基于深度强化学习种群优化的演化式分拣调度算法

Evolutionary job scheduling algorithm based on population optimization by deep reinforcement learning

机械制造中的产线分拣作业具有问题与数据的双重复杂性,为了对分拣操作进行优化以提高生产效率,设计了一套分拣作业的数据表示方法与一种基于种群优化的演化式算法,同时整理并公开了一个真实的工业数据集。数据表示方法通过借鉴词袋模型对原始作业数据进行抽象表示;演化式算法使用深度强化学习初始化遗传算法中的种群,同时引入了精英保留策略以提高算法的优化能力。最后,将提出的算法与其他算法在真实的工业数据集与旅行商问题数据集上进行了对比。结果表明,该算法能找到更优的分拣顺序与访问路径,验证了算法的有效性。

The sorting operation of the production line in mechanical manufacturing has the double complexity of the problem and data.To optimize the sorting operation and improve production efficiency,this paper designed a method for data representation and an evolutionary algorithm based on population optimization.At the same time,this paper arranged and disclosed a real industrial data set.The method for data representation abstracted the original job data by referring to the bag-of-words model.The evolutionary algorithm used deep reinforcement learning to initialize the population in the genetic algorithm and introduced the elite retention strategy,which improved the optimization ability of the algorithm.Finally,it compared the proposed algorithm with other algorithms on the real industrial data set and travelling salesman problem data set.The results show that the proposed algorithm can find a better sorting sequence and the access path,which verifies the effectiveness of the algorithm.