The Design And Implement Of Agent Modeling In Complex System Distributed Simulation Platform

The complex systems are found everywhere in our living. The research of complex system and complex science is concerned increasingly. The agent-based modeling and simulation methodology using the behavior and idea of agent models the simulating entities in complex systems. Through the modeling and simulation of basis elements in complex systems and interactions among them, the microcosmic behaviors can be organically combined with macroscopical phenomena of emergence. Since the agent-based modeling and simulation is a bottom-up efficacious simulating method, the research of agent modeling technology has great sense in theory and practice.Firstly, this paper introduces basis conceptions and steps on modeling and simulation of complex system. The workstation of this paper, which is named Java-based Complex Adaptive systems distributed Simulation platform (JCass), is also analyzed.Secondly, according to the different influence of different rote, a ranked multi-agent-based simulation framework is proposed. The core entity in community is modeled with deliberative agent, and the others are modeled with reactive agent, which are directed by deliberative ones. The ranked multi-agent-based simulation framework can balance the overhead of computation, the difficulty of modeling and the fidelity of simulation, and achieves the ideal result of simulation.Thirdly, going on the behavior pattern of agent in JCass, using XML technology, a rulebase of agent, which is easy to be modified and queried without cutting down the efficiency of simulation, is designed and completed. To deal with various fuzzy informations in complex systems and ensure the fidelity of simulation, the fuzzy reasoning is supplied to agents in JCass. And the reasoning based on traditional Sugeno model is advanced to fit the agent-based modeling and simulation.Fourthly, for laying out the wonderful evolvement process in complex system, the population evolvement based on genetic algorithm is proposed. With the algorithm, agent populations produce excellent generations adapting environment better through selection, crossover, and mutation, resulting the self-adapting character of agent population in the simulation of the complex system.Finally, based on the ranked multi-agent-based simulation framework, using XML rulebase and fuzzy reasoning, the isomorphic model of the public opinion system in Taiwan is designed and realized. The intelligent evolutive behavior based on genetic algorithm of agent population is also used. Through the analysis on the result of modeling and simulation, the reliablity and feasibility of agent modeling technology proposed front are proved.