管理工程学报2024，Vol.38Issue(3)：150-160,11.DOI:10.13587/j.cnki.jieem.2024.03.011
外卖智能派单的订单-骑手多目标匹配模型及其适应性算法研究
The multi-objective optimization model of order-driver matching and its online adaptive algorithm for the meal delivery platform
摘要
Abstract
With the rapid development of the meal delivery industry and the higher requirements for delivery service,smart delivery of orders has become an important operation method of the meal delivery platform.Ele.me and Meituan successively launched smart delivery systems for takeaway meal orders.However,many problems of the current smart delivery systems have been exposed and have attracted increasing attention.A primary reason is that the interests of different associated participants,such as the platform,driver,merchant,and customer,have not been fully considered or are not coordinated well.For example,traffic violations of delivery drivers are mainly caused by unsuitable driver-order matching;merchants' complaints about high order commissions can be due to improper order profit setting;and drivers' incomes can be greatly affected by their waiting times or no-load driving distances.Nevertheless,in most of the present literature,the meal delivery problem has been abstracted as a vehicle routing problem,and only the objectives of delivery time and routing cost have been considered.The multi-objectives of the meal delivery problem have not been given enough concern.Thus,many problems in the smart meal delivery system still exist and urgently need to be solved.Thus,to focus on these multi-objectives,the meal delivery problem is treated as an order-driver matching problem,which is quite different from the vehicle routing problem.Therefore,a multi-objective order-driver matching optimization model is proposed,and a real-time optimization algorithm based on an adaptive weight(AW)policy of objective weight is constructed.Furthermore,the model and algorithm are illustrated and verified by numerical simulation experiments. The theoretical contributions of this paper are mainly reflected in the following three aspects.First,a multi-objective optimization model of order-driver matching for smart meal delivery is constructed.Five primary objectives of the smart meal delivery are abstracted from the coordination practice,which are order profit,fulfillment time of order,waiting time of driver,no-load driving distance of driver,and service level of order.Simultaneously,four basic constraints of smart meal delivery are also proposed,which are promised delivery time,slack time of delivery,waiting time limitation of drivers,and no-load driving distance limitation of drivers.According to the above objectives and constraints,a multi-objective binary graph optimization model is constructed for each dispatching scenario.Second,a real-time optimization algorithm based on adaptive strategy is presented to solve the multi-objective order-driver matching model.The optimal objective(also called ideal value)from the single-objective optimization is set as a preset value of each objective.Next,the difference between the present value and the real value of every objective in each dispatch scenario is calculated as the difference of the objective.As well,the average objective difference from the historical dispatch scenario is taken as the weight of each objective to optimize the multi-objective order-driver matching model of the new dispatching scenario.Third,numerical simulation experiments of smart meal delivery are designed.In this paper,720 dispatch scenarios within the two-hour dispatch period are considered.The simulation data of each scenario include data of orders,data of riders,order-driver distance,etc.Three kinds of simulation experiments under three ratios of the order number and driver number are compared.The three ratios represent the following three situations:when the number of orders and drivers are approximately equal;when the former is greater than the latter;or when the former is less than the latter.The results of the numerical experiments not only illustrate the usability of the model and the algorithm but also verify the effective performance of the model and the algorithm. In addition,some useful conclusions could be summarized from the results of numerical simulation experiments:1)The result of the algorithm of AW strategy can simultaneously coordinate multiple objectives well when the importance of each objective could not be distinguished.2)The optimal value of each objective obtained from the algorithm of AW policy could reach or approach the preset objective value(ideal value)by dynamically adjusting the weight.3)More drivers could improve the ratio of successful matching of order and driver under the limitations of drivers' waiting time,drivers' no-load driving,and the promised delivery time. The results of this study could be applied to the practice of the smart meal delivery platform and help to solve the problems mentioned in the introduction of this paper.The main management implications could be summarized as follows:first,traffic violations,caused by rushing drivers,could be fundamentally reduced by setting some reasonable constraints for smart meal delivery;second,overtime of order fulfillment could be affected by not only the delivery driver but also many other reasons;third,the reward from an order delivery given to the driver should be related to the value of the order rather than a simple fixed reward;fourth,the waiting time and empty driving distance of the driver should be considered in smart meal delivery decisions.关键词
外卖平台/智能派单/多目标优化/实时优化算法/订单-骑手匹配Key words
Meal delivery platform/Smart dispatching/Multi-objective optimization/Online algorithm/Order-driver matching分类
管理科学引用本文复制引用
熊浩,鄢慧丽..外卖智能派单的订单-骑手多目标匹配模型及其适应性算法研究[J].管理工程学报,2024,38(3):150-160,11.基金项目
国家自然科学基金项目(71761009、72061010) （71761009、72061010）
海南省自科高层次人才项目(722RC646) The National Natural Science Foundation of China(71761009,72061010) （722RC646）
The High-level Talent Project of Natural Science Foundation of Hainan Province(722RC646) （722RC646）
