中文摘要：

针对柔性作业车间动态重调度问题，提出了一种将滚动窗口技术与基于Q学习的进化规划算法相结合的重调度算法。该算法设计了局部和全局两种滚动窗口更新方法，较好地缩小了问题求解规模；并在滚动机制的驱动下，为了能够较好地吸收和修复动态事件对调度的影响，重新设计了进化过程和每一次进化的Q学习过程；通过内嵌该算法提高了合同网协商机制的学习能力；最后，通过仿真实验验证了改进机制的有效性。研究结果表明，改进的合同网协商机制与基本合同网协商机制虽然都具有良好的反应能力，但是改进机制在最大完成时间和加工时间背离两项指标上更具有优势。

英文摘要：

Aiming at the problem of flexible job shop rescheduling, a rescheduling algorithm which mixed Q-learning based evolutionary programming with windows roiling technology was presented. In the algorithm, local and global roiling methods were designed in the purpose of reducing the size of the problem. The process of evolution and Q-learning driven was improved by the rolling mechanism, which was easier to absorb and repair the influence of dynamic events. The learning ability of the contract net consultation mechanism which was integrated with the algorithm was improved. Finally, the feasibility of the consultation mechanism was verified by simulation example. The result shows that the improved contract net consultation mechanism and basic contract net consultation mechanism have good response ability, but improved contract net consultation mechanism has more advantages in maximum completion time and processing departure time.