| 王强， 李仁旺*.基于改进DQN算法的柔性作业车间调度问题[J].轻工机械,2025,43(3):97-103 |
| 基于改进DQN算法的柔性作业车间调度问题 |
| Flexible Job Shop Scheduling Problem Based on Improved DQN Algorithm |
| |
| DOI：10.3969/j.issn.1005 2895.2025.03.014 |
| 中文关键词: 车间调度 柔性作业车间 多目标优化 深度强化学习 NSGA Ⅱ算法 |
| 英文关键词:job shop scheduling flexible job shop multi objective optimization deep reinforcement learning NSGA Ⅱ（Non Dominated Sorting Genetic Algorithm Ⅱ ） |
| 基金项目:浙江省2023年度“尖兵”“领燕”研发攻关计划（2022C01SA111123）；国家自然科学基金资助项目（51475434）。 |
|
| 摘要点击次数: 6 |
| 全文下载次数: 6 |
| 中文摘要: |
| 为了提高多目标柔性作业车间的生产效率并降低能耗，课题组提出了基于深度强化学习的多目标柔性作业车间调度方法，以最小化最大完工时间和最低能耗为优化目标，建立柔性作业车间调度模型，利用改进的深度Q网络（Deep Q Network，DQN）算法对目标函数进行求解。首先定义状态和动作空间，将调度问题的状态和动作信息嵌入DQN模型进行训练；对于每一代调度方案，利用非支配遗传算法（Non Dominated Sorting Genetic Algorithm Ⅱ，NSGA Ⅱ）将方案映射到多目标前沿，并更新经验回放中的奖励，进一步优化DQN的Q值更新；随着训练进行，DQN逐步优化调度方案，输出一个基于最小化最大完工时间和最低能耗的最优调度策略。实例验证结果表明改进DQN算法能够更快速、高效地找到符合多目标优化的调度方案。 |
| 英文摘要: |
| To improve production efficiency and reduce energy consumption of multi objective flexible job shop scheduling, the research group proposed a multi objective flexible job shop scheduling method based on deep reinforcement learning. To minimize the maximum completion time and energy consumption, a flexible job shop scheduling model was established, and the objective function was solved by an improved Deep Q Network (DQN) algorithm. Firstly, the state and action space were defined, and the state and action information of the scheduling problem were embedded into the DQN model for training. For each generation of scheduling schemes, Non Dominated Sorting Genetic Algorithm Ⅱ （NSGA Ⅱ）was used to map the schemes to multi objective frontiers and update the reward in experience playback to further optimize the Q values update of DQN. With the training progressed, DQN gradually optimized the scheduling scheme and outputted an optimal scheduling strategy based on minimizing the maximum completion time and the minimum energy consumption. The example shows that the improved DQN algorithm can find the scheduling scheme that meets the multi objective optimization more quickly and efficiently. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
