, IL.

Back to Swarmfest2010_program