Order Dispatching And Route Planning For Immediate Delivery Problem Based On Crowdsourcing Logistics Model

In the era of new retail,online and offline integration has promoted the rapid development of the immediate delivery,most notably in industries such as food delivery,fresh fruits and vegetables,retail convenience,and medical health.Meituan Dianping pointed out that in 2020,Gross Transaction Volume(GTV)of its food delivery business increased by 24.5%to RMB488.9 billion.The rapid growth in orders not only brings benefits to various platforms,but also causes a serious mismatch between supply and demand for drivers.In order to solve the shortage in the number of drivers,the crowdsourcing mode has been accepted and become the primary delivery mode in most delivery platforms for greatly ease the pressure on transportation capacity.However,the crowdsourcing mode still has potential problems such as low level of operating efficiency,customer satisfaction and order acceptance rate.This paper studies the routing optimization and order allocation,taking the logistics platform with the crowdsourcing mode as the research object,based on the research of Vehicle Routing Problems,Markov Decision Process and Scheduling Problems,providing reference for the platform's operation decision.The main research work is as follows:First,we establish the multi-vehicle pickup and delivery problem(PDPTW)model with the objective of minimizing the total travel cost.The Google optimization Research tools(OR-Tools)is used to compare the construction heuristic algorithm represented by path-cheapest-arc and the meta-heuristic algorithm represented by tabu search from the perspective of driver traversal cost and algorithm running time.The results show that compared to the construction heuristic algorithm,the meta-heuristic is designed more complex,but the quality of its solution increases over running time.However,when faced with a large number of orders and capacity supply,the platform is expected to respond in seconds or even milliseconds.Therefore,in the new scenario,the construction heuristic algorithm has feasibility and advantages to some extent.Second,the immediate delivery problem is a variant of Vehicle Routing Problem with stochastic demand,in which not all customers are known in advance but are revealed as the system progresses.In this work,we study a multi-period immediate delivery problem with a comprehensive indicator for order selection.In particular,we model the order selection as a sequential decision-making problem.Considering the immediacy of delivery tasks,the mismatch between demand and capacity,and the autonomy of the delivery staff,we propose a comprehensive indicator,a weighted combination of distance,time pressure,and order revenue,as the order selection criterion to help drivers choose the right order.Then the nearest neighbor algorithm is adopted to optimize routes dynamically.Finally,extensive simulation experiments show that the proposed indicator performs stable and better than the other single indicators.Because there exist no instances designed to examine the performance of model proposed,we conduct three sets of experiments in this section and carry out all experiments using C++.Comparative results and statistical analysis demonstrate its effectiveness and robustness.Third,considering the one-sidedness of traditional order scheduling,this paper introduces the comprehensive weight value as an important factor for scheduling process,and establishes a two-stage parallel machine scheduling problem.In order to maintain the uniformity of the experiment,this paper uses the classic data set of the pickup and delivery problem to test the model proposed,and the experiment is implemented using Python.The results show that the greedy insertion standard based on the comprehensive weight value is more stable and excellent than the other two standards in terms of algorithm running time dimension,order acceptance rate and total revenue.The research in this paper is an extension of previous vehicle routing problem and conducts the numerical simulation based on the actual background,which provides a reference for the expansion of dynamic examples.By using standard data sets,it partially expands the previous theories and provides new sights for future study.The model and algorithm proposed in this paper will help the food delivery platform to solve the problem of realistic route optimization and order scheduling,so as to achieve higher enterprise value.