Coevolutionary Game-Theoretic Multi-Agent Systems: the Application to Mapping and Scheduling Problems

TitleCoevolutionary Game-Theoretic Multi-Agent Systems: the Application to Mapping and Scheduling Problems
Publication TypeTechnical Report
Year of Publication1996
AuthorsSeredynski, F.
Other Numbers1054
Abstract

Multi-agent systems based on iterated, non cooperative N-person games with limited interaction are considered. Each player in the game has a payoff function and a set of actions. While each player acts to maximize his payoff, we are interested in the global behavior of the team of players, measured by the average payoff received by the team. To evolve a global behavior in the system, we propose two co-evolutionary schemes with evaluation only local fitness functions. The first scheme we call loosely coupled genetic algorithms, and the second one loosely coupled classifier systems. We present simulation results which indicate that the global behavior in both systems evolves, and is achieved only by a local cooperation between players acting without global information about the system. The models of co-evolutionary multi-agent systems are applied to develop parallel and distributed algorithms of dynamic mapping and scheduling tasks in parallel computers.

URLhttp://www.icsi.berkeley.edu/ftp/global/pub/techreports/1996/tr-96-045.pdf
Bibliographic Notes

ICSI Technical Report TR-96-045

Abbreviated Authors

F. Seredynski

ICSI Publication Type

Technical Report