Methodology And Application Of Production Scheduling Based On Multi-Agent System

Multi-Agent system is one of the important research fields of artificial intelligence. It has particular predominance for solving complex distributed problems, which is an important method full of applying value. Production scheduling is the joint of control and management. On the one hand, it will supply decisions for the enterprise; on the other hand, it will arrange production tasks and supervise the control layer. Because the most scheduling problem is NP hard, it is impossible to find out common algorithm with polynomial complexity. We combine the multi-agent technique and production scheduling and decompose complex task to units by the distributed characteristic of multi-agent system. We reduce the complexity of system designing through agents' cooperation.In this thesis, the Multi-Agent technology and its application in the production scheduling were introduced. The plans distributing and assignments scheduling were integrated into the whole system by Multi-Agent technology. The main research work is described as following:Firstly, we discuss the basic concept, characteristic and classification of an agent; analyze the communication and cooperation among agents. It will prepare for scheduling system based on Multi-Agent System.Secondly, we analyses the traditional job-shop scheduling system deeply, and creates a model of job-shop scheduling system based on Multi-Agent System. The intelligent character is embodied by inviting to public bidding; the dynamic, multi aims and global optimism character is realized by a set of reward-punish principle and synthesis index, of which mathematical model is introduced in thesis. The reward-punish principle and synthesis index can meet the needs of modern manufacture system, solve the complex job-shop scheduling problems and react on the intelligentization of job-shop scheduling, which based on Multi-Agent System.Thirdly, We suppose a dynamic scheduling model based on workshop MAS scheduling model and make simulations of machine uncertainty and task uncertainty. Combining ant colony algorithm and reinforcement learning, a new job-shop scheduling algorithm based on an adaptive agent was proposed. When the production environment changed, the artificial ant can make decision by the past and the immediate encouragement. Then the distribution of the assignments on the machines...