A strategy distribution of a self-repairing network in dynamic environments

• Authors: Masahiro Tokumitsu
• Venue: Proceedings of the International Conference on Management of Emergent Digital EcoSystems
• Year: 2011

This paper considers adaptation of autonomous distributed systems in uncertain environments. We investigate strategy distributions among agents against the environment and their convergence speed toward steady state. Adaptation to the environment is a central and challenging issue faced in a research on autonomous distributed systems. Cooperation among the agents is also a crucial issue for preventing absorbed states of the systems. We investigate the performance of a self-repairing network in dynamic environments by multi-agent simulations. We introduce two parameters for the systemic payoff: a connection weight and a neighborhood selection probability. Dynamic environments follow certain types of probability distributions, with failure rates changing as time evolves. We evaluate the performance for some distributions of the failure rates. We consider the obtained strategy distributions and the relationship between convergence speed and the shape of the strategy distributions. As a result, we obtain a long-tail exponential curve for the strategy distribution that can adapt to changes in the environment. Furthermore, we reveal that the neighborhood selection probability and connection weight affect the convergence speed toward steady state.