Intelligent vehicle dispatching system for logistics transportation

Abstract

The invention discloses an intelligent vehicle dispatching system for logistics transportation, and relates to the technical field of logistics transportation vehicle dispatching. Comprising a data acquisition and preprocessing module, a feature extraction and analysis module, a task classification and complexity evaluation module, a simple task solving module and a complex task dynamic solving module, and the data acquisition and preprocessing module firstly acquires real-time data information related to a current scheduling task and preprocesses the acquired data information; by introducing real-time data preprocessing, feature extraction and intelligent evaluation, the system can dynamically identify task complexity and reasonably distribute solving time, simple tasks can be rapidly solved, solving time of complex tasks can be prolonged, stable operation of the system is ensured, and logistics efficiency and service quality are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics transportation vehicle dispatching, and in particular to an intelligent vehicle dispatching system for logistics transportation. Background Art

[0002] The application of intelligent vehicle dispatching in logistics and transportation uses advanced technologies such as the Internet of Things, artificial intelligence, and big data analysis to monitor and optimize the management of logistics vehicles in real time to improve transportation efficiency, reduce costs, and ensure safety. Unlike traditional static dispatching methods that rely on manual or simple algorithms, intelligent systems can dynamically obtain and analyze information such as vehicles, goods, and road conditions, and automatically adjust dispatching strategies to achieve flexible and accurate resource allocation. For example, the system can obtain road condition data in real time, and intelligently plan the best route and dispatching plan based on the timeliness requirements of goods, the location and load of vehicles to maximize transportation efficiency and reduce empty driving rate and transportation costs. In addition, the intelligent dispatching system also continuously monitors and predicts the transportation process through big data analysis and machine learning, identifies potential risks in advance, and takes preventive measures. Combined with external information such as weather forecasts, traffic flow, and accident data, the system can dynamically adjust transportation plans, respond to emergencies, and optimize transportation timeliness. With the development of technology, intelligent vehicle dispatching can also be combined with unmanned driving technology to further improve the level of automation, reduce human intervention, and move towards full-process intelligent management. This technology not only improves logistics efficiency and reduces enterprise operating costs, but also improves customer satisfaction, and has broad application prospects and development potential.

[0003] Intelligent vehicle dispatching systems usually use optimization algorithms for dispatching. The role of optimization algorithms in dispatching is mainly to automatically find the optimal or approximately optimal dispatching scheme through mathematical models and calculation methods to improve transportation efficiency, reduce costs and ensure the rational allocation of resources. In complex logistics transportation, optimization algorithms can comprehensively consider multiple factors, such as vehicle load conditions, transportation routes, traffic conditions, cargo timeliness requirements, road restrictions, etc., and find the best dispatching strategy on the basis of satisfying all constraints. For example, commonly used optimization algorithms such as linear programming, integer programming, genetic algorithms, simulated annealing, etc. can help the system quickly calculate the optimal path and task allocation scheme, reduce empty driving rate, shorten transportation time and reduce fuel consumption, thereby greatly improving transportation efficiency and cost-effectiveness. In addition, optimization algorithms can also be dynamically adjusted according to real-time data, respond to emergencies (such as traffic accidents, weather changes, etc.) in a timely manner, and further improve the flexibility and adaptability of dispatching.

[0004] The prior art has the following deficiencies:

[0005] In practical applications, when using optimization algorithms for vehicle scheduling, in order to ensure efficient scheduling, the solution time of the optimization algorithm is usually limited. However, when the problem involves a large number of variables and constraints, especially multi-objective optimization problems, the solution process usually takes longer. In this case, especially in large-scale logistics scheduling, the limited solution time makes it impossible for the algorithm to fully explore the huge solution space. This results in the optimization algorithm often only being able to find a "good enough" solution within a limited time, and unable to conduct an extensive global search, which may cause it to stay near the local optimal solution and miss the global optimal solution.

[0006] When a system is stuck in a local optimal solution, although in the short term, the system seems to have optimized certain aspects (such as transportation time, cost, or certain specific resource allocations), the actual situation is that the system has ignored global constraints or potential optimization space. For example, the vehicle scheduling plan may improve the transportation efficiency of certain sections of the road in the short term, but in the long term, it may lead to some vehicles running idle, over-concentrated scheduling, or waste of certain key resources. In addition, the problem of local optimal solution is usually not easy to detect, and it often requires long-term and large-scale monitoring and feedback to show the impact on the overall system efficiency. This hidden nature means that once a problem occurs, it may lead to inefficiency and instability of the transportation network, and may even cause the collapse of the entire logistics system or a decline in service quality.

[0007] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0008] The purpose of the present invention is to provide an intelligent vehicle dispatching system for logistics transportation. By introducing real-time data preprocessing, feature extraction and intelligent evaluation, the system can dynamically identify the complexity of tasks and reasonably allocate solution time, quickly solve simple tasks to avoid resource waste; flexibly extend the solution time for complex tasks to increase the chance of finding the global optimal solution. This intelligent dispatching scheme is dynamically adjusted based on task complexity and resource availability to ensure the stable operation of the system in a complex environment and avoid the negative impact of local optimal solutions, thereby improving the overall efficiency and long-term service quality of the logistics system to solve the problems in the above-mentioned background technology.

[0009] In order to achieve the above-mentioned purpose, the present invention provides the following technical solutions: an intelligent vehicle dispatching system for logistics transportation, comprising a data acquisition and preprocessing module, a feature extraction and analysis module, a task classification and complexity evaluation module, a simple task solving module and a complex task dynamic solving module:

[0010] The data acquisition and preprocessing module, first, obtains the real-time data information related to the current scheduling task and preprocesses the acquired data information;

[0011] The feature extraction and analysis module extracts features from the pre-processed scheduling task data information. After extraction, the extracted features are analyzed and processed under the detection window, and the analyzed feature information is input into the pre-trained machine learning model to perform intelligent evaluation of the current scheduling task;

[0012] The task classification and complexity evaluation module divides the current scheduling tasks into complex tasks and simple tasks based on the evaluation results of the machine learning model;

[0013] Simple task solving module, for simple tasks, sets a standard solving time based on task characteristics and performs optimal solution within this time range;

[0014] The complex task dynamic solution module targets complex tasks, which involve a large number of variables and constraints. It takes the standard solution time as the basis and dynamically adjusts the solution time according to the task complexity to find the global optimal solution.

[0015] Preferably, feature extraction is performed on the preprocessed scheduling task data information, and the extracted features include the topological structure complexity in the task scheduling process and the degree of interaction between different constraints in the scheduling process. After extraction, the topological structure complexity in the task scheduling process and the degree of interaction between different constraints in the scheduling process are analyzed and processed under the detection window to generate a scheduling topology complexity index and a dynamic constraint interaction index. The scheduling topology complexity index quantifies the complexity of the dependencies, constraints and execution order between each task node in the scheduling task, and the dynamic constraint interaction index quantifies the degree of interaction and influence between different constraints in the task scheduling process.

[0016] Preferably, after obtaining the scheduling topology complexity index and the dynamic constraint interaction index generated after analyzing the extracted features, the scheduling topology complexity index and the dynamic constraint interaction index are input into a pre-learned machine learning model, and the task complexity coefficient is generated through the machine learning model, and the current task is intelligently evaluated through the task complexity coefficient.

[0017] Preferably, the task complexity coefficient generated after comprehensive analysis of the scheduling task information by the machine learning model is compared and analyzed with a preset task complexity coefficient reference threshold, and the current scheduling task is divided, and the division steps are as follows:

[0018] If the task complexity coefficient is greater than or equal to a preset task complexity coefficient reference threshold, the current scheduling task is classified as a complex task;

[0019] If the task complexity coefficient is less than a preset task complexity coefficient reference threshold, the current scheduled task is classified as a simple task.

[0020] Preferably, for complex tasks, based on the standard solution time, the specific steps of dynamically adjusting the solution time in combination with the task complexity to find the global optimal solution are as follows:

[0021] First, by comparing the task complexity coefficient TaCC with the task complexity coefficient reference threshold TaCC ref For comparison, the solution time is adjusted and the calculation expression is as follows:

[0022]

[0023] , where T solve is the adjusted complex task solving time, is the standard solution time based on simple task setting, and A is the adjustment factor;

[0024] After adjusting the solution time of complex tasks, the solution time is further corrected according to the current available resources and the urgency of the task to ensure that the adjusted solution time is reasonably allocated under resource constraints and time requirements. The calculation expression is as follows:

[0025]

[0026] , where T solveadjusted is the further adjusted complex task solving time, R avail is the currently available resource, R max is the maximum value of resources, E urgency is the urgency of the task, E max is the maximum urgency, D is the resource availability impact coefficient, and Z is the urgency impact coefficient;

[0027] Finally, according to the adjusted solution time, the optimization algorithm is used to solve the problem and check the optimality of the solution. If the current solution does not meet the requirements of the global optimal solution, the optimization is continued until the global optimal solution is found, which is expressed as follows:

[0028]

[0029] , where OptimalityCheck is the judgment condition, which is judged based on whether the solution has reached the global optimal solution. Optimal is the global optimal solution, and uboptimal is the solution that has not reached the global optimal solution.

[0030] If the solution reaches the global optimal solution, the result of OptimalityCheck is the global optimal solution Optimal, which means that the current solution has the best effect among all solutions and does not need further optimization;

[0031] If the solution does not reach the global optimal solution, the result of OptimalityCheck is that the global optimal solution Suboptimal is not reached, which means that the current solution is only a local optimal solution or a suboptimal solution, and there is a better solution. In this case, continue to try to optimize to find the global optimal solution Optimal.

[0032] Preferably, in the detection window, the topological complexity of the task scheduling process is analyzed, and the specific steps of generating the scheduling topological complexity index are as follows:

[0033] In the detection window, we first define the dependencies between tasks in the task scheduling process, and assume that the set of task nodes in the scheduling task is T = {t i}={t 1 ,t 2 ,t 3 ,……,t n}, where ti represents the ith independent scheduling task, n is the total number of tasks, and each task t i It has dependencies with other tasks, represented by a directed graph G = (V, E) in graph theory, where V = T, Represents a set of dependencies. If task t i Must be in task t j Before execution, there is an edge (t i , t j )∈E;

[0034] The dependency between tasks is reflected by the connectivity and hierarchy of the graph. The in-degree and out-degree are used to quantify the dependency of each task. The calculation expression is as follows:

[0035]

[0036] , where Indegree(t i ) is the task t i The in-degree of task t i The number of dependent tasks, t j is the jth independent scheduling task, ||(t j →t i ) is an indicator function. If there exists a j To task t i The dependency relationship between ||(t j →t i)=1, otherwise 0,

[0037]

[0038] , where Outdegree(t i ) is the task t i The out-degree of task t i The number of other tasks it depends on;

[0039] The topological complexity scoring function is introduced to reflect the complexity of each task through the in-degree, out-degree and the dependencies between tasks. i The complexity index is calculated as follows:

[0040]

[0041] , where C(t i ) is the task t i The complexity index is α, which is the exponential weight of in-degree and out-degree, N i It is the task i The set of adjacent tasks, w ij is the weight of the dependency, and β is the exponent of the dependency weight;

[0042] The weighted average of the complexity indexes of all tasks is used to comprehensively measure the topological complexity of the entire scheduling task and generate the scheduling topology complexity index. The calculation expression is as follows:

[0043]

[0044] , where STopoCI is the scheduling topology complexity index, γ is an adjustable weight factor used to adjust the influence of a single task complexity on the overall topology complexity, and Z is a normalization constant used to ensure that the scale of the index adapts to the entire task scheduling system.

[0045] Preferably, in the detection window, the interaction degree between different constraint conditions in the scheduling process is analyzed, and the specific steps of generating a dynamic constraint interaction index are as follows:

[0046] Under the detection window, first establish a constraint matrix M, M = {C s}={C 1 , C 2 , C 3 ,……,C m}, the constraint matrix M is used to represent the relationship between the various constraints in the scheduling task, C s is the sth constraint in the scheduling task, m is the total number of constraints in the scheduling task, and an m·m rectangle is constructed. The expression is as follows:

[0047]

[0048] , where the rows and columns of the matrix correspond to different constraints, and each matrix element M sx Represents the sth constraint condition C s and the xth constraint C x The intensity of interaction between

[0049] In order to quantify the coupling strength of each constraint in the overall task, the constraint coupling coefficient is defined and the calculation expression is as follows:

[0050]

[0051] , where K s is the constraint coupling coefficient, i.e., constraint C s The coupling coefficient, f(C s , C x ) is the weight function;

[0052] For each constraint C in the scheduling task s , define its constraint concentration, and the calculation expression is as follows:

[0053]

[0054] , where D s is the constraint concentration, that is, constraint C s The concentration, K x Yes Yes Constraint C x The coupling coefficient, ω is the smoothing coefficient;

[0055] In order to measure the mutual dependence between two constraints, the constraint mutual dependence coefficient is defined, and the calculation expression is as follows:

[0056] L sx =M sx ·(1+ln(1+|D s -D x |))

[0057] , where L sx is the constraint interdependence coefficient, i.e. constraint D s and constraint D x The mutual dependence coefficient between them;

[0058] The interdependence coefficients of all constraints are combined to generate the dynamic constraint interaction index. The calculation expression is as follows:

[0059]

[0060] , where DynConI is the dynamic constraint interaction index, g(Ks ,K x ) is a nonlinear function used to adjust the weights between different constraint coupling coefficients.

[0061] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0062] By introducing real-time data preprocessing, feature extraction and intelligent evaluation, the invention enables the system to dynamically perceive the complexity of the task and reasonably allocate the solution time. For simple tasks, the system can quickly obtain the optimal solution within the standard solution time to avoid wasting computing resources. For complex tasks, the system can flexibly extend the solution time based on the task complexity coefficient, providing more computing resources for the optimization algorithm to conduct more extensive exploration in the huge solution space, thereby having a better chance of finding the global optimal solution. This solution time allocation strategy based on task difficulty enables the system to reasonably allocate resources, which not only improves the overall scheduling efficiency, but also reduces resource waste and avoids suboptimal solutions or local optimal solutions due to insufficient standard time.

[0063] By combining the intelligent analysis of task complexity coefficients and machine learning models in task division and solution time allocation, the system can identify the long-term optimization needs of tasks, rather than just focusing on short-term efficiency. This intelligent scheduling solution can make dynamic adjustments based on the complexity of tasks and resource availability, ensuring that the system can operate stably even in complex environments and reducing potential problems caused by local optimal solutions. As the system continues to run, the complexity coefficients and other parameters of the tasks will also be optimized and updated through machine learning models, enhancing the system's adaptability to complex tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0065] Figure 1 The module diagram of an intelligent vehicle dispatching system for logistics transportation of the present invention is shown in FIG. DETAILED DESCRIPTION

[0066] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0067] The present invention provides Figure 1 An intelligent vehicle dispatching system for logistics transportation shown in the figure includes a data acquisition and preprocessing module, a feature extraction and analysis module, a task classification and complexity evaluation module, a simple task solving module and a complex task dynamic solving module:

[0068] The data acquisition and preprocessing module, first, obtains the real-time data information related to the current scheduling task and preprocesses the acquired data information;

[0069] Preprocessing is a crucial step in the data processing process. Its main purpose is to convert the original real-time data information into clean, standardized and consistent data that can be used for subsequent analysis and modeling. In the vehicle dispatch system, preprocessing includes multiple steps. The first is data cleaning to remove duplicate, erroneous or irrelevant data information to ensure the accuracy of the data. Next, missing value processing is performed, such as filling missing data caused by sensor failure or data transmission problems through interpolation, mean filling or other methods. Then, the formats of different data sources are unified, and data from different systems such as GPS, sensors, and traffic flow monitoring are standardized to ensure that they can be processed uniformly on the same platform. In addition, outlier detection is also required to identify abnormal data caused by measurement errors or abnormal events (such as traffic accidents or equipment failures), and to correct or delete them. The ultimate goal of preprocessing is to eliminate noise, ensure data quality, and provide an accurate, complete and unified data foundation for subsequent feature extraction, model training and decision analysis, so as to improve the accuracy and efficiency of system decision-making.

[0070] The feature extraction and analysis module extracts features from the pre-processed scheduling task data information. After extraction, the extracted features are analyzed and processed under the detection window, and the analyzed feature information is input into the pre-trained machine learning model to perform intelligent evaluation of the current scheduling task;

[0071] Feature extraction is performed on the preprocessed scheduling task data information. The extracted features include the topological complexity of the task scheduling process and the degree of interaction between different constraints in the scheduling process. After extraction, the topological complexity of the task scheduling process and the degree of interaction between different constraints in the scheduling process are analyzed and processed under the detection window to generate a scheduling topological complexity index and a dynamic constraint interaction index. The scheduling topological complexity index quantifies the complexity of the dependencies, constraints and execution order between each task node in the scheduling task. The dynamic constraint interaction index quantifies the degree of interaction and influence between different constraints in the task scheduling process.

[0072] After obtaining the scheduling topology complexity index and dynamic constraint interaction index generated by analyzing the extracted features, the scheduling topology complexity index and dynamic constraint interaction index are input into the pre-learned machine learning model, and the task complexity coefficient is generated by the machine learning model. The current task is intelligently evaluated by the task complexity coefficient;

[0073] A pre-trained machine learning model refers to a model that is trained using historical data and known scheduling tasks during the system development phase. The learned model can identify and predict the complexity of scheduling tasks and the relationship between related features. Specifically, machine learning models are usually trained using a large amount of labeled data (for example, the scheduling features of different tasks and the corresponding task complexity labels). By learning the patterns and rules in these data, the model gradually "masters" how to evaluate the complexity of a scheduling task based on input features (such as the scheduling topology complexity index and the dynamic constraint interaction index). Common training methods include classification and regression algorithms in supervised learning, such as decision trees, support vector machines (SVMs), neural networks, etc. In this way, the model can learn to predict the complexity of tasks based on changes in feature parameters (such as topological structures, constraint interactions, etc.), thereby providing guidance for subsequent scheduling optimization.

[0074] The trained machine learning model can not only handle a large amount of input data, but also cope with new and unknown task types or complexity patterns. For example, some scheduling tasks may involve new constraints or changing scenarios, and the pre-trained model can accurately evaluate these new tasks based on the knowledge it has learned. The core advantage of the model is that it can extract implicit information and patterns from historical data, which are usually not easily captured by manual rules. Therefore, the model can not only evaluate the immediate complexity of the task, but also dynamically adjust the evaluation criteria according to the changes in the task. By inputting the scheduling topology complexity index and the dynamic constraint interaction index into the pre-trained machine learning model, the system can derive the complexity coefficient of the task according to the different characteristics of the task, thereby performing an intelligent evaluation of the current scheduling task and providing a more accurate basis for subsequent scheduling decisions and solutions. This intelligent evaluation can help the optimization algorithm better understand the actual difficulty of the task, and then dynamically adjust the solution strategy, thereby improving the scheduling efficiency and the quality of the optimization results.

[0075] The linear increase in topological complexity during task scheduling usually indicates that the current scheduling task involves a large number of variables and constraints. When the complexity of the dependencies, constraints, and execution order between task nodes in task scheduling increases, the topological complexity of the scheduling task will also increase accordingly. If the topological complexity increases rapidly, it means that the dependencies between tasks become closer and more complex, and the scheduling variables and constraints involved also increase significantly. For example, factors such as interdependencies between multiple tasks, resource conflicts, and time window constraints may cause the system to consider the scheduling order and resource allocation of multiple tasks at the same time. This complex topological structure needs to handle a large number of tasks, and the constraints are many and intertwined, which increases the difficulty of scheduling decisions. Highly complex topological structures often require scheduling algorithms to require more computing resources and time when finding the optimal solution, and may also cause the feasible space of the solution to increase dramatically, making the algorithm more likely to fall into a local optimal solution. Therefore, the linear increase in topological complexity is an important indicator of the increase in task complexity, reflecting the surge in variables and constraints in the scheduling task.

[0076] In the detection window, the topological complexity of the task scheduling process is analyzed, and the specific steps to generate the scheduling topological complexity index are as follows:

[0077] In the detection window, we first define the dependencies between tasks in the task scheduling process, and assume that the set of task nodes in the scheduling task is T = {t i}={t 1 ,t 2 ,t 3 ,……,t n}, where t i represents the ith independent scheduling task, n is the total number of tasks, and each task t i It has dependencies with other tasks, represented by a directed graph G = (V, E) in graph theory, where V = T, Represents a set of dependencies. If task t i Must be in task t j Before execution, there is an edge (t i , t j )∈E;

[0078] A directed graph in graph theory, also known as a directed network, is a graph structure consisting of nodes (or vertices) and directed edges (or arcs). In a directed graph, each edge consists of a starting node and a target node, and the direction of the edge is clear, indicating that it points from one node to another. In mathematical terms, assuming that the graph G is a directed graph, it consists of a node set V and an edge set E, where each edge in E can be represented as an ordered pair (u, v), where (u, v) represents a node, and u is the starting node and v is the target node. This directional representation enables the directed graph to accurately reflect the causal or dependency relationship between nodes.

[0079] Directed graphs play an important role in many practical applications, especially in representing the dependencies and interactions between elements in complex systems. For example, in the task scheduling problem, nodes can represent individual tasks, while directed edges represent the execution order or dependency between tasks. The completion of a task may depend on the completion of other tasks, thus forming a highly directional network structure in the directed graph. By analyzing directed graphs, we can understand the process of task execution, optimize resource allocation, and reduce conflicts and time waste. Therefore, directed graphs are widely used in computer science, network analysis, project management, logistics scheduling and other fields, and can efficiently handle directional relationships and constraints.

[0080] The dependency between tasks in the task scheduling process refers to the situation where other tasks must be completed before a certain task is completed. This dependency usually reflects the order of task execution, resource sharing, and mutual constraints between tasks. In the scheduling process, the dependencies between tasks can be divided into many types, including pre-dependency, post-dependency, and bidirectional dependency. Pre-dependency means that one task must be started after another task is completed (for example, task B can only be started after task A is completed), and post-dependency means that task B can only be started after task A is completed. Bidirectional dependency means that task A and task B are dependent on each other and must be completed alternately or collaboratively. In addition, dependencies may also involve resource dependencies (multiple tasks need to share the same resources) and time window dependencies (tasks must be completed within a specific time period). These dependencies require task scheduling to reasonably arrange the execution order and resource allocation of tasks on the premise of ensuring dependencies, so as to optimize scheduling efficiency and avoid conflicts.

[0081] The dependency between tasks is reflected by the connectivity and hierarchy of the graph. The in-degree and out-degree are used to quantify the dependency of each task. The calculation expression is as follows:

[0082]

[0083] , where Indegree(t i ) is the task t i The in-degree of task t i The number of dependent tasks, t j is the jth independent scheduling task, ||(t j →t i ) is an indicator function. If there exists a j To task t i The dependency relationship between ||(t j →t i )=1, otherwise it is 0. The higher the in-degree, the better the task t i The more dependent tasks are,

[0084]

[0085] , where Outdegree(t i ) is the task t i The out-degree of task t i The number of other tasks that it depends on. The higher the out-degree, the stronger the task t i The more other tasks are affected;

[0086] Connectivity measures the direct dependencies between tasks. In-degree and out-degree are used to indicate how many other tasks a task node depends on to start execution, and how many tasks depend on its execution results.

[0087] Hierarchy reflects the complexity of the order in which tasks are executed. Hierarchy becomes more complex as the number of dependencies between tasks increases.

[0088] Indegree(t i ) refers to the task t i How many tasks depend on it to execute, indicating that other tasks need to wait for t when executing. i The tasks with larger in-degree are usually the core nodes that they depend on, and their connectivity is stronger.

[0089] Outdegree(t i ) refers to the task t i How many tasks need to be completed before the task can start. A task with a large outdegree means that it depends on the execution of multiple other tasks before it can start, usually requiring more resources or the completion of other prerequisites.

[0090] Through in-degree and out-degree, we can quantify the connectivity and hierarchy of each task node, and further understand the dependency structure and execution order complexity between tasks.

[0091] The topological complexity scoring function is introduced to reflect the complexity of each task through the in-degree, out-degree and the dependencies between tasks. i The complexity index is calculated as follows:

[0092]

[0093] , where C(t i ) is the task t i The complexity index is α, which is the exponential weight of in-degree and out-degree, N i It is the task i The set of adjacent tasks, w ij is the weight of the dependency, and β is the exponent of the dependency weight;

[0094] The exponential weight α of in-degree and out-degree is a parameter used to control the influence of in-degree and out-degree in the task complexity index. In-degree indicates the number of predecessor tasks that a task node depends on, and out-degree indicates the scope of influence of a task node on subsequent tasks. By adjusting the value of α, the weight of in-degree and out-degree in calculating the task complexity index can be magnified or reduced. For example, when the α value is large, the system will pay more attention to the connectivity of tasks, that is, the dependency and influence of tasks, thereby increasing the complexity index of tasks with higher in-degree and out-degree, reflecting their importance in scheduling. Conversely, a smaller α value will reduce the influence of in-degree and out-degree, and instead let other factors dominate the complexity index.

[0095] The weight of the dependency relationship w ij Represents task t i Its adjacent task t j The weight of each dependency relationship is w ij It can be set according to the relationship attributes between specific tasks, such as the resource sharing requirements between tasks, the urgency of the time window limit, or the priority level. This weight can reflect the closeness of the dependence between tasks. The larger the weight, the closer the task is to the task. i and t j The stronger the dependency, the greater the impact on the scheduling result. The weight of the dependency can effectively help the system quantify the complex relationship between task nodes, so as to more accurately calculate the complexity of task nodes, making it easier for the scheduling algorithm to prioritize or tilt resources for highly dependent tasks.

[0096] The exponent β of the dependency weight is used to control the relative importance of all dependency weights in the complexity index. Index adjustment can amplify or reduce the impact of these weights in the overall complexity index calculation. A higher β value will increase the impact of dependency strength in the complexity index, making those task nodes with high dependencies have a higher complexity index, so that the system gives priority to the dependencies of these tasks. In complex scheduling tasks, using β can help the system flexibly adjust the impact of dependencies, ensuring that under different scheduling conditions, the task complexity index can reasonably reflect the dependency structure and actual needs of the task.

[0097] Task set N i Represents the node t i The set of directly dependent tasks, including only those related to t i The task set N is directly connected, but not the indirectly dependent tasks. This definition simplifies the calculation and ensures that the complexity index mainly reflects the direct dependencies of the tasks. If you need to include indirect dependencies, you can further expand the set of tasks N i , the influence range parameter is introduced to adjust the weight of indirect dependencies.

[0098] The weighted average of the complexity indexes of all tasks is used to comprehensively measure the topological complexity of the entire scheduling task and generate the scheduling topology complexity index. The calculation expression is as follows:

[0099]

[0100] , where STopoCI is the scheduling topology complexity index, γ is an adjustable weight factor used to adjust the influence of a single task complexity on the overall topology complexity, and Z is a normalization constant used to ensure that the scale of the index is adapted to the entire task scheduling system;

[0101] The normalization constant Z is a parameter used to standardize the complexity index, making the complexity index of different task sizes comparable. It can be set according to the number of tasks and the density of dependencies, ensuring that the value of the scheduling complexity index does not lose its representativeness as the number of tasks changes.

[0102] The adjustable weight factor γ is used to adjust the influence of the complexity index of a single task node on the overall complexity index STopoCI. By adjusting the value of γ, the weight of complex tasks in the overall index can be controlled. A larger γ value will increase the influence of complex tasks, while a smaller γ value will balance the contributions of different tasks, making the system's response to task complexity more balanced. According to actual needs, the value of γ can be flexibly adjusted to optimize the scheduling strategy.

[0103] Under the detection window, the larger the value of the scheduling topology complexity index generated after analyzing the topological complexity of the task scheduling process, the more complex the current scheduling task is, involving a large number of variables and constraints. The topological complexity index reflects the complexity of the dependencies and constraints between the various task nodes in task scheduling. When the value of this index is large, it indicates that the dependencies between tasks are strong and may involve multiple intertwined constraints, such as time window constraints, resource allocation constraints, task priorities, etc. This complex task structure requires more computing resources and time to coordinate the relationship between different tasks and find the optimal scheduling solution. On the contrary, when the index value is small, it means that the relationship between tasks in task scheduling is relatively simple, there are fewer dependencies and constraints, the space for scheduling decisions is small, and the task is relatively easy to complete.

[0104] When the degree of interaction between different constraints in the scheduling process is highly cross-linked, it usually indicates that the current scheduling task is complex and involves a large number of variables and constraints. This is because in large-scale scheduling problems, various constraints (such as time windows, vehicle capacity, road conditions, resource allocation, etc.) often do not exist independently, but are interrelated and influence each other. For example, the time window constraint of a task may conflict with the vehicle load limit, road capacity, and the scheduling time of other tasks. This requires the scheduling system to dynamically adjust and coordinate other related constraints while considering each constraint. When the interaction between these constraints becomes more complex and tight, the difficulty of solving the scheduling problem also increases, and it is necessary to deal with the balance of multiple objectives and constraints at the same time. At this time, the scheduling of tasks is not just a simple time optimization problem, but a multi-dimensional, multi-objective, and complex constraint optimization problem, involving a greatly increased number of decision variables and constraints. The highly cross-linked constraint relationship causes the algorithm to search in a wider solution space and explore more complex solutions, which makes the entire scheduling process more difficult and computationally demanding. Therefore, the strong interactivity of constraints often means that the complexity of the scheduling task also increases.

[0105] In the detection window, the interaction degree between different constraints in the scheduling process is analyzed, and the specific steps to generate the dynamic constraint interaction index are as follows:

[0106] Under the detection window, first establish a constraint matrix M, M = {C s}={C 1 , C 2 , C 3 ,……,C m}, the constraint matrix M is used to represent the relationship between the various constraints in the scheduling task, C sis the sth constraint in the scheduling task, and m is the total number of constraints in the scheduling task. That is, the scheduling task involves m different constraints. We construct an m·m rectangle, and the expression is as follows:

[0107]

[0108] , where the rows and columns of the matrix correspond to different constraints, and each matrix element M sx Represents the sth constraint condition C s and the xth constraint C x The intensity of interaction between

[0109] In order to quantify the coupling strength of each constraint in the overall task, the constraint coupling coefficient is defined and the calculation expression is as follows:

[0110]

[0111] , where K s is the constraint coupling coefficient, i.e., constraint C s The coupling coefficient, f(C s , C x ) is the weight function;

[0112] Coupling strength refers to the degree of interrelationship and mutual influence between different constraints. In a scheduling task, multiple constraints (such as time windows, vehicle capacity, road conditions, etc.) may work together to form a complex interactive relationship. Quantifying the coupling strength of each constraint in the overall task can help us understand the degree of influence of the constraint on other constraints and its importance in the entire scheduling task. The higher the coupling strength, the closer the constraint is to other constraints and the greater its impact on the overall scheduling task. In scheduling optimization, identifying and quantifying coupling strength can provide data support for priority setting and resource allocation, ensuring that key constraints are met first, thereby improving the effectiveness of the scheduling plan.

[0113] The weight function is a function used to adjust and control the interaction strength between different constraints. It is usually set based on the specific requirements of the task, the priority of the constraint, or the business rules. The role of the weight function is to appropriately adjust the importance of different constraint interactions when calculating the coupling strength to reflect the importance of different constraints in actual scheduling. For example, some tasks may have higher requirements for time window constraints, so the time window constraint can be given a higher weight value through the weight function when calculating the coupling strength. The weight function dynamically adjusts the interaction relationship between constraints to make the scheduling model closer to actual needs, thereby optimizing task allocation and resource utilization.

[0114] For each constraint C in the scheduling task s, define its constraint concentration, and the calculation expression is as follows:

[0115]

[0116] , where D s is the constraint concentration, that is, constraint C s The concentration, K x Yes Yes Constraint C x The coupling coefficient, ω is the smoothing coefficient;

[0117] Constraint concentration indicates the core degree or importance of a constraint in the entire constraint network among all interactions. In scheduling tasks, some constraints are more critical in the network, have more interactions with other constraints, and affect more task decisions. The higher the constraint concentration, the more important the constraint is in the entire scheduling task. By calculating the constraint concentration, the constraints that have the greatest impact on the scheduling results can be identified, so that these key constraints can be given priority in the optimization process, ensuring that the scheduling solution can meet the core requirements and improve the execution efficiency and stability of the overall task.

[0118] The smoothing coefficient is a constant used to avoid calculation anomalies caused by the denominator being zero or excessive data fluctuations. Since the calculation of constraint concentration involves the sum of the coupling strengths of multiple constraints, the addition of the smoothing coefficient can ensure that the denominator will not be zero in extreme cases (such as when there is no interaction), thereby ensuring the stability of the calculation results. In addition, the smoothing coefficient can also smooth the calculation results to a certain extent, avoiding excessive fluctuations in the concentration index due to extreme values ​​of certain constraints, making the measurement of concentration more accurate and stable.

[0119] In order to measure the mutual dependence between two constraints, the constraint mutual dependence coefficient is defined, and the calculation expression is as follows:

[0120] L sx =M sx ·(1+ln(1+|D s -D x |))

[0121] , where L sx is the constraint interdependence coefficient, i.e. constraint D s and constraint D x The mutual dependence coefficient between them;

[0122] The constraint interdependence coefficient represents the degree of interdependence between two different constraints in the scheduling task, and measures the strength of the association between them. In scheduling optimization, different constraints are often not independent, but influence each other. For example, the time window constraint may be limited by vehicle capacity or road traffic conditions. By calculating the constraint interdependence coefficient, this dependency can be quantified and the constraint pairs that must be considered simultaneously during the optimization process can be identified. This helps the optimization algorithm coordinate these highly correlated constraints when scheduling, thereby ensuring that the scheduling scheme does not violate another associated constraint while satisfying one constraint, and improving the effectiveness and consistency of the overall scheduling.

[0123] A constraint pair refers to a group of two interrelated and mutually influential constraints in a scheduling task. Since the constraints in a scheduling task are often not independent but interdependent or mutually restrictive, these interrelated constraint pairs need to be considered simultaneously during the optimization process to ensure that when one constraint is satisfied, another related constraint will not be violated.

[0124] For example, in logistics scheduling, "time window constraint" and "vehicle capacity constraint" may form a constraint pair, because meeting the time window requirements of goods (completing delivery within the specified time) may be limited by vehicle capacity. If the time window is tight and the vehicle capacity is insufficient, these two constraints need to be coordinated in the scheduling plan to avoid affecting the other due to considering one. Identifying these constraint pairs helps the optimization algorithm to comprehensively consider multiple factors when making scheduling decisions, ensuring the practical feasibility and optimization effect of the scheduling plan.

[0125] The interdependence coefficients of all constraints are combined to generate the dynamic constraint interaction index. The calculation expression is as follows:

[0126]

[0127] , where DynConI is the dynamic constraint interaction index, g(K s ,K x ) is a nonlinear function used to adjust the weights between different constraint coupling coefficients;

[0128] Under the detection window, the larger the dynamic constraint interaction index performance value generated by analyzing the degree of interaction between different constraints in the scheduling process, the higher the complexity of the current scheduling task, the more variables and constraints involved, and the stronger the interaction between them. Because in complex scheduling tasks, various constraints (such as time windows, vehicle capacity, path selection, etc.) are often closely connected to each other, and multiple constraints will affect each other and jointly determine the scheduling plan. This high degree of interaction requires more variables and constraints to be considered comprehensively when solving the task, and the algorithm must explore in a wider solution space to find the balance point between multiple constraints. Therefore, a higher dynamic constraint interaction index indicates that there are more constraint conflicts and interactive relationships that need to be solved in the scheduling problem, and the optimization difficulty of the task increases. On the contrary, when the interactivity between constraints is low, the task is usually simpler, involving fewer decision variables, the constraints are more independent, and the problem is less difficult to solve.

[0129] The machine learning model is not limited here, and any machine learning model that can perform comprehensive analysis on the scheduling topology complexity index STopoCI and the dynamic constraint interaction index DynConI to generate the task complexity coefficient TaCC is acceptable. In order to implement the technical solution of the present invention, the present invention provides a specific implementation method;

[0130] The formula for generating the task complexity coefficient TaCC is as follows:

[0131]

[0132] , where f 1 、f 2 are the preset proportional coefficients of the scheduling topology complexity index STopoCI and the dynamic constraint interaction index DynConI, respectively, and f 1 、f 2 Both are greater than 0.

[0133] Preset proportionality factor (f 1 、f 2 ) is used to balance and adjust the weights of the scheduling topology complexity index STopoCI and the dynamic constraint interaction index DynConI in the calculation of the task complexity coefficient TaCC. Different task scenarios may have different requirements for the impact of these two indexes. Therefore, by presetting the proportional coefficient, the influence of a certain index can be appropriately enlarged or reduced when calculating the task complexity. In simple terms, f 1 and f 2 The value of reflects the relative importance of adjusting the scheduling topology complexity index STopoCI and the dynamic constraint interaction index DynConI in the task complexity evaluation. This flexible adjustment mechanism helps to more accurately reflect the actual complexity of the task and adapt to the complexity calculation under different scheduling requirements.

[0134] It can be seen from the calculation expression of the task complexity coefficient that under the detection window, the larger the scheduling topology complexity index performance value generated after analyzing the topological structure complexity in the task scheduling process, and the larger the dynamic constraint interaction index performance value generated after analyzing the degree of interaction between different constraints in the scheduling process, the larger the task complexity coefficient performance value generated under the detection window, indicating that the complexity of the current scheduling task is higher, otherwise it means that the current scheduling task is relatively simple.

[0135] The task classification and complexity evaluation module divides the current scheduling tasks into complex tasks and simple tasks based on the evaluation results of the machine learning model;

[0136] The task complexity coefficient generated by the comprehensive analysis of the scheduling task information through the machine learning model is compared with the preset task complexity coefficient reference threshold, and the current scheduling task is divided. The division steps are as follows:

[0137] If the task complexity coefficient is greater than or equal to a preset task complexity coefficient reference threshold, the current scheduling task is classified as a complex task;

[0138] If the task complexity coefficient is less than the preset task complexity coefficient reference threshold, the current scheduling task is divided into a simple task;

[0139] Complex tasks involve a large number of variables and constraints in the scheduling process, and these constraints have high interactivity and dependency, requiring comprehensive consideration of multiple factors to find the optimal solution. Simple tasks, on the other hand, involve relatively fewer variables and constraints, and the interactions between these constraints are low.

[0140] Simple task solving module, for simple tasks, sets a standard solving time based on task characteristics and performs optimal solution within this time range;

[0141] The purpose of setting a standard solution time and performing the optimal solution within this time range is to avoid wasting too much computing time while ensuring scheduling efficiency. When dealing with simple tasks, these tasks usually have fewer constraints and lower complexity, which means that their solution space is relatively small and the optimization process does not require too many computing resources or complex algorithms. By setting a standard solution time based on historical data and actual conditions, the system can quickly find a high-quality solution within this limited time without performing overly complex calculations, thereby saving time and computing resources. This method simplifies the processing and not only ensures that the task can be completed within a reasonable time, but also ensures that the scheduling plan for the task is reasonable and efficient. The setting of the standard solution time is usually tested and adjusted many times to ensure that it can provide a suitable solution without wasting computing resources or exceeding the real-time requirements of the system, which helps to improve the overall operating efficiency of the system.

[0142] Complex task dynamic solution module: For complex tasks, since they involve a large number of variables and constraints, it uses the standard solution time as the basis and dynamically adjusts the solution time according to the task complexity to find the global optimal solution;

[0143] For complex tasks, based on the standard solution time, the specific steps for dynamically adjusting the solution time according to the task complexity to find the global optimal solution are as follows:

[0144] First, by comparing the task complexity coefficient TaCC with the task complexity coefficient reference threshold Tacc ref For comparison, the solution time is adjusted and the calculation expression is as follows:

[0145]

[0146] , where T solve is the adjusted complex task solving time, is the standard solution time based on simple task setting, and A is the adjustment factor;

[0147] The adjustment factor A is a factor used to control the task complexity coefficient TaCC for the complex task solution time T solveA coefficient that indicates the degree of influence. In scheduling optimization, the complexity of different tasks will affect the demand for solution time, and the adjustment factor A enables the system to flexibly adjust the solution time by amplifying or reducing this influence. Specifically, the size of the adjustment factor A determines the extent to which the solution time is extended when the task complexity increases; when the A value is large, the change in task complexity has a more significant impact on the solution time, which is suitable for very complex tasks; when the A value is small, the impact of complexity on the solution time is weaker, which is more suitable for simple or standardized task scenarios. By adjusting A, the system can balance resource allocation and ensure that complex tasks get enough solution time without affecting the overall system efficiency.

[0148] After adjusting the solution time of complex tasks, the solution time is further corrected according to the current available resources and the urgency of the task to ensure that the adjusted solution time is reasonably allocated under resource constraints and time requirements. The calculation expression is as follows:

[0149]

[0150] , where T solveadjusted is the further adjusted complex task solution time. Based on the adjusted complex task solution time, the solution time is further dynamically adjusted according to factors such as resource availability and task urgency. avail is the currently available resource, R max is the maximum value of resources, indicating the maximum amount of resources that can be provided, E urgency is the urgency of the task, E max is the maximum urgency, D is the resource availability impact coefficient, and Z is the urgency impact coefficient;

[0151] The urgency of the task urgency Refers to the urgency of the task in terms of time to be completed as soon as possible. It is usually a quantitative indicator used to indicate the timeliness requirements of the task, for example, whether a transportation task needs to be completed within a specific time window, or how high the customer's delivery speed requirements for the task are. The higher the urgency, the faster the task needs to be completed. The role of this indicator is to help the system prioritize these urgent tasks when allocating solution time, ensuring that high-timeliness tasks can be solved in a shorter time, thereby improving the system's response speed and customer satisfaction.

[0152] Maximum urgency level E max It is the standardized upper limit of urgency in the system, indicating the urgency of the most urgent task. This value is usually set based on business needs or experience to provide a unified scale for urgency so that the urgency of all tasks can be compared and normalized. The role of the maximum urgency is to provide a benchmark for the impact of urgency. By comparing urgency with Emax In comparison, the system can standardize the urgency of different tasks, making the adjustment of solution time more balanced and consistent.

[0153] The resource availability impact coefficient D is an adjustment factor used to measure the impact of resource availability on solution time. When the task complexity is high and the system resources (such as computing power and number of vehicles) are relatively limited, the solution time needs to be adjusted according to the resource situation. The role of the resource availability impact coefficient is to control the weighted effect of resource sufficiency on solution time; the larger D is, the greater the adjustment of resources on solution time, thereby ensuring that the system can still reasonably allocate time when resources are insufficient. When resources are abundant, this coefficient can appropriately reduce the increase in solution time to improve the computational efficiency of the overall system.

[0154] The urgency impact coefficient Z is an adjustment factor used to control the impact of task urgency on solution time. The urgency impact coefficient determines the weighted effect of urgent tasks on solution time. Its function is to enable the system to flexibly adjust the solution time according to the urgency of the task. The larger Z is, the more significant the impact of task urgency on solution time, which is suitable for situations that require a quick response; when Z is small, the priority of urgent tasks is reduced, and the adjustment range of solution time is reduced accordingly. By adjusting Z, the system can achieve a balance between task priority and computing resources, ensuring that high-priority tasks can be solved in a timely manner without affecting overall efficiency.

[0155] Finally, according to the adjusted solution time, the optimization algorithm is used to solve the problem and check the optimality of the solution. If the current solution does not meet the requirements of the global optimal solution, the optimization is continued until the global optimal solution is found, which is expressed as follows:

[0156]

[0157] , where OptimalityCheck is the judgment condition, which is judged based on whether the solution has reached the global optimal solution. Optimal is the global optimal solution, and uboptimal is the solution that has not reached the global optimal solution.

[0158] If the solution reaches the global optimal solution, the result of OptimalityCheck is the global optimal solution Optimal, which means that the current solution has the best effect among all solutions and does not need further optimization;

[0159] If the solution does not reach the global optimal solution, the result of OptimalityCheck is "Suboptimal", which means that the current solution is only a local optimal solution or a suboptimal solution. If there is a better solution, continue to try to optimize to find the global optimal solution Optimal.

[0160] For complex tasks, the standard solution time is used as the basis, and the solution time is dynamically adjusted according to the task complexity, aiming to ensure that complex tasks obtain sufficient computing resources, thereby increasing the possibility of finding the global optimal solution. In scheduling optimization, complex tasks usually involve more variables, constraints, and interdependent factors, which makes the solution space huge and difficult to find the optimal solution. If processed according to the standard solution time, the system may not be able to fully explore the solution space within a limited time, so that only a local optimal solution can be obtained, affecting the overall efficiency and resource utilization. By dynamically adjusting the solution time, the system can adaptively increase or decrease the computing time according to the complexity coefficient of the task, allocate more solution time for more complex tasks, and enable the algorithm to have enough resources to explore a wider solution space. In addition, the dynamic adjustment of the solution time can also flexibly allocate time according to the availability of resources and the urgency of the task, ensuring that the system's computing resources are optimally configured. Such an adjustment mechanism can not only improve the solution quality of complex tasks, but also effectively prevent suboptimal solutions caused by insufficient time, thereby ensuring the stability and efficiency of the scheduling system in a complex environment, and improving the optimization effect and service quality of the overall scheduling process.

[0161] By introducing real-time data preprocessing, feature extraction and intelligent evaluation, the system can dynamically perceive the complexity of the task and reasonably allocate the solution time. For simple tasks, the system can quickly obtain the optimal solution within the standard solution time to avoid wasting computing resources. For complex tasks, the system can flexibly extend the solution time based on the task complexity coefficient, providing more computing resources for the optimization algorithm, so as to conduct more extensive exploration in the huge solution space, thereby having a better chance of finding the global optimal solution. This solution time allocation strategy based on task difficulty enables the system to reasonably allocate resources, which not only improves the overall scheduling efficiency, but also reduces resource waste and avoids suboptimal solutions or local optimal solutions caused by insufficient standard time.

[0162] By combining the intelligent analysis of task complexity coefficients and machine learning models in task division and solution time allocation, the system can identify the long-term optimization needs of tasks, rather than just focusing on short-term efficiency. This intelligent scheduling solution can make dynamic adjustments based on the complexity of tasks and resource availability, ensuring that the system can operate stably even in complex environments and reducing potential problems caused by local optimal solutions. As the system continues to run, the complexity coefficients and other parameters of the tasks will also be optimized and updated through machine learning models, enhancing the system's adaptability to complex tasks.

[0163] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. An intelligent vehicle dispatching system for logistics transportation, characterized in that: It includes data collection and preprocessing module, feature extraction and analysis module, task classification and complexity assessment module, simple task solution module and complex task dynamic solution module: The data acquisition and preprocessing module, first, obtains the real-time data information related to the current scheduling task and preprocesses the acquired data information; The feature extraction and analysis module extracts features from the pre-processed scheduling task data information. After extraction, the extracted features are analyzed and processed under the detection window, and the analyzed feature information is input into the pre-trained machine learning model to perform intelligent evaluation of the current scheduling task; The task classification and complexity evaluation module divides the current scheduling tasks into complex tasks and simple tasks based on the evaluation results of the machine learning model; Simple task solving module, for simple tasks, sets a standard solving time based on task characteristics and performs optimal solution within this time range; The complex task dynamic solution module targets complex tasks, which involve a large number of variables and constraints. It takes the standard solution time as the basis and dynamically adjusts the solution time according to the task complexity to find the global optimal solution.

2. The intelligent vehicle dispatching system for logistics transportation according to claim 1 is characterized by: Feature extraction is performed on the preprocessed scheduling task data information. The extracted features include the topological complexity of the task scheduling process and the degree of interaction between different constraints in the scheduling process. After extraction, the topological complexity of the task scheduling process and the degree of interaction between different constraints in the scheduling process are analyzed and processed under the detection window to generate a scheduling topology complexity index and a dynamic constraint interaction index. The scheduling topology complexity index quantifies the complexity of the dependencies, constraints and execution order between each task node in the scheduling task. The dynamic constraint interaction index quantifies the degree of interaction and influence between different constraints in the task scheduling process.

3. The intelligent vehicle dispatching system for logistics transportation according to claim 2 is characterized in that: After obtaining the scheduling topology complexity index and dynamic constraint interaction index generated by analyzing the extracted features, the scheduling topology complexity index and dynamic constraint interaction index are input into the pre-learned machine learning model, and the task complexity coefficient is generated through the machine learning model. Intelligently evaluate the current task through the task complexity coefficient.

4. The intelligent vehicle dispatching system for logistics transportation according to claim 3 is characterized in that: The task complexity coefficient generated by the comprehensive analysis of the scheduling task information through the machine learning model is compared with the preset task complexity coefficient reference threshold, and the current scheduling task is divided. The division steps are as follows: If the task complexity coefficient is greater than or equal to a preset task complexity coefficient reference threshold, the current scheduling task is classified as a complex task; If the task complexity coefficient is less than a preset task complexity coefficient reference threshold, the current scheduling task is classified as a simple task.

5. The intelligent vehicle dispatching system for logistics transportation according to claim 4 is characterized in that: For complex tasks, based on the standard solution time, the specific steps for dynamically adjusting the solution time according to the task complexity to find the global optimal solution are as follows: First, by comparing the task complexity coefficient TeCC with the task complexity coefficient reference threshold TaCC ref For comparison, the solution time is adjusted and the calculation expression is as follows: ， Where, T solve is the adjusted complex task solving time, is the standard solution time based on simple task setting, and A is the adjustment factor; After adjusting the solution time of complex tasks, the solution time is further corrected according to the current available resources and the urgency of the task to ensure that the adjusted solution time is reasonably allocated under resource constraints and time requirements. The calculation expression is as follows: ， Where, T solveadjusted is the further adjusted complex task solving time, R avail is the currently available resource, R max is the maximum value of resources, E urgency is the urgency of the task, E max is the maximum urgency, D is the resource availability impact coefficient, and Z is the urgency impact coefficient; Finally, according to the adjusted solution time, the optimization algorithm is used to solve the problem and check the optimality of the solution. If the current solution does not meet the requirements of the global optimal solution, the optimization is continued until the global optimal solution is found, which is expressed as follows: ， In the formula, OptimalityCheck is the judgment condition, which is judged based on whether the solution has reached the global optimal solution. Optimal is the global optimal solution, and uboptimal is the solution that has not reached the global optimal solution. If the solution reaches the global optimal solution, the result of OptimalityCheck is the global optimal solution Optimal, which means that the current solution has the best effect among all solutions and does not need further optimization; If the solution does not reach the global optimal solution, the result of OptimalityCheck is that the global optimal solution Suboptimal is not reached, which means that the current solution is only a local optimal solution or a suboptimal solution, and there is a better solution. In this case, continue to try to optimize to find the global optimal solution Optimal.

6. The intelligent vehicle dispatching system for logistics transportation according to claim 2 is characterized in that: In the detection window, the topological complexity of the task scheduling process is analyzed, and the specific steps to generate the scheduling topological complexity index are as follows: In the detection window, we first define the dependencies between tasks in the task scheduling process, and assume that the set of task nodes in the scheduling task is T = {t i }={t1, t2, t3,...,t n }, where t i represents the ith independent scheduling task, n is the total number of tasks, and each task t i It has dependencies with other tasks, represented by a directed graph G = (V, E) in graph theory, where V = T, V, represents the dependency set. If task t i Must be in task t j Before execution, there is an edge (t i , t j )∈E; The dependency between tasks is reflected by the connectivity and hierarchy of the graph. The in-degree and out-degree are used to quantify the dependency of each task. The calculation expression is as follows: ， Indegree(t i ) is the task t i The in-degree of task t i The number of dependent tasks, t j is the jth independent scheduling task, ||(t j →t i ) is an indicator function. If there exists a j To task t i The dependency relationship between ||(t j →t i )=1, otherwise 0, ， In the formula, Outdegree(t i ) is the task t i The out-degree of task t i The number of other tasks it depends on; The topological complexity scoring function is introduced to reflect the complexity of each task through the in-degree, out-degree and the dependencies between tasks. i The complexity index is calculated as follows: ， In the formula, C(t i ) is the task t i The complexity index is α, which is the exponential weight of in-degree and out-degree, N i It is the task i The set of adjacent tasks, w ij is the weight of the dependency, and β is the exponent of the dependency weight; The weighted average of the complexity indexes of all tasks is used to comprehensively measure the topological complexity of the entire scheduling task and generate the scheduling topology complexity index. The calculation expression is as follows: ， Where STopoCI is the scheduling topology complexity index, γ is an adjustable weight factor used to adjust the influence of a single task complexity on the overall topology complexity, and Z is a normalization constant used to ensure that the scale of the index adapts to the entire task scheduling system.

7. The intelligent vehicle dispatching system for logistics transportation according to claim 2 is characterized in that: In the detection window, the interaction degree between different constraints in the scheduling process is analyzed, and the specific steps to generate the dynamic constraint interaction index are as follows: Under the detection window, first establish a constraint matrix M, M = {C s }={C1, C2, C3,...,C m }, the constraint matrix M is used to represent the relationship between the various constraints in the scheduling task, C s is the sth constraint in the scheduling task, m is the total number of constraints in the scheduling task, and an m·m rectangle is constructed. The expression is as follows: ， In the formula, the rows and columns of the matrix correspond to different constraints, and each matrix element M sx Represents the sth constraint condition C s With the xth constraint C x The intensity of interaction between In order to quantify the coupling strength of each constraint in the overall task, the constraint coupling coefficient is defined and the calculation expression is as follows: ， In the formula, K s is the constraint coupling coefficient, i.e., constraint C s The coupling coefficient, f(C s , C x ) is the weight function; For each constraint C in the scheduling task s , define its constraint concentration, and the calculation expression is as follows: ， Where D s is the constraint concentration, that is, constraint C s The concentration, K x Yes Yes Constraint C x The coupling coefficient, ω is the smoothing coefficient; In order to measure the mutual dependence between two constraints, the constraint mutual dependence coefficient is defined, and the calculation expression is as follows: L sx =M sx ·(1+ln(1+|D s -D x |)), Where, L sx is the constraint interdependence coefficient, i.e. constraint D s and constraint D x The mutual dependence coefficient between them; The interdependence coefficients of all constraints are combined to generate the dynamic constraint interaction index. The calculation expression is as follows: ， Where DynConI is the dynamic constraint interaction index, g(K s ,E x ) is a nonlinear function used to adjust the weights between different constraint coupling coefficients.

Images