基于 DRL 的大规模定制装配车间调度研究

针对大规模定制装配车间中订单的随机性和偶然性问题, 文章提出一种基于深度强化学习(deep reinforcement learning, DRL)的大规模定制装配车间作业调度优化方法。建立以最小化产品组件更换次数和最小化订单提前/拖期惩罚为目标的大规模定制装配车间作业调度优化模型, 基于调度模型建立马尔科夫决策过程, 合理定义状态、动作和奖励函数; 将调度模型优化问题与 DRL 方法相结合, 并采用改进的 D3QN 算法进行模型求解; 最后进行仿真实验验证。结果表明, 文章所提方法能有效减少产品组件更换次数和降低订单提前/拖期惩罚。

Aiming at the randomness and contingency of orders in the assembly shop for mass customization, this paper proposes an assembly job shop scheduling optimization method based on deep reinforcement learning (DRL). Firstly, an assembly job shop scheduling optimization model is established to minimize the number of product component replacements and the penalties for order earliness or tardiness. Then, based on the scheduling model, a Markov decision process is established, and the state, action and reward functions are reasonably defined. The optimization problem of the scheduling model is solved by connecting it with the DRL method, and an improved D3QN algorithm is selected to solve the model. Finally, the simulation experiment is conducted, and the results show that the method effectively reduces the number of product component replacements and penalties for early or delayed orders.