智能仓货架返回位置分配动态优化策略研究  

Dynamic Optimization Strategy for Pod Return Location in Intelligent Warehouses

作　　者：李腾[1] 张茹兰 孙艳 LI Teng;ZHANG Rulan;SUN Yan(Management School,Harbin University of Commerce,Harbin 150028,China)

机构地区：[1]哈尔滨商业大学管理学院,黑龙江哈尔滨150028

出　　处：《工程管理科技前沿》2024年第5期41-48,共8页Frontiers of Science and Technology of Engineering Management

基　　金：国家科技支撑项目(2018YFB1402500);黑龙江省自然科学基金资助项目(LH2023G009);黑龙江省博士后科研启动金资助项目(LBH-Q21102)。

摘　　要：智能仓被广泛用于高频拣选场景,其效率受制于货架摆放位置。由于季节性需求变化、营销策略、促销活动等导致的订单波动,驱使货架摆放位置按需动态调整。本文利用智能仓特点,以移动机器人完成任务时间最短为目标,在任务周期中建立动态调度模型。构建由三种位置分配策略形成的策略空间,将货架热度、货架间关联度和货架现存位置综合形成状态值,设计加权双Q学习算法,得到适应订单状态波动的货架位置分配动态优化策略。仿真实验表明:与返回原位置相比,动态策略在大订单波动环境下的效率提升约31.36%,能够在感知订单波动后及时调整策略,提高智能仓拣选效率。Intelligent warehouses are extensively utilized in high-frequency picking scenarios,with their efficiency being significantly influenced by pod placement.Fluctuations in orders,caused by seasonal demand changes,marketing strategies,promotional activities,and other factors,drive the dynamic adjustment of pod placement according to demand.Leveraging the feature of intelligent warehouses,a dynamic scheduling model is established during the task cycle with the objective of minimizing completion time of mobile robots.The strategy space,formulated by three pod placement allocation strategies,is constructed.And pod heat,inter pod correlation,existing pod placement are integrated to form a comprehensive state value.A weighted double Q-learning algorithm is designed to derive a dynamic optimization strategy for pod placement allocation,ensuring adaptability to fluctuations in order status.Simulation experiments demonstrate that,the dynamic strategy achieves an efficiency improvement of approximately 31.36% compare to returning to the original placement in the presence of large order fluctuations.The dynamic strategy is capable of promptly adjusting its approach upon order fluctuations,thereby enhancing the picking efficiency of intelligent warehouses.

关 键 词：智能仓 货架返回位置分配 加权双Q学习 两阶段分配策略 动态优化 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]