Genetic fuzzy systems to evolve coordination strategies in competitive distributed systems

This paper suggests an evolutionary approach to design coordination strategies, a key issue in distributed intelligent systems. We focus on competitive strategies in the form of fuzzy rule-based models. The aim is to evolve data and rule bases to improve agent performance when playing in a competitive environment. In this situation, data for learning and tuning are rare and rule base must jointly evolve with the database. We suggest a genetic algorithm whose operators use variable length chromosome, a hierarchical relationship among individuals through fitness, and a scheme that successively explore and exploits the search space along generations. Evolution of coordination strategies uncovers unknown and unexpected agent behaviors and allows a richer analysis of negotiation mechanisms and their role as a coordination protocol. An application concerning an electric power market illustrates the effectiveness of the approach.