深度强化学习求解动态柔性作业车间调度问题  

A dynamic flexible job shop scheduling method based on deep reinforcement learning

作　　者：杨丹[1,2] 舒先涛 余震 鲁光涛[1,2] 纪松霖[1,3] 王家兵 YANG Dan;SHU Xiantao;YU Zhen;LU Guangtao;JI Songlin;WANG Jiabing(Key Laboratory for Metallurgical Equipment and Control of Ministry of Education,Wuhan University of Science and Technology,Wuhan 430081,China;Hubei Key Laboratory of Mechanical Transmission and Manufacturing Engineering,Wuhan University of Science and Technology,Wuhan 430081,China;Precision Manufacturing Institute,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]武汉科技大学冶金装备及其控制省部共建教育部重点实验室,武汉430081 [2]武汉科技大学机械传动与制造工程湖北省重点实验室,武汉430081 [3]武汉科技大学精密制造研究院,武汉430081

出　　处：《现代制造工程》2025年第2期10-16,共7页Modern Manufacturing Engineering

基　　金：国家自然科学基金项目(51808417)。

摘　　要：随着智慧车间等智能制造技术的不断发展,人工智能算法在解决车间调度问题上的研究备受关注,其中车间运行过程中的动态事件是影响调度效果的一个重要扰动因素,为此提出一种采用深度强化学习方法来解决含有工件随机抵达的动态柔性作业车间调度问题。首先以最小化总延迟为目标建立动态柔性作业车间的数学模型,然后提取8个车间状态特征,建立6个复合型调度规则,采用ε-greedy动作选择策略并对奖励函数进行设计,最后利用先进的D3QN算法进行求解并在不同规模车间算例上进行了有效性验证。结果表明,提出的D3QN算法能非常有效地解决含有工件随机抵达的动态柔性作业车间调度问题,在所有车间算例中的求优胜率为58.3%,相较于传统的DQN和DDQN算法车间延迟分别降低了11.0%和15.4%,进一步提升车间的生产制造效率。The study of the artificial intelligence algorithms for job shop scheduling has gained attention due to the advancements in intelligent manufacturing technologies like smart factories.Dynamic events in the job shop are crucial factors affecting scheduling effectiveness.To this end,it proposes a novel approach employing the deep reinforcement learning to solve the dynamic flexible job shop scheduling problem with random job arrival.Initially,a mathematical model is formulated for the dynamic job shop scheduling problem with the objective of minimizing the total tardiness.Subsequently,eight job shop state features are extracted,and six composite scheduling rules are designed.Anε-greedy action selection strategy is adopted,and the reward function is designed.Finally,the advanced D3QN algorithm is introduced to solve the problem and the effectiveness of this method is verified on different scale of instances.The results show that the D3QN algorithm effectively solves the dynamic flexible job shop scheduling problem with random job arrival,and the winning rate in all instances is 58.3%.Compared with traditional DQN and DDQN algorithm,the total tardiness is reduced by 11.0%and 15.4%respectively,which proves that this method further enhances the production efficiency of the job shop.

关 键 词：深度强化学习 D3QN算法 工件随机抵达 柔性作业车间调度 动态调度 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]