The Methods of Model Integration Using Agents

摘要

Models are the most important components of forecasting support systems (FSS), decision support systems (DSS) and halls for workshop of meta-synthetic engineering (HWME). As we use these systems to deal with a complex problem, one simple model will not be competent for problems solving. The huge complicated models, which are hard to build and to utilize, are needed on this occasion. Fortunately, the model integration provides us another selection. The model integration is to combine some simple models for solving a certain complex problem But the model integration is difficult to implement in the traditional FSS, DSS. In multiagent based model system the units are agents, including a management agent, many interface agents and a lot of model agents. Every model is encapsulated by an agent, which can be called by user utilizing interface agents and other agents. In this paper, the model integration method using the cooperation of model agents is proposed. The cooperation of model agent is classified to three types: serial, parallel, and extended of serial and parallel. In the serial cooperation, inputs of a model agent are outputs of another. In the parallel cooperation, two agents share the same data, compute with their own algorithm and get results. And in the extended cooperation of serial and parallel, there will be at least three model agents, both serial and parallel cooperation exist among them. We can use interface agent to define cooperation of model agents.