关键词: Q-学习, 合同网机制, 基于Agent的制造调度, 柔性制造系统, 柔性作业车间调度问题

Abstract: To intensify the learning ability of Contract Net Protocol(CNP), an adaptive coordination mechanism named CNP-QL which integrated Q-learning with the basic CNP was presented. The CNP-QL was expressed in Unified Modeling Language(UML) sequence diagram and its decision process as well as learning process were illustrated in detail.key elements in the embedded Q-learning algorithm were designed to realize the CNP-QL. Computational experiments on tasks dynamic assignment among alternative cells in flexible manufacturing workshop were conducted. Simulation results showed that CNP-QL outperformed the basic CNP in terms of mean tardiness and workload balance among cells.

Key words: Q-learning, contract net protocol, Agent-based manufacturing scheduling, flexible manufacturing systems, flexible Job Shop scheduling problem