An intelligent logistics scheduling method based on big data analysis

Abstract

The application discloses an intelligent logistics scheduling method based on big data analysis, and relates to the technical field of intelligent logistics scheduling, which comprises the following steps: collecting orders, vehicles, warehouses, customers, roads and historical performance data based on a scheduling period, dividing a delivery area, calculating area order arrival rate, average service time and dynamic load factor, and constructing a historical logistics event gene library; generating an order gene based on a current order, performing similar matching with historical event genes, calculating order delay risk and high-pending abnormal risk; performing capacity judgment and time window screening on vehicles, forming a candidate scheduling scheme set, and calculating p-norm three-party waiting cost; constructing a comprehensive scheduling objective function based on time penalty, stable correction waiting cost and uncertainty budget, selecting a target scheduling scheme, and updating the historical logistics event gene library. The application improves logistics scheduling precision, order performance stability, regional congestion resistance and overall platform operation efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent logistics scheduling technology, and in particular to an intelligent logistics scheduling method based on big data analysis. Background Technology

[0002] With the rapid development of e-commerce, instant delivery, urban collaborative delivery, cold chain transportation, and multi-warehouse collaborative operations, logistics and delivery orders are characterized by large volume, strong spatial and temporal fluctuations, high timeliness requirements, and complex business constraints. Existing intelligent logistics scheduling technologies typically rely on order management systems, warehouse management systems, vehicle positioning systems, and road traffic information platforms to collect data such as order location, cargo weight, vehicle capacity, warehouse loading and unloading status, customer delivery time windows, and road traffic speed. They then employ methods such as vehicle route planning, heuristic search, rolling optimization, or data prediction to complete order allocation and delivery route generation. The optimization goals of logistics scheduling have also evolved from the traditional focus on shortest transportation distance and lowest vehicle operating costs to a multi-objective collaborative optimization that considers order timeliness, vehicle utilization, warehouse operational efficiency, and customer service quality.

[0003] However, existing logistics scheduling methods are insufficient in mining the order delay patterns, abnormal waiting characteristics, and regional congestion evolution patterns contained in historical fulfillment data. They often struggle to identify order delay risks and high-waiting-risk anomalies in advance based on the correlation between current orders and similar historical logistics scenarios. Furthermore, some methods primarily focus on customer delays or the waiting costs of a single entity, lacking a unified quantitative evaluation of warehouse waiting, vehicle waiting, and customer waiting. This can easily lead to situations where the overall average cost is low, but individual orders remain stuck for extended periods. In addition, traditional scheduling models typically use fixed load parameters and deterministic estimated arrival times, making it difficult to reflect the uncertainties caused by concentrated order arrivals, changes in regional capacity, backlogs of unfulfilled orders, and fluctuations in road conditions. Locking the entire delivery route at once reduces the system's responsiveness to new orders, while frequent rescheduling of all tasks affects driver execution stability, thus making it difficult to balance short-term fulfillment efficiency with long-term regional operational stability. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an intelligent logistics scheduling method based on big data analysis to solve the problems of insufficient risk identification, unstable path execution, and difficulty in resource collaborative optimization in existing logistics scheduling.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides an intelligent logistics scheduling method based on big data analysis, which includes:

[0008] Collect basic logistics scheduling data, divide delivery areas and calculate regional order arrival rates, calculate regional average order service time, and calculate regional dynamic load factors based on regional order arrival rates and average order service time to build a historical logistics event gene library;

[0009] Calculate potential waiting time, generate current event genes, calculate order risk based on historical event genes, assess vehicle capacity, and then perform time window filtering based on order risk and regional dynamic load factors to form a set of candidate scheduling schemes. Finally, calculate the performance of each candidate scheme. - Norm three-party waiting cost;

[0010] Calculate the time penalty for candidate solutions, obtain the number of uncompleted orders in the current region, calculate the stable adjusted waiting cost based on the time penalty and the third-party waiting cost, calculate the uncertainty budget for candidate solutions, construct a comprehensive scheduling objective function based on the stable adjusted waiting cost and the uncertainty budget, select the solution with the smallest comprehensive scheduling objective function value, and finally output the results and update the historical logistics event gene library.

[0011] As a preferred embodiment of the intelligent logistics scheduling method based on big data analysis described in this invention, the steps of collecting basic logistics scheduling data, dividing delivery areas, and calculating regional order arrival rates refer to a fixed scheduling cycle. Collect logistics data;

[0012] At each scheduling time Collect logistics data to form a dataset ;

[0013] The current order will be assigned to the delivery area based on the warehouse service area corresponding to the order's delivery address. ;

[0014] Get area Number of new orders in the last 30 minutes Based on the number of new orders Calculate the regional order arrival rate .

[0015] As a preferred embodiment of the intelligent logistics scheduling method based on big data analysis described in this invention, the method involves: calculating the average order service time for a given region, calculating the regional dynamic load factor based on the regional order arrival rate and average order service time, and constructing a historical logistics event gene library from historical fulfillment data. Extracting regions The driving time and node service time of orders completed within the last 7 days, and the first Each historical order defines its order service time. ;

[0016] For the region Completed within the last 7 days Calculate the average service time for each order ;

[0017] Get area Number of vehicles currently available ;

[0018] Based on regional order arrival rate Average order service time and the number of vehicles currently available. Calculate the dynamic load factor of the region ;

[0019] Define the basic performance genes Potential waiting genes Regional stable genes Genes of contract performance results , combined into the first Genes of historical events .

[0020] As a preferred embodiment of the intelligent logistics scheduling method based on big data analysis described in this invention, the steps include: calculating potential waiting time, generating current event genes, and calculating order risk index for current orders to be scheduled based on historical event genes. Get the order creation time and the current moment Calculate the potential waiting time ;

[0021] Generate the current event gene based on the current order. ;

[0022] Calculate the current event gene and the historical event gene Comprehensive distance ;

[0023] Based on comprehensive distance Calculate the similarity between the current event gene and the historical event genes. ;

[0024] Select the top results based on similarity from highest to lowest. A collection of historical events. ;

[0025] Calculate the probability of delay in similar events based on order similarity. ;

[0026] Calculate the probability of high waiting anomalies among similar events ;

[0027] Weighted Delay Probability and abnormal probability Calculate order risk .

[0028] As a preferred embodiment of the intelligent logistics scheduling method based on big data analysis described in this invention, the step of determining vehicle capacity and then filtering by time window based on order risk and regional dynamic load factors to form a candidate scheduling scheme set refers to the process of determining the capacity of each available vehicle and then filtering by time window based on order risk and regional dynamic load factors to form a candidate scheduling scheme set. Capacity assessment and time window filtering are performed to establish a candidate order set corresponding to each candidate vehicle. ;

[0029] Generate vehicles Candidate order sequence ;

[0030] Based on the candidate order sequence, vehicles are formed. Corresponding candidate path ;

[0031] The first part consists of candidate vehicles, candidate order sequences, and their corresponding candidate paths. One candidate scheduling scheme ;

[0032] Form a set of candidate scheduling schemes for vehicle v. .

[0033] As a preferred embodiment of the intelligent logistics scheduling method based on big data analysis described in this invention, wherein: the calculation of candidate solutions... - Norm of three-party waiting cost refers to the calculation of vehicle Number of orders locked in this round ;

[0034] The travel time of the path node for the l-th order Add them together to get the order travel time , obtain the Node service time for each order Calculate orders In candidate solutions The resulting third-party execution waiting costs ;

[0035] Potential waiting time Execution waiting cost Travel time Service Time Synthesized as candidate solutions - Normative third-party waiting costs ;

[0036] Calculate candidate solutions Normalized three-party waiting costs .

[0037] As a preferred embodiment of the intelligent logistics scheduling method based on big data analysis described in this invention, wherein: the time penalty for calculating candidate solutions refers to obtaining candidate solutions. Actual path length ;

[0038] Calculate the historical path growth coefficient of the region and the maximum node switching time in the region ;

[0039] Calculate the upper bound of the path length of the candidate solution ;

[0040] If the actual path length Not exceeding the upper limit Time penalty Set to 0; otherwise, the excess will be converted into a time penalty. .

[0041] As a preferred embodiment of the intelligent logistics scheduling method based on big data analysis described in this invention, wherein: obtaining the number of uncompleted orders in the current region and calculating a stable correction waiting cost based on time penalties and third-party waiting costs refers to obtaining the number of uncompleted orders in the current region. ;

[0042] Calculate candidate solutions Locked segment execution time ;

[0043] Prediction uses candidate solutions Number of unfulfilled orders in the next scheduling cycle ;

[0044] Calculate the regional stability coefficient ;

[0045] The waiting costs for all three parties Regional stability coefficient and time penalty Synthesizing as a stable correction waiting cost .

[0046] As a preferred embodiment of the intelligent logistics scheduling method based on big data analysis described in this invention, wherein: the uncertainty budget for calculating the candidate solution refers to the ETA fluctuation radius for calculating the candidate solution. ;

[0047] Get the maximum number of unfulfilled orders allowed in the region. ;

[0048] ETA fluctuation radius and the number of unfulfilled orders in the next scheduling cycle region Uncertainty budget calculated by weighting after normalization .

[0049] As a preferred embodiment of the intelligent logistics scheduling method based on big data analysis described in this invention, the following steps are taken: A comprehensive scheduling objective function is constructed based on the stable correction waiting cost and the uncertainty budget; the scheme with the smallest comprehensive scheduling objective function value is selected; and finally, the result is output and the historical logistics event gene database is updated to match the candidate schemes. Calculate the normalized value of transportation costs ;

[0050] Calculate the normalized value of the delay ;

[0051] Based on normalized transportation cost values Delay normalized value Stability correction waiting costs Normalized values ​​and uncertain budgets Constructing a comprehensive scheduling objective function ;

[0052] Select the comprehensive scheduling objective function The solution with the minimum value ;

[0053] Output target scheduling scheme ;

[0054] After order fulfillment is completed, the actual delivery time, actual waiting time, and whether there was a delay are obtained to generate a fulfillment result gene. ;

[0055] The basic fulfillment genes, potential waiting genes, regional stability genes, and fulfillment result genes of the current order are merged to form a new complete event gene and written into the historical logistics event gene library.

[0056] The beneficial effects of this invention are as follows: By collecting order, vehicle, warehouse, customer, road, and historical fulfillment data, this invention divides delivery areas and calculates order arrival rate, average service time, and dynamic load factor, constructing a historical logistics event gene library to achieve a structured expression of logistics fulfillment scenarios; by generating current event genes and matching them with similar historical event genes, it identifies delay risks and high waiting anomaly risks in advance; through capacity judgment, time window screening, and candidate solution construction, it eliminates unfeasible solutions in advance; by calculating the p-norm three-party waiting cost, time penalty, stable correction waiting cost, and uncertainty budget, it constructs a comprehensive scheduling objective function to achieve synergistic optimization of cost, timeliness, waiting pressure, and regional backlog risks; ultimately, it improves logistics scheduling accuracy, order on-time fulfillment rate, regional anti-congestion capability, and overall platform operational efficiency. Attached Figure Description

[0057] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart of an intelligent logistics scheduling method based on big data analysis.

[0059] Figure 2 A flowchart for generating the current event gene and calculating order risk.

[0060] Figure 3 The flowchart shows the branch decision process for forming a set of candidate scheduling schemes for vehicles. Detailed Implementation

[0061] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0062] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0063] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0064] Reference Figures 1-3 This is one embodiment of the present invention, which provides an intelligent logistics scheduling method based on big data analysis, including the following steps:

[0065] S1. Collect basic logistics scheduling data, divide delivery areas and calculate regional order arrival rates; calculate regional average order service time, and calculate regional dynamic load factors based on regional order arrival rates and average order service time, and construct a historical logistics event gene library;

[0066] S1.1, Collect basic data for logistics scheduling;

[0067] According to a fixed scheduling cycle Collect logistics data, including This represents the time interval between two consecutive scheduling calculations in the system, preferably 5 minutes, which can reflect vehicle location, road speed, warehouse queuing and new order status, while not causing frequent changes to the driver's route;

[0068] At each scheduling time Collect logistics data to form a dataset :

[0069]

[0070] in, This represents the data set at the current scheduling moment; This represents order data, including order number and order creation time. Goods preparation time Promised delivery time (The promised delivery time is the end of the customer's selected available delivery time), order weight ; This represents vehicle data, including vehicle number, current location, and maximum capacity. Current occupied capacity Delivery area; This represents warehouse data, including the number of vehicles queuing at the platform. Loading and unloading efficiency of a single platform Number of available platforms ; This represents customer data, obtained from customer order records and historical delivery records, including the customer's available delivery time. Customer waiting tolerance (Through customer statistics) For historical orders within the most recent preset period, calculate the percentage of orders that were still signed for normally after a delay without complaints, the average acceptable delay time, the percentage of orders accepted for contract revisions, and the percentage of delay complaints. Then, normalize the data and sum them using a weighted average. This represents road data, including road distances between delivery nodes. Effective speed of road section ; This represents historical performance data, including historical delay markers, historical waiting times, and historical high waiting time anomaly markers.

[0071] S1.2, Divide delivery areas and calculate the regional order arrival rate;

[0072] The current order will be assigned to the delivery area based on the warehouse service area corresponding to the order's delivery address. ;

[0073] Get area Number of new orders in the last 30 minutes ,in, Indicates the area At the present moment Backwards, the number of new orders placed within the last 30 minutes;

[0074] Based on the number of new orders Calculate the regional order arrival rate :

[0075]

[0076] in, Indicates the area The order arrival rate is expressed in units of orders per minute. The denominator of 30 indicates that the statistical window is 30 minutes, which can reflect the current order peaks or troughs in the region without being affected by abnormal fluctuations in a single minute.

[0077] S1.3 Calculate the average order service time for the region and calculate the dynamic load factor for the region;

[0078] From historical performance data Extracting regions The driving time and node service time of orders completed within the last 7 days, and the first Each historical order defines its order service time. :

[0079]

[0080] in, This indicates the actual travel time for the order; This indicates the service time for each stage of the order, including loading, unloading, signing for receipt, and communication regarding any issues.

[0081] For the region Completed within the last 7 days Calculate the average service time for each order :

[0082]

[0083] in, Indicates the area Number of orders completed in the last 7 days;

[0084] Get area Number of vehicles currently available ,in, Indicates the area The number of vehicles currently online, not fully loaded, and available to continue accepting orders;

[0085] Based on regional order arrival rate Average order service time and the number of vehicles currently available. Calculate the dynamic load factor of the region :

[0086]

[0087] in, Indicates the area The dynamic load factor, if This indicates that the region theoretically has the capacity to process orders stably. This indicates that the current order arrival speed has reached or exceeded the region's transportation capacity.

[0088] S1.4 Construct a historical logistics event gene bank;

[0089] Transform completed historical orders into logistics event genes; specifically, define basic fulfillment genes. Potential waiting genes Regional stable genes Genes of contract performance results , combined into the first Genes of historical events :

[0090]

[0091] in, Indicates the basic performance genes; Indicates a potential waiting gene; Indicates a region-stable gene; Genes representing the outcome of contract fulfillment;

[0092] Basic performance genes for:

[0093] in, This represents the normalized order attribute characteristics, including weight, volume, waiting time, and remaining promised delivery time. This represents the normalized vehicle status characteristics, including occupied volume, remaining pickup distance, pending orders, and whether the vehicle and order are in the same area. This represents the normalized warehouse status characteristics, including the number of vehicles queuing at platforms, the loading / unloading efficiency of a single platform, and the number of available platforms. This represents the normalized customer status characteristics, including customer delivery time, customer waiting tolerance, and historical delivery rate. This represents the normalized road condition features, including the distance from the warehouse to the customer, the current effective speed, and the estimated travel time. It should be noted that these basic data are all normalized before being used to form the features.

[0094] Potential waiting genes for:

[0095]

[0096] in, This indicates the waiting time from when a historical order enters the dispatch pool to when it is assigned a vehicle; This indicates the waiting time from when the goods for a historical order are ready until they are loaded onto the vehicle. This indicates the waiting time for historical orders that has accumulated over the period specified by the customer.

[0097] Regional stable genes for:

[0098]

[0099] in, This indicates the regional dynamic load factor of the region to which the historical orders belong at the time of scheduling. This indicates the number of unfulfilled orders in the region at that time.

[0100] Performance outcome genes for:

[0101]

[0102] in, This indicates whether a historical order was delayed; a value of 1 indicates a delayed order, and a value of 0 indicates no delayed order. This indicates the actual total waiting time for historical orders; This indicates whether a historical order belongs to the high-waiting-time abnormal order category. If the actual total waiting time of the order exceeds the 95th percentile waiting time of the same type of order in the same region, then it is set to 1; otherwise, it is set to 0. For example, historically, about 95% of orders have a waiting time of no more than 48 minutes, so the 95th percentile waiting time is 48 minutes.

[0103] Logistics event geneization transforms historical fulfillment orders into logistics event genes containing basic fulfillment genes, potential waiting genes, regional stability genes, and fulfillment outcome genes. This enables a structured expression of logistics fulfillment scenarios, moving beyond scheduling based solely on distance, time, or cost rules. Instead, it comprehensively identifies the similarity between the current order's fulfillment scenario and historical risk scenarios, thereby more accurately predicting delay risks and high-waiting-point anomaly risks, improving the targeting of scheduling decisions, and reducing the occurrence of long-waiting orders and regional backlogs.

[0104] S2. Calculate potential waiting time, generate current event genes, and calculate order risk based on historical event genes; assess vehicle capacity, then filter by time window based on order risk and regional dynamic load factor to form a set of candidate scheduling schemes; calculate the candidate schemes'... - Norm three-party waiting cost;

[0105] S2.1 Calculate the potential waiting time;

[0106] For the current pending orders Get the order creation time and the current moment Further calculate the potential waiting time :

[0107]

[0108] in, Indicates order The potential waiting time already incurred at the current scheduling moment; it should be noted that at the current moment... It is the moment that is acquired in real time;

[0109] S2.2 Generate the current event gene and calculate the order risk based on the historical event genes;

[0110] Generate the current event gene based on the current order. :

[0111]

[0112] in, This represents the underlying fulfillment mechanism of the current order; This indicates the potential waiting genes for the current order; Indicates the stable genes of the region to which the current order belongs;

[0113] Calculate the current event gene and the historical event gene Comprehensive distance :

[0114]

[0115] in, This indicates the distance between the current order's basic fulfillment data and the historical order's basic fulfillment data. This indicates the distance between the potential waiting genes of the current order and the potential waiting genes of historical orders; This represents the distance between the current stable genes and the historical stable genes. All three distances are calculated from the normalized field, with values ​​ranging from 0 to 1, so they can be directly synthesized with equal weights.

[0116] Basic performance gene distance The calculation is as follows: sum the root mean square of the differences between the features corresponding to the current order's basic fulfillment genes and the historical order's basic fulfillment genes, and divide by the number of features.

[0117] Potential waiting gene distance The calculation is as follows: sum the root mean square of the differences between the features corresponding to the potential waiting genes of the current order and the potential waiting genes of the historical orders, and divide by the number of features.

[0118] Regional stable gene distance The calculation is as follows: sum the root mean square of the differences between the features corresponding to the stable genes in the current region and the stable genes in the historical region, and divide by the number of features.

[0119] Based on comprehensive distance Calculate the similarity between the current event gene and the historical event genes. :

[0120]

[0121] Among them, order similarity The closer the value is to 1, the more similar the two values ​​are.

[0122] Select the top results based on similarity from highest to lowest. A collection of historical events. :

[0123]

[0124] in, Represents the number of similar historical events, preferably Too few samples are easily affected by occasional anomalies, while too many samples will reduce similarity. 50 similar events can balance stability and scene similarity.

[0125] Calculate the probability of delay in similar events based on order similarity. :

[0126]

[0127] in, Indicates the current event gene Genes of historical events similarity, This represents the probability of delay in similar historical events; Representing historical events Whether there is a delay: 1 for delay, 0 for no delay;

[0128] Calculate the probability of high waiting anomalies among similar events :

[0129]

[0130] in, This indicates the probability of a high-wait anomaly occurring among similar historical events; Indicate whether the historical event belongs to the high-wait-time abnormal order category;

[0131] Weighted Delay Probability and abnormal probability Calculate order risk :

[0132]

[0133] in, Indicates order The overall risk value, Indicates the delay risk weight. This indicates a high waiting anomaly risk weight, preferably, =0.6, =0.4, the delay risk weight is higher than the high waiting anomaly risk weight because on-time delivery of orders is the main fulfillment goal of logistics scheduling. At the same time, retaining the high waiting anomaly risk weight can identify and suppress the long tail waiting caused by multiple delays of a small number of orders, so that the comprehensive risk assessment of orders takes into account both on-time fulfillment and waiting fairness.

[0134] S2.3. Determine the vehicle capacity and then filter by time window to form a set of candidate scheduling schemes;

[0135] For each available vehicle Perform capacity determination:

[0136]

[0137] in, Indicates the maximum capacity of the vehicle; This indicates that the vehicle is currently using its capacity. Indicates order The required capacity is determined by adding vehicles that meet the capacity requirements to the candidate vehicle set.

[0138] Subsequently, time window filtering is performed; specifically, vehicle order execution is calculated. Estimated delivery time ,in, It is calculated by combining the estimated vehicle arrival time at the warehouse, the estimated warehouse loading time, the estimated road travel time, and the customer's signing and delivery time.

[0139] Calculate the buffer time for this order. :

[0140]

[0141] in, Indicates order The time window buffer is in minutes; 10 is the basic buffer time, which can cover most minor time disturbances in the routine performance process, so that the system can avoid frequent risk warnings or scheduling adjustments due to slight time fluctuations.

[0142] Then perform time window filtering:

[0143]

[0144] in, Indicates the promised delivery time for the order;

[0145] For the vehicle-order combinations that have undergone capacity assessment and time window filtering, establish a candidate order set for each candidate vehicle. :

[0146]

[0147] in, Indicates candidate vehicles The set of candidate orders that can be executed; This indicates that the order has been fulfilled. A set of candidate vehicles that meet both capacity and time window requirements;

[0148] For candidate vehicles Based on the total potential waiting time of the orders already formed Candidate orders are sorted from largest to smallest. When two orders have the same overall potential waiting time, their overall risk value is prioritized. For larger orders, when the combined potential waiting time and combined risk value are the same, the promised delivery time is prioritized. For earlier orders, the initial insertion order of candidate orders is obtained. Following this initial insertion order, candidate orders are sequentially inserted into the order execution sequence of the candidate vehicles. For orders to be inserted... Each of these can be inserted into an available position in the current order execution sequence to obtain multiple temporary candidate order sequences (e.g., candidate vehicles). The current order execution sequence is , order By inserting the data at different positions in the order execution sequence, we obtain... , , For each order in the temporary candidate order sequence, a time window check is performed. Temporary candidate order sequences that do not meet vehicle capacity requirements, order pickup time requirements, or customer delivery time requirements are deleted. Order sequences that pass all constraint checks are used as vehicle queues. Candidate order sequence :

[0149]

[0150] in, Indicates the first Vehicles in the candidate scheduling schemes The corresponding candidate order sequence; Indicates vehicle In the Among the candidate scheduling schemes, the first one is... One executed order; Indicates vehicle The number of orders included in the candidate scheduling scheme;

[0151] If there are multiple candidate order sequences Combine them into a candidate order sequence set ;

[0152] Based on the candidate order sequence, the vehicle's current location, the pickup warehouse for each order, and the customer's location are connected sequentially to form a vehicle... Corresponding candidate path :

[0153]

[0154] in, Indicates vehicle The current location; Indicates vehicle In candidate scheduling schemes The first one passed in sequence There are several path nodes, including the order pickup warehouse, the order customer point, and the transit nodes that need to be passed through. Indicates vehicle The number of road segments included in the candidate path;

[0155] The first part consists of candidate vehicles, candidate order sequences, and their corresponding candidate paths. One candidate scheduling scheme :

[0156]

[0157] It should be noted that if a vehicle corresponds to multiple candidate order sequences, each order sequence will be assigned to the corresponding vehicle, thereby generating multiple candidate scheduling schemes for the same vehicle.

[0158] This ultimately forms a set of candidate scheduling schemes for vehicle v. :

[0159]

[0160] in, This represents the number of candidate scheduling schemes that pass the verification of capacity constraints, time window constraints, and unique order allocation constraints.

[0161] S2.4 Calculate candidate solutions - Norm three-party waiting cost;

[0162] Calculate vehicles Number of orders locked in this round :

[0163]

[0164] in, Indicates vehicle The number of orders locked for execution in this round; Indicates to Round up; n is the number of vehicles. The total number of orders in the candidate order sequence; This means that even if there are few candidate orders, at least one order will be locked in; Indicates the proportion of the scrolling execution segment, preferably This means that only about 10% of the orders in the candidate sequence are locked in each round, and the remaining orders can be rearranged in the next scheduling cycle, which can balance dynamic response capability and driver execution stability.

[0165] Calculate vehicles From the candidate path, the first The path node to the first Travel time to each path node :

[0166]

[0167] in, Indicates the first The path node and the first The actual distance between each path node; Indicates vehicle The effective speed on this section of road;

[0168] The first Travel time of each order's path nodes Add them together to get the order travel time , obtain the Node service time for each order Calculate orders In candidate solutions The resulting third-party execution waiting costs :

[0169]

[0170] in, Indicates order The waiting costs for all three parties during the execution phase; Indicates the warehouse waiting time; Indicates vehicle waiting time; This represents customer waiting time; it should be noted that the three-party waiting cost simultaneously evaluates the interests of the warehouse, vehicle, and customer, allowing for overlapping time intervals; warehouse waiting cost weighting. The optimal value is 0.20 because warehouses typically have a certain capacity for temporary storage and buffering of goods, and the marginal cost of goods waiting for a short period is usually lower than the cost of vehicle idleness and the loss due to customer lateness; vehicle waiting cost weighting The optimal value is 0.30 because vehicle waiting occupies limited transport capacity and incurs costs for driver hours, energy consumption, and vehicle depreciation. It may also compress the execution time of subsequent orders, causing a chain reaction of delays. (Customer waiting cost weighting) The preferred value is 0.50 because customer waiting time directly affects the on-time fulfillment rate of orders and may lead to complaints, compensation, order cancellations and customer loss.

[0171] Warehouse waiting time is calculated as follows:

[0172]

[0173] in, Indicates the time it takes for the goods to be ready for order; Indicates the vehicle in the candidate scheme The estimated time to reach the warehouse is obtained by adding the current time to the actual distance from the vehicle's current location to the warehouse and dividing the effective speed of the corresponding road segment. This indicates the number of vehicles queuing on the platform when a vehicle arrives at the warehouse. The loading and unloading efficiency of a single platform is expressed in vehicles per minute. The loading and unloading time is calculated by subtracting the start time from the completion time. The loading and unloading efficiency is obtained by taking the reciprocal of the average loading and unloading time of the vehicles. Indicates the number of currently available platforms;

[0174] Vehicle waiting time is calculated as follows:

[0175]

[0176] in, Indicates the start time of loading; Indicates the vehicle's arrival time at the warehouse;

[0177] Customer waiting time is calculated as follows:

[0178]

[0179] in, Indicates the estimated time the vehicle will arrive at the customer's location; Indicates the promised delivery time for the order; This indicates the customer's tolerance for waiting;

[0180] Potential waiting time Execution waiting cost Travel time Service Time Synthesized as candidate solutions - Normative third-party waiting costs :

[0181]

[0182] in, Indicate candidate solutions of - Norm three-party waiting cost; This indicates that the vehicles arrive at the first candidate route in order. The cumulative driving time and service time generated by the preceding nodes before each order; Indicate candidate solutions The number of orders included; This represents the waiting penalty index, which is preferred. It is used for non-linear amplification of long-waiting orders, but it will not be as large as... That would lead to excessive route detours. It should be noted that although the travel time of the preceding road segment and the service time of the node are included in the fulfillment time cost of multiple subsequent orders, this cumulative inclusion is used to characterize the continuous transmission impact of the consumption of the preceding time on the arrival time of subsequent orders, and is not a repeated statistical analysis of the actual running time of the vehicle.

[0183] The purpose of calculating third-party waiting costs is to: if an order has been waiting for a long time, or is ranked low in the candidate path, resulting in an excessively long waiting time, its cost will be calculated. - This amplification prompts the system to prioritize scheduling schemes that can reduce the pressure of long-waiting orders;

[0184] To compare candidate solutions with different order quantities, candidate solutions are calculated. Normalized three-party waiting costs :

[0185]

[0186] in, The smaller the value, the lower the waiting pressure on the warehouse, vehicles, and customers.

[0187] S3. Calculate the time penalty for candidate solutions; obtain the number of uncompleted orders in the current region and calculate the stable correction waiting cost based on the time penalty; calculate the uncertainty budget for candidate solutions; construct a comprehensive scheduling objective function based on the stable correction waiting cost and the uncertainty budget, select the solution with the smallest comprehensive scheduling objective function value, and finally output the results and update the historical logistics event gene library.

[0188] S3.1 Calculate the time penalty for candidate solutions;

[0189] Obtaining candidate solutions Actual path length ,in, It is obtained by adding the road distances of all adjacent nodes in the candidate solution;

[0190] Calculate the historical path growth coefficient of the region Specifically, extract completed contracts within region z from historical performance data. Get the number of orders for the k-th historical delivery. and actual path length Establish a fitting relationship The least squares method was used to calculate the result. Among them, the path growth index This indicates that the path length increases with the number of orders, but the growth rate is lower than linear growth. Here, a historical delivery batch refers to a delivery vehicle starting from a preset starting node according to a predetermined order execution sequence, and sequentially completing the tasks of picking up, transporting, unloading, and signing for each order in the order execution sequence until the order execution sequence is completed.

[0191] Calculate the maximum node switching time in the region Specifically, for the region For any two adjacent execution nodes in the historical delivery route, obtain the service time of the previous node and the travel time from the previous node to the next node, and use the sum of the two as the corresponding node switching time. Calculate the node switching time of each adjacent node in the historical delivery route, and take the maximum value as the maximum node switching time for the region. ;

[0192] Calculate the upper bound of the path length of the candidate solution :

[0193]

[0194] in, Indicate candidate solutions The upper bound of the allowed path length; This represents the upper bound of the path length based on regional historical experience; Indicates based on - The norm awaits the upper bound of the path length obtained from the cost control. This represents the average effective driving speed in the area, which is obtained by averaging the effective speeds of each road.

[0195] If the actual path length Not exceeding the upper limit Time penalty Set to 0; otherwise, the excess will be converted into a time penalty. :

[0196]

[0197] By simultaneously setting an upper bound for path length based on regional historical path patterns and waiting costs, and converting the path portion exceeding the upper bound into a time penalty, the priority guarantee for long-waiting orders is constrained by reasonable detour boundaries. Compared to simply increasing order priority based on waiting time or waiting penalty value, which easily leads to cross-regional order insertion, repeated path backtracking, and a significant increase in overall mileage, this invention guarantees high-waiting orders while restricting uneconomical detours. It avoids the increase in overall transportation time, mileage costs, and other order delays caused by local waiting improvements, thereby achieving a balance between waiting fairness and path efficiency.

[0198] S3.2 Obtain the number of uncompleted orders in the current region and calculate the stability correction waiting cost;

[0199] Get the number of incomplete orders in the current region ,in, Indicates the region In the current number The number of unfinished orders at the next scheduling moment; it should be noted that a fixed scheduling cycle of 5 minutes is used to re-collect the status of orders, vehicles, warehouses and roads and perform a rolling scheduling once;

[0200] Calculate candidate solutions Locked segment execution time :

[0201]

[0202] in, This indicates the estimated execution time of the locked execution segment in the candidate solution for this round; This indicates the number of orders locked in this round; Indicates travel time; Indicates the node service time; This represents the execution waiting time. For the order pickup warehouse node, the execution waiting time is the difference between the start loading time and the estimated arrival time of the vehicle at the warehouse. For the customer node, the execution waiting time is the difference between the earliest time the customer can receive the goods and the estimated arrival time of the vehicle at the customer node.

[0203] Prediction uses candidate solutions Number of unfulfilled orders in the next scheduling cycle :

[0204]

[0205] in, Indicates the adoption of candidate solutions The estimated number of unfulfilled orders for the next cycle in the following region; Indicates the regional order arrival rate. Indicating in candidate solutions During the expected execution period of the locked execution fragment, the region The expected number of new orders; The segment ratio represents the proportion of the segment, which can strike a balance between the dynamic response capability of the scheduling and the stability of driver execution, so that the system can both reserve the space for reordering subsequent orders and ensure that the current execution segment has sufficient continuity. This represents an estimate of the number of orders in the current order pool that were not locked for execution in this round;

[0206] Calculate the regional stability coefficient :

[0207]

[0208] in, Indicates the regional scheduling stability coefficient; Indicates the regional dynamic load factor, if The closer it is to 1, the slower the regional order pool is cleared;

[0209] The waiting costs for all three parties Regional stability coefficient and time penalty Synthesizing as a stable correction waiting cost :

[0210]

[0211] in, Indicate candidate solutions The cost of stabilization correction waiting; 0.95 is the preset high-load stability threshold, when When this occurs, it indicates a low risk of regional backlog. Not magnified, when If this occurs, it indicates a high risk of backlog in the regional order pool, and the waiting cost for that candidate solution will be automatically increased.

[0212] By adaptively adjusting the waiting costs of all three parties using a regional stability coefficient, and amplifying the costs of candidate solutions that may exacerbate backlogs under high load conditions, the same order waiting status can obtain different scheduling evaluations under different regional load conditions. Compared with the problem of using fixed waiting weights or fixed penalty coefficients, which cannot reflect changes in the order pool clearing capacity, this invention can retain cost efficiency when the region is stable, and actively favor solutions that are conducive to clearing backlogs when the region is close to saturation. This prevents local optimal solutions from continuously occupying capacity and amplifying order pool fluctuations, thereby improving the stability, recovery capacity, and anti-congestion capacity of peak-period scheduling.

[0213] S3.3 Calculate the uncertainty budget for candidate solutions;

[0214] Calculate the ETA fluctuation radius of the candidate scheme ,in, It is calculated from the ETA error in similar historical events, specifically based on the set of similar historical events. The mean absolute error between historical actual arrival times and historical predicted arrival times is calculated and used as the ETA fluctuation radius of the candidate route. The maximum ETA fluctuation radius of the historical candidate routes is recorded as . ;

[0215] Get the maximum number of unfulfilled orders allowed in the region. ,in, It can be set to the number of available vehicles in the area. The maximum number of orders that can be processed within a service cycle, for example, a region. Number of vehicles currently available Average order service time in the region Minutes / order, set a service cycle Each vehicle can complete a maximum of 3 orders within a service cycle, and the maximum number of uncompleted orders allowed in a region is [number missing]. Therefore, 24 units are designated as the region. The maximum number of orders that can be processed within a service cycle under current capacity conditions;

[0216] Calculate Uncertainty Budget :

[0217]

[0218] in, Indicate candidate solutions The uncertainty budget; the first term represents the time forecast uncertainty of the candidate solutions; the second term represents the uncertainty impact of the candidate solutions on the regional order pool backlog. Indicates the weight of time forecast uncertainty; This represents the uncertainty weight of the regional order pool backlog, preferably... , This indicates that the system simultaneously considers the ETA stability of candidate solutions and the stability of the regional order pool. As the value increases, the system places greater emphasis on the stability of the arrival time predictions of candidate solutions. As the number of users increases, the system places greater emphasis on the ability of candidate solutions to suppress regional order backlogs.

[0219] S3.4 Construct a comprehensive scheduling objective function and select the scheme with the smallest comprehensive objective value;

[0220] For candidate solutions Calculate the normalized value of transportation costs :

[0221]

[0222] in, This indicates the transportation cost of the candidate routes, based on the vehicle. The path length is obtained by summing the actual road network distances between adjacent nodes based on the corresponding order execution sequence and candidate paths. The cumulative travel time is calculated based on the effective speed of the corresponding road. And obtain the service time of each node. and the three-party execution waiting Then multiply by the vehicle unit mileage cost. Unit usage time cost (This may include vehicle depreciation, operating costs, etc.) and driver unit labor costs. Finally, the transportation cost of the candidate solution is obtained by summing the costs of all vehicles in the candidate solution. ; This represents the maximum transportation cost among the current set of candidate solutions;

[0223] Calculate the normalized value of the delay :

[0224]

[0225] in, Indicates the expected delivery time of the candidate proposal; Indicates the promised delivery time; This represents the normalized benchmark for maximum delay, set at 120 minutes. It can cover most of the more severe delays in logistics and delivery that can still be recovered through vehicle adjustments, order rearrangement, or cross-regional capacity supplementation.

[0226] Constructing a comprehensive scheduling objective function :

[0227]

[0228] in, Indicate candidate solutions The overall target value is such that the smaller the value, the better the solution. This represents the normalized value of transportation costs; This represents the normalized value of the delay; This indicates the cost of waiting for a stable correction. Indicates an uncertain budget; This represents the normalized benchmark for waiting costs, set at 180 minutes. This can cover typical high-waiting scenarios while avoiding excessive amplification of waiting metrics by a small number of extreme abnormal orders. Represents the weights of the objective function. This represents the weight of transportation costs, set to 0.20, so that the system can control vehicle mileage, execution time and labor costs while avoiding sacrificing order timeliness in the pursuit of low cost. The delay weight is set to 0.25 because on-time order fulfillment is a crucial objective of logistics scheduling, and its importance outweighs that of general transportation costs. The stable correction waiting cost weight is set to 0.40 and is set as the highest weight to highlight the collaborative optimization of the invention on the waiting pressure of warehouses, vehicles and customers, as well as long-tail waiting orders; The uncertainty budget weight is set to 0.15, which is used to suppress ETA fluctuations and regional order pool backlog risks, but avoids the forecast uncertainty from exceeding the actual transportation costs, delays and waiting status and dominating the scheduling results;

[0229] The solution with the smallest overall objective value was ultimately selected. :

[0230]

[0231] Among them, establishing the objective function The relationship between the scheduling scheme f and the objective function is used to select the scheduling scheme that minimizes the objective function value as the objective scheduling scheme. , Represents the set of candidate solutions;

[0232] Output target scheduling scheme The output includes vehicle number, driver number, pickup warehouse, locked order sequence, estimated arrival time, estimated delivery time, and stability correction waiting cost. Regional dynamic load factor Regional stability coefficient Uncertain budget ;

[0233] After order fulfillment is completed, the actual delivery time, actual waiting time, and whether there was a delay are obtained to generate a fulfillment result gene. :

[0234]

[0235] in, Indicate whether there is a delay; This indicates the actual total waiting time for the order; Indicate whether it belongs to the high-waiting-time abnormal order category;

[0236] The basic fulfillment genes, potential waiting genes, regional stability genes, and fulfillment result genes of the current order are merged to form a new complete event gene and written into the historical logistics event gene library.

[0237] In summary, this invention collects order, vehicle, warehouse, customer, road, and historical fulfillment data to divide delivery areas and calculate order arrival rates, average service times, and dynamic load factors, constructing a historical logistics event gene library to achieve a structured expression of logistics fulfillment scenarios. By generating current event genes and matching them with similar historical event genes, it identifies delay risks and high waiting anomaly risks in advance. Through capacity judgment, time window screening, and candidate solution construction, it eliminates unfeasible solutions in advance. By calculating the p-norm three-party waiting cost, time penalty, stable correction waiting cost, and uncertainty budget, it constructs a comprehensive scheduling objective function to achieve synergistic optimization of cost, timeliness, waiting pressure, and regional backlog risks. Ultimately, it improves logistics scheduling accuracy, order on-time fulfillment rate, regional anti-congestion capability, and overall platform operational efficiency.

[0238] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A smart logistics scheduling method based on big data analysis, characterized in that, include: Collect basic logistics scheduling data, divide delivery areas and calculate regional order arrival rates, calculate regional average order service time, and calculate regional dynamic load factors based on regional order arrival rates and average order service time to build a historical logistics event gene library; Calculate potential waiting time, generate current event genes, calculate order risk based on historical event genes, assess vehicle capacity, and then perform time window filtering based on order risk and regional dynamic load factors to form a set of candidate scheduling schemes. Finally, calculate the performance of each candidate scheme. - Norm three-party waiting cost; Calculate the time penalty for candidate solutions, obtain the number of uncompleted orders in the current region, calculate the stable adjusted waiting cost based on the time penalty and the third-party waiting cost, calculate the uncertainty budget for candidate solutions, construct a comprehensive scheduling objective function based on the stable adjusted waiting cost and the uncertainty budget, select the solution with the smallest comprehensive scheduling objective function value, and finally output the results and update the historical logistics event gene library.

2. The intelligent logistics scheduling method based on big data analysis as described in claim 1, characterized in that: The collection of basic logistics scheduling data, the division of delivery areas, and the calculation of regional order arrival rates refer to the process according to a fixed scheduling cycle. Collect logistics data; At each scheduling time Collect logistics data to form a dataset ; The current order will be assigned to the delivery area based on the warehouse service area corresponding to the order's delivery address. ; Get area Number of new orders in the last 30 minutes Based on the number of new orders Calculate the regional order arrival rate .

3. The intelligent logistics scheduling method based on big data analysis as described in claim 2, characterized in that: The calculation involves determining the average order service time for the region and calculating the regional dynamic load factor based on the regional order arrival rate and average order service time. A historical logistics event gene library is then constructed, drawing upon historical fulfillment data. Extracting regions The driving time and node service time of orders completed within the last 7 days, and the first Each historical order defines its order service time. ; For the region Completed within the last 7 days Calculate the average service time for each order ; Get area Number of vehicles currently available ; Based on regional order arrival rate Average order service time and the number of vehicles currently available Calculate the dynamic load factor of the region ; Define the basic performance genes Potential waiting genes Regional stable genes Genes of contract performance results , combined into the first Genes of historical events .

4. The intelligent logistics scheduling method based on big data analysis as described in claim 3, characterized in that: The process involves calculating the potential waiting time, generating the current event gene, and calculating the order risk index for the current pending order based on historical event genes. Get the order creation time and the current moment Calculate the potential waiting time ; Generate the current event gene based on the current order. ; Calculate the current event gene and the historical event gene Comprehensive distance ; Based on comprehensive distance Calculate the similarity between the current event gene and the historical event genes. ; Select the top results based on similarity from highest to lowest. A collection of historical events. ; Calculate the probability of delay in similar events based on order similarity. ; Calculate the probability of high waiting anomalies among similar events ; Weighted Delay Probability and abnormal probability Calculate order risk .

5. The intelligent logistics scheduling method based on big data analysis as described in claim 4, characterized in that: The process of determining vehicle capacity, then filtering by time window based on order risk and regional dynamic load factors to form a set of candidate scheduling schemes refers to the process of determining the capacity of each available vehicle. Capacity assessment and time window filtering are performed to establish a candidate order set corresponding to each candidate vehicle. ; Generate vehicles Candidate order sequence ; Based on the candidate order sequence, vehicles are formed. Corresponding candidate path ; The first part consists of candidate vehicles, candidate order sequences, and their corresponding candidate paths. One candidate scheduling scheme ; Form a set of candidate scheduling schemes for vehicle v .

6. The intelligent logistics scheduling method based on big data analysis as described in claim 5, characterized in that: The computational candidate scheme - Norm of three-party waiting cost refers to the calculation of vehicle Number of orders locked in this round ; The travel time of the path node for the l-th order Add them together to get the order travel time , obtain the Node service time for each order Calculate orders In candidate solutions The resulting third-party execution waiting costs ; Potential waiting time Execution waiting cost Travel time Service Hours Synthesized as candidate solutions - Normative third-party waiting costs ; Calculate candidate solutions Normalized three-party waiting costs .

7. The intelligent logistics scheduling method based on big data analysis as described in claim 6, characterized in that: The time penalty for calculating candidate solutions refers to the time penalty for obtaining candidate solutions. Actual path length ; Calculate the historical path growth coefficient of the region and the maximum node switching time in the region ; Calculate the upper bound of the path length of the candidate solution ; If the actual path length Not exceeding the upper limit Time penalty Set to 0; otherwise, the excess will be converted into a time penalty. .

8. The intelligent logistics scheduling method based on big data analysis as described in claim 7, characterized in that: The process of obtaining the number of incomplete orders in the current region and calculating the stable adjustment waiting cost based on time penalties and third-party waiting costs refers to obtaining the number of incomplete orders in the current region. ; Calculate candidate solutions Locked segment execution time ; Prediction uses candidate solutions Number of unfulfilled orders in the next scheduling cycle ; Calculate the regional stability coefficient ; The waiting costs for all three parties Regional stability coefficient and time penalty Synthesizing as a stable correction waiting cost .

9. The intelligent logistics scheduling method based on big data analysis as described in claim 8, characterized in that: The uncertainty budget for calculating the candidate solution refers to the ETA fluctuation radius of the candidate solution. ; Get the maximum number of unfulfilled orders allowed in the region. ; ETA fluctuation radius and the number of unfulfilled orders in the next scheduling cycle region Uncertainty budget calculated by weighting after normalization .

10. The intelligent logistics scheduling method based on big data analysis as described in claim 9, characterized in that: The process involves constructing a comprehensive scheduling objective function based on the stable correction waiting cost and the uncertainty budget, selecting the scheme with the smallest comprehensive scheduling objective function value, and finally outputting the result and updating the historical logistics event gene pool to match the candidate schemes. Calculate the normalized value of transportation costs ; Calculate the normalized value of the delay ; Based on normalized transportation cost Delay normalized value Stability correction waiting costs Normalized values ​​and uncertain budgets Constructing a comprehensive scheduling objective function ; Select the comprehensive scheduling objective function The solution with the minimum value ; Output target scheduling scheme ; After order fulfillment is completed, the actual delivery time, actual waiting time, and whether there was a delay are obtained to generate a fulfillment result gene. ; The basic fulfillment genes, potential waiting genes, regional stability genes, and fulfillment result genes of the current order are merged to form a new complete event gene and written into the historical logistics event gene library.

Images