A logistics transportation route optimization method and system based on digital twin

Abstract

The present invention relates to the field of logistics and transportation technology, and in particular to a logistics and transportation route optimization method and system based on digital twins. The method comprises the following steps: constructing a digital twin logistics base through multi-source heterogeneous data fusion and dynamic logistics network modeling; extracting regional logistics distribution characteristics and dynamic time series prediction based on the digital twin base to obtain predicted regional transportation demand; performing real-time status evaluation on the line, generating a dynamic line performance matrix, and allocating vehicle resources in combination with order task priority; performing multi-objective solution and dynamic adjustment on the transportation scheduling plan, and finally obtaining the optimal transportation scheduling plan; iteratively optimizing the digital twin model through real-time deviation analysis and feedback mechanism, and dynamically adjusting the transportation plan to form a closed-loop optimization process. The present invention significantly improves the efficiency of logistics and transportation, reduces costs, and enhances the resilience of the system through logistics and transportation route optimization technology.

Description

Technical Field

[0001] The present invention relates to the field of logistics and transportation technology, and in particular to a logistics and transportation route optimization method and system based on digital twins. Background Art

[0002] The role of logistics and transportation in economic and social development is becoming increasingly important. How to complete logistics and transportation tasks efficiently, at low cost and reliably has become a goal that logistics companies continue to pursue. Traditional logistics and transportation route optimization methods, such as those based on mathematical programming and heuristic algorithms, have improved efficiency to a certain extent. However, in the face of an increasingly dynamic and complex logistics environment, these methods have exposed obvious limitations.

[0003] The rise of digital twin technology has brought new ideas to solve these problems. Digital twin refers to building one or more virtual copies of physical entities (here refers to the logistics system), collecting data of physical entities in real time through sensors and other means, and simulating, monitoring, analyzing and predicting in virtual space. Applying digital twin technology to the optimization of logistics transportation routes can achieve real-time perception, intelligent decision-making and dynamic optimization of the logistics system.

[0004] However, the existing logistics transportation route optimization methods based on digital twins have the disadvantages of relying on static data, lacking real-time performance, and having difficulty in handling complex dynamic factors. The specific disadvantages are as follows:

[0005] Reliance on static data: Some methods rely mainly on static historical data when building digital twin models, such as road network structure, average vehicle speed, historical order data, etc. These static data cannot reflect the impact of dynamic factors such as real-time road conditions, weather changes, and emergencies, resulting in deviations between the optimization results and the actual situation, reducing the optimization effect. For example, path planning based on static traffic data cannot adjust the route in time when encountering real-time congested sections, resulting in delays and increased costs.

[0006] Lack of real-time performance: Although some methods can access real-time data, the data processing and model update speeds are slow and cannot achieve true real-time optimization. For example, when traffic conditions change, the system takes a certain amount of time to recalculate the optimal path, missing the best time for adjustment. In addition, some methods have high computational complexity and are difficult to handle real-time optimization problems in large-scale logistics networks.

[0007] Difficulty in handling complex dynamic factors: There are many complex dynamic factors in the logistics transportation process, such as the propagation effect of traffic congestion, the impact of weather changes on road conditions, the chain reaction of emergencies, etc. Some existing methods have difficulty in effectively modeling and analyzing the interactions between these complex dynamic factors, resulting in insufficient accuracy and reliability of optimization results. Summary of the invention

[0008] Based on this, it is necessary to provide a logistics transportation route optimization method and system based on digital twins to solve at least one of the above technical problems.

[0009] To achieve the above purpose, a logistics transportation route optimization method based on digital twins includes the following steps:

[0010] Step S1: Acquire multi-source logistics and transportation data, and perform dynamic logistics network modeling to obtain a dynamic logistics network model; perform three-dimensional modeling of transportation vehicles based on the dynamic logistics network model and multi-source logistics and transportation data, and perform vehicle capacity analysis to obtain a vehicle digital twin; initialize the logistics digital twin model based on the dynamic logistics network model and the vehicle digital twin to obtain a digital twin logistics base;

[0011] Step S2: Extract the logistics distribution characteristics of each region from the digital twin logistics base to obtain regional characteristic data; perform dynamic time series prediction modeling based on the regional transportation characteristic model to obtain the predicted regional transportation demand; generate dynamic transportation distribution based on the predicted regional transportation demand to obtain the dynamic transportation demand distribution;

[0012] Step S3: Modeling the initial state of the line according to the dynamic transportation demand distribution to obtain the initial state model of the line; performing a real-time line state evaluation based on the physical characteristics of the vehicle according to the initial state model of the line, and generating a dynamic line performance matrix to obtain a dynamic line performance matrix;

[0013] Step S4: According to the dynamic line performance matrix, vehicle resource allocation based on order task priority is performed to obtain a vehicle allocation plan; according to the vehicle allocation plan and the dynamic line performance matrix, a collaborative optimization algorithm is designed to obtain a collaborative optimization target; multi-objective solutions are performed on the collaborative optimization target to obtain a preliminary transportation scheduling plan; the preliminary transportation scheduling plan is input into the digital twin environment for dynamic scheduling adjustment, and the plan is globally integrated to obtain the optimal transportation scheduling plan;

[0014] Step S5: Perform transport execution deviation analysis based on the optimal transport scheduling plan to obtain a classification result of the cause of the deviation; perform iterative optimization of the digital twin model based on the classification result of the cause of the deviation, and adjust and optimize the transport plan for the optimal transport scheduling plan to obtain real-time operation deviation feedback data.

[0015] The present invention constructs a digital twin logistics base by integrating multi-source data and dynamic modeling, realizes the virtualization and real-time mapping of the logistics system, and lays a solid foundation for subsequent analysis, prediction and optimization. This effectively avoids the information lag and decision-making bias caused by relying on static data, enabling the system to more accurately reflect and respond to the dynamic changes in the real world. Through regional feature extraction, time series prediction modeling and dynamic correction, accurate prediction of future transportation demand is achieved, providing reliable data support for resource pre-configuration and scheduling optimization. This helps to improve resource utilization, reduce transportation costs, and enhance the ability to respond quickly to changes in market demand. Through real-time line status evaluation, a dynamic line performance matrix is ​​constructed, which comprehensively considers multiple factors such as timeliness, cost and risk, and provides a comprehensive evaluation basis for subsequent path optimization and scheduling decisions. This helps to select the best transportation route, reduce transportation risks, and strike a balance between efficiency and cost. Through priority-based vehicle resource allocation, collaborative optimization algorithm and digital twin simulation, an optimal transportation scheduling solution that takes into account efficiency, cost and risk is generated, achieving optimal resource allocation and improved scheduling efficiency. This helps to maximize the use of existing resources, reduce operating costs, and improve the on-time delivery rate of orders. Through real-time deviation analysis, cause classification and feedback mechanism, the iterative optimization of the digital twin model and the dynamic adjustment of the transportation plan are realized, forming a closed-loop optimization process, and continuously improving the prediction accuracy and adaptability of the system. This helps to enhance the system's ability to respond to emergencies and continuously improve logistics and transportation efficiency. Therefore, the present invention provides a logistics and transportation route optimization method based on digital twins, which effectively solves the shortcomings of the existing logistics and transportation route optimization method based on digital twins, which relies on static data, lacks real-time performance, and is difficult to handle complex dynamic factors, and realizes true real-time, dynamic, and adaptive optimization, thereby significantly improving logistics and transportation efficiency, reducing costs, and enhancing the resilience of the system.

[0016] Preferably, step S1 comprises the following steps:

[0017] Step S11: Acquire multi-source logistics and transportation data through a multi-source heterogeneous data fusion interface to obtain an initial data set;

[0018] Step S12: performing data cleaning and standardization processing on the initial data set to obtain standardized logistics data; constructing a vehicle three-dimensional model on the standardized logistics data to obtain a transport vehicle three-dimensional model;

[0019] Step S13: Perform dynamic logistics network modeling based on standardized logistics data to obtain a dynamic logistics network model; perform capacity parameter calculation based on the dynamic logistics network model and the three-dimensional model of the transport vehicle to obtain the transport vehicle capacity parameter;

[0020] Step S14: Generate a digital twin environment according to the dynamic logistics network model to obtain a digital twin initial environment; generate a vehicle digital twin according to the transport vehicle capacity parameters to obtain a vehicle digital twin;

[0021] Step S15: Perform real-time dynamic optimization and base generation based on the digital twin initial environment and the vehicle digital twin to obtain the digital twin flow base.

[0022] The present invention constructs a more comprehensive and refined initial data set by integrating multimodal and high-precision data from the vehicle's own sensors, roadside environmental perception equipment and infrastructure. This not only eliminates information islands, but also provides a richer and more reliable data foundation for subsequent refined modeling, high-precision simulation and intelligent decision-making, significantly improving the system's perception and prediction accuracy of logistics and transportation status. Advanced deep learning and geometric modeling technologies are used to intelligently process and reconstruct the original multimodal data with high precision, significantly improving the quality and information density of the data. By constructing a three-dimensional model of the vehicle itself and the surrounding environment and performing semantic understanding, it provides key data support for subsequent physical simulation, collision detection and environmental perception. The spatiotemporal synchronization and fusion of multi-sensor data ensure the consistency and relevance of the data, laying a solid foundation for subsequent analysis and decision-making. By integrating the vehicle's three-dimensional model and real-time status information, the model can simulate the vehicle's motion behavior and interaction with the environment more finely. The capacity parameters calculated based on the vehicle's three-dimensional model and the environment's three-dimensional reconstruction model, as well as the road condition parameters based on multispectral image analysis. By building a virtual digital twin environment, a platform is provided for the simulation and analysis of the logistics and transportation system. This allows various experiments and tests to be conducted in a virtual environment, such as simulating different transportation options and evaluating the effects of different strategies, thereby reducing risks and costs in actual operations and providing a basis for optimizing decisions. By integrating real-time optimization algorithms, the digital twin environment has dynamic optimization capabilities and can continuously optimize logistics and transportation plans based on real-time data. This enables the system to better adapt to dynamically changing environments and continuously improve the efficiency and performance of logistics and transportation, ultimately forming a digital twin logistics base that integrates real-time data, dynamic models, and optimization algorithms, providing strong support for subsequent steps.

[0023] Preferably, step S2 comprises the following steps:

[0024] Step S21: partitioning the digital twin logistics base into regional logistics data to obtain partitioned logistics data;

[0025] Step S22: performing intra-regional data clustering analysis on the partitioned logistics data to obtain regional logistics distribution;

[0026] Step S23: extracting regional features of regional logistics distribution to obtain regional feature data;

[0027] Step S24: performing regional transportation characteristic modeling according to the regional characteristic data to obtain a regional transportation characteristic model;

[0028] Step S25: Perform dynamic time series forecasting modeling according to the regional transportation characteristic model to obtain the predicted regional transportation demand;

[0029] Step S26: dynamically modify the predicted regional transportation demand to obtain a modified regional transportation demand;

[0030] Step S27: Generate dynamic transportation distribution according to the modified regional transportation demand to obtain dynamic transportation demand distribution.

[0031] The present invention helps to reduce the computational complexity and improve the data processing efficiency by dividing the huge logistics network into smaller and easier-to-manage regions. The partitioned data is also more targeted, which facilitates the subsequent differentiated analysis and optimization based on the characteristics of different regions. The logistics hotspots in the region are identified through cluster analysis, revealing the concentration trend and spatial distribution law of logistics activities in the region. This helps to better understand the characteristics of regional logistics demand and provide a basis for subsequent accurate prediction and resource allocation. The complex logistics distribution information is converted into quantifiable feature data, such as order density, traffic flow index, etc., which provides the necessary input for subsequent modeling. The extracted feature data can effectively summarize the characteristics of regional logistics activities and facilitate quantitative analysis and prediction. By establishing a relationship model between regional characteristics and transportation indicators, the transportation time and cost can be predicted according to the regional characteristics. This provides an important decision-making basis for subsequent route optimization and resource scheduling, which helps to improve transportation efficiency and reduce costs. Through the dynamic time series prediction model, accurate prediction of future transportation demand is achieved, which provides key information for planning transportation resources in advance and optimizing scheduling plans. This helps to improve resource utilization, avoid idle or insufficient resources, and thus improve overall logistics efficiency. By taking into account dynamic factors such as real-time traffic events and weather changes, the forecast results are revised to improve the accuracy and reliability of the forecast. This enables the system to more accurately reflect the actual situation and make timely adjustments based on real-time changes, thereby enhancing the adaptability and robustness of the system. The revised forecast demand is converted into a visual dynamic transportation demand distribution map, providing intuitive decision support for dispatchers. The dynamically updated distribution map can reflect the changing trend of transportation demand in real time, making it easier for dispatchers to adjust dispatch strategies and optimize resource allocation in a timely manner.

[0032] Preferably, step S25 comprises the following steps:

[0033] Step S251: extracting the time series of order quantities in each region from the regional transportation characteristic model to obtain the time series data of order quantities in each region;

[0034] Step S252: constructing a regional demand time series forecasting model based on the time series data of the order volume in each region to obtain a regional demand time series forecasting model set;

[0035] Step S253: constructing a regional interaction relationship diagram according to the regional logistics distribution to obtain a dynamic regional interaction relationship diagram;

[0036] Step S254: using the prediction results of the regional time series prediction model set as the initial features of the corresponding nodes in the dynamic regional interaction relationship graph, performing graph neural network modeling that integrates the time series features, and obtaining the regional interaction enhanced prediction demand;

[0037] Step S255: Integrate the forecast results and calculate the indicators of the regional interactive enhanced forecast demand to obtain the forecast regional transportation demand.

[0038] The present invention extracts the historical order volume data of each region to form a time series, which provides basic data for the subsequent construction of a time series prediction model. This enables the model to learn the changing rules of historical order volumes, thereby more accurately predicting future demand. By constructing a time series prediction model for each region separately, the unique pattern of order volume changes in different regions can be captured. The use of the LSTM model can effectively process the long-term dependencies in the time series data, thereby improving the accuracy of the prediction. The construction of a model set also provides a basis for the subsequent fusion of regional interaction information. The logistics interaction relationship between regions is represented in the form of a graph, which captures the mutual influence between regions and provides structured data for the subsequent modeling using a graph neural network. The dynamically updated interaction relationship graph can reflect the impact of real-time traffic conditions on the transportation efficiency between regions, making the prediction closer to the actual situation. Using the graph neural network model, the interaction information between regions is integrated into the demand forecast, which effectively improves the prediction accuracy. Through the graph convolution operation, the model can learn the dependencies between regions and reflect them in the final prediction results, thereby obtaining more accurate prediction results than using the time series prediction model alone. The output results of the graph neural network are converted into the final predicted regional transportation demand, and related indicators such as the distribution and urgency of order types are calculated. This provides complete and actionable forecast data for subsequent vehicle scheduling and route planning, making scheduling decisions more scientific and reasonable.

[0039] Preferably, step S3 comprises the following steps:

[0040] Step S31: Analyze the physical characteristics of the vehicle according to the dynamic transportation demand distribution and the vehicle digital twin to obtain vehicle dynamics data, vehicle energy consumption data, and vehicle safety data; extract the initial state of the route according to the dynamic transportation demand distribution and the vehicle digital twin, and perform route initial state modeling to obtain a route initial state model;

[0041] Step S32: performing a line timeliness evaluation according to the line initial state model and the vehicle dynamics data to obtain a line timeliness evaluation result;

[0042] Step S33: performing a line cost assessment based on the line timeliness assessment result and the vehicle energy consumption data to obtain a line cost assessment result;

[0043] Step S34: performing a route risk assessment based on the route cost assessment result and the vehicle safety data to obtain a route risk assessment result;

[0044] Step S35: Generate a dynamic line performance matrix according to the line timeliness evaluation result, the line cost evaluation result and the line risk evaluation result to obtain a dynamic line performance matrix.

[0045] By extracting the vehicle's dynamics, energy consumption and safety data, more refined and targeted vehicle parameters are provided for subsequent route evaluation, making route evaluation no longer universal, but customized according to different vehicle types and states, thereby significantly improving the accuracy and rationality of the evaluation. Integrating the vehicle's dynamics data into the route timeliness evaluation makes the evaluation results more accurate. The driving time of different vehicles on the same route will vary due to differences in their dynamic performance. Considering factors such as the vehicle's acceleration performance, braking performance, and climbing ability, the driving time of the vehicle on different routes can be more accurately predicted, providing a basis for selecting the optimal route that better meets the vehicle's characteristics. Integrating the vehicle's energy consumption data into the route cost evaluation makes the cost evaluation more refined. The energy consumption of different vehicles on the same route will vary due to differences in their own energy consumption characteristics. Considering factors such as the vehicle's fuel consumption model and energy recovery efficiency, the energy consumption of the vehicle on different routes can be more accurately predicted, providing a basis for selecting a more economical transportation route. By combining vehicle safety data with route risk assessment, we can more comprehensively consider factors affecting transportation safety, identify potential risks, and select safer transportation routes, thereby maximizing the safety of goods and personnel.

[0046] Integrating the timeliness, cost and risk assessment results of the route into the dynamic route performance matrix provides a comprehensive decision-making basis for subsequent transportation scheduling optimization. The dynamically updated performance matrix can reflect the changes in route status in real time, making scheduling decisions more scientific and efficient.

[0047] Preferably, step S4 comprises the following steps:

[0048] Step S41: performing order task analysis according to the dynamic transportation demand distribution and the dynamic line performance matrix to obtain order task data;

[0049] Step S42: assigning demand priorities to the order task data to obtain a demand priority task list;

[0050] Step S43: Allocate vehicle resources according to the demand priority task list to obtain a vehicle allocation plan;

[0051] Step S44: designing a collaborative optimization algorithm according to the vehicle allocation plan and the dynamic line performance matrix to obtain a collaborative optimization target; performing multi-objective solution on the collaborative optimization target to obtain a preliminary transportation scheduling plan;

[0052] Step S45: inputting the preliminary transport scheduling plan into the digital twin environment for dynamic scheduling adjustment to obtain an optimized transport scheduling plan;

[0053] Step S46: globally integrate the optimized transport scheduling solutions to obtain the optimal transport scheduling solution.

[0054] The present invention combines order information with line performance indicators to form order task data, which provides a basis for subsequent priority allocation and vehicle resource allocation. This enables the system to comprehensively consider order requirements and line conditions, thereby making more reasonable scheduling decisions. Orders are prioritized according to factors such as the urgency of the order, the type of goods, and time requirements, ensuring that urgent orders can be processed first. This helps to improve customer satisfaction and optimize overall logistics and transportation efficiency. According to the priority of the order and the real-time status of the vehicle, suitable vehicles are allocated to the corresponding orders to maximize the utilization of vehicle resources. This helps to reduce transportation costs and ensure that orders can be processed in a timely manner. The transportation scheduling scheme is optimized using a multi-objective genetic algorithm, which comprehensively considers multiple objectives such as transportation time, cost, and risk, and finds a preliminary optimal solution under the premise of meeting the constraints. This provides a basis for subsequent dynamic adjustments and helps to strike a balance between multiple objectives. By performing simulation in a digital twin environment, the preliminary scheduling scheme is dynamically adjusted, taking into account dynamic factors such as real-time traffic flow and weather changes, so that the scheduling scheme is closer to the actual situation. This helps to improve the robustness and adaptability of the scheduling scheme, thereby better responding to emergencies. The optimized scheduling scheme is globally integrated, taking into account the dependencies between orders and the scheduling of vehicles, and further optimizing the overall performance of the scheduling scheme. This helps to maximize the efficiency of logistics transportation and reduce overall transportation costs.

[0055] Preferably, step S44 includes the following steps:

[0056] Step S441: constructing a multi-objective optimization problem instance according to the vehicle allocation scheme to obtain a multi-objective optimization problem instance;

[0057] Step S442: Initializing the multi-objective genetic algorithm population for the multi-objective optimization problem instance to obtain an initial scheduling solution population;

[0058] Step S443: Designing a dynamic weighted fitness function for the initial scheduling solution population to obtain a scheduling solution population with a fitness score;

[0059] Step S444: performing genetic operations and population update execution according to the scheduling scheme population with fitness scores to obtain a new generation of scheduling scheme population;

[0060] Step S445: performing population update iteration according to the new generation scheduling plan population to obtain the last generation scheduling plan population; generating a preliminary transport scheduling plan for the last generation scheduling plan population to obtain a preliminary transport scheduling plan.

[0061] The present invention converts the transportation scheduling problem into a clear multi-objective optimization problem instance, defines the objective function and constraints, and provides a framework for the subsequent use of the optimization algorithm to solve the problem. The clear objective function and constraints can guide the search direction of the optimization algorithm and ensure that a solution that meets the actual needs is found. By randomly generating multiple feasible scheduling solutions, the population of the multi-objective genetic algorithm is initialized, and an initial solution set is provided for the iterative optimization of the algorithm. The diverse initial population helps the algorithm explore a broader solution space and avoid falling into a local optimum. By designing a dynamic weight fitness function, the weights of different optimization objectives can be dynamically adjusted according to actual conditions, such as paying more attention to timeliness during peak hours and paying more attention to cost during non-peak hours. This enables the optimization algorithm to better adapt to different operating scenarios and find solutions that better meet actual needs. Through genetic operations such as selection, crossover and mutation, the scheduling solution population is continuously evolved, the overall fitness of the population is gradually improved, and the search is made in the direction of the Pareto optimal solution set. This enables the algorithm to effectively explore the solution space and find a better scheduling solution. The population is evolved through multiple iterations until the termination condition is met, such as reaching the maximum number of iterations or the fitness of the population is no longer significantly improved. Finally, the best individual is selected from the last generation of the population to generate a preliminary transportation scheduling plan, which achieves a good balance between multiple objectives.

[0062] Preferably, step S45 includes the following steps:

[0063] Step S451: construct a preliminary transportation scheduling simulation scenario according to the preliminary transportation scheduling plan and the digital twin logistics base to obtain a preliminary scheduling simulation scenario;

[0064] Step S452: performing virtual vehicle driving simulation on the preliminary scheduling simulation scene to obtain preliminary scheduling simulation results;

[0065] Step S453: performing a preliminary scheduling scheme performance evaluation on the preliminary transportation scheduling scheme according to the preliminary scheduling simulation result to obtain a preliminary scheduling scheme evaluation report;

[0066] Step S454: According to the preliminary scheduling plan evaluation report, the preliminary scheduling plan is dynamically adjusted based on reinforcement learning to obtain an optimized and adjusted scheduling plan;

[0067] Step S455: Generate an optimized transport scheduling plan based on the optimized and adjusted scheduling plan to obtain an optimized transport scheduling plan.

[0068] The present invention constructs a simulation scenario in a digital twin environment, maps the information such as the vehicle driving route and order delivery order in the preliminary scheduling plan into a virtual environment, and provides a realistic experimental platform for subsequent simulation and evaluation. This makes it possible to safely test the performance of the scheduling plan in a virtual environment without actual operation, thereby reducing the cost and risk of testing. By simulating the operation of virtual vehicles, the performance data of the scheduling plan under different traffic conditions and weather conditions are collected, providing a basis for subsequent performance evaluation. This makes it possible to have a more comprehensive understanding of the advantages and disadvantages of the scheduling plan and provide direction for subsequent optimization and adjustment. By quantitatively evaluating the various performance indicators of the preliminary scheduling plan, such as total transportation time, total transportation cost, order fulfillment rate, etc., and generating an evaluation report, a reference is provided for subsequent optimization and adjustment. This makes it possible to clearly understand the shortcomings of the scheduling plan and make targeted improvements. The preliminary scheduling plan is dynamically adjusted using a reinforcement learning algorithm, and through interaction with the simulation environment, it is learned how to make the optimal scheduling decision under different conditions, thereby improving the overall performance of the scheduling plan. The introduction of reinforcement learning enables the scheduling plan to be adaptively adjusted to better cope with the dynamically changing logistics environment. The optimization strategy generated by the reinforcement learning algorithm was applied to the preliminary scheduling plan, generating the final optimized transportation scheduling plan, which can better adapt to real-time traffic conditions and weather changes, improve order fulfillment rate and reduce transportation costs. This ultimately realizes the intelligence and dynamic nature of the scheduling plan, making it more practical.

[0069] Preferably, step S5 comprises the following steps:

[0070] Step S51: collecting transport execution data according to the optimal transport scheduling plan to obtain real-time transport execution data;

[0071] Step S52: Detect and extract operation deviations of the real-time transport execution data and the optimal transport scheduling plan to obtain operation deviation information;

[0072] Step S53: Analyze and classify the causes of the deviation according to the operation deviation information to obtain the classification result of the causes of the deviation;

[0073] Step S54: performing feedback update and digital twin model iteration according to the deviation cause classification result and the digital twin flow base to obtain the iteratively optimized digital twin base;

[0074] Step S55: Adjust and optimize the transportation plan of the optimal transportation scheduling solution based on the iteratively optimized digital twin base to obtain real-time operation deviation feedback data.

[0075] The present invention collects data such as vehicle location, cargo status and order status in real time through various channels, such as GPS devices, IoT sensors and TMS systems, and provides real-time data support for subsequent deviation analysis and model optimization. This enables the system to grasp the dynamic changes in the transportation process in a timely manner and provide a basis for making timely adjustments and optimizations. By comparing real-time data and planned data, the deviation information in the transportation process, such as time deviation, route deviation and cargo status deviation, is identified and extracted, providing a specific target for subsequent deviation cause analysis. This enables the system to quickly discover problems that occur during transportation and provide a direction for solving the problems. The detected deviations are analyzed in depth to find out the root causes of the deviations and classify them, such as traffic factors, weather factors, equipment failures or human factors, providing guidance for subsequent model optimization. This helps the system understand the mechanism of deviation generation and take corresponding improvement measures for different types of deviations. The deviation cause analysis results are fed back to the digital twin base for updating model parameters and rules, such as adjusting section weights, updating cargo transportation specifications, etc., so as to continuously improve the accuracy and predictive ability of the model. This enables the digital twin model to continuously learn and evolve and better adapt to the actual logistics and transportation environment. Using the optimized digital twin base, the current transportation plan is dynamically adjusted and optimized, such as re-planning routes, adjusting the distribution sequence, etc., and real-time operation deviation feedback data is generated to guide actual transportation operations and form a closed-loop optimization process. This enables the system to dynamically adjust the transportation plan according to real-time conditions, improve the system's response speed and robustness, and ultimately improve the overall logistics and transportation efficiency.

[0076] Preferably, the present invention also provides a logistics transportation route optimization system based on digital twins, which is used to execute the logistics transportation route optimization method based on digital twins as described above. The logistics transportation route optimization system based on digital twins includes:

[0077] The digital twin model initialization module is used to obtain multi-source logistics and transportation data, and to perform dynamic logistics network modeling to obtain a dynamic logistics network model; to perform three-dimensional modeling of transportation vehicles based on the dynamic logistics network model and multi-source logistics and transportation data, and to perform vehicle capacity analysis to obtain a vehicle digital twin; to initialize the logistics digital twin model based on the dynamic logistics network model and the vehicle digital twin to obtain a digital twin logistics base;

[0078] Dynamic transportation demand prediction extracts the logistics distribution characteristics of each region from the digital twin logistics base to obtain regional characteristic data; performs dynamic time series prediction modeling based on the regional transportation characteristic model to obtain the predicted regional transportation demand; generates dynamic transportation distribution based on the predicted regional transportation demand to obtain the dynamic transportation demand distribution;

[0079] The real-time line state evaluation module is used to model the line initial state according to the dynamic transportation demand distribution to obtain the line initial state model; perform real-time line state evaluation based on the physical characteristics of the vehicle according to the line initial state model, and generate a dynamic line performance matrix to obtain the dynamic line performance matrix;

[0080] The transport scheduling optimization module is used to allocate vehicle resources based on the order task priority according to the dynamic line performance matrix to obtain the vehicle allocation plan; design the collaborative optimization algorithm according to the vehicle allocation plan and the dynamic line performance matrix to obtain the collaborative optimization target; perform multi-objective solution on the collaborative optimization target to obtain the preliminary transport scheduling plan; input the preliminary transport scheduling plan into the digital twin environment for dynamic scheduling adjustment, and perform global integration of the plan to obtain the optimal transport scheduling plan;

[0081] The real-time feedback and iterative optimization module is used to analyze the transportation execution deviation according to the optimal transportation scheduling plan and obtain the classification results of the deviation causes; iteratively optimize the digital twin model according to the classification results of the deviation causes, and adjust and optimize the transportation plan of the optimal transportation scheduling plan to obtain real-time operation deviation feedback data. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 A schematic diagram of the steps of a logistics transportation route optimization method based on digital twins;

[0083] Figure 2 Detailed implementation flow chart of step S2 in the present invention;

[0084] Figure 3 It is a schematic diagram of the detailed implementation steps of step S4 in the present invention.

[0085] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0086] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0087] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0088] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0089] To achieve this, please refer to Figures 1 to 3 , a logistics transportation route optimization method based on digital twin, comprising the following steps:

[0090] Step S1: Acquire multi-source logistics and transportation data, and perform dynamic logistics network modeling to obtain a dynamic logistics network model; perform three-dimensional modeling of transportation vehicles based on the dynamic logistics network model and multi-source logistics and transportation data, and perform vehicle capacity analysis to obtain a vehicle digital twin; initialize the logistics digital twin model based on the dynamic logistics network model and the vehicle digital twin to obtain a digital twin logistics base;

[0091] Step S2: Extract the logistics distribution characteristics of each region from the digital twin logistics base to obtain regional characteristic data; perform dynamic time series prediction modeling based on the regional transportation characteristic model to obtain the predicted regional transportation demand; generate dynamic transportation distribution based on the predicted regional transportation demand to obtain the dynamic transportation demand distribution;

[0092] Step S3: Model the initial state of the line according to the dynamic transportation demand distribution to obtain the initial line state model; perform real-time line state evaluation based on the vehicle physical characteristics according to the initial line state model, and generate a dynamic line performance matrix to obtain the dynamic line performance matrix;

[0093] Step S4: Allocate vehicle resources based on the order task priority according to the dynamic line performance matrix to obtain a vehicle allocation plan; design a collaborative optimization algorithm according to the vehicle allocation plan and the dynamic line performance matrix to obtain a collaborative optimization objective; perform multi-objective solution on the collaborative optimization objective to obtain a preliminary transportation scheduling plan; input the preliminary transportation scheduling plan into the digital twin environment for dynamic scheduling adjustment, and perform global integration of the plan to obtain the optimal transportation scheduling plan;

[0094] Step S5: Analyze the transportation execution deviation according to the optimal transportation scheduling plan to obtain the classification result of the deviation cause; perform iterative optimization of the digital twin model according to the classification result of the deviation cause, and adjust and optimize the transportation plan of the optimal transportation scheduling plan to obtain the real-time operation deviation feedback data.

[0095] In the embodiment of the present invention, with reference to Figure 1 As shown, it is a schematic diagram of the step flow of the logistics transportation line optimization method based on digital twin of the present invention. In this example, the logistics transportation line optimization method based on digital twin includes the following steps:

[0096] Step S1: Obtain multi-source logistics transportation data, and perform dynamic logistics network modeling to obtain a dynamic logistics network model; perform three-dimensional modeling of transportation vehicles according to the dynamic logistics network model and the multi-source logistics transportation data, and perform vehicle passing capacity analysis to obtain a vehicle digital twin; initialize the logistics digital twin model according to the dynamic logistics network model and the vehicle digital twin to obtain a digital twin logistics base;

[0097] In the embodiment of the present invention, multi-source data is obtained and a digital twin base is constructed. This step first obtains real-time data from the TMS, WMS, GPS terminal, traffic monitoring system, and meteorological platform through a multi-source heterogeneous data fusion interface, cleans and standardizes it, and stores it in a distributed database. Then, a dynamic logistics network model is constructed using the graph database Neo4j, and a digital twin initial environment is created based on this model and the Unity3D engine. Finally, real-time optimization algorithms, such as path planning and load balancing algorithms, are integrated to generate a digital twin logistics base that includes the optimized logistics network model, vehicle status, order information, and digital twin environment.

[0098] Step S2: Extract the logistics distribution characteristics of each region from the digital twin logistics base to obtain regional characteristic data; perform dynamic time series prediction modeling based on the regional transportation characteristic model to obtain the predicted regional transportation demand; generate dynamic transportation distribution based on the predicted regional transportation demand to obtain the dynamic transportation demand distribution;

[0099] In an embodiment of the present invention, the dynamic transportation demand distribution is predicted. This step first performs regional division and cluster analysis on the digital twin logistics base, identifies logistics hot spots and extracts regional characteristics, such as order density, average order size, traffic flow index, warehouse load rate and weather influencing factors. Then, the regional transportation characteristics are modeled using a multivariate linear regression model, and a LSTM model is used for dynamic time series prediction to obtain the predicted regional transportation demand. Finally, the prediction results are dynamically corrected according to real-time traffic events and weather changes to generate a dynamic transportation demand distribution map.

[0100] Step S3: Modeling the initial state of the line according to the dynamic transportation demand distribution to obtain the initial state model of the line; performing a real-time line state evaluation based on the physical characteristics of the vehicle according to the initial state model of the line, and generating a dynamic line performance matrix to obtain a dynamic line performance matrix;

[0101] In an embodiment of the present invention, the line status is evaluated in real time and a dynamic performance matrix is ​​generated. This step first uses the Dijkstra algorithm to calculate candidate lines based on the dynamic transportation demand distribution and road network information, and constructs a line initial state model, which includes the static and dynamic attributes of the line. Then, based on the line initial state model, the timeliness (ETA), cost, and risk level of the line are evaluated respectively. The timeliness evaluation takes into account real-time traffic flow and weather conditions, the cost evaluation takes into account the line length and estimated transportation time, and the risk evaluation takes into account congestion status, weather conditions, and historical accident data. Finally, the evaluation results are integrated to generate a dynamic line performance matrix DRPM and stored in Redis.

[0102] Step S4: According to the dynamic line performance matrix, vehicle resource allocation based on order task priority is performed to obtain a vehicle allocation plan; according to the vehicle allocation plan and the dynamic line performance matrix, a collaborative optimization algorithm is designed to obtain a collaborative optimization target; multi-objective solutions are performed on the collaborative optimization target to obtain a preliminary transportation scheduling plan; the preliminary transportation scheduling plan is input into the digital twin environment for dynamic scheduling adjustment, and the plan is globally integrated to obtain the optimal transportation scheduling plan;

[0103] In an embodiment of the present invention, transport scheduling is performed based on collaborative optimization. This step first parses the order tasks according to the dynamic transport demand distribution and the dynamic line performance matrix, assigns priorities according to order attributes and expected delivery time, and generates a demand priority task list. Then, vehicle resources are allocated according to the task list and the real-time status of the vehicle. After that, based on the vehicle allocation plan and the dynamic line performance matrix, a multi-objective genetic algorithm is used for collaborative optimization, with the objectives including minimizing the total transport time, total transport cost, and total transport risk. The optimized plan is input into the digital twin environment for simulation, and dynamic adjustments are made according to the simulation results to finally generate the optimal transport scheduling plan.

[0104] Step S5: Perform transportation execution deviation analysis according to the optimal transportation scheduling plan to obtain the classification results of the deviation causes; perform iterative optimization of the digital twin model according to the classification results of the deviation causes, and adjust and optimize the transportation plan of the optimal transportation scheduling plan to obtain real-time operation deviation feedback data;

[0105] In an embodiment of the present invention, real-time feedback and iterative optimization are performed. This step first collects real-time data on transportation execution, including vehicle location, cargo status, and order status, through GPS devices, IoT sensors, and TMS. Then, the real-time data is compared with the optimal transportation scheduling plan to detect and extract operational deviations, including time deviations, route deviations, and cargo status deviations. Next, the causes of the deviations are analyzed and classified, such as traffic factors, weather factors, equipment failures, or human factors. The classification results of the causes of the deviations are fed back to the digital twin logistics base to update model parameters and safety specifications. Finally, the transportation plan is dynamically adjusted and optimized based on the iteratively optimized digital twin base to generate real-time operational deviation feedback data for guiding actual operations and subsequent optimization.

[0106] Preferably, step S1 comprises the following steps:

[0107] Step S11: Acquire multi-source logistics and transportation data through a multi-source heterogeneous data fusion interface to obtain an initial data set;

[0108] Step S12: performing data cleaning and standardization processing on the initial data set to obtain standardized logistics data; constructing a vehicle three-dimensional model on the standardized logistics data to obtain a transport vehicle three-dimensional model;

[0109] Step S13: Perform dynamic logistics network modeling based on standardized logistics data to obtain a dynamic logistics network model; perform capacity parameter calculation based on the dynamic logistics network model and the three-dimensional model of the transport vehicle to obtain the transport vehicle capacity parameter;

[0110] Step S14: Generate a digital twin environment according to the dynamic logistics network model to obtain a digital twin initial environment; generate a vehicle digital twin according to the transport vehicle capacity parameters to obtain a vehicle digital twin;

[0111] Step S15: Perform real-time dynamic optimization and base generation based on the digital twin initial environment and the vehicle digital twin to obtain the digital twin flow base.

[0112] In the embodiment of the present invention, a pre-built multi-source heterogeneous data fusion interface is used to obtain real-time data from a logistics management system (TMS), a warehouse management system (WMS), a vehicle-mounted GPS terminal, a vehicle-mounted laser radar, a road traffic monitoring system, and a meteorological service platform. The logistics management system (TMS) provides order information, including order ID, starting point, end point, cargo type, weight, volume, expected delivery time, etc. The warehouse management system (WMS) provides warehouse inventory information, including warehouse ID, cargo type, quantity, storage location, etc. The vehicle-mounted GPS terminal provides vehicle real-time location, speed, direction, fuel consumption, and other information. The vehicle-mounted laser radar scans and obtains three-dimensional point cloud data of the surrounding environment at a frequency of 10Hz; the multispectral camera collects road and traffic condition images at a frequency of 5Hz; the high-precision IMU provides vehicle posture information at a frequency of 100Hz; the GNSS receiver provides centimeter-level positioning at a frequency of 1Hz; the OBD-II interface reads parameters such as engine speed and fuel consumption in real time; and the intelligent cargo tracking tag reports temperature, humidity, etc. in real time. The road traffic monitoring system provides real-time road condition information, including road section ID, traffic flow, average speed, congestion status, etc. The meteorological service platform provides weather forecast information, including temperature, humidity, rainfall, wind speed, etc. The data fusion interface uses a message queue mechanism, such as Kafka or RabbitMQ, to achieve asynchronous data transmission, and converts data in different formats into a unified JSON format, which is stored in the distributed database MongoDB to form an initial data set. The MongoDB schema design includes key fields for each data source, and uses a timestamp field to ensure the time consistency of the data.

[0113] The Apache Spark distributed computing framework was used to clean and standardize the initial data set. The cleaning process included removing duplicate data, filling missing values, and detecting and processing outliers. Different filling strategies were used for missing values ​​according to the data type: numerical data was filled with the mean or median, time data was filled with the previous valid value, and categorical data was filled with the mode. Outlier detection used a method based on the 3σ criterion to identify data that exceeded the mean ±3 times the standard deviation as outliers and replace or delete them. The data standardization process included: converting timestamps to UTC time format, converting geographic location information to a unified latitude and longitude coordinate system, converting speed units to kilometers per hour, and converting cargo weight units to kilograms. The processed standardized logistics data was stored in the Hadoop Distributed File System (HDFS) in Parquet format for efficient reading and processing in subsequent steps. The PointNet++ deep learning model is used to filter the vehicle-mounted lidar point cloud data and extract key points. The ICP algorithm is used to align the point cloud data of consecutive frames to build a dense three-dimensional point cloud map of the vehicle itself and its surroundings. The Mask R-CNN model is used to perform semantic segmentation on multispectral images, and objects such as vehicles, pedestrians, and traffic signs in the point cloud model are identified and labeled.

[0114] Neo4j graph database is used to represent the logistics network. Static nodes such as warehouses, distribution centers, and traffic intersections store geographical locations, capacity, and other attributes. Intelligent transportation vehicles are used as dynamic nodes to store their ID, location, speed, posture, 3D model, sensor data, and other attributes in real time. Transportation routes are used as edges to store static attributes such as length, width, slope, curvature, and real-time traffic flow. The 3D model of the vehicle and the 3D reconstruction model of the environment are used to calculate the minimum turning radius requirements for different types of vehicles, the maximum passable height, and other traffic capacity parameters for each route. Based on multispectral image analysis, the friction coefficient and water depth of the road surface are evaluated. Traffic light timing data is used to estimate the average delay time at each intersection. By subscribing to real-time sensor data in the Kafka message queue, the location, speed, posture, and other attributes of the vehicle node are dynamically updated. The traffic flow and environmental condition attributes of the route edge are dynamically updated using the traffic data of the road traffic monitoring system and the weather information of the meteorological service platform.

[0115] The digital twin initial environment is constructed based on the Unity3D game engine and the dynamic logistics network model. The node and edge data in the Neo4j graph database are imported into Unity3D to generate a visualized three-dimensional logistics network scene. The road section color is dynamically adjusted using real-time traffic flow data. For example, green represents smooth traffic, yellow represents slow traffic, and red represents congestion. Real-time weather data is used to simulate weather changes. For example, rainy and snowy weather will affect the friction coefficient of the road section and the vehicle speed. The digital twin initial environment provides a virtual logistics transportation scene for simulating and analyzing the logistics transportation process. The high-precision three-dimensional model of the vehicle reconstructed in step S12 is imported into Unity3D, and the state of the model is dynamically updated according to the real-time status data of the vehicle (for example: position, speed, door status); the sensor data of the vehicle is integrated to visualize the sensor's perception range and data in the virtual environment.

[0116] On the basis of the initial digital twin environment, real-time optimization algorithms are integrated, such as the path planning algorithm based on the A* algorithm, to calculate the optimal path between any two points, and the calculation results are fed back to the dynamic logistics network model to update the estimated transportation time and cost of the route. At the same time, based on the real-time order information and vehicle location information, the load balancing algorithm is used to dynamically adjust the task allocation of the vehicle, and the adjustment results are fed back to the digital twin environment to update the location and status of the vehicle. The final digital twin logistics base includes a real-time optimized logistics network model, vehicle status information, order information, and a digital twin environment, providing basic data and simulation environment for subsequent transportation demand forecasting and route optimization.

[0117] Preferably, step S2 comprises the following steps:

[0118] Step S21: partitioning the digital twin logistics base into regional logistics data to obtain partitioned logistics data;

[0119] Step S22: performing intra-regional data clustering analysis on the partitioned logistics data to obtain regional logistics distribution;

[0120] Step S23: extracting regional features of regional logistics distribution to obtain regional feature data;

[0121] Step S24: performing regional transportation characteristic modeling according to the regional characteristic data to obtain a regional transportation characteristic model;

[0122] Step S25: Perform dynamic time series forecasting modeling according to the regional transportation characteristic model to obtain the predicted regional transportation demand;

[0123] Step S26: dynamically modify the predicted regional transportation demand to obtain a modified regional transportation demand;

[0124] Step S27: Generate dynamic transportation distribution according to the modified regional transportation demand to obtain dynamic transportation demand distribution.

[0125] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0126] Step S21: partitioning the digital twin logistics base into regional logistics data to obtain partitioned logistics data;

[0127] In an embodiment of the present invention, the geographical location information in the digital twin logistics base is used in combination with the K-means clustering algorithm to divide the logistics network into regions. First, the starting and ending coordinates of all orders, as well as the location coordinates of warehouses and distribution centers, are extracted from the digital twin logistics base. Then, a predetermined number of regions K is set, for example, according to the number of urban area divisions. These coordinate points are clustered into K regions using the K-means algorithm. Each region contains several order starting points, end points, warehouses, and distribution centers. The order data, vehicle data, traffic flow data, etc. in the digital twin logistics base are divided according to the regions to which they belong, and K partitioned logistics data sets are generated. Each partitioned logistics data set contains all relevant data in the region and is marked with a regional ID.

[0128] Step S22: performing intra-regional data clustering analysis on the partitioned logistics data to obtain regional logistics distribution;

[0129] In an embodiment of the present invention, DBSCAN cluster analysis is performed on each partitioned logistics data to identify logistics hot spots in the region. The coordinates of the order start and end points in the partitioned logistics data are extracted, and the DBSCAN algorithm is used to cluster these coordinates. The parameter epsilon (neighborhood radius) of the DBSCAN algorithm is set to 1 km, and minPts (minimum number of points) is set to 5. The clustering results cluster the order start and end points into different density areas, representing the logistics hot spots in the region. The clustering results are associated with the location information of warehouses and distribution centers in the region to generate a regional logistics distribution map. The regional logistics distribution map shows the density distribution of different logistics activities in the region, as well as the spatial relationship with warehouses and distribution centers.

[0130] Step S23: extracting regional features of regional logistics distribution to obtain regional feature data;

[0131] In the embodiment of the present invention, the following features are extracted from the regional logistics distribution: order density, which calculates the number of orders per unit area in each clustering area; average order size, which calculates the average weight and volume of ordered goods in each clustering area; traffic flow index, which calculates the average traffic flow and speed of the main roads in each area; warehouse load rate, which calculates the average inventory and capacity ratio of warehouses in each area; weather impact factor, which calculates the impact of rainfall, temperature, etc. on logistics transportation based on the weather forecast data in the area. The extracted feature data is stored as structured data, with one record for each area, including the area ID and various feature values.

[0132] Step S24: performing regional transportation characteristic modeling according to the regional characteristic data to obtain a regional transportation characteristic model;

[0133] In the embodiment of the present invention, a multiple linear regression model is used to model regional transportation characteristics. The regional characteristic data extracted in step S23 is used as an input variable, and the regional transportation time, cost, etc. in the historical transportation data are used as output variables. The multiple linear regression model is trained to learn the relationship between regional characteristics and transportation indicators. After the model training is completed, the transportation time and cost of the region can be predicted based on the regional characteristics. The regional transportation characteristic model can be regularly updated based on new data to maintain the accuracy of the model.

[0134] Step S25: Perform dynamic time series forecasting modeling according to the regional transportation characteristic model to obtain the predicted regional transportation demand;

[0135] In an embodiment of the present invention, a long short-term memory network (LSTM) model is used to predict the future transportation demand of each region. The historical order data of each region is arranged in time series as the input of the LSTM model. The output of the LSTM model is the predicted number of orders for each region in the future. During the model training process, historical traffic flow data and weather data are used as auxiliary inputs to improve the prediction accuracy. The prediction results are output in the form of a time series, for example, the number of orders per hour in the next 24 hours is predicted.

[0136] Step S26: dynamically modify the predicted regional transportation demand to obtain a modified regional transportation demand;

[0137] In the embodiment of the present invention, the predicted regional transportation demand is dynamically revised according to real-time traffic events and weather changes. For example, if a traffic accident occurs in a certain area and the road is closed, the predicted number of orders in the area is reduced; if heavy rain occurs in a certain area, the predicted value of the transportation time in the area is increased. The dynamic correction process is adjusted according to preset rules and parameters. For example, if a traffic accident causes a road to be closed, the predicted number of orders in the area is reduced by 20%.

[0138] Step S27: Generate dynamic transportation distribution according to the modified regional transportation demand to obtain dynamic transportation demand distribution;

[0139] In the embodiment of the present invention, the corrected regional transportation demand data is integrated into the digital twin environment to generate a dynamic transportation demand distribution map. The dynamic transportation demand distribution map displays the predicted number of orders for each region in the form of a heat map, and the darker the color, the more orders. The dynamic transportation demand distribution map is updated in real time to reflect the latest changes in transportation demand, providing a basis for subsequent path planning and scheduling optimization.

[0140] Preferably, step S25 comprises the following steps:

[0141] Step S251: extracting the time series of order quantities in each region from the regional transportation characteristic model to obtain the time series data of order quantities in each region;

[0142] Step S252: constructing a regional demand time series forecasting model based on the time series data of the order volume in each region to obtain a regional demand time series forecasting model set;

[0143] Step S253: constructing a regional interaction relationship diagram according to the regional logistics distribution to obtain a dynamic regional interaction relationship diagram;

[0144] Step S254: using the prediction results of the regional time series prediction model set as the initial features of the corresponding nodes in the dynamic regional interaction relationship graph, performing graph neural network modeling that integrates the time series features, and obtaining the regional interaction enhanced prediction demand;

[0145] Step S255: Integrate the forecast results and calculate the indicators of the regional interactive enhanced forecast demand to obtain the forecast regional transportation demand.

[0146] In the embodiment of the present invention, the historical order volume data of each region is extracted from the regional transportation feature model, arranged in chronological order, and the time series data of the order volume of each region is generated. The time series data is in hours, and records the number of orders per hour in each region in the past period of time (for example, the past year). Missing data values ​​are filled using linear interpolation. The time series data is normalized, such as Min-Max scaling, to scale the data range to the [0, 1] interval to avoid the impact of large differences in order volumes in different regions on model training.

[0147] For each region's order volume time series data, an independent LSTM time series forecasting model is constructed. The input of the LSTM model is the order volume data of the past period (for example, the past 24 hours), and the output is the order volume forecast value of the future period (for example, the next 24 hours). The model is trained using the Adam optimizer, and the loss function is the mean square error (MSE). The LSTM model of each region is independently trained using the data of the respective region, and finally a regional demand time series forecasting model set containing all regional LSTM models is obtained.

[0148] A regional interaction relationship diagram is constructed based on regional logistics distribution and road network information. Each region is abstracted as a node in the diagram. If there is a direct road connection between two regions, an edge is established between the corresponding nodes. The weight of the edge represents the degree of traffic convenience between the two regions. For example, the inverse of the road distance or the average transportation time can be used as the weight. The degree of traffic convenience is dynamically updated based on the real-time traffic flow data in the digital twin logistics base, and the edge weight corresponding to the congested section will be reduced.

[0149] The regional interaction relationship graph is modeled using a graph convolutional neural network (GCN) model. The prediction results of the LSTM model for each region in step S252 are used as the initial feature vector of the corresponding node. The GCN model propagates and aggregates node features along the edges of the graph through a message passing mechanism, thereby capturing the mutual influence between regions. For example, if the predicted order volume of a region increases, the predicted order volume of the adjacent regions will also be affected. After multiple layers of GCN convolution operations, a new feature vector is obtained for each node, which incorporates the information of neighboring nodes and represents the predicted demand after enhanced regional interaction.

[0150] The feature vector of each node output by the GCN model in step S254 is converted into a predicted order volume. The specific method can be to perform a linear transformation on the feature vector, or to use a small neural network for prediction. Finally, the predicted order volume for each region in the future period of time and other related indicators, such as the distribution of order types, the urgency of orders, etc., are obtained. These prediction results are integrated together to form the predicted regional transportation demand data for subsequent vehicle scheduling and path planning.

[0151] Preferably, step S3 comprises the following steps:

[0152] Step S31: Analyze the physical characteristics of the vehicle according to the dynamic transportation demand distribution and the vehicle digital twin to obtain vehicle dynamics data, vehicle energy consumption data, and vehicle safety data; extract the initial state of the route according to the dynamic transportation demand distribution and the vehicle digital twin, and perform route initial state modeling to obtain a route initial state model;

[0153] Step S32: performing a line timeliness evaluation according to the line initial state model and the vehicle dynamics data to obtain a line timeliness evaluation result;

[0154] Step S33: performing a line cost assessment based on the line timeliness assessment result and the vehicle energy consumption data to obtain a line cost assessment result;

[0155] Step S34: performing a route risk assessment based on the route cost assessment result and the vehicle safety data to obtain a route risk assessment result;

[0156] Step S35: Generate a dynamic line performance matrix according to the line timeliness evaluation result, the line cost evaluation result and the line risk evaluation result to obtain a dynamic line performance matrix.

[0157] Extract the static attributes of the vehicle, such as the 3D model, weight, wheelbase, and tire parameters, from the vehicle digital twin. Use the vehicle's historical operating data and the physical engine of the digital twin environment to simulate the vehicle's driving state under different working conditions and calculate the vehicle's acceleration performance, braking performance, climbing ability, and other dynamic data. Based on the vehicle's 3D model, engine / motor efficiency, driving speed and other parameters, establish a vehicle energy consumption model to predict energy consumption data under different routes. Analyze the vehicle's structural parameters, sensor configuration, historical accident data, etc., evaluate the vehicle's safety performance, and obtain vehicle safety data. Extract the transportation demand between each region and the coordinates of the center point of each region from the dynamic transportation demand distribution (DDD). Based on the road network information in the digital twin's logistics base (TLB), use the Dijkstra algorithm to calculate the shortest path between the center points of the region as a candidate route. The route initial state model is represented by a graph data structure, with nodes representing the center points of the region and edges representing candidate routes. Each edge stores the static attributes of the route, such as the route ID, the starting area ID, the end area ID, the route length, and the list of road segment IDs along the way. At the same time, the analyzed vehicle dynamics data, vehicle energy consumption data and vehicle safety data are associated with the corresponding vehicle to provide vehicle-specific parameters for subsequent evaluation.

[0158] Based on the static attributes (line length, list of road segment IDs along the way) and dynamic attributes (current average speed, congestion status, weather conditions) of the line in the line initial state model, and the vehicle dynamics data obtained in step S31, the estimated transportation time (ETA) of each line is calculated. The ETA calculation method is as follows: the line is divided into multiple sections, and the driving time of the vehicle in each section is more accurately calculated based on the length, speed limit, traffic flow, and acceleration and braking performance of each section. For example, for a climbing section, the impact of the vehicle's climbing ability on the speed is considered; for a curved road section, the impact of the vehicle's braking performance on the safe speed is considered. Bad weather (such as heavy rain and heavy snow) will reduce the speed of the section, and the congestion status will be multiplied by the corresponding delay coefficient according to the congestion level. Finally, the ID of each line and the corresponding ETA are stored as the line timeliness evaluation result. Based on the estimated transportation time (ETA) in the line timeliness evaluation result (RTE) and the line length in the line initial state model, and the vehicle energy consumption data obtained in step S31, the transportation cost of each line is calculated. The formula for calculating transportation cost is as follows: cost = energy consumption cost + other costs. Among them, energy consumption cost is calculated based on the vehicle's energy consumption data and energy unit price, for example, fuel consumption * fuel unit price or electricity consumption * electricity price. Other costs include unit distance cost * route length + unit time cost * ETA. Unit distance cost includes fixed costs such as vehicle wear and tear costs. Unit time cost includes time-related costs such as driver wages and vehicle operating costs. Different types of vehicles have different unit distance costs, unit time costs and energy consumption characteristics. Finally, the ID of each route and the corresponding transportation cost are stored as the route cost evaluation result. Based on the route dynamic attributes (congestion status, weather conditions), historical accident data and vehicle safety data in the route initial state model, the risk level of each route is evaluated. For example, based on the size and stability of the vehicle, the rollover risk of the vehicle on a curve or slope is evaluated; based on the safety performance of the vehicle, such as airbags, anti-lock braking systems, etc., the risk assessment coefficient is adjusted. The route timeliness evaluation results (RTE), route cost evaluation results (RCE) and route risk evaluation results (RRE) are integrated to generate a dynamic route performance matrix (DRPM). Each row of DRPM represents a route, including route ID, estimated transportation time, transportation cost, risk level, etc. DRPM data will be updated in real time based on real-time traffic flow, weather conditions and other dynamic factors, providing dynamic route performance evaluation data for subsequent transportation scheduling optimization. DRPM uses the distributed cache system Redis for storage to achieve fast reading and updating.

[0159] Preferably, step S4 comprises the following steps:

[0160] Step S41: performing order task analysis according to the dynamic transportation demand distribution and the dynamic line performance matrix to obtain order task data;

[0161] Step S42: assigning demand priorities to the order task data to obtain a demand priority task list;

[0162] Step S43: Allocate vehicle resources according to the demand priority task list to obtain a vehicle allocation plan;

[0163] Step S44: designing a collaborative optimization algorithm according to the vehicle allocation plan and the dynamic line performance matrix to obtain a collaborative optimization target; performing multi-objective solution on the collaborative optimization target to obtain a preliminary transportation scheduling plan;

[0164] Step S45: inputting the preliminary transport scheduling plan into the digital twin environment for dynamic scheduling adjustment to obtain an optimized transport scheduling plan;

[0165] Step S46: globally integrate the optimized transport scheduling solutions to obtain the optimal transport scheduling solution.

[0166] As an example of the present invention, refer to Figure 3 As shown, in this example, step S4 includes:

[0167] Step S41: performing order task analysis according to the dynamic transportation demand distribution and the dynamic line performance matrix to obtain order task data;

[0168] In an embodiment of the present invention, the order information to be processed is extracted from the dynamic transportation demand distribution (DDD), including the order ID, the starting area ID, the end area ID, the cargo type, weight, volume, expected delivery time, etc. Combined with the dynamic route performance matrix (DRPM), a feasible transportation route is matched for each order. The routes connecting the order starting area and the end area in the DRPM are screened, and the order information and route performance indicators (estimated transportation time, transportation cost, risk level) are integrated together to generate order task data. The order task data is stored in the distributed database Cassandra, with the order ID as the primary key.

[0169] Step S42: assigning demand priorities to the order task data to obtain a demand priority task list;

[0170] In an embodiment of the present invention, order task data is prioritized according to factors such as the expected delivery time of the order, the type of goods, and the importance of the order, and a demand priority task list is generated. The order priority calculation formula is as follows: Priority = Weight 1 * (Deadline - Current Time) + Weight 2 * Order Importance + Weight 3 * Goods Type Coefficient. Among them, Weight 1, Weight 2, and Weight 3 are preset parameters, the deadline is the expected delivery time of the order, the current time is the current time of the system, the order importance is set according to the order attributes (for example, the importance of urgent orders is set to 1, and the ordinary order is set to 0), and the goods type coefficient is set according to the goods type (for example, the perishable goods coefficient is set to 1, and the ordinary goods are set to 0). The smaller the priority value, the higher the order priority. The demand priority task list is arranged in ascending order of priority.

[0171] Step S43: Allocate vehicle resources according to the demand priority task list to obtain a vehicle allocation plan;

[0172] In an embodiment of the present invention, real-time status information of available vehicles is obtained from the digital twin logistics base (TLB), including vehicle ID, current location, load capacity, remaining fuel, etc. According to the demand priority task list, vehicles are assigned to each order in turn. The vehicle allocation strategy is as follows: give priority to vehicles that are close to the order starting point and whose load capacity meets the order cargo requirements. If multiple vehicles meet the conditions, the vehicle with more remaining fuel is selected. The allocation results are recorded in the vehicle allocation plan, which includes the vehicle ID, the list of assigned order IDs, and the estimated departure time.

[0173] Step S44: designing a collaborative optimization algorithm according to the vehicle allocation plan and the dynamic line performance matrix to obtain a collaborative optimization target; performing multi-objective solution on the collaborative optimization target to obtain a preliminary transportation scheduling plan;

[0174] In an embodiment of the present invention, a multi-objective genetic algorithm is used to optimize the transportation scheduling scheme. The collaborative optimization objectives include minimizing the total transportation time, minimizing the total transportation cost, and minimizing the total transportation risk. An initial population is generated based on the vehicle allocation plan, and each individual represents a transportation scheduling plan, including the driving route of each vehicle and the order delivery sequence. The fitness function is calculated based on the weighted sum of the three optimization objectives. Genetic operations include selection, crossover and mutation, and the population is iteratively updated until a preset termination condition (such as the maximum number of iterations) is reached. Finally, the optimal solution on the Pareto frontier is selected as the preliminary transportation scheduling plan. The preliminary transportation scheduling plan includes the driving route of each vehicle, the order delivery sequence, and the estimated time of arrival at each delivery point.

[0175] Step S45: inputting the preliminary transport scheduling plan into the digital twin environment for dynamic scheduling adjustment to obtain an optimized transport scheduling plan;

[0176] In the embodiment of the present invention, the preliminary transport scheduling plan is imported into the digital twin environment for simulation. During the simulation, the traffic flow data and weather data in the digital twin environment are obtained in real time, and the driving speed and route of the vehicle are dynamically adjusted according to these data. If the vehicle encounters traffic congestion or other emergencies, the driving route is replanned and the estimated arrival time is updated. After the simulation is completed, the adjusted scheduling plan is output as the optimized transport scheduling plan.

[0177] Step S46: globally integrating the optimized transport scheduling scheme to obtain the optimal transport scheduling scheme;

[0178] In the embodiment of the present invention, the optimized transportation scheduling plan is globally optimized and adjusted. For example, the order delivery sequence is adjusted to reduce the empty driving distance of the vehicle by considering the dependency between orders. Finally, the optimal transportation scheduling plan is generated, which includes the final driving route of each vehicle, the order delivery sequence, and the estimated time of arrival at each delivery point. The optimal transportation scheduling plan will be sent to the terminal system of each vehicle to guide the actual transportation operation.

[0179] Preferably, step S44 includes the following steps:

[0180] Step S441: constructing a multi-objective optimization problem instance according to the vehicle allocation scheme to obtain a multi-objective optimization problem instance;

[0181] Step S442: Initializing the multi-objective genetic algorithm population for the multi-objective optimization problem instance to obtain an initial scheduling solution population;

[0182] Step S443: Designing a dynamic weighted fitness function for the initial scheduling solution population to obtain a scheduling solution population with a fitness score;

[0183] Step S444: performing genetic operations and population update execution according to the scheduling scheme population with fitness scores to obtain a new generation of scheduling scheme population;

[0184] Step S445: performing population update iteration according to the new generation scheduling plan population to obtain the last generation scheduling plan population; generating a preliminary transport scheduling plan for the last generation scheduling plan population to obtain a preliminary transport scheduling plan.

[0185] In an embodiment of the present invention, a multi-objective optimization problem instance is constructed based on a vehicle allocation scheme and a dynamic route performance matrix (DRPM). The objective functions of the multi-objective optimization problem instance include minimizing the total transportation time, minimizing the total transportation cost, and minimizing the total transportation risk. The calculation method of each objective function is as follows: the total transportation time is the sum of the total time for all vehicles to complete all assigned tasks; the total transportation cost is the sum of the total cost of all vehicles to complete all assigned tasks, and the cost is calculated based on the cost of each route in the DRPM; the total transportation risk is the sum of the risk levels of all vehicle routes, and the risk level is calculated based on the risk level of each route in the DRPM. Constraints include: the total load of each vehicle cannot exceed its maximum load; each order must be delivered before the expected delivery time. The decision variables are the driving route of each vehicle and the order delivery order.

[0186] 100 feasible transport scheduling plans are randomly generated to form the initial scheduling plan population. Each scheduling plan contains the driving route of each vehicle and the order delivery order. When generating a scheduling plan, first determine the orders that each vehicle is responsible for according to the vehicle allocation plan, and then randomly generate the order delivery order. After that, according to the order delivery order and the dynamic route performance matrix (DRPM), the Dijkstra algorithm is used to calculate the shortest path between different delivery points for each vehicle to form the vehicle's driving route. Ensure that each generated scheduling plan meets the constraints defined in step S441, such as vehicle load constraints and order delivery time constraints.

[0187] A dynamic weighted fitness function is designed to evaluate the pros and cons of each scheduling scheme. The fitness function is calculated as follows: Fitness = Weight 1*Total Transportation Time + Weight 2*Total Transportation Cost + Weight 3*Total Transportation Risk. Among them, Weight 1, Weight 2, and Weight 3 are dynamically adjusted according to the current logistics operation status. For example, during peak hours, the value of Weight 1 (total transportation time) is increased to give priority to timeliness; during non-peak hours, the value of Weight 2 (total transportation cost) is increased to give priority to cost reduction. The weight adjustment strategy can be learned and optimized based on historical data and real-time data. The fitness value of each scheduling scheme is calculated according to the fitness function, and a scheduling scheme population with fitness scores is generated.

[0188] The scheduling scheme population with fitness scores is selected, crossover and mutated to generate a new generation of scheduling scheme population. The selection operation adopts the roulette selection method, giving priority to scheduling schemes with higher fitness values. The crossover operation adopts the partial mapping crossover (PMX) method to exchange the delivery order of some orders of the two parent scheduling schemes to generate two new child scheduling schemes. The mutation operation adopts the exchange mutation method to randomly select two orders in the scheduling scheme and exchange their delivery order. The crossover probability is set to 0.8 and the mutation probability is set to 0.1. The size of the new generation of scheduling scheme population is the same as the initial population size, both of which are 100.

[0189] Repeat steps S443 and S444 to perform population update iterations. During the iteration process, record the average fitness value of each generation of population. When the average fitness value no longer increases significantly, or reaches the preset maximum number of iterations (e.g., 1000 times), stop the iteration. Select the scheduling plan with the highest fitness value from the last generation of scheduling plan population as the preliminary transportation scheduling plan.

[0190] Preferably, step S45 includes the following steps:

[0191] Step S451: construct a preliminary transportation scheduling simulation scenario according to the preliminary transportation scheduling plan and the digital twin logistics base to obtain a preliminary scheduling simulation scenario;

[0192] Step S452: performing virtual vehicle driving simulation on the preliminary scheduling simulation scene to obtain preliminary scheduling simulation results;

[0193] Step S453: performing a preliminary scheduling scheme performance evaluation on the preliminary transportation scheduling scheme according to the preliminary scheduling simulation result to obtain a preliminary scheduling scheme evaluation report;

[0194] Step S454: According to the preliminary scheduling plan evaluation report, the preliminary scheduling plan is dynamically adjusted based on reinforcement learning to obtain an optimized and adjusted scheduling plan;

[0195] Step S455: Generate an optimized transport scheduling plan based on the optimized and adjusted scheduling plan to obtain an optimized transport scheduling plan.

[0196] In an embodiment of the present invention, a preliminary scheduling simulation scenario is constructed in a simulation platform (e.g., Prescan) using road network information, real-time traffic flow data, real-time weather data, and location information of warehouses and distribution centers in the digital twin logistics base (TLB). The driving route, order delivery sequence, and estimated arrival time of each vehicle in the preliminary transportation scheduling plan are imported into the simulation scenario. In the simulation scenario, a virtual vehicle object corresponding to the actual vehicle type is created, and the initial position, cargo status, and driving speed of the virtual vehicle are set. Real-time traffic flow data is used to simulate emergencies such as road congestion and traffic accidents, and real-time weather data is used to simulate the impact of weather changes on road conditions and vehicle driving. The simulation scenario is simulated using a time step of 1 second.

[0197] Simulate the operation of virtual vehicles in the simulation scenario. Virtual vehicles travel according to the route and delivery sequence set in the preliminary transportation scheduling plan. At each time step, the speed and route of the virtual vehicles are dynamically adjusted according to real-time traffic flow data and weather data. For example, if there is traffic congestion on the road ahead, the virtual vehicle will automatically choose a detour route. Record the actual arrival time of the virtual vehicle at each delivery point, the completion status of each order (whether it was delivered on time), and the total transportation time, total transportation cost, and total transportation risk during the entire transportation process. After the simulation run is completed, the recorded data is output as the preliminary scheduling simulation results.

[0198] According to the preliminary scheduling simulation results, the performance indicators of the preliminary transportation scheduling plan are calculated to generate a preliminary scheduling plan evaluation report. The performance indicators include: total transportation time, the total time for all vehicles to complete all tasks; total transportation cost, calculated based on the mileage of each vehicle and the unit mileage cost; total transportation risk, calculated based on the weighted risk level of the road section passed by each vehicle; order fulfillment rate, the proportion of the number of orders delivered on time to the total number of orders; and average delay time, the average delay time of all delayed orders. The preliminary scheduling plan evaluation report contains the specific values ​​of each performance indicator.

[0199] Use deep reinforcement learning algorithms (such as DeepQ-Network, DQN) to dynamically adjust the preliminary transportation scheduling plan. The state space of the reinforcement learning agent includes the current vehicle location, order status, traffic conditions and weather conditions; the action space is the next destination selection of the vehicle (such as selecting the next delivery point or selecting an alternative route); the reward function is a weighted combination of order fulfillment rate, total transportation time, total transportation cost and total transportation risk, and the weights are set according to actual business needs. Through interaction with the simulation environment, the reinforcement learning agent learns to select the best action under different states to maximize the reward. The DQN algorithm uses an experience replay mechanism and a target network to improve training stability. After training, the trained DQN model is applied to the preliminary transportation scheduling plan, the best action is selected according to the current state, and the optimized and adjusted scheduling plan is generated, including the adjusted driving route and order delivery sequence of each vehicle.

[0200] The optimized and adjusted scheduling plan generated in step S454 is converted into an executable transportation scheduling plan. Based on the optimized vehicle driving routes and order delivery sequence, combined with the road network information and real-time traffic flow data in the digital twin logistics base, the estimated arrival time of each vehicle at each delivery point is recalculated. The optimized transportation scheduling plan finally generated includes the driving route of each vehicle, the order delivery sequence, and the estimated arrival time at each delivery point.

[0201] Preferably, step S5 comprises the following steps:

[0202] Step S51: collecting transport execution data according to the optimal transport scheduling plan to obtain real-time transport execution data;

[0203] Step S52: Detect and extract operation deviations of the transport execution real-time data and the optimal transport scheduling plan to obtain operation deviation information;

[0204] Step S53: Analyze and classify the causes of the deviation according to the operation deviation information to obtain the classification result of the causes of the deviation;

[0205] Step S54: performing feedback update and digital twin model iteration according to the deviation cause classification result and the digital twin flow base to obtain the iteratively optimized digital twin base;

[0206] Step S55: Adjust and optimize the transportation plan of the optimal transportation scheduling solution based on the iteratively optimized digital twin base to obtain real-time operation deviation feedback data.

[0207] In an embodiment of the present invention, transport execution data is collected in real time through a vehicle-mounted GPS device, an IoT sensor, and a logistics management system (TMS). The vehicle-mounted GPS device uploads information such as the real-time location, speed, and direction of the vehicle every 10 seconds. The IoT sensor monitors the status of the goods in real time, such as temperature, humidity, vibration, etc., and uploads the data to the cloud server. The logistics management system (TMS) records changes in order status, such as order receipt time, order delivery start time, order completion time, etc. The collected multi-source data is integrated to generate real-time transport execution data, including the real-time location trajectory of each vehicle, cargo status information, and order status information. The data is stored in the time series database InfluxDB to facilitate real-time query and analysis.

[0208] Compare the real-time data of transportation execution with the optimal transportation scheduling plan, detect and extract operation deviation information. Operation deviation includes time deviation, route deviation and cargo status deviation. Time deviation calculation method: Compare the actual arrival time of the vehicle at each distribution point with the estimated arrival time set in the optimal transportation scheduling plan, and calculate the time difference. Route deviation calculation method: Compare the actual driving trajectory of the vehicle with the driving route planned in the optimal transportation scheduling plan, and calculate the trajectory deviation distance. Cargo status deviation calculation method: Compare the cargo status data collected by the sensor with the preset safety threshold to determine whether the cargo status is abnormal. The detected deviation information is recorded in the operation deviation information, including the deviation type, deviation value, occurrence time, and related vehicle and order information.

[0209] According to the operation deviation information and the real-time traffic data, weather data and other information in the digital twin logistics base (TLB), the causes of the deviation are analyzed and classified. For example, if the vehicle is delayed, analyze whether the delay is caused by traffic congestion, traffic accidents or other reasons. The analysis method includes: querying the real-time traffic event data in the TLB, such as traffic accidents, road construction, etc.; analyzing the historical traffic flow data of the vehicle driving section to determine whether congestion often occurs; querying the real-time weather data in the TLB to determine whether it is affected by bad weather. The causes of the deviation are classified into categories such as traffic factors, weather factors, equipment failures, and human factors. The deviation cause classification results include the deviation type, the deviation cause category, and a detailed analysis description.

[0210] The classification results of the deviation causes are fed back to the digital twin logistics base (TLB) to update and optimize the model parameters in the TLB. For example, if a certain section of road is often congested and causes vehicle delays, the weight of the section in the path planning algorithm is increased to reduce the probability of it being selected. If a certain type of cargo is prone to damage during transportation, the cargo transportation safety specifications are updated, such as adjusting the temperature control range or adding buffering materials. Through continuous iterative optimization, the prediction accuracy and simulation effect of the TLB are improved.

[0211] The optimal transportation scheduling plan is dynamically adjusted and optimized based on the iteratively optimized digital twin logistics base (TLB) and the current transportation execution status. For example, if a vehicle is delayed, the delivery order and estimated arrival time of the remaining orders are recalculated and the relevant drivers and customers are notified. If the cargo status is abnormal, the emergency plan is activated, such as changing the delivery route or contacting the nearest maintenance point. The adjusted transportation plan and optimization suggestions are output as real-time operation deviation feedback data to guide actual transportation operations and provide reference for subsequent scheduling optimization.

[0212] Preferably, the present invention also provides a logistics transportation route optimization system based on digital twins, which is used to execute the logistics transportation route optimization method based on digital twins as described above. The logistics transportation route optimization system based on digital twins includes:

[0213] The digital twin model initialization module is used to obtain multi-source logistics and transportation data, and to perform dynamic logistics network modeling to obtain a dynamic logistics network model; to perform three-dimensional modeling of transportation vehicles based on the dynamic logistics network model and multi-source logistics and transportation data, and to perform vehicle capacity analysis to obtain a vehicle digital twin; to initialize the logistics digital twin model based on the dynamic logistics network model and the vehicle digital twin to obtain a digital twin logistics base;

[0214] Dynamic transportation demand prediction extracts the logistics distribution characteristics of each region from the digital twin logistics base to obtain regional characteristic data; performs dynamic time series prediction modeling based on the regional transportation characteristic model to obtain the predicted regional transportation demand; generates dynamic transportation distribution based on the predicted regional transportation demand to obtain the dynamic transportation demand distribution;

[0215] The real-time line state evaluation module is used to model the line initial state according to the dynamic transportation demand distribution to obtain the line initial state model; perform real-time line state evaluation based on the physical characteristics of the vehicle according to the line initial state model, and generate a dynamic line performance matrix to obtain the dynamic line performance matrix;

[0216] The transport scheduling optimization module is used to allocate vehicle resources based on the order task priority according to the dynamic line performance matrix to obtain the vehicle allocation plan; design the collaborative optimization algorithm according to the vehicle allocation plan and the dynamic line performance matrix to obtain the collaborative optimization target; perform multi-objective solution on the collaborative optimization target to obtain the preliminary transport scheduling plan; input the preliminary transport scheduling plan into the digital twin environment for dynamic scheduling adjustment, and perform global integration of the plan to obtain the optimal transport scheduling plan;

[0217] The real-time feedback and iterative optimization module is used to analyze the transportation execution deviation according to the optimal transportation scheduling plan and obtain the classification results of the deviation causes; iteratively optimize the digital twin model according to the classification results of the deviation causes, and adjust and optimize the transportation plan of the optimal transportation scheduling plan to obtain real-time operation deviation feedback data.

[0218] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0219] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A logistics transportation route optimization method based on digital twin, characterized in that: The following steps are involved: Step S1: Acquire multi-source logistics and transportation data, and perform dynamic logistics network modeling to obtain a dynamic logistics network model; perform three-dimensional modeling of transportation vehicles based on the dynamic logistics network model and multi-source logistics and transportation data, and perform vehicle capacity analysis to obtain a vehicle digital twin; The logistics digital twin model is initialized according to the dynamic logistics network model and the vehicle digital twin to obtain the digital twin logistics base; Step S2: extracting the logistics distribution characteristics of each region from the digital twin logistics base to obtain regional characteristic data; Based on the regional transportation characteristic model, dynamic time series forecasting modeling is performed to obtain the predicted regional transportation demand; based on the predicted regional transportation demand, dynamic transportation distribution is generated to obtain the dynamic transportation demand distribution; Step S3: Modeling the initial state of the line according to the dynamic transportation demand distribution to obtain the initial state model of the line; performing a real-time line state evaluation based on the physical characteristics of the vehicle according to the initial state model of the line, and generating a dynamic line performance matrix to obtain a dynamic line performance matrix; Step S4: performing vehicle resource allocation based on order task priority according to the dynamic line performance matrix to obtain a vehicle allocation plan; Design collaborative optimization algorithm based on vehicle allocation scheme and dynamic line performance matrix to obtain collaborative optimization target; Perform multi-objective solution for collaborative optimization objectives and obtain a preliminary transportation scheduling plan; Input the preliminary transportation scheduling plan into the digital twin environment for dynamic scheduling adjustment, and perform global integration of the plan to obtain the optimal transportation scheduling plan; Step S5: Perform transport execution deviation analysis based on the optimal transport scheduling plan to obtain a classification result of the cause of the deviation; perform iterative optimization of the digital twin model based on the classification result of the cause of the deviation, and adjust and optimize the transport plan for the optimal transport scheduling plan to obtain real-time operation deviation feedback data.

2. The method for optimizing logistics transportation routes based on digital twins according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: Acquire multi-source logistics and transportation data through a multi-source heterogeneous data fusion interface to obtain an initial data set; Step S12: performing data cleaning and standardization processing on the initial data set to obtain standardized logistics data; constructing a vehicle three-dimensional model on the standardized logistics data to obtain a transport vehicle three-dimensional model; Step S13: Perform dynamic logistics network modeling based on standardized logistics data to obtain a dynamic logistics network model; perform capacity parameter calculation based on the dynamic logistics network model and the three-dimensional model of the transport vehicle to obtain the transport vehicle capacity parameter; Step S14: Generate a digital twin environment according to the dynamic logistics network model to obtain a digital twin initial environment; generate a vehicle digital twin according to the transport vehicle capacity parameters to obtain a vehicle digital twin; Step S15: Perform real-time dynamic optimization and base generation based on the digital twin initial environment and the vehicle digital twin to obtain the digital twin flow base.

3. The method for optimizing logistics transportation routes based on digital twins according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: partitioning the digital twin logistics base into regional logistics data to obtain partitioned logistics data; Step S22: performing intra-regional data clustering analysis on the partitioned logistics data to obtain regional logistics distribution; Step S23: extracting regional features of regional logistics distribution to obtain regional feature data; Step S24: performing regional transportation characteristic modeling according to the regional characteristic data to obtain a regional transportation characteristic model; Step S25: Perform dynamic time series forecasting modeling according to the regional transportation characteristic model to obtain the predicted regional transportation demand; Step S26: dynamically modify the predicted regional transportation demand to obtain a modified regional transportation demand; Step S27: Generate dynamic transportation distribution according to the modified regional transportation demand to obtain dynamic transportation demand distribution.

4. The method for optimizing logistics transportation routes based on digital twins according to claim 3 is characterized in that: Step S25 includes the following steps: Step S251: extracting the time series of order quantities in each region from the regional transportation characteristic model to obtain the time series data of order quantities in each region; Step S252: constructing a regional demand time series forecasting model based on the time series data of the order volume in each region to obtain a regional demand time series forecasting model set; Step S253: constructing a regional interaction relationship diagram according to the regional logistics distribution to obtain a dynamic regional interaction relationship diagram; Step S254: using the prediction results of the regional time series prediction model set as the initial features of the corresponding nodes in the dynamic regional interaction relationship graph, and performing graph neural network modeling that integrates the time series features to obtain the regional interaction enhanced prediction demand; Step S255: Integrate the forecast results and calculate the indicators of the regional interactive enhanced forecast demand to obtain the forecast regional transportation demand.

5. The method for optimizing logistics transportation routes based on digital twins according to claim 1 is characterized in that: Step S3 includes the following steps: Step S31: Analyze the physical characteristics of the vehicle according to the dynamic transportation demand distribution and the vehicle digital twin to obtain vehicle dynamics data, vehicle energy consumption data, and vehicle safety data; extract the initial state of the route according to the dynamic transportation demand distribution and the vehicle digital twin, and perform route initial state modeling to obtain a route initial state model; Step S32: performing a line timeliness evaluation according to the line initial state model and the vehicle dynamics data to obtain a line timeliness evaluation result; Step S33: performing a line cost assessment based on the line timeliness assessment result and the vehicle energy consumption data to obtain a line cost assessment result; Step S34: performing a route risk assessment based on the route cost assessment result and the vehicle safety data to obtain a route risk assessment result; Step S35: Generate a dynamic line performance matrix according to the line timeliness evaluation result, the line cost evaluation result and the line risk evaluation result to obtain a dynamic line performance matrix.

6. The method for optimizing logistics transportation routes based on digital twins according to claim 1 is characterized in that: Step S4 includes the following steps: Step S41: performing order task analysis according to the dynamic transportation demand distribution and the dynamic line performance matrix to obtain order task data; Step S42: assigning demand priorities to the order task data to obtain a demand priority task list; Step S43: Allocate vehicle resources according to the demand priority task list to obtain a vehicle allocation plan; Step S44: designing a collaborative optimization algorithm according to the vehicle allocation plan and the dynamic line performance matrix to obtain a collaborative optimization target; performing multi-objective solution on the collaborative optimization target to obtain a preliminary transportation scheduling plan; Step S45: inputting the preliminary transport scheduling plan into the digital twin environment for dynamic scheduling adjustment to obtain an optimized transport scheduling plan; Step S46: globally integrate the optimized transport scheduling solutions to obtain the optimal transport scheduling solution.

7. The method for optimizing logistics transportation routes based on digital twins according to claim 6 is characterized in that: Step S44 includes the following steps: Step S441: constructing a multi-objective optimization problem instance according to the vehicle allocation scheme to obtain a multi-objective optimization problem instance; Step S442: Initializing the multi-objective genetic algorithm population for the multi-objective optimization problem instance to obtain an initial scheduling solution population; Step S443: Designing a dynamic weighted fitness function for the initial scheduling solution population to obtain a scheduling solution population with a fitness score; Step S444: performing genetic operations and population update execution according to the scheduling scheme population with fitness scores to obtain a new generation of scheduling scheme population; Step S445: performing population update iteration according to the new generation scheduling plan population to obtain the last generation scheduling plan population; generating a preliminary transport scheduling plan for the last generation scheduling plan population to obtain a preliminary transport scheduling plan.

8. The method for optimizing logistics transportation routes based on digital twins according to claim 6 is characterized in that: Step S45 includes the following steps: Step S451: construct a preliminary transportation scheduling simulation scenario according to the preliminary transportation scheduling plan and the digital twin logistics base to obtain a preliminary scheduling simulation scenario; Step S452: performing virtual vehicle driving simulation on the preliminary scheduling simulation scene to obtain preliminary scheduling simulation results; Step S453: performing a preliminary scheduling scheme performance evaluation on the preliminary transportation scheduling scheme according to the preliminary scheduling simulation result to obtain a preliminary scheduling scheme evaluation report; Step S454: According to the preliminary scheduling plan evaluation report, the preliminary scheduling plan is dynamically adjusted based on reinforcement learning to obtain an optimized and adjusted scheduling plan; Step S455: Generate an optimized transport scheduling plan based on the optimized and adjusted scheduling plan to obtain an optimized transport scheduling plan.

9. The method for optimizing logistics transportation routes based on digital twins according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: collecting transport execution data according to the optimal transport scheduling plan to obtain real-time transport execution data; Step S52: Detect and extract operation deviations of the real-time transport execution data and the optimal transport scheduling plan to obtain operation deviation information; Step S53: Analyze and classify the causes of the deviation according to the operation deviation information to obtain the classification result of the causes of the deviation; Step S54: performing feedback update and digital twin model iteration according to the deviation cause classification result and the digital twin flow base to obtain the iteratively optimized digital twin base; Step S55: Adjust and optimize the transportation plan of the optimal transportation scheduling solution based on the iteratively optimized digital twin base to obtain real-time operation deviation feedback data.

10. A logistics transportation route optimization system based on digital twins, characterized in that: Used to execute the logistics transportation route optimization method based on digital twin as claimed in claim 1, the logistics transportation route optimization system based on digital twin includes: The digital twin model initialization module is used to obtain multi-source logistics and transportation data, and to perform dynamic logistics network modeling to obtain a dynamic logistics network model; to perform three-dimensional modeling of transportation vehicles based on the dynamic logistics network model and multi-source logistics and transportation data, and to perform vehicle capacity analysis to obtain a vehicle digital twin; to initialize the logistics digital twin model based on the dynamic logistics network model and the vehicle digital twin to obtain a digital twin logistics base; Dynamic transportation demand prediction extracts the logistics distribution characteristics of each region from the digital twin logistics base to obtain regional characteristic data; performs dynamic time series prediction modeling based on the regional transportation characteristic model to obtain the predicted regional transportation demand; generates dynamic transportation distribution based on the predicted regional transportation demand to obtain the dynamic transportation demand distribution; The real-time line state evaluation module is used to model the line initial state according to the dynamic transportation demand distribution to obtain the line initial state model; perform real-time line state evaluation based on the physical characteristics of the vehicle according to the line initial state model, and generate a dynamic line performance matrix to obtain the dynamic line performance matrix; The transport scheduling optimization module is used to allocate vehicle resources based on the order task priority according to the dynamic line performance matrix to obtain the vehicle allocation plan; design the collaborative optimization algorithm according to the vehicle allocation plan and the dynamic line performance matrix to obtain the collaborative optimization target; perform multi-objective solution on the collaborative optimization target to obtain the preliminary transport scheduling plan; input the preliminary transport scheduling plan into the digital twin environment for dynamic scheduling adjustment, and perform global integration of the plan to obtain the optimal transport scheduling plan; The real-time feedback and iterative optimization module is used to analyze the transportation execution deviation according to the optimal transportation scheduling plan and obtain the classification results of the deviation causes; iteratively optimize the digital twin model according to the classification results of the deviation causes, and adjust and optimize the transportation plan of the optimal transportation scheduling plan to obtain real-time operation deviation feedback data.

Images