摘要
车间资源分配知识是以专家历史经验为基础,由粗模糊集理论获取调度规则知识,从而来进行资源的分配。这种规则知识在短时间内有较高的可靠性,但伴随现在制造环境的急剧变化,知识陈旧化是一个必然的问题。对此提出了一种具有自学习能力的动态调度决策机制,其知识迭代的更新采用一种具有双向学习能力的改进算法,从而使系统在运行过程中能自动感知环境的变化,不断进行自适应、自学习的知识更新;最后,描述动态调度决策系统的决策及知识更新过程算法,并通过仿真进行了分析验证。
The resource distributions in JSP are based on the experience of experts,the scheduling rules are obtained by the rough and fuzzy sets. These rules have a higher reliability in a short time. But with the rapidly changing manufacturing environment,obsolescence of knowledge is an inevitable problem. This paper presented a dynamic scheduling mechanism with self-learning ability. The knowledge was updated by a two-way learning algorithm,so that the system could sense the changes in environment automatically and update the knowledge. Finally,the decision-making and knowledge update algorithm were described and analyzed by simulation.
出处
《计算机应用研究》
CSCD
北大核心
2010年第10期3834-3836,3840,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(60874075)
江苏省高校自然科学基础研究面上(07KJB520139)
关键词
动态调度
合同网
强化学习
自学习
dynamic scheduling
contract net protocol
reinforcement learning
self-learning
