【摘要】 设计了一个强化学习和仿真相结合的动态实时车间作业排序系统.首先引入多个随机变量,将车间作业排序问题转换成序贯决策问题;然后通过仿真手段构建车间作业排序问题的模型环境,求取系统性能指标并保证解的可行性;接着设计了一个多智能体Q-学习算法和仿真集成解决作业排序问题;最后通过仿真优化实验验证了该系统的有效性.更多还原

【Abstract】 A dynamic and real-time system integrating reinforcement learning with simulation is designed for job-shop scheduling.Several stochastic variables are introduced to transform the job-shop scheduling problem into sequential decision problem.The model environment of job-shop scheduling is built by simulation for obtaining the system performance indices and ensuring the feasibility of the solution.Then,a multi-agent Q-learning algorithm integrated with simulation is developed to solve the job-shop problem.Finally,simulation and optimization experiments show the effectiveness of the system.更多还原