Abstract: To reduce the influence of dynamic uncertainties on the production performance of smart shop floor,a self-adaptive scheduling method based on Q-learning was proposed.A self-adaptive scheduling framework based on reinforcement learning was designed for the smart shop floor.The Q-learning algorithm was adopted in this framework to train the scheduling model autonomously through the agent-environment interactive trial and error mechanism.In addition,the scheduling model would be dynamically updated according to the changes in the production environment to generate the optimal decision trajectory,which could guide the operation of the shop floor.The proposed method was validated on a semiconductor production line model,namely the MiniFab model.Experimental results demonstrated that the proposed method could effectively respond to the changes in the production environment of the smart shop floor,adjusting the scheduling decision in real time throughout the production process.As a result,the comprehensive performance was optimized,and time and labor costs were significantly reduced.

Key words: smart shop floor, self-adaptive scheduling, reinforcement learning, Q-learning

摘要： 为降低智能车间中动态不确定因素对生产性能的影响,提出一种基于Q学习的智能车间自适应调度方法。该方法设计基于强化学习的智能车间自适应调度框架,采用Q学习算法,通过智能体—环境交互试错机制,自主训练调度模型,并根据生产车间环境变化动态更新调度模型,以支持能够指导车间运行的最优决策轨迹的生成。所提方法在MiniFab半导体生产线模型上进行了验证,结果证明该方法能够有效应对智能车间生产环境变化,在生产全过程中能对调度决策进行实时调整,优化车间综合性能指标,同时显著降低时间与人力成本。

关键词: 智能车间, 自适应调度, 强化学习, Q学习