Logistics path planning method and device based on deep learning, equipment and medium

Abstract

The invention relates to a logistics path planning method and device based on deep learning, equipment and a medium. The method comprises the following steps: obtaining logistics demand expected topological information and logistics demand expected verification information; verifying the logistics demand expected topological information based on the logistics demand expected verification information to obtain logistics demand real-time topological information; carrying out logistics demand planning on the logistics demand real-time topology information, and generating logistics vehicle scheduling information of each logistics node; screening available path information of each logistics vehicle based on the real-time departure node information, the real-time target node information and the logistics label information, and obtaining logistics road section expected information of an available path corresponding to the available path information; and inputting the logistics road section expected information and the logistics label information into the logistics path planning deep learning model, carrying out logistics path planning, and generating a logistics path planning result. By adopting the method, the accuracy, the reliability, the adaptability and the intelligent degree of logistics path planning can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of logistics path planning, and in particular relates to a logistics path planning method, device, equipment and medium based on deep learning. Background Art

[0002] With the rapid development of the logistics industry, logistics route planning has become a critical factor affecting logistics efficiency and costs. Traditional logistics route planning methods primarily rely on manual experience or simple heuristic algorithms. Manual experience-based planning struggles with large-scale, complex data, while simple heuristic algorithms are prone to getting stuck in local optima and are unable to adapt to the dynamic changes in logistics operations. In recent years, the rise of deep learning technology has brought new breakthroughs in logistics route planning. Deep learning builds multi-layer neural networks to automatically learn complex patterns and deep features in data, demonstrating powerful nonlinear modeling capabilities. Currently, logistics route planning methods based on deep learning have become a hot topic of research.

[0003] However, deep learning-based logistics route planning methods still face several challenges. Deep learning model training requires large amounts of high-quality data, but logistics data often contains noise and missing values. Effectively managing this data to improve quality is crucial. Furthermore, deep learning models lack interpretability, making it difficult to intuitively understand the decision-making process and rationale. Furthermore, when using deep learning models for logistics route planning, there is a risk of delays in real-time data acquisition, which can result in route planning results that fail to reflect actual road conditions. Furthermore, deep learning models' reliance on historical data can hinder their ability to quickly adapt to environmental changes when responding to emergencies, making them less adaptable to sudden events. Summary of the Invention

[0004] Based on this, it is necessary to provide a logistics path planning method, device, equipment and medium based on deep learning that can structuredly manage basic data, intelligent dynamic path screening and refined demand matching to address the above technical problems.

[0005] In a first aspect, the present application provides a logistics path planning method based on deep learning, comprising:

[0006] Obtaining expected logistics demand topology information and expected logistics demand verification information;

[0007] Verify the expected logistics demand topology information based on the expected logistics demand verification information to obtain the real-time logistics demand topology information. The real-time logistics demand topology information includes the real-time departure node information and real-time destination node information of each cargo.

[0008] Perform logistics demand planning based on real-time logistics demand topology information and generate logistics vehicle scheduling information for each logistics node. The logistics vehicle scheduling information includes logistics tag information for each logistics vehicle. The logistics tag information is used to represent the carrying capacity of the logistics vehicle and the transportation requirements of the goods carried by the logistics vehicle;

[0009] Filter the available path information of each logistics vehicle based on the real-time departure node information, real-time destination node information and logistics tag information, and obtain the expected logistics section information of the available path corresponding to the available path information;

[0010] The expected information of logistics sections and logistics label information are input into the logistics path planning deep learning model to perform logistics path planning and generate logistics path planning results.

[0011] In a second aspect, the present application also provides a logistics path planning device based on deep learning, comprising:

[0012] Logistics demand data acquisition module, used to obtain logistics demand expected topology information and logistics demand expected verification information;

[0013] The logistics demand data verification module is used to verify the expected logistics demand topology information based on the expected logistics demand verification information to obtain the real-time logistics demand topology information. The real-time logistics demand topology information includes the real-time departure node information and real-time destination node information of each cargo;

[0014] The logistics demand planning and scheduling module is used to plan logistics demand based on real-time logistics demand topology information and generate logistics vehicle scheduling information for each logistics node. The logistics vehicle scheduling information includes the logistics tag information of each logistics vehicle, which is used to represent the carrying capacity of the logistics vehicle and the transportation requirements of the goods carried by the logistics vehicle;

[0015] The available path information screening module is used to screen the available path information of each logistics vehicle based on the real-time departure node information, real-time destination node information and logistics tag information, and obtain the expected logistics section information of the available path corresponding to the available path information;

[0016] The logistics path planning calculation module is used to input the expected information of logistics sections and logistics label information into the logistics path planning deep learning model to perform logistics path planning and generate logistics path planning results.

[0017] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any method of the first aspect of the present application are implemented.

[0018] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any method of the first aspect of the present application.

[0019] The above-described deep learning-based logistics path planning method, apparatus, device, and medium, by acquiring expected logistics demand topology information and combining it with verification information for real-time verification, can ensure that logistics path node information is dynamically synchronized with actual business needs, avoiding path deviations caused by data lag in traditional planning, and improving the adaptability of path planning to real-time order demands, site status changes, and even e-commerce promotions. By generating logistics vehicle scheduling information and logistics tag information based on real-time topology information, the vehicle carrying capacity is deeply linked to cargo carrying requirements, achieving precise matching of vehicles and cargo, reducing resource waste caused by mismatches between vehicle models and cargo demand, and improving the consistency of timeliness and quality requirements for cargo within the same logistics vehicle, thereby improving the efficiency of logistics resource utilization and the effectiveness of logistics path planning. By combining real-time node information and logistics tag information to screen available paths, it can consider both vehicle physical constraints and cargo timeliness requirements, filter out inefficient paths, and thus ensure the feasibility and reliability of the planned path. Through the deep learning model, it can automatically learn the optimal path pattern from historical distribution data, dynamically balance complex constraints such as distance, timeliness, and cost, and generate planning results that balance efficiency and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 A flowchart of a logistics path planning method based on deep learning provided in one embodiment of the present application;

[0022] Figure 2 A schematic diagram of a process for generating logistics vehicle scheduling information provided in one embodiment of the present application;

[0023] Figure 3 A flowchart of another deep learning-based logistics path planning method provided in one embodiment of the present application;

[0024] Figure 4 A schematic diagram of a process for generating expected road speed information provided by one embodiment of the present application;

[0025] Figure 5A schematic diagram of a process for generating expected verification information of logistics demand provided in one embodiment of the present application;

[0026] Figure 6 A schematic diagram of a process for generating logistics demand coding information for goods provided in one embodiment of the present application;

[0027] Figure 7 A schematic structural diagram of a deep learning-based logistics path planning device provided for one embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0029] In an exemplary embodiment of the present application, Figure 1 As shown, a logistics path planning method based on deep learning is provided. This embodiment uses the method applied to a server as an example. It is understandable that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0030] Step S101: Obtain expected logistics demand topology information and expected logistics demand verification information.

[0031] Specifically, the server can generate expected logistics demand topology information based on the preset logistics site connection topology information and the historical logistics demand information of each logistics site. This can improve the stability of the expected logistics demand topology information through the highly stable structured logistics site connection topology information and statistical information, effectively reducing the impact of missing values ​​and errors in individual goods on logistics planning. The server can also generate expected logistics demand verification information based on the logistics node path topology information and transportation requirements of newly added goods according to the preset cargo demand analysis cycle.

[0032] Illustratively, expected logistics demand topology information can be used to represent the expected logistics demand topology information obtained through analysis of historical logistics demand topology information. The server can pre-allocate logistics resources at each logistics site based on the expected logistics demand topology information. Expected logistics demand verification information can be used to represent the actual logistics demand topology information corresponding to the expected logistics demand topology information. Based on the pre-allocation of logistics resources at each logistics site based on the expected logistics demand topology information, the server can further calibrate the scheduling plan for each logistics site's logistics resources and achieve precise detailed allocation of logistics resources at each logistics site. This detailed allocation can include, but is not limited to, the allocation of cargo to logistics vehicles.

[0033] Step S102 : verifying expected logistics demand topology information based on expected logistics demand verification information to obtain real-time logistics demand topology information.

[0034] Optionally, the server may, but is not limited to, perform a verification analysis on the expected logistics demand topology information based on the expected logistics demand verification information, including an error verification analysis of the total amount of cargo to be shipped at the starting station, an error verification analysis of the total amount of cargo to be shipped at the target station, and an error verification analysis of the cargo carrying requirements, to obtain a verification analysis result of the error verification analysis of the total amount of cargo to be shipped at the starting station, an error verification analysis result of the total amount of cargo to be shipped at the target station, and an error verification analysis result of the cargo carrying requirements. The server may update and correct the expected logistics demand topology information based on the verification analysis result of the error verification analysis of the total amount of cargo to be shipped at the starting station, the error verification analysis result of the total amount of cargo to be shipped at the target station, and the error verification analysis result of the cargo carrying requirements, to obtain real-time logistics demand topology information.

[0035] Furthermore, the error verification analysis results may include, but are not limited to, acceptable error levels, error levels that require adjustment, and severe error levels. If the error verification analysis results for the total amount of cargo to be transported at the starting site, the error verification analysis results for the total amount of cargo to be transported at the target site, and the error verification analysis results for the transportation requirements of the cargo to be transported are all at acceptable error levels, the server may update the transportation requirements of each actual cargo to be transported based on the expected topology information of the logistics demand, and obtain real-time topology information of the logistics demand. If the error verification analysis results for the total amount of cargo to be transported at the starting site, the error verification analysis results for the total amount of cargo to be transported at the target site, and the error verification analysis results for the transportation requirements of the cargo to be transported contain error verification results that require adjustment, the server may update the portion of the error verification analysis results in the expected topology information of the logistics demand that is at the error level that requires adjustment based on the expected verification information of the logistics demand, and on this basis, update the transportation requirements of each actual cargo to be transported, and obtain real-time topology information of the logistics demand. If there are error verification analysis results of severe error level in the error verification analysis results of the total amount of cargo to be transported at the starting site, the error verification analysis results of the total amount of cargo to be transported at the target site, and the error verification analysis results of the transportation requirements of cargo to be transported, the server can generate a severe error result report and regenerate the real-time topology information of the logistics demand based on the expected verification information of the logistics demand and the logistics node connection topology information.

[0036] Schematically, the real-time topology information of logistics demand includes the real-time departure node information and real-time destination node information of each cargo.

[0037] Step S103 , performing logistics demand planning on the real-time topology information of logistics demand, and generating logistics vehicle scheduling information for each logistics node.

[0038] Specifically, the server can analyze the cargo distribution status of each logistics node based on the real-time topological information of logistics demand, and combine the carrying capacity of logistics vehicles preliminarily allocated based on the expected topological information of logistics demand, use preset matching rules to plan logistics demand, and generate logistics vehicle scheduling information for each logistics node.

[0039] Schematically, the logistics vehicle dispatch information includes the logistics label information of each logistics vehicle, and the logistics label information is used to characterize the carrying capacity of the logistics vehicle and the carrying requirements of the goods carried by the logistics vehicle.

[0040] Optionally, the server may assign cargo to be transported that has the same real-time departure node information and real-time destination node information to the same logistics vehicle, and set the same real-time departure node information and real-time destination node information of the cargo to be transported as the real-time departure node information and real-time destination node information of the same logistics vehicle. The server may generate logistics label information for the same logistics vehicle by combining the transportation requirements of the cargo to be transported in the same logistics vehicle and the real-time physical properties of the same logistics vehicle.

[0041] Step S104, filtering the available path information of each logistics vehicle based on the real-time departure node information, the real-time target node information and the logistics tag information, and obtaining the expected logistics section information of the available path corresponding to the available path information.

[0042] Specifically, the server can filter the available path information of each logistics vehicle based on the real-time departure node information, real-time target node information and logistics label information of each logistics vehicle, and obtain the expected information of the logistics section of the available path corresponding to the available path information.

[0043] Optionally, the expected logistics route information may include, but is not limited to, long-term logistics route information and short-term logistics route information. The long-term logistics route information may include, but is not limited to, holiday logistics route information. The short-term logistics route information may include, but is not limited to, accident logistics route information, daily peak logistics route information, weekly peak logistics route information, construction logistics route information, and weather logistics route information.

[0044] Step S105: Input the expected logistics route information and logistics label information into the logistics path planning deep learning model to perform logistics path planning and generate logistics path planning results.

[0045] Optionally, the logistics path planning deep learning model can be constructed based on, but not limited to, one or more of the following models, including the Asynchronous Advantage Actor-Critic (A3C), Deep Q Network (DQN), Proximal Policy Optimization (PPO), Multilayer Perceptron (MLP), Pointer Networks (Ptr-Net), Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), Transformer, and Graph Neural Network (GNN), or their improved models.

[0046] Illustratively, the logistics path planning result may include an initial logistics path result from real-time departure node information to real-time destination node information, and may also include a real-time logistics path result obtained from real-time updated expected logistics route information. The real-time logistics path result can be obtained by a server inputting the real-time updated expected logistics route information into a logistics path planning deep learning model in combination with relatively stable logistics label information generated based on logistics demand planning to perform logistics path planning.

[0047] In the above-mentioned logistics path planning method based on deep learning, by comprehensively analyzing the expected topological information of logistics demand and the expected verification information of logistics demand, it is possible to quickly update the logistics demand information, ensure that the logistics path planning meets the latest actual logistics demand, and thus solve the transportation delays and resource waste problems caused by inaccurate or untimely information in traditional logistics path planning; by updating logistics demand information in real time, it can provide reliable data support for logistics path planning and logistics demand planning, thereby improving the accuracy and timeliness of logistics path planning; by generating logistics vehicle scheduling information for each logistics node, it can ensure that logistics vehicles transport goods according to the optimal loading content, sequence and time, and thus solve the problems of low transportation efficiency and excessive cost caused by mismatch between vehicles and goods or unreasonable vehicle scheduling in traditional logistics scheduling; through precise vehicle scheduling, the vehicle's empty driving rate and waiting time can be reduced, and the vehicle utilization rate can be improved, thereby reducing operating costs and improving the overall efficiency of logistics transportation.

[0048] Furthermore, in the above-mentioned logistics path planning method based on deep learning, the deep learning model comprehensively analyzes the expected information of the logistics section and the logistics label information, which can realize the intelligent planning of the logistics path, so that the logistics path planning results can be generated according to factors such as real-time road conditions, vehicle carrying capacity and cargo requirements, thereby solving the problems of unstable transportation time and large fluctuations in transportation costs caused by unreasonable path selection in traditional logistics path planning; through the intelligent planning of the deep learning model, the efficiency and economic benefits of logistics transportation can be improved, ensuring that the goods are delivered to the destination on time and safely; through the multi-source data fusion capability of the deep learning model, comprehensive monitoring and intelligent decision-making of the logistics process can be realized, risks can be predicted in advance and measures can be taken, thereby improving the quality of logistics services and customer satisfaction.

[0049] In an optional embodiment of the present application, the real-time topology information of logistics demand may further include global departure node information, global destination node information, first local timeliness demand information and global timeliness demand information, and the logistics tag information includes second local timeliness demand information, such as Figure 2 As shown, logistics demand planning is performed on the logistics demand information to be planned, and logistics vehicle scheduling information for each logistics node is generated, including:

[0050] Step S201 , calculate and allocate the first local time efficiency demand information of each cargo based on the real-time departure node information, real-time destination node information, global departure node information, global destination node information and global time efficiency demand information of each cargo in the real-time logistics demand topology information.

[0051] Optionally, the server may assign a logistics transportation time ratio coefficient to each cargo between the logistics nodes it needs to pass through based on the loading and unloading time and logistics diversion time between the logistics nodes, as well as the historical average logistics transportation time between the logistics nodes. The server may calculate and assign the first local time efficiency requirement information for each cargo based on the logistics transportation time ratio coefficient and the global time efficiency requirement information.

[0052] Step S202 : performing time-efficiency cluster analysis on the goods based on the local time-efficiency demand information to generate local time-efficiency cluster data of the goods.

[0053] Step S203: obtain the dispatchable logistics vehicle information, match the dispatchable logistics vehicle information with the local time-efficiency clustering data of the goods, and calculate the second local time-efficiency demand information of the logistics vehicle and the deviation value of the second local time-efficiency demand information.

[0054] Illustratively, the second local time requirement information deviation value may be used to represent the degree of dispersion of the first local time requirement information of each cargo in the logistics vehicle.

[0055] Step S204 , performing logistics demand planning with the second local time-sensitive demand information deviation value and the vehicle carrying space vacancy rate as a loss function to generate logistics vehicle scheduling information.

[0056] Illustratively, the server may, but is not limited to, perform logistics demand planning based on the second local time-sensitive demand information deviation value and the vehicle carrying space vacancy rate based on conventional dynamic programming algorithms and deep learning models to generate logistics vehicle scheduling information. When the server uses the logistics demand planning deep learning model to perform logistics demand planning based on the second local time-sensitive demand information deviation value and the vehicle carrying space vacancy rate to generate logistics vehicle scheduling information, since the input, output, and planning objectives of the logistics demand planning deep learning model and the logistics path planning deep learning model are different, the logistics demand planning deep learning model and the logistics path planning deep learning model may be different.

[0057] The above-mentioned deep learning-based logistics path planning method can solve the problem of low transportation efficiency caused by differences in cargo time requirements through the refined description and classification of cargo time requirements, thereby achieving more accurate cargo time management. By calculating and allocating the first local time requirement information of each cargo, it can achieve a reasonable allocation of cargo time requirements, avoid cargo transportation delays, and improve the timeliness and reliability of cargo transportation. By classifying cargo with similar time requirements, it can improve the efficiency of cargo scheduling and the accuracy of logistics vehicle scheduling. By calculating the second local time requirement information of logistics vehicles and the deviation value of the second local time requirement information, it can achieve accurate matching and evaluation of the time requirements of cargo carried by logistics vehicles, thereby improving the time satisfaction rate and vehicle utilization rate of cargo transportation. By using the deviation value of the second local time requirement information and the vehicle carrying space vacancy rate as the loss function for logistics demand planning, it can achieve optimized scheduling that comprehensively considers the matching degree of cargo time requirements and vehicle carrying space utilization, thereby improving the overall operational efficiency and economic benefits of the logistics system.

[0058] In an optional embodiment of the present application, the expected logistics route information includes expected route speed information. Figure 3 , logistics path planning methods based on deep learning also include:

[0059] Step S306 : Calculate a time efficiency reward function value based on the second local time efficiency demand information and the expected road speed information in combination with the time efficiency reward function.

[0060] Step S307 : Calculating an economic loss function value based on the vehicle fuel consumption information, path tolls, and vehicle loss information of the available paths in combination with the economic loss function.

[0061] Step S308: The time efficiency reward function value and the economic loss function value are integrated to update and optimize the logistics path planning deep learning model.

[0062] Among them, the expressions of time reward function and economic loss function can be:

[0063]

[0064] Where R T is the time-effective reward function, κ and τ are the proportional coefficient and exponential term coefficient of the time-effective reward function respectively, is the second local time requirement information, T E is the expected speed information of the available path, N is the total number of goods in the logistics vehicle, α is the exponential coefficient of the second local time demand information, is the first local time requirement information of the i-th item, is the average first local time efficiency demand information of all goods in the logistics vehicle, LC is the economic loss function, M is the total number of available path segments, is the fuel consumption base of the j-th road section, P F is the fuel consumption coefficient of logistics vehicles, C j is the toll of the j-th road section, is the vehicle loss base of the j-th road section, P W is the vehicle loss coefficient of logistics vehicles.

[0065] For example, the toll C of the jth road segment j It may include, but is not limited to, the highway toll for the jth road section.

[0066] Schematically, the expected speed information T of the available path section E The smaller the value of T The larger the value can be, the more the expected speed information T of the available path segment is. E The value is greater than the second local time requirement information When the value of T The value can be negative.

[0067] Optionally, it can be based on the fuel consumption coefficient P of the logistics vehicle F and vehicle loss coefficient P W The average value and the toll cost C of the road section j , fuel consumption base and vehicle loss base The average value of , sets the proportional coefficient κ of the time-effective reward function.

[0068] Optionally, it can be based on the second local time requirement information and the second local time requirement information The average logistics transportation time of the corresponding logistics node Set the time-sensitive reward function R T The exponential term coefficient τ, the second local time-sensitive demand information Compared with the average logistics transportation time The smaller the ratio, the larger the exponential term coefficient τ can be.

[0069] In the above-mentioned logistics path planning method based on deep learning, by introducing the timeliness reward function to obtain the timeliness reward function value, it is possible to realize the quantitative evaluation of the timeliness of logistics transportation, improve the ability to accurately control the timeliness of logistics transportation, and realize the refined management of the timeliness of cargo transportation, thereby improving the timeliness and reliability of logistics services; by combining the vehicle fuel consumption information, path tolls and vehicle loss information based on the available paths with the economic loss function to obtain the economic loss function value, it is possible to realize a comprehensive quantitative analysis of logistics transportation costs, thereby improving the accuracy of logistics transportation cost accounting, and helping to reduce logistics operating costs; by updating and optimizing the logistics path planning deep learning model through the comprehensive timeliness reward function value and the economic loss function value, the logistics path planning deep learning model can continuously learn and adapt to the complex logistics environment, thereby further improving the scientificity and rationality of logistics path planning, thereby improving the overall efficiency of the logistics system.

[0070] In an optional embodiment of the present application, the logistics tag information also includes physical attribute information of the logistics vehicle and cargo bumping restriction information. The available path information of each logistics vehicle is filtered based on the real-time departure node information, real-time destination node information, and logistics tag information, which may include:

[0071] Obtain a historical maximum feasible road set between a departure node corresponding to the real-time departure node information and a target node corresponding to the real-time target node information.

[0072] Obtain the road section restriction information, road section bump information and road section time efficiency information of the roads in the historical most extensive feasible road set. The road section time efficiency information is used to characterize the efficiency of logistics vehicles when passing through the road.

[0073] Based on the physical property information of logistics vehicles, cargo bumpy restriction information and the second local time requirement information, the road section restriction information, road section bumpy information and road section time requirement information are matched and verified respectively, and the available path information of each logistics vehicle is screened from the historically widest feasible road set.

[0074] In the above-mentioned logistics path planning method based on deep learning, the physical property information of logistics vehicles and the bumpy restriction information of cargo can be used to achieve a more comprehensive and accurate matching of logistics vehicles and cargo, thereby improving the safety of logistics transportation and the integrity of cargo; by obtaining the historical most extensive feasible road set between the departure node and the target node and its related road section information, it is possible to deeply mine and utilize road information, thereby providing logistics vehicles with richer path choices and improving the flexibility and reliability of path planning; by matching and verifying the road section restriction information, road section bumpy information and road section timeliness information respectively, it is possible to conduct a refined evaluation of road suitability, thereby optimizing the path selection of logistics vehicles and further improving the efficiency of logistics transportation and the stability of cargo delivery.

[0075] In an optional embodiment of the present application, Figure 4 As shown, obtaining the expected logistics section information of the available path corresponding to the available path information may include:

[0076] Step S401 : predicting holiday long-term impact information of sections of available paths based on holiday time series information.

[0077] Step S402: Acquire weather information, accident information, and construction information of available routes.

[0078] Step S403 : generating short-term weather impact information, short-term accident impact information and short-term construction impact information of the available road sections respectively based on the weather information, accident information and construction information.

[0079] Step S404, obtain basic speed information of available paths, and generate expected speed information of sections by combining long-term holiday impact information, short-term weather impact information, short-term accident impact information and short-term construction impact information.

[0080] Optionally, the server can obtain the basic speed information of the available paths, and superimpose the long-term impact information of holidays on the road sections, the short-term impact information of weather on the road sections, the short-term impact information of accidents on the road sections, and the short-term impact information of construction on the road sections on the basis of the obtained basic speed information of the available paths to obtain the expected speed information of the road sections.

[0081] Furthermore, the server can use the acquired basic speed information for available routes, as well as information about the long-term impact of holidays, short-term weather conditions, short-term accidents, and short-term construction as inputs to a neural network to obtain expected speed information for the road section. The neural network architecture is not limited here.

[0082] In the above-mentioned logistics route planning method based on deep learning, the long-term impact information of road sections during holidays and the short-term impact information of road sections due to weather, accidents and construction can be used to achieve a multi-dimensional dynamic assessment of road conditions, thereby solving the problem of incomplete consideration of road dynamic factors in traditional route planning; by combining basic speed information and multi-period impact information of each road section to generate expected speed information of the road section, it is possible to accurately predict the road traffic efficiency, thereby reducing the deviation in transportation timeliness caused by changes in road conditions, thereby improving the timeliness and reliability of logistics transportation.

[0083] In an optional embodiment of the present application, Figure 5 As shown, obtaining expected logistics demand topology information and expected logistics demand verification information may include:

[0084] Step S501: Obtain logistics site connection topology information and logistics demand coding information of newly added goods.

[0085] Step S502: construct a logistics demand prediction model based on historical logistics demand coding information and historical e-commerce activity time series information.

[0086] Step S503: Input the expected e-commerce activity timing information into the logistics demand forecasting model to generate expected logistics demand information, and combine the expected logistics demand information with the logistics site connection topology information to obtain the expected logistics demand topology information.

[0087] Step S504: Generate logistics demand expectation verification information based on the real-time logistics demand coding.

[0088] In the above-mentioned logistics route planning method based on deep learning, by obtaining the connection topology information of logistics sites and the logistics demand coding information of new goods, and constructing a logistics demand forecasting model, it is possible to achieve accurate prediction of logistics demand and construction of topological structure, providing more reliable data support for subsequent logistics route planning; by inputting the expected e-commerce activity timing information into the logistics demand forecasting model to generate expected logistics demand information, it is possible to achieve forward-looking prediction of logistics demand, thereby reducing the impact of logistics demand fluctuations in special periods such as e-commerce activities on route planning, and improving the ability of the logistics system to cope with sudden changes in demand; through technical means of generating logistics demand expectation verification information, it is possible to achieve real-time verification and correction of logistics demand forecasts, ensuring the accuracy and timeliness of logistics demand information, thereby further improving the accuracy and reliability of logistics route planning.

[0089] In an optional embodiment of the present application, Figure 6 As shown, obtain the logistics demand coding information of the newly added goods, including:

[0090] Step S601, obtaining the global departure node information, transit logistics node information, global destination node information and transportation requirement information of the goods.

[0091] Illustratively, the logistics node corresponding to the real-time departure node information can be the logistics node corresponding to the global departure node information or the logistics node corresponding to the transit logistics node information; the real-time destination node information can be the logistics node corresponding to the global destination node information or the logistics node corresponding to the transit logistics node information. The transit logistics node information can correspond to one or more logistics nodes. When there is only one logistics node corresponding to the transit logistics node information, only one of the logistics nodes corresponding to the real-time departure node information and the real-time destination node information can be the logistics node corresponding to the transit logistics node information.

[0092] Step S602: Set the starting logistics node code of the goods to the logistics site code of the logistics site corresponding to the global departure node information.

[0093] Step S603: Set the transit logistics node code of the goods to the logistics site code of the logistics site corresponding to the transit logistics node information.

[0094] Step S604: Set the end logistics node code of the goods to the logistics site code of the logistics site corresponding to the global target node information.

[0095] Step S605 , generating a logistics transportation requirement code based on the transportation requirement information according to a preset transportation requirement determination rule.

[0096] Schematically, the logistics transportation requirement code can be used to represent the transportation requirements of the goods.

[0097] Optionally, the transportation requirements may include, but are not limited to, integrity transportation requirements and timeliness transportation requirements.

[0098] Step S606: Summarize the starting logistics node code, transit logistics node code, ending logistics node code and logistics transportation requirement code to obtain the logistics requirement code information of the goods.

[0099] In the above-mentioned deep learning-based logistics path planning method, by encoding the starting, transit, and ending logistics node information of the goods into corresponding logistics site codes respectively, it is possible to achieve standardization and normalization of logistics node information, thereby improving the intelligent decision-making level and operational management efficiency of the logistics system; by generating logistics transportation requirement codes, it is possible to achieve standardized expression of cargo transportation requirements, ensuring that the carrying capacity of logistics vehicles matches the cargo transportation requirements, thereby improving the safety and reliability of cargo transportation and reducing transportation risks.

[0100] In an illustrative manner, the above-mentioned deep learning-based logistics path planning method generates logistics transportation requirement codes and separates logistics demand information from cargo content information and customer profile information, thereby improving the security of logistics data and enhancing the protection of the privacy security of logistics customers.

[0101] In an exemplary embodiment of the present application, Figure 3 As shown in the figure, a logistics path planning method based on deep learning is given, including:

[0102] Step S301: Obtain expected logistics demand topology information and expected logistics demand verification information.

[0103] Step S302 : verifying expected logistics demand topology information based on expected logistics demand verification information to obtain real-time logistics demand topology information.

[0104] Step S303: perform logistics demand planning on the real-time topology information of logistics demand, and generate logistics vehicle scheduling information for each logistics node.

[0105] Step S304: Filter the available path information of each logistics vehicle based on the real-time departure node information, the real-time target node information, and the logistics tag information, and obtain the expected logistics section information of the available path corresponding to the available path information.

[0106] Step S305: Input the expected logistics route information and logistics label information into the logistics path planning deep learning model to perform logistics path planning and generate logistics path planning results.

[0107] Step S306 : Calculate a time efficiency reward function value based on the second local time efficiency demand information and the expected road speed information in combination with the time efficiency reward function.

[0108] Step S307 : Calculating an economic loss function value based on the vehicle fuel consumption information, path tolls, and vehicle loss information of the available paths in combination with the economic loss function.

[0109] Step S308: The time efficiency reward function value and the economic loss function value are integrated to update and optimize the logistics path planning deep learning model.

[0110] In the above-mentioned deep learning-based logistics path planning method, by obtaining and verifying logistics demand information, it is possible to ensure the accuracy and timeliness of logistics demand data, laying a solid foundation for subsequent planning; through vehicle scheduling information, the distribution of vehicles and goods can be optimized; by using time-effectiveness rewards and economic loss functions to comprehensively evaluate and update the optimization model, the scientificity and rationality of path planning can be improved.

[0111] Specifically, the above-mentioned logistics path planning method based on deep learning can effectively improve the rationality of vehicle and cargo allocation, path selection adaptability and model dynamic adaptability in logistics path planning, thereby realizing the economical and reliable operation of logistics transportation.

[0112] It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0113] Based on the same inventive concept, the embodiments of the present application also provide a deep learning-based logistics path planning device for implementing the deep learning-based logistics path planning method mentioned above. The implementation solution provided by this device is similar to the implementation solution described in the above method. Therefore, the specific limitations of one or more embodiments of the deep learning-based logistics path planning device provided below can be found in the above limitations of the deep learning-based logistics path planning method and will not be repeated here.

[0114] In an exemplary embodiment, Figure 7 As shown, a logistics path planning device 700 based on deep learning is provided, including:

[0115] The logistics demand data acquisition module 701 can be used to obtain logistics demand expected topology information and logistics demand expected verification information.

[0116] The logistics demand data verification module 702 can be used to verify the expected logistics demand topology information based on the expected logistics demand verification information to obtain the real-time logistics demand topology information. The real-time logistics demand topology information includes the real-time departure node information and real-time destination node information of each cargo.

[0117] The logistics demand planning and scheduling module 703 can be used to perform logistics demand planning based on the real-time topological information of logistics demand, and generate logistics vehicle scheduling information for each logistics node. The logistics vehicle scheduling information includes the logistics label information of each logistics vehicle. The logistics label information is used to characterize the carrying capacity of the logistics vehicle and the carrying requirements of the goods carried by the logistics vehicle.

[0118] The available path information screening module 704 can be used to screen the available path information of each logistics vehicle based on real-time departure node information, real-time target node information and logistics tag information, and obtain the expected logistics section information of the available path corresponding to the available path information.

[0119] The logistics path planning calculation module 705 can be used to input the expected logistics route information and logistics label information into the logistics path planning deep learning model to perform logistics path planning and generate logistics path planning results.

[0120] In an optional embodiment of the present application, the logistics demand planning and scheduling module 703 may also be used to:

[0121] According to the real-time departure node information, real-time destination node information, global departure node information, global destination node information and global time requirement information of each cargo in the real-time topology information of logistics demand, the first local time requirement information of each cargo is calculated and allocated.

[0122] Based on the local timeliness demand information, the goods are clustered for timeliness analysis to generate local timeliness clustering data for the goods.

[0123] Obtain the dispatchable logistics vehicle information, match the dispatchable logistics vehicle information and the local time-efficiency clustering data of the goods, calculate the second local time-efficiency demand information of the logistics vehicle and the deviation value of the second local time-efficiency demand information, and the second local time-efficiency demand information deviation value is used to characterize the degree of discreteness of the first local time-efficiency demand information of each cargo in the logistics vehicle.

[0124] Logistics demand planning is performed using the second local time-sensitive demand information deviation value and the vehicle carrying space vacancy rate as the loss function to generate logistics vehicle scheduling information.

[0125] In an optional embodiment of the present application, the deep learning-based logistics path planning device 700 may also be used to:

[0126] Based on the second local time demand information and the expected road speed information combined with the time reward function, the time reward function value is obtained.

[0127] The economic loss function value is obtained based on the vehicle fuel consumption information, path toll and vehicle loss information of the available paths combined with the economic loss function.

[0128] The time-efficiency reward function value and the economic loss function value are integrated to update and optimize the logistics path planning deep learning model.

[0129] In an optional embodiment of the present application, the available path information screening module 704 may also be used to:

[0130] Obtain a historical maximum feasible road set between a departure node corresponding to the real-time departure node information and a target node corresponding to the real-time target node information.

[0131] Obtain the road section restriction information, road section bump information and road section time efficiency information of the roads in the historical largest feasible road set. The road section time efficiency information is used to characterize the efficiency of logistics vehicles when passing through the road.

[0132] Based on the physical property information of logistics vehicles, cargo bumpy restriction information and the second local time requirement information, the road section restriction information, road section bumpy information and road section time requirement information are matched and verified respectively, and the available path information of each logistics vehicle is screened from the historically widest feasible road set.

[0133] In an optional embodiment of the present application, the available path information screening module 704 may also be used to:

[0134] Predict the long-term holiday impact information of available routes based on holiday time series information.

[0135] Get weather information, accident information, and construction information for available routes.

[0136] According to the meteorological information, accident information and construction information, the short-term meteorological impact information, the short-term accident impact information and the short-term construction impact information of the available road sections are generated respectively.

[0137] Obtain basic speed information of available paths, and generate expected speed information for sections by combining long-term holiday impact information, short-term weather impact information, short-term accident impact information, and short-term construction impact information.

[0138] In an optional embodiment of the present application, the logistics demand data acquisition module 701 may also be used to:

[0139] Obtain the logistics site connection topology information and the logistics demand coding information of the newly added goods.

[0140] A logistics demand forecasting model is constructed based on historical logistics demand coding information and historical e-commerce activity time series information.

[0141] The expected e-commerce activity time series information is input into the logistics demand forecasting model to generate expected logistics demand information, and the expected logistics demand topology information is obtained by combining the expected logistics demand information with the logistics site connection topology information.

[0142] Generate logistics demand expectation verification information based on real-time logistics demand coding.

[0143] In an optional embodiment of the present application, the logistics demand data acquisition module 701 may also be used to:

[0144] Obtain the global departure node information, transit logistics node information, global destination node information and transportation requirement information of the goods.

[0145] The starting logistics node code of the goods is set to the logistics site code of the logistics site corresponding to the global departure node information.

[0146] The transit logistics node code of the goods is set to the logistics site code of the logistics site corresponding to the transit logistics node information.

[0147] The ending logistics node code of the goods is set to the logistics site code of the logistics site corresponding to the global target node information.

[0148] According to the preset transportation requirement determination rules, a logistics transportation requirement code is generated based on the transportation requirement information, and the logistics transportation requirement code is used to represent the transportation requirements of the goods.

[0149] The starting logistics node code, transit logistics node code, ending logistics node code and logistics transportation requirement code are summarized to obtain the logistics requirement code information of the goods.

[0150] In one embodiment, a computer device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the logistics path planning method based on deep learning as described above are implemented.

[0151] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0152] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0153] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A logistics path planning method based on deep learning, characterized in that: The method comprises: Obtaining expected logistics demand topology information and expected logistics demand verification information; Verify the expected logistics demand topology information based on the expected logistics demand verification information to obtain real-time logistics demand topology information, where the real-time logistics demand topology information includes real-time departure node information and real-time destination node information of each cargo; Performing logistics demand planning on the real-time logistics demand topology information to generate logistics vehicle scheduling information for each logistics node, wherein the logistics vehicle scheduling information includes logistics tag information of each logistics vehicle, and the logistics tag information is used to characterize the carrying capacity of the logistics vehicle and the carrying requirements of the goods carried by the logistics vehicle; Filtering available path information of each logistics vehicle based on the real-time departure node information, the real-time destination node information, and the logistics tag information, and obtaining expected logistics section information of the available path corresponding to the available path information; The expected information of the logistics route and the logistics label information are input into the logistics path planning deep learning model to perform logistics path planning and generate logistics path planning results.

2. The method according to claim 1, characterized in that The real-time logistics demand topology information also includes global departure node information, global destination node information, first local time-limit demand information, and global time-limit demand information. The logistics tag information includes second local time-limit demand information. The logistics demand planning is performed on the logistics demand to be planned information to generate logistics vehicle scheduling information for each logistics node, including: Calculate and allocate first local timeliness requirement information for each of the goods based on the real-time departure node information, the real-time destination node information, the global departure node information, the global destination node information, and the global timeliness requirement information of each of the goods in the real-time logistics demand topology information; Performing a time-efficiency cluster analysis on the goods based on the local time-efficiency demand information to generate local time-efficiency cluster data of the goods; Obtaining dispatchable logistics vehicle information, matching the dispatchable logistics vehicle information with the local time-limit clustering data of the goods, and calculating the second local time-limit demand information of the logistics vehicle and the second local time-limit demand information deviation value, where the second local time-limit demand information deviation value is used to represent the degree of dispersion of the first local time-limit demand information of each of the goods in the logistics vehicle; Logistics demand planning is performed using the second local time-sensitive demand information deviation value and the vehicle carrying space vacancy rate as a loss function to generate the logistics vehicle scheduling information.

3. The method according to claim 2, characterized in that The expected logistics section information includes expected section speed information, and the method further includes: Calculating a time efficiency reward function value based on the second local time efficiency demand information and the expected road speed information in combination with a time efficiency reward function; Calculating an economic loss function value based on the vehicle fuel consumption information, path tolls, and vehicle loss information of the available paths in combination with the economic loss function; Integrating the time-efficiency reward function value and the economic loss function value, updating and optimizing the logistics path planning deep learning model; The expressions of the time-effect reward function and the economic loss function are: Where R T is the time-effective reward function, κ and τ are the proportional coefficient and exponential term coefficient of the time-effective reward function respectively, is the second local time requirement information, T E is the expected speed information of the road section of the available path, N is the total number of the goods of the logistics vehicle, α is the exponential term coefficient of the second local time efficiency demand information, is the first local time requirement information of the i-th item of goods, is the average first local time requirement information of all the goods of the logistics vehicle, L C is the economic loss function, M is the total number of road segments of the available path, is the fuel consumption base of the j-th road section, P F is the fuel consumption coefficient of the logistics vehicle, C j is the toll for the section described in section j, is the vehicle loss base of the j-th section, P W is the vehicle loss coefficient of the logistics vehicle.

4. The method according to claim 2, characterized in that The logistics label information also includes physical attribute information of the logistics vehicle and cargo bumping restriction information. The screening of available path information for each logistics vehicle based on the real-time departure node information, the real-time destination node information, and the logistics label information includes: Obtaining a historically widest feasible road set between a departure node corresponding to the real-time departure node information and a target node corresponding to the real-time target node information; Obtaining road section restriction information, road section bump information, and road section time efficiency information of the roads in the historically widest feasible road set, wherein the road section time efficiency information is used to characterize the efficiency of the logistics vehicle when passing through the road; Based on the physical property information of the logistics vehicle, the cargo bumpy restriction information and the second local time requirement information, the road section restriction information, the road section bumpy information and the road section time requirement information are matched and verified respectively, and the available path information of each logistics vehicle is filtered from the historically widest feasible road set.

5. The method according to claim 1, wherein The obtaining of expected logistics route information of the available route corresponding to the available route information includes: Predicting holiday long-term impact information of the available path sections based on holiday time series information; Obtaining weather information, accident information, and construction information for the available path; Generating, based on the meteorological information, the accident information, and the construction information, short-term meteorological impact information, short-term accident impact information, and short-term construction impact information of the available road section, respectively; The basic speed information of the available path is obtained, and the expected speed information of the section is generated by combining the long-term impact information of holidays on the section, the short-term impact information of weather on the section, the short-term impact information of accidents on the section, and the short-term impact information of construction on the section.

6. The method according to any one of claims 2 to 5, characterized in that The obtaining of expected logistics demand topology information and expected logistics demand verification information includes: Obtaining logistics site connection topology information and newly added logistics demand coding information of the goods; Constructing a logistics demand forecasting model based on the historical logistics demand coding information and historical e-commerce activity time series information; Inputting the expected e-commerce activity time series information into the logistics demand forecasting model to generate the expected logistics demand information, and combining the expected logistics demand information with the logistics site connection topology information to obtain the expected logistics demand topology information; The logistics demand expectation verification information is generated based on the real-time logistics demand coding.

7. The method according to claim 6, characterized in that The acquiring of the newly added logistics demand coding information of the goods includes: Obtaining the global departure node information, transit logistics node information, global destination node information and transportation requirement information of the cargo; The starting logistics node code of the goods is set to the logistics site code of the logistics site corresponding to the global departure node information; The transit logistics node code of the goods is set to the logistics site code of the logistics site corresponding to the transit logistics node information; The end logistics node code of the goods is set to the logistics site code of the logistics site corresponding to the global target node information; generating a logistics transportation requirement code based on the transportation requirement information according to a preset transportation requirement determination rule, wherein the logistics transportation requirement code is used to represent the transportation requirement of the cargo; The starting logistics node code, the transit logistics node code, the ending logistics node code and the logistics transportation requirement code are summarized to obtain the logistics requirement coding information of the goods.

8. A logistics path planning device based on deep learning, characterized in that: The device comprises: Logistics demand data acquisition module, used to obtain logistics demand expected topology information and logistics demand expected verification information; A logistics demand data verification module is used to verify the expected logistics demand topology information based on the expected logistics demand verification information to obtain real-time logistics demand topology information, wherein the real-time logistics demand topology information includes real-time departure node information and real-time destination node information of each cargo; A logistics demand planning and scheduling module is used to perform logistics demand planning based on the real-time logistics demand topology information and generate logistics vehicle scheduling information for each logistics node. The logistics vehicle scheduling information includes logistics tag information of each logistics vehicle, and the logistics tag information is used to represent the carrying capacity of the logistics vehicle and the carrying requirements of the goods carried by the logistics vehicle; An available path information screening module is used to screen the available path information of each logistics vehicle based on the real-time departure node information, the real-time destination node information and the logistics tag information, and obtain the expected logistics section information of the available path corresponding to the available path information; The logistics path planning calculation module is used to input the expected information of the logistics section and the logistics label information into the logistics path planning deep learning model, perform logistics path planning, and generate logistics path planning results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images