A dynamic distribution order allocation optimization method considering supply and demand state of transport capacity system

Abstract

The application relates to the technical field of order allocation, in particular to a dynamic distribution order allocation optimization method considering the supply-demand state of a transport capacity system.S1, a dynamic pickup and delivery problem model is constructed, the dynamic pickup and delivery problem system state and the action set are determined, and the dynamic pickup and delivery problem model is constructed;S2, a rolling time sequence mechanism is adopted to describe the dynamic nature of the urban dynamic pickup and delivery problem; and S3, a mixed allocation strategy is adopted to allocate the dynamic distribution order.The dynamic distribution order allocation optimization method considering the supply-demand state of the transport capacity system optimizes the design of the order allocation, efficiently allocates a scheme and a solution algorithm, realizes the matching of efficient dynamic time requirements and distribution resources, reduces the distribution cost and improves the distribution punctuality rate; meanwhile, the efficient order allocation strategy and the distribution optimization search operator based on the transport capacity supply-demand are designed, the response speed of the distribution system to the dynamic time requirements is improved, the distribution service quality of the distribution enterprise to the market and the customer requirements can be improved, and the distribution cost is optimized.

Description

Technical Field

[0001] This invention relates to the field of order allocation technology, and in particular to a dynamic delivery order allocation optimization method that takes into account the supply and demand status of the transportation system. Background Technology

[0002] With the development of urban delivery platforms and the application of online apps by delivery companies, the volume of dynamic pickup and delivery services in cities is increasing year by year. Delivery companies are paying more and more attention to the service and management strategies of real-time dynamic delivery. Dynamic delivery costs and delivery timeliness are closely related. The faster the dynamic delivery response time, the higher the average delivery cost and the higher the customer satisfaction; conversely, the lower the delivery timeliness, the lower the average delivery cost and the lower the customer satisfaction. Data shows that the on-time delivery rate directly affects customer satisfaction and loyalty; 89% of customers highly value on-time delivery. When a package is delivered on time, the probability of a customer making another purchase is 85%; if the delivery is not timely, the probability is only 33%. Real-time delivery demand exhibits strong randomness, mainly characterized by point-to-point delivery, high timeliness, discontinuity, and unevenness. Delivery service companies need to complete delivery services within a short promised timeframe. These characteristics increase the difficulty of providing real-time delivery services. Meanwhile, customers have significantly different needs regarding the timeliness of express delivery services. For example, different customers exhibit considerable individual differences in their sensitivity to both timeliness and price; even the same customer's needs for delivery timeliness vary considerably when purchasing different goods. Specifically, for those with weak timeliness needs, customers have lower requirements for delivery timeliness and are accepting of the quality of standard express delivery services; for those with strong timeliness needs, customers are less price-sensitive and require rapid response and on-time delivery. The stronger the timeliness of the delivery service chosen by the customer, the higher the price they will pay. Secondly, the composition of on-demand delivery capacity is complex, including merchant capacity, platform capacity, and crowdsourced capacity. Therefore, on-demand delivery services present new challenges for delivery companies. Designing a reasonable and efficient dynamic demand order allocation strategy for on-demand delivery is currently a key focus and challenge for the delivery industry.

[0003] The on-demand delivery problem is characterized by its dynamic nature and time constraints, making it a typical dynamic optimization problem. Dynamic optimization problems have always been a challenging area in operations research, and on-demand delivery is a dynamic pickup and delivery problem (DPDP) with multiple pick-up and delivery points.

[0004] Current technologies for solving the problem of dynamic urban delivery order allocation mainly include waiting strategies, greedy strategies, or breaking down the dynamic problem into several independent static sub-problems.

[0005] Waiting strategy: When a new order is generated, the order demand is calculated and accumulated until the accumulated demand reaches the delivery threshold, and then the delivery is started.

[0006] Greed strategy: When a new order is generated, the order is assigned to the closest delivery capacity in the alternative delivery capacity.

[0007] Static sub-problem strategy: The dynamic delivery problem is divided into several independent static delivery problems, and the static sub-problems are solved in turn, and finally the sub-problem results are combined to obtain the delivery scheme.

[0008] The following disadvantages exist when in use:

[0009] 1. The prior art method lacks analysis of the demand timeliness characteristics and the analysis of the state of the delivery capacity supply and demand system, and the allocation strategy is single.

[0010] 2. The prior art method (order allocation method) has poor adaptability and cannot adapt to fluctuations in order demand, pickup and delivery points and time. The solution stability is poor in different scenarios.

[0011] 3. The prior art method (dynamic path planning) does not solve the dynamic pickup and delivery constraints. When the delivery goods are large or heavy, the solution feasibility is poor. SUMMARY

[0012] Therefore, it is necessary to provide a dynamic delivery order allocation optimization method considering the state of the delivery capacity system supply and demand to solve the technical problems proposed in the background art.

[0013] In order to solve the above technical problems, the following technical solutions are adopted:

[0014] A dynamic delivery order allocation optimization method considering the state of the delivery capacity system supply and demand, the steps are as follows:

[0015] S1, constructing a dynamic pickup and delivery problem model;

[0016] By defining the state of the dynamic pickup and delivery problem system, the action set, and constructing the dynamic pickup and delivery problem model;

[0017] S2, using a rolling time mechanism to depict the dynamics of the city dynamic pickup and delivery problem, taking time sequence as the benchmark, and applying time slicing to deconstruct the dynamic delivery problem, constructing a pickup and delivery sub-problem model to form a dynamic pickup and delivery problem model;

[0018] S3, using a mixed allocation strategy to allocate dynamic delivery orders.

[0019] As one optimization modeling method of the dynamic distribution order allocation optimization method considering the supply and demand state of the transportation system provided by the application, in the S1 step, the dynamic pickup and delivery problem model is determined, and the steps are as follows:

[0020] The tuple data (t h ,O h ,K h ) represents the system state of the decision point h;

[0021] Wherein, t h represents the time of the decision point h, O h represents the demand set at the decision point h, K h represents the vehicle state information of the decision point h, and K h is represented as Wherein and represent the current position and destination of the vehicle k respectively, represents the time when the vehicle is expected to arrive at the destination, represents the current demand list carried by the vehicle; by The load Q ik when the vehicle k leaves the logistics point i can be calculated; represents the distribution route of the vehicle k;

[0022] The tuple data (q i ,p i ,d i , ) represents the information of the dynamic demand of the customer;

[0023] Wherein, q i is the order weight; p i is the pickup point, d i is the delivery point; is the order creation time; is the order promised arrival time; is the order pickup time; is the order delivery time.

[0024] As one optimization modeling method of the dynamic distribution order allocation optimization method considering the supply and demand state of the transportation system provided by the application, in the S1 step, the action set step is as follows:

[0025] The distribution optimization decision is composed of the action set A h based on the time sequence of each decision point h, that is, A = ∑ h∈ H A h ;

[0026] Wherein, h is a decision point, A h is the action set of decision point h;

[0027] Action set A of decision point h h is composed of a series of selectable actions a.

[0028] As an optimization modeling method of the dynamic distribution order allocation optimization method considering the supply and demand state of the transportation system provided by the application, in the S1 step, a dynamic pickup and delivery problem model objective function is constructed, and the steps are as follows:

[0029] The difference of the distribution optimization target in different scenarios is embodied by setting the weight parameter value of different targets, and the formula is as follows:

[0030] minF(O)=λ1f1+λ2f2

[0031]

[0032]

[0033] Wherein, f1 is the total of order delivery overtime; f2 represents the average delivery distance of each vehicle; λ1 and λ2 represent the weight coefficients of different sub-targets; The delivery time of the order; The order commitment arrival time; x ijk 1 when vehicle k travels arc (i, j), otherwise 0.

[0034] As an optimization solving method of the dynamic distribution order allocation optimization method considering the supply and demand state of the transportation system provided by the application, in the S2 step, the dynamic distribution problem is decomposed, and the steps are as follows:

[0035] According to the distribution requirement, the time interval Δt is set, that is, every interval Δt triggers the time series based decision point h (h ∈ H), and the formula is as follows

[0036] t h =t h-1 +Δt

[0037] Calculate the latest departure time of each order The formula is as follows:

[0038]

[0039] Wherein, The latest departure time; The order commitment arrival time; The order pickup time; The delivery time of the order; p i The order i pickup address; d iOrder i delivery address.

[0040] As an optimization solving method of the dynamic distribution order allocation optimization method considering the supply and demand state of the transport system provided by the application, in the S2 step, a pickup and delivery sub-problem model is constructed, and the steps are as follows:

[0041] The objective function of the sub-problem model at each decision point h is obtained by calculating the order set, and the calculation formula is as follows:

[0042]

[0043] As a mixed order allocation method of the dynamic distribution order allocation optimization method considering the supply and demand state of the transport system provided by the application, the S3 step is an instant distribution order allocation step, and the steps are as follows:

[0044] Step 1: initialization setting of dynamic time limit demand pickup and delivery problem simulation environment parameters;

[0045] Step 2: according to the time interval Δt and the decision point time t h Update customer demand information;

[0046] Step 3: according to the decision point time t h And the vehicle path Update all vehicle loading demand list Vehicle current position, vehicle destination And vehicle driving route

[0047] Step 4: read demand information, calculate the latest departure time of demand And based on According to the establishment of the order pool;

[0048] Step 5: demand pickup and drop-off point identification, obtain hot pickup and drop-off point;

[0049] Step 6: emergency single strategy is adopted for the emergency demand in the demand pool;

[0050] Step 7: the non-emergency demand in the demand pool is adopted by the wind order strategy and the order clustering strategy;

[0051] Step 8: using local search operator to optimize the vehicle planning distribution route;

[0052] Step 9: feed back the vehicle distribution route to the simulation environment;

[0053] Step 10: judge whether the simulation environment has new order; yes, turn to step 2; no, turn to step 11;

[0054] Step 11: Determine if all deliveries have been completed; if yes, proceed to step 12; if no, proceed to step 3.

[0055] Step 12: Output all vehicle routes and objective function values.

[0056] As a hybrid order allocation method in the dynamic delivery order allocation optimization method considering the supply and demand status of the transportation system provided by the present invention, the emergency order strategy in step 6 is as follows:

[0057] Step ①: Read emergency order information and vehicle information;

[0058] Step 2: Read the urgent orders based on the latest departure time sequence;

[0059] Step 3: Calculate the vehicle load capacity and obtain a set of available vehicles;

[0060] Step 4: Read the vehicle location / destination and select the vehicle with the lowest distance cost;

[0061] Step 5: Calculate the minimum marginal cost of inserting an emergency order and record the result;

[0062] Step 6: Should all available vehicles be traversed? If no, go to step 4; if yes, go to step 7.

[0063] Step 7: Output the vehicle with the minimum marginal cost and insert an emergency order;

[0064] Step 8: Should all urgent orders be iterated? If no, go to step 2; if yes, the program ends.

[0065] As a hybrid order allocation method in the dynamic delivery order allocation optimization method considering the supply and demand status of the transportation system provided by the present invention, the following steps are taken in step 7 regarding the courier order strategy:

[0066] The order vehicles in the ride-sharing strategy are calculated using two-point matching operators, single-point matching operators, and approximate matching operators;

[0067] Two-point matching operator: Selects the demand where both the pickup and delivery points match two adjacent points in the vehicle's planned route, and inserts the demand between the two points;

[0068] Single-point matching operator: Select a pickup point that is the same as the vehicle's destination, and the distance between the delivery point and the next node on the delivery route is acceptable, and then insert the demand between the two points;

[0069] Approximate matching operator: Selects orders where the pickup and delivery points are close to the vehicle's destination and subsequent points adjacent to the destination.

[0070] As the mixed order allocation mode of the dynamic distribution order allocation optimization method considering the supply and demand state of the transport system provided by the application, the order clustering strategy in step 7 is as follows:

[0071] According to the similarity and related constraints of the time and space of the order, the unallocated demand is clustered and allocated to the same order package for processing at the same time.

[0072] It can be seen without doubt that the above technical solutions of the application can solve the technical problems to be solved by the application.

[0073] Meanwhile, through the above technical solutions, the application at least has the following beneficial effects:

[0074] The dynamic distribution order allocation optimization method considering the supply and demand state of the transport system provided by the application optimizes the design of order allocation, efficiently allocates schemes and solving algorithms, realizes efficient matching of dynamic time-sensitive demand and distribution resources, reduces distribution cost and improves distribution punctuality rate; meanwhile, the efficient order allocation strategy and distribution optimization search operator based on the supply and demand of transport capacity are designed, the response speed of the distribution system to strong time-sensitive dynamic demand is improved, the distribution service quality of the distribution enterprise to market and customer demand can be improved, and the distribution cost is optimized. BRIEF DESCRIPTION OF DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0076] Figure 1 The dynamic taking and delivering problem of the application is shown in the schematic diagram.

[0077] Figure 2 The dynamic taking and delivering optimization method of the application is shown in the flowchart.

[0078] Figure 3 The order timeline of the application is shown in the schematic diagram.

[0079] Figure 4 The order pool of the application is shown in the schematic diagram.

[0080] Figure 5 The pseudo code of the order pool construction of the application is shown in the schematic diagram.

[0081] Figure 6 The pseudo code of the emergency order strategy of the application is shown in the schematic diagram.

[0082] Figure 7 The schematic diagram of the wind order strategy of the application is shown in the schematic diagram.

[0083] Figure 8 Fig. 1 is a schematic diagram of unassigned orders of the present application;

[0084] Figure 9 Fig. 2 is a schematic diagram of order clustering and packing of the present application;

[0085] Figure 10 Fig. 3 is a schematic diagram of Block-path of the present application;

[0086] Figure 11 Fig. 4 is a schematic diagram of single-point insertion operator of the present application;

[0087] Figure 12 Fig. 5 is a schematic diagram of Block insertion operator of the present application;

[0088] Figure 13 Fig. 6 is a schematic diagram of crossover operation factor of the present application;

[0089] Figure 14 Fig. 7 is a schematic diagram of comparison of solving results of DPDP-50-5 example group of the present application;

[0090] Figure 15 Fig. 8 is a schematic diagram of comparison of solving results of DPDP-300-20 example group of the present application. DETAILED DESCRIPTION

[0091] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0092] It should be noted that the embodiments in the present application and the features and technical solutions in the embodiments can be combined with each other without conflict.

[0093] It should be noted that: similar labels and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0094] Reference Figures 1-15 , a dynamic distribution order allocation optimization method considering the supply and demand state of the transport system, the steps are as follows:

[0095] S1, constructing a dynamic pickup and delivery problem model;

[0096] I. Dynamic pickup and delivery problem state

[0097] In the dynamic demand pickup and delivery optimization problem, the current state of the system at each decision point h (h∈H) is the necessary information of demand and distribution path planning, as shown in Figure 1 First, the tuple data (th ,O h ,K h ) represents the system state at decision point h, where t h represents the time at decision point h, O h represents the demand set at decision point h, K h represents the vehicle state information at decision point h.

[0098] where the vehicle attribute K h is represented as where and represent the current location and destination of vehicle k, respectively, and are set to be the same if the vehicle is in transit. represents the time when vehicle k is expected to arrive at the destination, represents the current vehicle load list; by the load Q ik when vehicle k leaves logistics point i can be calculated. represents the delivery route of vehicle k.

[0099] Secondly, the tuple data (q i , p i , d i , ) represents the information of customer dynamic demand , specifically, each demand has a set of pickup points p i and delivery points d i , the pickup point p i may be a central warehouse, a store or a temporarily designated location, and the delivery point d i may be a store or a customer designated point. The related attributes of each demand i also include the order weight q i , the order creation time , the order promised arrival time , the order pickup time , and the order delivery time Finally, the state of the dynamic pickup and delivery system at decision point h can be represented as:

[0100] II. Action set

[0101] The delivery optimization decision is composed of a time series based action set A h at each decision point h, i.e. A = ∑ h∈ H A h . The action set A h at decision point h is composed of a series of selectable actions a, and the meaning of each action a is to select appropriate orders o (o ∈ OU ) is assigned to a suitable vehicle k. For example, the action represents the unassigned order o is assigned to vehicle k at decision point h. The action set A h must satisfy the demand assignment constraint, the vehicle capacity constraint and the order last-in-first-out constraint.

[0102] III. Constructing the dynamic pickup and delivery problem model

[0103] The dynamic pickup and delivery problem model includes two parts, one is to minimize the order delivery overtime, and the other is to minimize the delivery distance. The difference of the delivery optimization goal in different scenarios can be reflected by setting different weight parameter values of the goal. The dynamic demand pickup and delivery optimization model goal formula is established as follows:

[0104] min F(O) = λ1f1+ λ2f2

[0105]

[0106]

[0107] In the model application, the size of the weight value can be adjusted according to the importance of different factors.

[0108] S2, the rolling time mechanism is used to describe the dynamics of the city dynamic pickup and delivery problem. Time series is used as the benchmark, and time slicing is used to deconstruct the dynamic delivery problem. The time interval Δt is set, that is, every interval Δt triggers the decision point h (h ∈ H) based on time series, as shown in Figure 2 .

[0109] The sub-problem model objective function at each decision point h is the same as the solution method of the total objective function of the dynamic pickup and delivery problem, and the order set solved is different, as follows:

[0110]

[0111] Time interval: in the dynamic pickup and delivery optimization simulation environment, the generation and update of orders are dynamic, and the time interval for each order update is Δt, that is, the "time interval". According to the characteristics of the problem, the "time interval" can be set according to the delivery requirements, and its value will directly affect the response speed and re-optimization frequency of dynamic decision-making.

[0112] Order distribution: the distribution law of orders in location and generation time.

[0113] Average order per vehicle (AO.PV): refers to the ratio of order quantity to vehicle number, which represents the supply and demand relationship of the pickup and delivery system.

[0114] Dock time: refers to the time it takes for each vehicle to enter the node's pallet when it enters the logistics point.

[0115] Latest departure time: Each customer's request requires the latest departure time. The order must depart before the designated time; otherwise, it will inevitably incur overtime costs, the solution of which is as follows: Figure 3 And as described in the following formula:

[0116]

[0117] in, This is the latest departure time; Order delivery time commitment; Order pickup time; Order delivery time; p i Order i: Pickup address; d: i Order and delivery address.

[0118] Latest end time: Based on the vehicle loading demand list and planned route, the earliest time for each vehicle to complete all onboard demand can be calculated in real time.

[0119] Demand package coordinates: These refer to the coordinates of the starting point in the demand list formed by demand clustering under the demand clustering and packaging strategy.

[0120] Demand Stack: A "stack" is a linear list with restricted operations. A demand stack is a linear list used to store dynamic, time-sensitive demands; the first demand is inserted at the bottom of the stack, and the last demand is at the top. When operating on the demand stack, a demand is popped from the top. An order stack uses a Last-In-First-Out (LIFO) standard to store demand data.

[0121] S3. Allocate instant delivery orders.

[0122] Because customer demand is dynamic and random, the supply-demand ratio of vehicles in the pickup and delivery system changes dynamically over time. A single demand allocation strategy cannot cope with the ever-changing dynamic time-sensitivity pickup and delivery needs in the city. Therefore, based on the different time-sensitivity characteristics of customer demand, multiple demand allocation strategies are designed, such as... Figure 2 As shown. The steps are as follows:

[0123] Step 1: Initialize the environmental parameters for simulating dynamic time-sensitive delivery and pickup issues;

[0124] Step 2: Based on the time interval Δt and the decision point time t h Update customer needs information;

[0125] Step 3: Based on the decision point time t h and vehicle routes Update all vehicle loading requirements list current vehicle location, vehicle destination and vehicle travel route

[0126] Step 4: Read demand information, calculate the latest departure time of demand and based on According to the establishment of the order pool;

[0127] Step 5: Demand pickup and drop-off point identification, obtain hot pickup and drop-off point;

[0128] Step 6: Emergency single strategy for emergency demand in demand pool;

[0129] Step 7: Use the wind single strategy and order clustering strategy for non-urgent demand in the demand pool;

[0130] Step 8: Use local search operator to optimize vehicle planning distribution route;

[0131] Step 9: Feedback the vehicle distribution route to the simulation environment;

[0132] Step 10: Determine whether the simulation environment has new orders; Yes, go to step 2; No, go to step 11;

[0133] Step 11: Determine whether all demand distribution is completed; Yes, go to step 12; No, go to step 3;

[0134] Step 12: Output all vehicle routes and objective function values.

[0135] Order identification and order pool construction

[0136] According to the latest departure time of each order Unassigned demand is classified and stored in the order pool. If the latest departure time of unassigned demand i Demand is classified as emergency demand; if the demand latest departure time Demand is classified as sub-urgent demand; if Demand is classified as non-urgent demand, such as Figure 4 ; The order pool algorithm step code is shown in Figure 5 .

[0137] Emergency single distribution

[0138] To reduce the timeout costs in dynamic delivery optimization, urgent orders in the order pool should be prioritized; otherwise, they will inevitably incur timeout costs over time. Therefore, the "urgent order strategy" first identifies urgent orders in the order pool; second, it calculates vehicle load capacity to obtain a set of available vehicles; then, it selects the vehicle with the lowest marginal cost for inserting the urgent order, allocates the urgent order, and replans the delivery route. The execution code for the urgent order strategy is as follows: Figure 6 As shown, the steps are as follows:

[0139] Step ①: Read emergency order information and vehicle information;

[0140] Step 2: Read the urgent orders based on the latest departure time sequence;

[0141] Step 3: Calculate the vehicle load capacity and obtain a set of available vehicles;

[0142] Step 4: Read the vehicle location / destination and select the vehicle with the lowest distance cost;

[0143] Step 5: Calculate the minimum marginal cost of inserting an emergency order and record the result;

[0144] Step 6: Should all available vehicles be traversed? If no, go to step 4; if yes, go to step 7.

[0145] Step 7: Output the vehicle with the minimum marginal cost and insert an emergency order;

[0146] Step 8: Should all urgent orders be iterated? If no, go to step 2; if yes, the program ends.

[0147] SF Express delivery strategy

[0148] like Figure 7 As shown, in the dynamic pickup and delivery optimization problem, considering whether non-urgent needs can be inserted into the vehicle's planned route is an effective way to reduce the total delivery cost. "Ride-sharing" and "carpooling" services can effectively reduce travel costs. This invention borrows the "ride-sharing" concept to design a "ride-sharing strategy" for urban delivery. This strategy searches for needs with optimal marginal costs based on the vehicle's planned route, inserts non-urgent needs into the vehicle's delivery route through ride-sharing strategy operators, and performs local search optimization on the delivery route. To adapt to different system states in the dynamic pickup and delivery optimization problem, ride-sharing strategy operators with different constraint strengths are designed, satisfying the L IFO constraint. These mainly include two-point matching operators, single-point matching operators, and approximate matching operators.

[0149] Two-point matching operator: Selects the demand where both the pickup and delivery points match two adjacent points in the vehicle's planned route, and inserts the demand between the two points.

[0150] Single point matching operator: Select the demand whose pickup point is the same as the vehicle's destination and whose delivery point is close to the next node on the delivery route, and then insert the demand between the two points.

[0151] Approximate matching operator: Select the demand whose pickup point and delivery point are close to the vehicle's destination and the destination's next node.

[0152] Order clustering delivery strategy

[0153] The order clustering strategy is to cluster and assign the unassigned demands to the same order package for processing according to the similarity of time and space and related constraints. The advantages of this strategy are: first, similar orders are processed together to improve the solution efficiency; second, it helps to cluster the demands with more hot logistics points to reduce the dock time caused by frequent vehicle access.

[0154] First, the strategy selects a demand that is close to departure time from the order pool according to the latest departure time, then loads the information of the selected demand, and sets it as the basic information of the order package. Finally, the similarity and matching degree of other orders and the order package are calculated and compared to determine whether to insert them into the order. The demands in the order package will be stored in the order stack in the form of a stack to comply with the last-in first-out constraint. To improve the applicability of the strategy and the response to changes in the dynamic pickup and delivery system, different constraint level packing operation operators are designed, including same pickup-delivery (SPD), same pickup (SP), and same delivery (SD).

[0155] Figure 8 Taking five unassigned demands as an example, the pickup point, delivery point, latest departure time, and decision point time of each demand are shown. Figure 9 Figures (A), (B), and (C) are the clustering and packaging of unassigned demands and the planned delivery routes under different order clustering and packaging operation operators. With the SPD operation operator, only demands 1 and 3 are successfully packaged because they have the same pickup and delivery points. With the SP operation operator, demands 1, 2, and 3 are successfully packaged, and the order delivery sequence is 1, 3, 2. With the SD operation operator, demands 1, 3, and 5 are successfully packaged. The packaging strategy can select and execute different operation operators according to different address attributes and vehicle demand-supply ratios.

[0156] Neighborhood search

[0157] Block-path is a continuous path node sequence, whose start node and end node are the pickup point and delivery point of the same demand, and contains the pickup point and delivery point of each order. The number of Block-paths contained in the vehicle path is used to represent the number of Block-paths. There are mainly three kinds of Block-paths in this application, as shown in Figure 10 If the local search operation operator optimizes the distribution path based on the Block-path meeting the LIFO constraint, the newly generated path also meets the LIFO constraint.

[0158] Single-point insertion operator: select demand o, vehicle k and vehicle distribution route Single-point insertion can select any point in the path , and then insert the pickup point and delivery point of the selected demand to obtain a new feasible route. As shown in Figure 11 , in Block (P1, D1), the pickup and delivery points (P, D) of the new demand have three points that can perform single-point insertion operator to obtain three different new paths.

[0159] Block-based insertion operator: select demand o, vehicle k and vehicle distribution route The Block-based insertion operator can select one Block in the path , and then insert the pickup point and delivery point of the selected demand at both ends of the Block to obtain a new path. As shown in Figure 12 , the pickup and delivery points (P, D) of the new demand can be inserted into the path across (P2, D2) or (P3, D3) to obtain two new paths.

[0160] Repeat the above operation operators to obtain multiple new paths, and finally compare the changes in the cost of the new routes to determine the best optimization scheme.

[0161] The crossover operation operator generates a new path meeting the LIFO constraint by exchanging the positions of two Blocks. As shown in Figure 13 , exchanging Blocks P2-D2 and P4-D4 obtains new route 1, or exchanging Blocks P4-D4 and P1-D1 obtains new route 2.

[0162] Technical effects

[0163] Example naming rule: DPDP-i-k-n represents the type of example as DPDP problem, where there are i orders, k vehicles, and the same type of example number is n.

[0164] 1) Small-scale example

[0165] The first group of small-scale example group contains 8 examples, each with 50 demand orders and 5 vehicles, named DPDP-50-5-n. The second group of small-scale example group contains 8 examples, each with 100 demand orders and 5 vehicles, named DPDP-100-5-n. The present application uses three methods to calculate the 8 DPDP-50-5 examples and the 8 DPDP-100-5 examples respectively, namely the classic greedy strategy, the waiting strategy and the hybrid allocation strategy designed by the present application. The results are shown in Figure 14 .

[0166] Figure 14 The results of solving the two groups of small-scale examples by the three methods are shown respectively. Among them, the point-filled column chart represents the greedy strategy, the vertical line-filled column chart represents the waiting strategy, and the diagonal column chart represents the hybrid optimization method proposed by the present application. In Figure 14 , the vertical coordinate uses logarithmic scale in order to more comprehensively and accurately show the advantages and disadvantages of different methods. The results show that for the DPDP-50-5 example group and the DPDP-100-5 example group, the hybrid optimization method obtains the best result, and the greedy strategy is the second. Due to the threshold constraint effect of the waiting strategy, the waiting strategy cannot effectively solve the small-scale example problem. At the same time, it is found that in Figure 14 , the solving results of the same size examples also have great differences; the order distribution law will directly affect the distribution cost.

[0167] 2) Medium-scale example

[0168] The first group of medium-scale example group has 300 orders and 20 vehicles in each example, named DPDP-300-20-n. The second group of medium-scale example group has 500 orders and 20 vehicles in each example, named DPDP-500-20-n. The three methods are applied to solve the 8 DPDP-300-20 examples and the 8 DPDP-500-20 examples respectively. Figure 15 The results of solving the medium examples by the three methods are shown respectively.

[0169] As shown in Figure 15 , for the DPDP-300-20 example group, the hybrid optimization method obtains the best result in 6 examples, the waiting strategy obtains the best result in 4 examples, and two instances are basically the same as the multiple strategy. The result obtained by the greedy strategy in the 7th example is the same as that of the hybrid optimization method. However, the adaptability of the greedy strategy is poor, and its solving result in example 6 is extremely large. In Figure 15 , for the 8 examples of 500 orders, the hybrid optimization method obtains the best result in 5 examples. The waiting strategy obtains the best result in 3 examples. In summary, the hybrid optimization method performs best in medium-scale examples.

[0170] 3) Large-scale instances

[0171] The present application selects 9 instances with demand greater than or equal to 1000 as large-scale instances for testing, including instances with 1000 demand orders (50 vehicles), 2000 demand orders (50 vehicles) and 3000 demand orders (100 vehicles), wherein each scale has three different instances.

[0172] Table 1 shows the results of solving large-scale instances

[0173]

[0174] Table 1 shows the results of solving large-scale instances by the three methods. The mixed allocation strategy gets the best results in the 6 large-scale instances, and its performance is better than that of the greedy strategy and the waiting strategy.

[0175] The preferred embodiments of the present application disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details and limit the present application to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of the present application. The present application selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can well understand and use the present application. The present application is limited only by the claims and their full scope and equivalents.

Claims

1. A dynamic distribution order allocation optimization method considering the supply and demand state of a transportation system, characterized in that, The steps are as follows: S1, constructing a dynamic picking and delivery problem model; By defining the state of the dynamic picking and delivery problem, the steps are as follows: In S1, the dynamic picking and delivery problem model objective function is constructed, and the steps are as follows: By setting different target weight parameter values to reflect the differences between different scenarios, the formula is as follows: minF(O)=λ1f1+λ2f2 by the tuple data (t h ,O h ,K h ) representing the system state at decision point h; wherein, t h denotes the time of decision point h, O h denotes the set of demands at decision point h, K h denotes the vehicle state information at decision point h, K h is represented as wherein and denote the current location and destination of vehicle k, respectively, denotes the time vehicle k is expected to reach the destination, denotes the current list of demands carried by the vehicle; by calculating the load Q ik when vehicle k leaves logistics point i; denotes the delivery route of vehicle k; Applying tuple data Information representing dynamic demand of customers ​ wherein q i is the order weight; p i is the pickup point, d i is the delivery point; is the order creation time; is the order promised arrival time; is the order pickup time; is the order delivery time; In the S2 step, the dynamic distribution problem is decomposed, and the steps are as follows: According to the distribution requirements, set the time interval Δt, that is, every interval Δt triggers a decision point h based on the time sequence, and the formula is as follows In the S2 step, the picking and delivery sub-problem model is constructed, and the steps are as follows: wherein f1 is the sum of order delivery overtime; f2 represents the average delivery distance of each vehicle; λ1 and λ2 represent the weight coefficients of different sub-goals; Delivery time of the order; Order commitment arrival time; x ijk 1 if vehicle k travels arc (i, j), otherwise 0; Through the calculation of the order set, the sub-problem model objective function at each decision point h is obtained, and the calculation formula is as follows: In S1, the action set is as follows: t h = t h-1 + Δt Calculating the latest departure time for each order The formula is as follows: wherein, is the latest departure time; order promise arrival time; order pickup time; order delivery time;p i order i pickup address;d i order i delivery address; In S3, the immediate delivery order allocation steps are as follows: Step 1: Dynamic time efficiency demand picking and delivery problem simulation environment parameter initialization setting; 2.The dynamic distribution order allocation optimization method considering the supply and demand state of the transportation system according to claim 1, wherein, Step 5: Demand picking and delivery point identification, obtain hot picking and delivery point; The distribution optimization decisions are a set of actions A based on time series at each decision point h h constituted, i.e. A =∑ h∈H A h ; where h is a decision point, A h a set of actions for decision point h; The action set A at decision point h h is composed of a series of alternative actions a. 3.The dynamic distribution order allocation optimization method considering the state of supply and demand of a transportation system according to claim 1, wherein, Step 6: For the urgent demand in the demand pool, adopt the urgent single strategy; Step 7: For the non-urgent demand in the demand pool, adopt the wind single strategy and order clustering strategy; Step 2: Based on the time interval Δt and the decision point time t h updating customer demand information; Step 3: Based on decision point time t h and vehicle path Update all vehicle loading demand list Vehicle current location, vehicle destination and vehicle travel route Step 4: Read demand information, calculate demand latest departure time and based on According to the order pool is established; Step 8: Use local search operator to optimize the vehicle planning distribution route; Step 9: Feed back the vehicle distribution route to the simulation environment; Step 10: Determine whether there are new orders in the simulation environment; Yes, go to step 2; No, go to step 11; Step 11: Determine whether all demands are completed; Yes, go to step 12; No, go to step 3; Step 12: Output all vehicle routes and objective function values. In the step 6, the urgent single strategy, the steps are as follows: Step ①: Read the urgent order information and vehicle information; Step ②: Sort the urgent orders according to the latest departure time sequence; 4.The method of claim 3, wherein, Step ③: Calculate the vehicle load to obtain the available vehicle set; Step ④: Read the vehicle location / destination, and select the vehicle according to the distance cost in turn; Step ⑤: Calculate the minimum marginal cost of inserting the urgent order, and record the result; Step ⑥: Whether all available vehicles are traversed; No, go to step 4; Yes, go to step 7; Step ⑦: Output the vehicle with the minimum marginal cost, and insert the urgent order; Step ⑧: Whether all urgent orders are traversed; No, go to step 2; Yes, the program ends. In the step 7, the wind single strategy, the steps are as follows: Through the double-point matching operator, single-point matching operator and approximate matching operator, the order vehicle in the wind single strategy is calculated; Double-point matching operator: Select the demand that matches the two adjacent points in the vehicle planning route, and insert the demand between the two points; 5.The dynamic distribution order allocation optimization method considering the state of supply and demand of a transportation system according to claim 3, wherein, ​ ​ ​ Single-point matching operator: select the pickup point and the destination of the vehicle, and the distance between the delivery point and the next node of the distribution route is the accepted demand, and then insert the demand between the two points; Approximate matching operator: select the pickup point and the delivery point and the order close to the destination and the destination of the subsequent point. 6.The method of claim 3, wherein, The order clustering strategy in step 7 is as follows: According to the hot point pickup and delivery point order quantity, the time and space similarity of the order and the related constraint conditions, the unallocated demand is clustered and allocated to the same order package for processing.

Images