规则驱动的自适应多代理系统研究

项目名称： 规则驱动的自适应多代理系统研究

项目编号： No.61202101

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 肖亮

作者单位： 湖北工业大学

项目金额： 26万元

中文摘要： 当我们设计一个软件系统时，对需求的了解注定是不全面的，因为这方面的知识很大一部分将在未来慢慢明了。因此我们必须依靠好的设计：一种可以提高系统自适应性的技术，而不是试图掌握全部需求知识来减轻因其连续改变而造成的软件维护负担。我们提出结合多代理技术和模型驱动体系结构的研究方法来提高软件的自适应能力。在尊重现有基于组件与服务的基础设施投资，充分复用这些构件的前提下，引入多代理系统作为高层抽象，同时用可反复配置的规则模型捕捉需求知识，实时驱动代理行为。当需求发生变化时，业务专家动态配置规则，通过代理的实时解释和执行，系统自动改变其行为模式并适应新的需求，这有可能体现在建立新的业务协作关系，作出新的业务决策，处理新的业务任务，完成新的业务流程等。同时业已开发的组件与服务等可复用计算资源被代理按需重新选择、绑定、调用，来支持新的业务行为。通过对需求模型与组件服务的持续配置与复用，大大提高软件自适应性。

中文关键词： 软件自适应；多代理系统；需求建模；规则表达与解析；基于多代理系统的临床决策支持

英文摘要： Our understanding of the requirements is doomed to be incomplete and will never be perfect when we design a software system, simply because a large part of them will come into clearer recognition in the future. For this reason, we must rely on good design, one that can improve software adaptivity, rather than attempt to capture as much requirements knowledge as possible to relieve the burden of software maintenance that caused by the continuous requirements change. To this end, we propose a research method that combines Multi-Agent System and Model-Driven Architecture to improve software adaptivity. Recognising the vast investment already made on software components and services by the industry, our method will try to make full use of them. At the same time, we introduce Multi-Agent System as the high level abstraction. Rule-based knowledge models are used to capture requirements. These are configurable and able to drive agent behaviour at runtime. Whenever requirements change, domain experts reconfigure the business rule models, agents interpret and execute the new rules dynamically, and new requirements are adopted. This may be the establishment of new business partnership, making of new business decisions, processing of new business tasks, and completion of new business processes. At the same time, the compon

英文关键词： Software Adaptivity；Multi-Agent System；Requirements Modelling；Rule Representation and Interpretation；Agent-oriented Clinical Decision Support