基于多Agent的分布式模型管理与组合方法

Multi-agent based approach to model management and composition

为在分布网络环境下快速准确地进行科学决策,提出了可实现分布式的模型管理与组合的多Agent模型系统。设计了系统组成成分———管理Agent和模型Agent的结构,分析了它们的功能和行为。利用逆向推理技术实现模型Agent的动态自动组合,给出了利用多个模型Agent组合求解复杂问题的过程。最后,以某企业产品开发中的投产决策问题为例,论述了模型系统的工作过程。与传统模型系统相比,该系统具有模型远程调用和运行、模型动态自动组合等功能,为研究开发复杂问题的决策支持系统提供了新的模型系统建模方法。

To satisfy requirements of making scientific decision quickly and correctly in distributed network environment, a Multi-Agent Model System (MAMS) was proposed which could implement distributed model management and combination. The structures of MAMS constituents-management agent and model agent were designed, and their functions and actions were analyzed. Utilizing reverse reasoning technology to combine model agents dynamically and automatically, a process was then provided which applied several model agents to solve a complex problem. Finally, taking a decision making in product development as an example, the working process of MAMS was discussed. Compared with traditional model systems, MAMS endowed with functions of model remote calling & operation, automatic combining which provided a new approach to build model system for decision supporting system in resolving complex problems.