Logistics efficient transportation method and system based on improved graph neural network

Abstract

The application discloses a logistics efficient transportation method and system based on an improved graph neural network, relates to the field of logistics transportation computer systems, and improves initial path planning accuracy by filling real-time Internet of Things data and constructing a globally visualized logistics transportation graph; further trains a space-time graph convolution network based on historical data, fuses space-time feature analysis, and greatly improves the path planning accuracy of a final space-time graph convolution network model for commodity transportation; by adjusting an initial collection interval, controls the overtime rate within a preset threshold range, and greatly reduces the average transportation time; finally, by setting an initial hop value of a sudden event propagation level, screening backup nodes, and iteratively optimizing the hop value, the proportion of time-consuming data for remediation and overall overtime data for the to-be-transported commodities is reduced; by adjusting the initial collection interval of the to-be-transported commodity transportation data, the waste of computing power caused by high-frequency collection is avoided, and the data update efficiency is ensured.

Description

Technical Field

[0001] This invention belongs to the field of logistics and transportation computer systems, and more specifically, relates to a method and system for efficient logistics transportation based on an improved graph neural network. Background Technology

[0002] In current logistics transportation, the data collection intervals used for optimal route planning are often fixed and cannot be dynamically adjusted according to the current logistics transportation situation (such as transportation routes, types of transported goods, etc.). This can easily lead to problems with the currently designated optimal route, resulting in logistics transportation timeouts. In addition, when logistics transportation encounters emergencies, the settings for backup transportation nodes are often fixed, which increases the time required to recover goods awaiting shipment. Summary of the Invention

[0003] In response to the problems in related technologies, this invention proposes an efficient logistics transportation method and system based on an improved graph neural network to overcome the aforementioned technical problems in existing related technologies.

[0004] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0005] This invention relates to an efficient logistics transportation method based on an improved graph neural network, comprising the following steps:

[0006] S1. Construct a logistics transportation map for the goods to be shipped;

[0007] S2. Based on the logistics transportation map of the goods to be transported, collect the logistics transportation maps and corresponding optimal transportation route data of the surrounding areas of multiple goods that have been transported in history, and construct the final spatiotemporal graph convolutional network model of the goods transportation.

[0008] S3. Set the initial collection interval for the transportation data of goods to be transported. Adjust the initial collection interval for the transportation data of goods to be transported based on the transportation timeout rate data of goods to be transported until the transportation timeout rate data of goods to be transported is less than the preset threshold, and obtain the final collection interval for the transportation data of goods to be transported.

[0009] S4. Based on the final collection interval of the goods transportation data to be transported, the goods transportation data to be transported are collected in real time and the optimal route is output in real time in combination with the final spatiotemporal graph convolutional network model of goods transportation, so as to carry out logistics transportation of the goods to be transported.

[0010] S5. In the logistics transportation of goods to be shipped, an initial jump value for the propagation level of an emergency is set; then, by screening backup nodes and combining the recovery time and timeout ratio, the initial jump value for the propagation level of an emergency is iteratively optimized.

[0011] Preferably, step S1 includes the following steps:

[0012] S11. Construct a logistics transportation graph for goods to be shipped; the logistics transportation graph for goods to be shipped includes nodes and edges; the node types include shipping warehouses, goods distribution centers, and transportation vehicles; the edge types include physical connection edges, logical dependency edges, and temporal constraint edges; set the feature types corresponding to each node type and edge type in the logistics transportation graph for goods to be shipped, and obtain the feature type set of shipping warehouses, the feature type set of goods distribution centers, and the feature type set of transportation vehicles;

[0013] S12. Set the weight calculation rules for the physical connection edges, logical dependency edges, and temporal constraint edges to obtain the physical weight calculation rules, logical weight calculation rules, and temporal weight calculation rules.

[0014] The physical weight calculation rule is actual transportation route distance × road condition coefficient; the logical weight calculation rule is order relevance; the time-series weight calculation rule is time window priority.

[0015] S13. Based on the set of features of the shipping warehouse, the set of features of the commodity distribution center, the set of features of the transportation vehicle, the physical weight calculation rules, the logical weight calculation rules, and the time-series weight calculation rules, and by synchronizing IoT device data in real time through the MQTT protocol, calculate and fill the data in the logistics transportation map of the goods to be shipped.

[0016] By constructing a logistics transportation map for goods to be shipped, ineffective transportation can be eliminated first. Through topological modeling of physical connection edges, convection and detour phenomena in transportation paths can be identified and eliminated, reducing unnecessary travel mileage. This provides input for subsequent dynamic calculation of the shortest path using graph neural networks. At the same time, combined with real-time location updates of transportation vehicles, cross-regional delivery routes can be optimized. In addition, based on the temporal weight calculation rules of temporal constraint edges, path replanning can be automatically triggered, improving response timeliness.

[0017] Subsequently, the graph neural network model was trained and tested using the logistics transportation map of the goods to be shipped. This enabled the graph neural network model to combine the load / location data of the transportation vehicles with logical dependency edges (order collaboration relationships), thereby optimizing the multi-vehicle collaborative loading and unloading strategy. Finally, the road condition coefficient of the physical connection edge can be dynamically adjusted (such as road interruption caused by heavy rain) to trigger a rapid switch to alternative routes.

[0018] Preferably, step S2 includes the following steps:

[0019] S21. Set several types of goods that were transported under normal historical conditions and the boundaries of the logistics surrounding areas to obtain a historical transported goods set; based on the logistics transport map of the goods to be transported and the boundaries of the logistics surrounding areas, obtain the logistics transport map and the optimal transport path data of each item in the historical transported goods set to obtain a historical goods logistics transport map dataset and a historical goods logistics transport optimal path dataset.

[0020] S22. Construct an initial spatiotemporal graph convolutional network model for commodity transportation;

[0021] S23. The initial spatiotemporal graph convolutional network model for commodity transportation is trained and tested using the historical commodity logistics transportation graph dataset and the historical commodity logistics transportation optimal path dataset; after training and testing, the final spatiotemporal graph convolutional network model for commodity transportation is obtained.

[0022] Historical transportation graph data contains multi-dimensional information such as physical topology, dynamic road conditions, and resource load rate, enabling the model to learn the deep interaction patterns between nodes (warehouses / distribution centers) and edges (transportation paths / temporal constraints) in the transportation network. Among them, the historical optimal route dataset for commodity logistics transportation provides decision feedback in real-world scenarios, enhancing the model's ability to comprehensively weigh objectives such as transportation costs and time window constraints. In addition, the spatiotemporal graph convolution module can simultaneously process static topological features (warehouse capacity, road distance) and dynamic temporal features (real-time vehicle location, road condition changes), thereby achieving dynamic optimal route prediction for the transportation network.

[0023] Preferably, step S3 includes the following steps:

[0024] S31. Set the data collection period for the transportation of goods to be shipped and the initial data collection interval for the transportation of goods to be shipped. Using the method of S13, the data in the logistics transportation map of goods to be shipped is collected and filled in real time multiple times within the data collection period for the transportation of goods to be shipped according to the initial data collection interval for the transportation of goods to be shipped, so as to obtain the initial real-time logistics transportation map matrix of goods to be shipped.

[0025] S32. Set a timeout rate threshold for goods transportation; input each real-time goods logistics transportation map in the real-time goods logistics transportation map matrix into the final spatiotemporal graph convolutional network model of goods transportation to perform optimal path mapping, and obtain the optimal transportation path matrix of real-time goods waiting to be transported.

[0026] S33. Adjust the initial collection interval of the commodity transportation data according to the commodity transportation timeout rate threshold and the real-time optimal transportation path matrix of the commodity to be transported;

[0027] Preferably, step S33 includes the following steps:

[0028] S331. Transport the corresponding goods according to each row of data in the real-time optimal transportation path matrix for goods to be transported; after all goods have been transported, calculate the transportation timeout rate data for goods to be transported.

[0029] S332. When the timeout rate data of the goods to be shipped is greater than or equal to the timeout rate threshold, the initial collection interval of the goods to be shipped data is adjusted until the timeout rate of the goods to be shipped is less than the timeout rate threshold, and the final collection interval of the goods to be shipped data is obtained.

[0030] Based on the complexity of the transportation task, a baseline is set (initial data collection interval for goods to be transported: e.g., 5 minutes / time) to avoid the waste of computing power caused by high-frequency collection (reducing redundant data transmission by 30%-40%); the interval is adjusted according to the timeout rate (shortened to 2 minutes or extended to 10 minutes) to ensure data update efficiency, thereby ensuring the normal transportation of goods.

[0031] Preferably, the Dragonfly optimization algorithm is used to adjust the initial collection interval of the transportation data of the goods to be transported in S332;

[0032] The Dragonfly Optimization Algorithm, by simulating swarm intelligence behavior, can quickly find the optimal solution under multi-dimensional constraints (such as transportation routes, capacity resources, and time windows). This characteristic is suitable for dynamically changing timeout problems in logistics transportation, allowing for real-time adjustment of data collection frequency to balance monitoring accuracy and system load. Furthermore, compared to traditional gradient descent algorithms, the Dragonfly Algorithm has stronger global search capabilities. When faced with multiple local optima in the transportation network (such as the correlation between transportation delays in different regions), it can avoid getting trapped in local optima and find a more reasonable data collection interval adjustment scheme. Moreover, transportation timeout rate and data collection interval often exhibit a non-linear relationship (such as a sudden change in the curve after a specific threshold). The iterative mechanism of the Dragonfly Algorithm can effectively capture this non-linear characteristic, setting the critical data collection frequency more accurately than linear models. Compared to similar swarm intelligence algorithms such as genetic algorithms, the Dragonfly Algorithm has a faster convergence speed and simpler parameter tuning. This is particularly important for real-time logistics systems requiring high-frequency adjustments (such as air freight scheduling), as it can reduce the impact of computational latency on decision-making.

[0033] Preferably, step S4 includes the following steps:

[0034] S41. Set mutation thresholds for corresponding features in the set of feature types of shipping warehouses, the set of feature types of commodity distribution centers, and the set of feature types of transportation vehicles to obtain the mutation threshold sets for the features of shipping warehouses, commodity distribution centers, and transportation vehicles.

[0035] S42. The data in the logistics transportation map of the goods to be transported is collected in real time using the final collection interval of the transportation data of the goods to be transported, and the optimal route is output in real time by combining it with the convolutional network model of the final spatiotemporal graph of the goods transportation; the optimal route is used to carry out logistics transportation of the goods to be transported.

[0036] Among them, when the absolute value of the difference between the data in the logistics transportation map of the goods to be transported collected at adjacent intervals is greater than or equal to the data corresponding to the set of thresholds for mutation of the characteristics of the shipping warehouse, the set of thresholds for mutation of the characteristics of the goods distribution center, and the set of thresholds for mutation of the characteristics of the transportation vehicle, it is recorded as a sudden event in the transportation of goods.

[0037] By setting mutation thresholds for features corresponding to each node type in the logistics transportation map of goods to be transported, a quantitative basis is provided for determining whether the data difference of the corresponding feature in adjacent data collection intervals constitutes a mutation. Then, corresponding remedial measures are taken in the event of a data mutation, thereby ensuring the normal operation of goods transportation.

[0038] Preferably, step S5 includes the following steps:

[0039] S51. Set the initial jump value for the propagation level of an emergency, denoted as... N The scope of impact of the aforementioned unforeseen event in commodity transportation is defined as starting from the unforeseen event itself. N Skip the propagation level; when the aforementioned unforeseen event in goods transportation occurs, set the propagation level starting from the unforeseen event in goods transportation. N Nodes in the logistics transportation graph of goods that are of the same type as sudden events in goods transportation and are outside the scope of the propagation level are recorded as backup nodes; the time taken to remedy sudden events in goods transportation using the backup nodes and the timeout time of the final goods transportation are recorded multiple times to obtain the data set of remedy time and the data set of final goods transportation time.

[0040] S52. Set a threshold for the time consumption of pending recovery and a threshold for the proportion of timeout in the final pending goods logistics transportation. When the proportion of data in the pending recovery time dataset that is greater than or equal to the threshold for the time consumption of pending recovery is greater than or equal to 50%, or the proportion of timeout data in the final pending goods logistics transportation timeout dataset is greater than or equal to the threshold for the proportion of timeout in the final pending goods logistics transportation timeout, adjust the initial jump value of the emergency propagation level until the proportion of data in the pending recovery time dataset that is greater than or equal to the threshold for the time consumption of pending recovery is less than 50% and the proportion of timeout data in the final pending goods logistics transportation timeout dataset is less than the threshold for the proportion of timeout in the final pending goods logistics transportation timeout, and obtain the final jump value of the emergency propagation level.

[0041] By dynamically adjusting the number of hops (setting a baseline value (e.g., recovery time ≤ 4 hours, triggering hop adjustment if the percentage exceeds the threshold > 50%), N is automatically expanded or contracted based on the event recovery effect, avoiding excessive expansion or contraction of the emergency scope leading to low emergency efficiency; emergency nodes are screened based on: distance ≥ N hops; load safety threshold (CPU < 70% + storage capacity > emergency demand 120%); and connection redundancy (≥ 2 independent physical links, such as highway + railway in parallel), the response success rate of backup nodes that meet the conditions is greatly improved.

[0042] Preferably, adjusting the initial jump value of the sudden event propagation level in S52 includes the following steps:

[0043] S521. Set the value range of the initial jump value of the propagation level of the sudden event and generate the propagation level jump number to adjust the initial position of each dragonfly in the dragonfly population to obtain the second initial position set.

[0044] S522. Construct the fitness function for adjusting the hop count of the propagation hierarchy for the dragonfly population;

[0045] S523. Start the iteration; in each iteration, use the fitness function of the propagation level hop count to adjust the fitness value of the position of each dragonfly in the propagation level hop count adjusted dragonfly population obtained in the previous iteration to calculate the fitness value of the position of each dragonfly in the propagation level hop count adjusted dragonfly population obtained in the previous iteration and update the position of each dragonfly in the propagation level hop count adjusted dragonfly population obtained in the previous iteration.

[0046] S524. When the maximum number of iterations is reached, stop the iteration and obtain the second final global optimal position; otherwise, continue the iteration until the maximum number of iterations is reached; after the iteration is completed, if the proportion of data in the pending recovery time dataset that is greater than or equal to the pending recovery time threshold is less than 50% and the proportion of timeout data in the final pending goods logistics transportation timeout dataset is less than the final pending goods logistics transportation timeout proportion threshold, the second final global optimal position is used as the final jump value of the sudden event propagation level; otherwise, return to S523 to continue the iteration.

[0047] The Dragonfly Optimization Algorithm simulates the predation and migration behavior of dragonfly populations, achieving a dual balance between global search and local exploitation within the solution space. Compared to genetic algorithms, it can quickly locate the optimal hop count in logistics network propagation path optimization, reducing the risk of adjustment failure due to local optima (fault tolerance rate increased to 99.2%). By combining a logistics timeout ratio threshold with a dynamic adjustment strategy based on individual distance weights, the iteration process of the hierarchical hop count N has the ability to adapt to sudden traffic surges. Experiments show that the Dragonfly Optimization Algorithm can shorten the initial hop count adjustment cycle to 1 / 3 of the traditional method (average time reduced from 45 minutes to 15 minutes), ensuring adjustment efficiency.

[0048] The efficient logistics transportation system based on improved graph neural networks includes a commodity logistics transportation graph construction module, a logistics transportation graph initial data filling module, a historical commodity transportation data acquisition module, a spatiotemporal graph convolutional network model construction module, a real-time acquisition parameter setting module for goods to be transported, a real-time acquisition interval adjustment module, a goods logistics transportation module for goods to be transported, and an event recovery level jump number adjustment module.

[0049] The present invention has the following beneficial effects:

[0050] 1. This invention constructs a globally visualized logistics transportation map by filling it with real-time IoT data, thereby improving the accuracy of initial path planning. Then, a spatiotemporal graph convolutional network is trained based on historical data, and spatiotemporal feature analysis is integrated to significantly improve the accuracy of the final spatiotemporal graph convolutional network model for path planning in commodity transportation. By adjusting the initial collection interval, the timeout rate is controlled within a preset threshold range, and data is collected at the optimal interval to update the path in real time, greatly reducing the average transportation time. Finally, by setting the initial hop value for the propagation level of sudden events, backup nodes are selected, and the number of hops is iteratively optimized using the Dragonfly algorithm, reducing the proportion of data on waiting-to-ship recovery time and the overall timeout data for waiting-to-ship goods transportation.

[0051] 2. In this invention, a baseline is set based on the complexity of the transportation task, i.e., the initial collection interval of the transportation data of the goods to be transported, to avoid the waste of computing power caused by high-frequency collection; the interval is adjusted according to the timeout rate, which also ensures the efficiency of data update, thereby ensuring the normal transportation of goods.

[0052] 3. In this invention, dynamic hop count adjustment is set. If the proportion exceeds the threshold > 50%, the hop count adjustment is triggered. The number N is automatically expanded or contracted according to the event recovery effect, avoiding excessive expansion or contraction of the emergency scope, which would lead to low emergency efficiency. Emergency nodes are screened based on the following criteria: distance ≥ N hops; load safety threshold (CPU < 70% + storage capacity > emergency demand 120%); and connection redundancy (≥ 2 independent physical links). This greatly improves the response success rate of backup nodes that meet the conditions.

[0053] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of the overall process of the efficient logistics transportation method based on the improved graph neural network of the present invention;

[0056] Figure 2 This is a schematic diagram of the process of the efficient logistics transportation method based on the improved graph neural network of the present invention;

[0057] Figure 3 This is a schematic diagram illustrating the process of adjusting the initial collection interval of transportation data for goods to be transported according to the present invention;

[0058] Figure 4 This is a schematic diagram illustrating the process of adjusting the initial jump value of the propagation level of an emergency according to the present invention;

[0059] Figure 5 This is a schematic diagram of the modules of the efficient logistics transportation system based on the improved graph neural network of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the invention.

[0061] Example 1

[0062] Please see Figure 1-4 This embodiment describes an efficient logistics transportation method based on an improved graph neural network, which includes the following steps:

[0063] S1. Construct a logistics transportation map for the goods to be shipped;

[0064] S1 includes the following steps:

[0065] S11. Construct a logistics transportation graph for goods awaiting shipment; the logistics transportation graph for goods awaiting shipment includes nodes and edges; the node types include shipping warehouses, goods distribution centers, and transportation vehicles; the edge types include physical connection edges, logical dependency edges, and temporal constraint edges; define the feature types corresponding to each node type and edge type in the logistics transportation graph for goods awaiting shipment to obtain a feature type set for shipping warehouses, a feature type set for goods distribution centers, and a feature type set for transportation vehicles; the feature type set for shipping warehouses includes geographical coordinates (latitude and longitude), maximum capacity (tons), and current load rate (%), etc.; the feature type set for goods distribution centers includes radiation radius (km), average daily order volume (orders), and cold chain equipment ratio (%), etc.; the feature type set for transportation vehicles includes real-time location (GPS coordinates), remaining load (tons), and maximum speed (km / h), etc.

[0066] S12. Set the weight calculation rules for the physical connection edges, logical dependency edges, and temporal constraint edges to obtain the physical weight calculation rules, logical weight calculation rules, and temporal weight calculation rules.

[0067] The physical weight calculation rule is the actual transportation route distance (km) × road condition coefficient (0-1); for example: Warehouse A → Distribution Center B (distance 200km, road condition 0.8, then the corresponding physical weight is 200×0.8=160); the logical weight calculation rule is the order correlation (number of shared orders / total number of orders); for example: Distribution Center C E-commerce platform D (collaborative orders account for 35%); the time sequence weight calculation rule is time window priority (remaining time / deadline); for example: fresh food transportation task E (4 hours remaining / total time limit 8 hours);

[0068] S13. Based on the set of features of the shipping warehouse, the set of features of the commodity distribution center, the set of features of the transportation vehicle, the physical weight calculation rules, the logical weight calculation rules, and the time-series weight calculation rules, and by synchronizing IoT device data in real time through the MQTT protocol, calculate and fill the data in the logistics transportation map of the goods to be shipped.

[0069] S2. Based on the logistics transportation map of the goods to be transported, collect the logistics transportation maps and corresponding optimal transportation route data of the surrounding areas of multiple goods that have been transported in history, and construct the final spatiotemporal graph convolutional network model of the goods transportation.

[0070] S2 includes the following steps:

[0071] S21. Set several types of goods that were transported under normal historical conditions and the boundaries of the logistics surrounding areas to obtain a historical transported goods set; based on the logistics transport map of the goods to be transported and the boundaries of the logistics surrounding areas, obtain the logistics transport map and the optimal transport path data of each item in the historical transported goods set to obtain a historical goods logistics transport map dataset and a historical goods logistics transport optimal path dataset.

[0072] S22. Construct an initial spatiotemporal graph convolutional network model for commodity transportation; the initial spatiotemporal graph convolutional network model for commodity transportation includes an input layer for receiving spatiotemporal graph data (node ​​feature matrix + adjacency matrix); two spatiotemporal convolutional blocks: each layer contains an alternating combination of temporal convolution (TCN) and spatial convolution (GCN) and a fully connected output layer; and Dijkstra's algorithm is embedded after the fully connected output layer to calculate the optimal path;

[0073] S23. The initial spatiotemporal graph convolutional network model for commodity transportation is trained and tested using the historical commodity logistics transportation graph dataset and the historical commodity logistics transportation optimal path dataset; after training and testing, the final spatiotemporal graph convolutional network model for commodity transportation is obtained.

[0074] S23 includes the following steps:

[0075] S231. Set the proportion of historical transportation training data; use the proportion of historical transportation training data to divide the historical commodity logistics transportation map dataset and the historical commodity logistics transportation optimal path dataset to obtain the historical commodity logistics transportation map training dataset, the historical commodity logistics transportation optimal path training dataset, the historical commodity logistics transportation map test dataset, and the historical commodity logistics transportation optimal path test dataset.

[0076] S232. Set a training error threshold; input the historical commodity logistics transportation map training dataset and the historical commodity logistics transportation optimal path training dataset as training data and training label data respectively into the commodity transportation initial spatiotemporal graph convolutional network model for training; during the training process, when the training error is less than the training error threshold, stop training and obtain the trained spatiotemporal graph convolutional network model; otherwise, continue training until the training error is less than the training error threshold.

[0077] S233. Set a test accuracy threshold; input the historical commodity logistics transportation map test dataset and the historical commodity logistics transportation optimal path test dataset as test data and test label data, respectively, into the trained spatiotemporal graph convolutional network model for testing; after the test is completed, obtain the test accuracy data; when the test accuracy data is greater than or equal to the test accuracy threshold, use the trained spatiotemporal graph convolutional network model as the final spatiotemporal graph convolutional network model for commodity transportation; otherwise, return to S232 to continue training the trained spatiotemporal graph convolutional network model until the test accuracy data is greater than or equal to the test accuracy threshold.

[0078] S3. Set the initial collection interval for the transportation data of goods to be transported. Adjust the initial collection interval for the transportation data of goods to be transported based on the transportation timeout rate data of goods to be transported until the transportation timeout rate data of goods to be transported is less than the preset threshold, and obtain the final collection interval for the transportation data of goods to be transported.

[0079] S3 includes the following steps:

[0080] S31. Set the data collection period for the transportation of goods to be shipped and the initial data collection interval for the transportation of goods to be shipped. Using the method of S13, the data in the logistics transportation map of goods to be shipped is collected and filled in real time multiple times within the data collection period for the transportation of goods to be shipped according to the initial data collection interval for the transportation of goods to be shipped, so as to obtain the initial real-time logistics transportation map matrix of goods to be shipped.

[0081] S32. Set a timeout rate threshold for goods transportation; input each real-time goods logistics transportation map in the real-time goods logistics transportation map matrix into the final spatiotemporal graph convolutional network model of goods transportation to perform optimal path mapping, and obtain the optimal transportation path matrix of real-time goods waiting to be transported.

[0082] S33. Adjust the initial collection interval of the commodity transportation data according to the commodity transportation timeout rate threshold and the real-time optimal transportation path matrix of the commodity to be transported;

[0083] S33 includes the following steps:

[0084] S331. Transport the corresponding goods according to each row of data in the real-time optimal transportation path matrix for goods to be transported; after all goods have been transported, calculate the transportation timeout rate data for goods to be transported.

[0085] S332. When the timeout rate data of the goods to be shipped is greater than or equal to the timeout rate threshold, the initial collection interval of the goods to be shipped data is adjusted until the timeout rate of the goods to be shipped is less than the timeout rate threshold, and the final collection interval of the goods to be shipped data is obtained.

[0086] S332 includes the following steps in adjusting the initial collection interval of the transportation data of the goods to be shipped:

[0087] S3321. Set the value range of the initial collection interval for the transportation data of the goods to be transported, and obtain the value range of the initial collection interval for transportation data. , , These represent the lower and upper limits of the initial data collection interval for the goods to be transported; a dragonfly population for adjusting the transport data collection interval is constructed; and the maximum number of iterations for the dragonfly population for adjusting the transport data collection interval is set to [value missing]. And the current iteration number is , denoted as the maximum number of iterations for interval adjustment and the current number of iterations for interval adjustment, respectively; the search space dimension of the dragonfly population for the transportation data acquisition interval adjustment is 1;

[0088] S3322. Based on the initial data acquisition interval range, generate a data acquisition interval adjustment for each dragonfly in the dragonfly population to obtain a first initial position set. , This indicates that the adjustment of the transportation data collection interval in the dragonfly population is the [number]th [number]. i The initial position of the dragonfly. The interval for collecting the transport data indicates that the size of the dragonfly population is adjusted; the generation formula is as follows:

[0089]

[0090] In the formula, rand 1i Indicating targeting Generate random numbers between 0 and 1;

[0091] S3323. Construct the fitness function for adjusting the dragonfly population based on the transportation data acquisition interval. b 1; as follows,

[0092]

[0093] In the formula, This indicates that the timeout rate data for the transportation of goods to be shipped is obtained by applying the data collection interval of the goods transportation obtained in each iteration to S31, S32 and S33;

[0094] S3324. Start the iteration. Before the iteration, set the current iteration count of the interval adjustment to 1. During the first iteration, use the fitness function of the dragonfly population adjusted by the transportation data collection interval to calculate the fitness value of the initial position of each dragonfly in the first initial position set, and obtain the first fitness value set. Take the largest fitness value in the first fitness value set and the corresponding initial position of the dragonfly as the first global best fitness and the first global best position, respectively. Update the initial position of each dragonfly in the first initial position set according to the first global best fitness and the first global best position. After the update is completed, increment the current iteration count of the interval adjustment by 1 and enter the next iteration.

[0095] In each iteration, the fitness function of the dragonfly population adjusted by the transportation data acquisition interval is used to calculate the fitness value of the position of each dragonfly in the transportation data acquisition interval adjusted dragonfly population updated in the previous iteration, resulting in a second fitness value set. The maximum fitness value in the second fitness value set and the corresponding dragonfly position are respectively taken as the second global best fitness and the second global best position. The position of each dragonfly in the transportation data acquisition interval adjusted dragonfly population updated in the previous iteration is updated according to the second global best fitness and the second global best position. After the update is completed, the current iteration number of the interval adjustment is incremented by 1 and the next iteration is started.

[0096] S3325, when If the first final global optimum is reached, stop the iteration and obtain the first final global optimum position and the first final global optimum fitness; otherwise, continue the iteration until... Until the time limit is reached; when the reciprocal of the first final global best fitness is less than the commodity transportation timeout rate threshold, the first final global best position is taken as the final collection interval of the commodity transportation data; otherwise, return to S3324 to continue iterating until the reciprocal of the first final global best fitness is less than the commodity transportation timeout rate threshold.

[0097] The Dragonfly Optimization Algorithm, by simulating swarm intelligence behavior, can quickly find the optimal solution under multi-dimensional constraints (such as transportation routes, capacity resources, and time windows). This characteristic is suitable for dynamically changing timeout problems in logistics transportation, allowing for real-time adjustment of data collection frequency to balance monitoring accuracy and system load. Furthermore, compared to traditional gradient descent algorithms, the Dragonfly Algorithm has stronger global search capabilities. When faced with multiple local optima in the transportation network (such as the correlation between transportation delays in different regions), it can avoid getting trapped in local optima and find a more reasonable data collection interval adjustment scheme. Moreover, transportation timeout rate and data collection interval often exhibit a non-linear relationship (such as a sudden change in the curve after a specific threshold). The iterative mechanism of the Dragonfly Algorithm can effectively capture this non-linear characteristic, setting the critical data collection frequency more accurately than linear models. Compared to similar swarm intelligence algorithms such as genetic algorithms, the Dragonfly Algorithm has a faster convergence speed and simpler parameter tuning. This is particularly important for real-time logistics systems requiring high-frequency adjustments (such as air freight scheduling), as it can reduce the impact of computational latency on decision-making.

[0098] S4. Based on the final collection interval of the goods transportation data to be transported, the goods transportation data to be transported are collected in real time and the optimal route is output in real time in combination with the final spatiotemporal graph convolutional network model of goods transportation, so as to carry out logistics transportation of the goods to be transported.

[0099] S4 includes the following steps:

[0100] S41. Set mutation thresholds for corresponding features in the set of feature types of shipping warehouses, the set of feature types of commodity distribution centers, and the set of feature types of transportation vehicles to obtain the mutation threshold sets for the features of shipping warehouses, commodity distribution centers, and transportation vehicles.

[0101] S42. The data in the logistics transportation map of the goods to be transported is collected in real time using the final collection interval of the transportation data of the goods to be transported, and the optimal route is output in real time by combining it with the convolutional network model of the final spatiotemporal graph of the goods transportation; the optimal route is used to carry out logistics transportation of the goods to be transported.

[0102] Among them, when the absolute value of the difference between the data in the logistics transportation map of the goods to be transported collected at adjacent intervals is greater than or equal to the data corresponding to the set of thresholds for mutation of the characteristics of the shipping warehouse, the set of thresholds for mutation of the characteristics of the goods distribution center, and the set of thresholds for mutation of the characteristics of the transportation vehicle, it is recorded as a sudden event in the transportation of goods.

[0103] S5. In the logistics transportation of the goods to be shipped, an initial jump value for the propagation level of an emergency is set; then, by screening backup nodes and combining the recovery time and timeout ratio, the initial jump value for the propagation level of an emergency is iteratively optimized.

[0104] S5 includes the following steps:

[0105] S51. Set the initial jump value for the propagation level of an emergency, denoted as... N The scope of impact of the aforementioned unforeseen event in commodity transportation is defined as starting from the unforeseen event itself. N The propagation hierarchy is hopped (each hop covers a radius of approximately 50-100 kilometers); when the aforementioned goods transportation emergency occurs, the propagation is set at the point where the goods transportation emergency originates. N Nodes in the logistics transportation graph of goods transported in the same type as the sudden incident in goods transportation, located outside the hop propagation level range, are designated as backup nodes. The time taken to remedy the sudden incident using these backup nodes and the final logistics transportation timeout for the goods are recorded multiple times to obtain a backup remedy time dataset and a final logistics transportation time dataset for the goods. The backup nodes must meet the following requirements: located outside the disaster area and within a safe radiation range (distance from the event center ≥ N hop propagation levels); have a load safety threshold (CPU < 70% + storage capacity > emergency demand ≥ 120%); and have at least two physical connection edges that can directly reach the area where the sudden incident in goods transportation occurred.

[0106] S52. Set a threshold for the time consumption of pending recovery and a threshold for the proportion of timeout in the final pending goods logistics transportation. When the proportion of data in the pending recovery time dataset that is greater than or equal to the threshold for the time consumption of pending recovery is greater than or equal to 50%, or the proportion of timeout data in the final pending goods logistics transportation timeout dataset is greater than or equal to the threshold for the proportion of timeout in the final pending goods logistics transportation timeout, adjust the initial jump value of the emergency propagation level until the proportion of data in the pending recovery time dataset that is greater than or equal to the threshold for the time consumption of pending recovery is less than 50% and the proportion of timeout data in the final pending goods logistics transportation timeout dataset is less than the threshold for the proportion of timeout in the final pending goods logistics transportation timeout, and obtain the final jump value of the emergency propagation level.

[0107] The adjustment of the initial jump value of the sudden event propagation level in S52 includes the following steps:

[0108] S521. Set the value range of the initial hop count for the propagation level of the sudden event, and obtain the value range of the propagation level hop count. , , These represent the lower and upper limits of the initial hop count for the propagation level of the sudden event, respectively; a dragonfly population for adjusting the hop count of the propagation level is constructed; and the maximum number of iterations for the dragonfly population for adjusting the hop count of the propagation level is set to [value missing]. And the current iteration number is , respectively denoted as the maximum number of iterations for hop count adjustment and the current number of iterations for hop count adjustment; the search space dimension of the dragonfly population for hop count adjustment at the propagation level is 1;

[0109] S522. Based on the range of propagation level hop count values, generate a propagation level hop count to adjust the initial position of each dragonfly in the dragonfly population, thus obtaining a second initial position set. , This indicates that the adjustment of the transportation data collection interval in the dragonfly population is the [number]th [number]. i The initial position of the dragonfly. The interval for collecting the transport data indicates that the size of the dragonfly population is adjusted; the generation formula is as follows:

[0110] In the formula, rand 2i Indicating targeting Generate random numbers between 0 and 1;

[0111] S523. Construct the fitness function of the dragonfly population to adjust the hop count of the propagation hierarchy. b 2; as follows,

[0112]

[0113] In the formula, This indicates the proportion of data in the pending recovery time dataset that is greater than or equal to the pending recovery time threshold. This represents the proportion of timeout data in the final shipment logistics timeout dataset;

[0114] S524. Begin iteration. Before each iteration, set the hop count to 1 for the current iteration count. During the first iteration, use the propagation level hop count adjustment function to calculate the fitness value of the initial position of each dragonfly in the second initial position set, obtaining the third fitness value set. Use the maximum fitness value in the third fitness value set and the corresponding initial position of the dragonfly as the third global best fitness and the third global best position, respectively. Update the initial position of each dragonfly in the second initial position set according to the third global best fitness and the third global best position. After the update, increment the hop count adjustment for the current iteration count by 1 and proceed to the next iteration.

[0115] In each iteration, the fitness function of the dragonfly population adjusted by the propagation hierarchy hop count is used to calculate the fitness value of each dragonfly in the propagation hierarchy hop count adjusted dragonfly population obtained in the previous iteration, resulting in a fourth fitness value set. The largest fitness value in the fourth fitness value set and the corresponding dragonfly position are taken as the fourth global best fitness and the fourth global best position, respectively. The position of each dragonfly in the propagation hierarchy hop count adjusted dragonfly population obtained in the previous iteration is updated according to the fourth global best fitness and the fourth global best position. After the update is completed, the hop count adjustment current iteration number is incremented by 1 and the next iteration is started.

[0116] S525, when If the condition is met, stop iterating to obtain the second final global optimal position; otherwise, continue iterating until... Until the iteration is complete, if the proportion of data in the pending recovery time dataset that is greater than or equal to the pending recovery time threshold is less than 50% and the proportion of timeout data in the final pending goods logistics transportation timeout dataset is less than the final pending goods logistics transportation timeout proportion threshold, then the second final global optimal position is taken as the final jump value of the sudden event propagation level; otherwise, return to S524 to continue the iteration.

[0117] Example 2

[0118] Please see Figure 5 This embodiment discloses an efficient logistics transportation system based on an improved graph neural network. The system can implement the method of the above embodiment, including a commodity logistics transportation graph construction module, a logistics transportation graph initial data filling module, a historical commodity transportation data collection module, a spatiotemporal graph convolutional network model construction module, a real-time collection parameter setting module for goods to be transported, a real-time collection interval adjustment module, a goods logistics transportation module for goods to be transported, and an event recovery level jump number adjustment module.

[0119] The commodity logistics transportation map construction module constructs a logistics transportation map of the goods to be transported.

[0120] The initial data filling module for the logistics transportation map calculates and fills the data in the logistics transportation map of the goods to be transported by synchronizing data from IoT devices in real time.

[0121] The historical commodity transportation data acquisition module collects logistics transportation maps and corresponding optimal transportation route data of multiple commodities that have been transported in the past, and obtains historical commodity logistics transportation map dataset and historical commodity logistics transportation optimal route dataset.

[0122] The spatiotemporal graph convolutional network model building module uses the historical commodity logistics transportation graph dataset and the historical commodity logistics transportation optimal path dataset to build the final spatiotemporal graph convolutional network model for commodity transportation.

[0123] The real-time data collection parameter setting module for goods awaiting shipment sets the initial data collection interval for the goods awaiting shipment.

[0124] The real-time acquisition interval adjustment module transports multiple goods to be transported according to the initial acquisition interval of the goods transportation data and counts the corresponding goods transportation timeout rate data. Then, it adjusts the initial acquisition interval of the goods transportation data according to the goods transportation timeout rate data until the goods transportation timeout rate data is less than the preset threshold, and obtains the final acquisition interval of the goods transportation data.

[0125] The logistics transportation module for goods awaiting shipment collects data in real time from the logistics transportation map of goods awaiting shipment according to the final collection interval of the goods transportation data, and outputs the optimal route in real time in combination with the final spatiotemporal graph convolutional network model of goods transportation, so as to carry out logistics transportation of goods awaiting shipment.

[0126] The event recovery level jump number adjustment module is based on the logistics transportation of goods to be shipped in S4, and sets the initial jump value of the emergency event propagation level; then, by screening backup nodes and combining the recovery time and timeout ratio, the initial jump value of the emergency event propagation level is iteratively optimized to obtain the final jump value of the emergency event propagation level.

[0127] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0128] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A high-efficiency logistics transportation method based on an improved graph neural network, characterized in that, Includes the following steps: S1. Construct a logistics transportation map for the goods to be shipped; Specifically, this includes: constructing a logistics transportation map for goods awaiting shipment; the logistics transportation map for goods awaiting shipment includes nodes and edges; node types include shipping warehouses, goods distribution centers, and transportation vehicles; edge types include physical connection edges, logical dependency edges, and temporal constraint edges; setting feature types corresponding to each node type and edge type in the logistics transportation map for goods awaiting shipment, obtaining a feature type set for shipping warehouses, a feature type set for goods distribution centers, and a feature type set for transportation vehicles; setting weight calculation rules for the physical connection edges, logical dependency edges, and temporal constraint edges, obtaining physical weight calculation rules, logical weight calculation rules, and temporal weight calculation rules; the physical weight calculation rule is actual transportation path distance × road condition coefficient; the logical weight calculation rule is order correlation degree; the temporal weight calculation rule is time window priority; and calculating and filling the data in the logistics transportation map for goods awaiting shipment based on the feature type set for shipping warehouses, the feature type set for goods distribution centers, the feature type set for transportation vehicles, the physical weight calculation rules, the logical weight calculation rules, and the temporal weight calculation rules, and by synchronizing IoT device data in real time via the MQTT protocol. S2. Based on the logistics transportation map of the goods to be transported, collect the logistics transportation maps and corresponding optimal transportation route data of the surrounding areas of multiple goods that have been transported in history, and construct the final spatiotemporal graph convolutional network model of the goods transportation. S3. Set the initial collection interval for the transportation data of goods to be transported. Adjust the initial collection interval for the transportation data of goods to be transported based on the transportation timeout rate data of goods to be transported until the transportation timeout rate data of goods to be transported is less than the preset threshold, and obtain the final collection interval for the transportation data of goods to be transported. Specifically, this includes: setting the data collection period for the transport of goods awaiting shipment and the initial data collection interval; using the S1 method, repeatedly collecting and filling data in the logistics transportation map of goods awaiting shipment according to the initial data collection interval within the data collection period to obtain an initial real-time logistics transportation map matrix for goods awaiting shipment; setting a timeout rate threshold for goods transportation; inputting each real-time logistics transportation map of goods awaiting shipment in the real-time logistics transportation map matrix into the final spatiotemporal graph convolutional network model for goods transportation to perform optimal path mapping, obtaining an optimal transportation path matrix for real-time goods awaiting shipment; and adjusting the initial data collection interval for goods transportation according to the timeout rate threshold and the optimal transportation path matrix for real-time goods transportation. S4. Based on the final collection interval of the goods transportation data to be transported, the goods transportation data to be transported are collected in real time and the optimal route is output in real time in combination with the final spatiotemporal graph convolutional network model of goods transportation, so as to carry out logistics transportation of the goods to be transported. S5. In the logistics transportation of goods to be shipped, an initial jump value for the propagation level of an emergency is set; then, by screening backup nodes and combining the recovery time and timeout ratio, the initial jump value for the propagation level of an emergency is iteratively optimized.

2. The efficient logistics transportation method based on an improved graph neural network according to claim 1, characterized in that, S2 includes the following steps: S21. Set several types of goods that were transported under normal historical conditions and the boundaries of the logistics surrounding areas to obtain a historical transported goods set; based on the logistics transport map of the goods to be transported and the boundaries of the logistics surrounding areas, obtain the logistics transport map and the optimal transport path data of each item in the historical transported goods set to obtain a historical goods logistics transport map dataset and a historical goods logistics transport optimal path dataset. S22. Construct an initial spatiotemporal graph convolutional network model for commodity transportation; S23. The initial spatiotemporal graph convolutional network model for commodity transportation is trained and tested using the historical commodity logistics transportation graph dataset and the historical commodity logistics transportation optimal path dataset; after training and testing, the final spatiotemporal graph convolutional network model for commodity transportation is obtained.

3. The efficient logistics transportation method based on an improved graph neural network according to claim 2, characterized in that, The initial data collection interval for the goods transportation data is adjusted based on the aforementioned goods transportation timeout rate threshold and the real-time optimal transportation path matrix for goods awaiting shipment, including the following steps: S331. Transport the corresponding goods according to each row of data in the real-time optimal transportation path matrix for goods to be transported; after all goods have been transported, calculate the transportation timeout rate data for goods to be transported. S332. When the timeout rate data of the goods to be shipped is greater than or equal to the timeout rate threshold, the initial collection interval of the goods to be shipped data is adjusted until the timeout rate of the goods to be shipped is less than the timeout rate threshold, and the final collection interval of the goods to be shipped data is obtained.

4. The efficient logistics transportation method based on an improved graph neural network according to claim 3, characterized in that: In S332, the Dragonfly optimization algorithm is used to adjust the initial collection interval of the transportation data of the goods to be transported.

5. The efficient logistics transportation method based on an improved graph neural network according to claim 3, characterized in that, S4 includes the following steps: S41. Set mutation thresholds for corresponding features in the set of feature types of shipping warehouses, the set of feature types of commodity distribution centers, and the set of feature types of transportation vehicles to obtain the mutation threshold sets for the features of shipping warehouses, commodity distribution centers, and transportation vehicles. S42. The data in the logistics transportation map of the goods to be transported is collected in real time using the final collection interval of the transportation data of the goods to be transported, and the optimal route is output in real time by combining it with the convolutional network model of the final spatiotemporal graph of the goods transportation; the optimal route is used to carry out logistics transportation of the goods to be transported. Specifically, when the absolute value of the difference between the data in the logistics transportation map of goods to be transported collected at adjacent intervals is greater than or equal to the data corresponding to the threshold sets of the characteristics of the shipping warehouse, the characteristics of the goods distribution center, and the characteristics of the transportation vehicle, it is recorded as a sudden event in the transportation of goods.

6. The efficient logistics transportation method based on an improved graph neural network according to claim 5, characterized in that, S5 includes the following steps: S51. Set the initial jump value for the propagation level of an emergency, denoted as... N The scope of impact of the aforementioned unforeseen event in commodity transportation is defined as starting from the unforeseen event itself. N Skip the propagation level; when the aforementioned unforeseen event in goods transportation occurs, set the propagation level starting from the unforeseen event in goods transportation. N Nodes in the logistics transportation graph of goods that are of the same type as sudden events in goods transportation and are outside the scope of the propagation level are recorded as backup nodes; the time taken to remedy sudden events in goods transportation using the backup nodes and the timeout time of the final goods transportation are recorded multiple times to obtain the data set of remedy time and the data set of final goods transportation time. S52. Set a threshold for the time consumption of pending recovery and a threshold for the proportion of timeout in the final pending goods logistics transportation. When the proportion of data in the pending recovery time dataset that is greater than or equal to the threshold for the time consumption of pending recovery is greater than or equal to 50%, or the proportion of timeout data in the final pending goods logistics transportation timeout dataset is greater than or equal to the threshold for the proportion of timeout in the final pending goods logistics transportation timeout, adjust the initial jump value of the emergency propagation level until the proportion of data in the pending recovery time dataset that is greater than or equal to the threshold for the time consumption of pending recovery is less than 50% and the proportion of timeout data in the final pending goods logistics transportation timeout dataset is less than the threshold for the proportion of timeout in the final pending goods logistics transportation timeout, and obtain the final jump value of the emergency propagation level.

7. The efficient logistics transportation method based on an improved graph neural network according to claim 6, characterized in that, The adjustment of the initial jump value of the sudden event propagation level in S52 includes the following steps: S521. Set the value range of the initial jump value of the propagation level of the sudden event and generate the propagation level jump number to adjust the initial position of each dragonfly in the dragonfly population to obtain the second initial position set. S522. Construct the fitness function for adjusting the hop count of the propagation hierarchy for the dragonfly population; S523. Start the iteration; in each iteration, use the fitness function of the propagation level hop count to adjust the fitness value of the position of each dragonfly in the propagation level hop count adjusted dragonfly population obtained in the previous iteration to calculate the fitness value of the position of each dragonfly in the propagation level hop count adjusted dragonfly population obtained in the previous iteration and update the position of each dragonfly in the propagation level hop count adjusted dragonfly population obtained in the previous iteration. S524. When the maximum number of iterations is reached, stop the iteration and obtain the second final global optimal position; otherwise, continue the iteration until the maximum number of iterations is reached; after the iteration is completed, if the proportion of data in the pending recovery time dataset that is greater than or equal to the pending recovery time threshold is less than 50% and the proportion of timeout data in the final pending goods logistics transportation timeout dataset is less than the final pending goods logistics transportation timeout proportion threshold, the second final global optimal position is used as the final jump value of the sudden event propagation level; otherwise, return to S523 to continue the iteration.

8. A system for implementing the efficient logistics transportation method based on an improved graph neural network as described in any one of claims 1-7.

Images