An emergency material distribution vehicle scheduling method based on cooperative guarantee

Abstract

The application discloses an emergency material distribution vehicle scheduling method based on cooperative guarantee, and relates to the technical field of traffic transportation and logistics management, and comprises the following steps: uniformly accessing instructions from different platforms into a centrally managed scheduling system, collecting and aggregating scheduling requests and resource information of each department; performing standardized analysis and preprocessing operations on the obtained scheduling requests and resource information, and storing all the processed effective scheduling information into a unified data set according to a predetermined data structure to support subsequent feature extraction and intelligent analysis. The application realizes conflict intelligent identification and dynamic arbitration by uniformly accessing and standardizing multiple source scheduling instructions in combination with a machine learning model, avoids resource repeated allocation, performs task priority allocation through an arbitration window mechanism, and improves the continuity, scheduling stability and cooperative efficiency of emergency material distribution.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of transportation and logistics management, and particularly relates to an emergency material distribution vehicle scheduling method based on collaborative guarantee. BACKGROUND

[0002] The "emergency material distribution vehicle scheduling based on collaborative guarantee" refers to a method for scientifically and reasonably scheduling and path optimizing the distribution vehicles through multi-party collaboration and resource guarantee mechanism to ensure that the emergency materials (such as food, medicine, medical devices, rescue equipment, etc.) can be efficiently and timely delivered to the disaster-stricken areas or designated locations when a sudden event (such as natural disasters, major accidents, public health events, etc.) occurs. This method not only considers the path planning and transportation efficiency of a single vehicle, but also introduces the information sharing and collaboration mechanism among government agencies, warehouse centers, traffic management, front-line rescue teams and other subjects, realizes the collaborative linkage of people, vehicles, goods, roads and warehouses, dynamically responds to various sudden situations and transportation obstacles, and ensures the reliability, timeliness and safety of the material distribution task. It is a comprehensive scheduling optimization strategy for sudden public events.

[0003] The prior art has the following disadvantages: Although the existing emergency material distribution system usually establishes a unified scheduling coordination mechanism for integrating the scheduling requirements of multiple departments and centrally managing transportation resources, in actual high-pressure emergency response scenarios, the scheduling system still lacks fine-grained conflict perception and dynamic processing capability for concurrent scheduling instructions. Specifically, when multiple functional agencies almost simultaneously initiate scheduling requests for the same batch of vehicle resources according to the first-line disaster situation dynamics they master, the system can receive all scheduling instructions, but due to the coarse processing granularity in task priority judgment, vehicle resource state identification, task occupation time reasoning, etc., or information update lag, the same vehicle resource may still be repeatedly allocated, forming a "conflict scheduling" phenomenon. Such conflicts not only cause the execution of scheduling instructions to fail or overlap, but also may cause frequent switching of vehicle tasks, increase of path reconstruction, affect the continuity and stability of material distribution, and further significantly interfere with the overall emergency response efficiency and disaster area resource guarantee capability.

[0004] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The application aims to provide an emergency material distribution vehicle scheduling method based on cooperative guarantee, which unifies access and standardizes processing of multi-source scheduling instructions, combines a machine learning model to realize intelligent conflict identification and dynamic arbitration, and avoids repeated allocation of resources; through an arbitration window mechanism, task priority is deployed to improve the continuity, scheduling stability and cooperative efficiency of emergency material distribution, thereby solving the problems in the background art.

[0006] To achieve the above-mentioned purpose, the application provides the following technical solution: an emergency material distribution vehicle scheduling method based on cooperative guarantee, comprising the following steps:

[0007] The instructions from different platforms are uniformly accessed into a centrally managed scheduling system, and scheduling requests and resource information of various departments are collected and aggregated;

[0008] The obtained scheduling requests and resource information are subjected to standardization analysis and preprocessing operations, and all the processed effective scheduling information is stored in a unified data set according to a predetermined data structure to support subsequent feature extraction and intelligent analysis;

[0009] Key indicators reflecting potential resource conflicts between scheduling tasks are extracted from the constructed data set through feature engineering technology, and the extracted key indicators are subjected to multi-dimensional comprehensive analysis to quantify the potential conflict state in the current scheduling task;

[0010] The scheduling conflict key indicators subjected to quantization processing are input into a pre-trained machine learning model as feature vectors, and the machine learning model is used to intelligently evaluate the conflict risk of the current scheduling task to determine whether there is a conflict risk in the current scheduling task;

[0011] When the machine learning model determines that multiple scheduling tasks have conflict risks, all scheduling instructions are not executed, and an arbitration window with a short time delay is entered, the emergency degree, influence range and vehicle resource occupation state of each scheduling task are dynamically evaluated in the arbitration window, and based on the preset task priority and scheduling influence factor, real-time conflict resolution and task fusion processing are performed to realize optimal allocation and efficient scheduling of limited resources.

[0012] Preferably, the instructions from different platforms are uniformly accessed into a centrally managed scheduling system, comprising the following key steps:

[0013] First, an interface module supporting multi-data source access is built, so that each functional agency can send scheduling instructions to the unified platform through API calling, data subscription and message bus mode;

[0014] Second, corresponding conversion rules and field mapping relationships are established for the instruction formats of each agency, and the instruction contents in different systems are converted into a unified standard data model;

[0015] Then, all the accessed and standardized converted scheduling instructions are unified into the core instruction center of the scheduling system, and are associated with the vehicle, material and personnel resource database to form a complete task-resource association view.

[0016] Preferably, the key indicators reflecting potential resource conflicts between scheduling tasks are extracted from the constructed data set through feature engineering technology, the extracted key indicators include the probability of different agencies issuing the same scheduling target to the same resource category and the boundary division clarity degree between scheduling tasks of multiple functional agencies, and the probability of different agencies issuing the same scheduling target to the same resource category and the boundary division clarity degree between scheduling tasks of multiple functional agencies are analyzed under the detection window to respectively generate an organization command convergence reference value and a scheduling domain boundary ambiguity reference value, and the organization command convergence reference value and the scheduling domain boundary ambiguity reference value are used to quantify the potential conflict state in the current scheduling task.

[0017] Preferably, the specific steps of analyzing the probability of different agencies issuing the same scheduling target to the same resource category under the detection window to generate the organization command convergence reference value are as follows:

[0018] First, the scheduling targets of different agencies are grouped according to the resource category, for each resource category, the scheduling target set of each agency is constructed, then, the intersection strength ratio index of all agency scheduling target sets is generated, which represents the coincidence degree of scheduling targets of multiple agencies under the same resource category, and the generation formula is as follows:

[0019]

[0020] , in the formula, ISR is the intersection strength ratio index, G i and G j are the scheduling target sets issued by the i th and j th scheduling agencies respectively within the detection window for the resource category, and n is the total number of agencies participating in scheduling;

[0021] Further enhance the sensitivity of the model to the consistency trend of the target, introduce a discrete structure convergence function to measure the tension relationship between the distribution dispersion of the overall scheduling target and the concentration degree of the instruction set, and construct the final organization command convergence reference value, the formula is as follows:

[0022]

[0023] , in the formula, OCC-RV is the organization command convergence reference value, U R is the number of unique scheduling targets.

[0024] Preferably, the specific steps of analyzing the boundary division clarity degree between scheduling tasks of multiple functional agencies under the detection window to generate the scheduling domain boundary ambiguity reference value are as follows:

[0025] A plurality of functions of the agencies under a certain detection window are constructed into a task scheduling domain cross graph, wherein each node represents a scheduling task, and the edges represent the existence of target area repetition relationship between scheduling tasks. Then, the cross index of each scheduling task is calculated, and the calculation expression is as follows:

[0026] The formula is as follows:

[0027]

[0028] , CI q is the cross index of the scheduling task q, R q and R p are the resource sets of the scheduling task q and the scheduling task p respectively, δ(g q , g p ) is the function of the difference between the agencies, which is used to determine whether two tasks come from different agencies, g q and g p represent the agency identification of the scheduling task q and the scheduling task p respectively; Γ(q) is the task set having resource intersection with the task q.

[0029] After obtaining the cross index of all tasks, the boundary ambiguity degree of the scheduling domain is further extracted at the global level to generate a scheduling domain boundary ambiguity reference value to quantify the "instruction overlap confusion degree" of the overall scheduling domain, and the generation formula is as follows:

[0030]

[0031] In the formula, DDBF-RV is the scheduling domain boundary ambiguity reference value, Φ is the cross index set of all tasks, Φ={CI q}={CI1, CI2, …, CI m}, m is the total number of all scheduling tasks published by the multiple agencies within the current detection window, max(Φ) is the maximum value of the cross index of all scheduling tasks, which finds the most serious resource conflict task, n is the total number of agencies participating in scheduling, and D split is the scheduling responsibility domain fragmentation degree.

[0032] Preferably, the quantized organization command convergence reference value and the scheduling domain boundary ambiguity reference value are input into the machine learning model which has been trained in advance as a feature vector, and the scheduling task conflict risk coefficient is generated by the machine learning model, and the conflict risk of the current scheduling task is intelligently evaluated by the scheduling task conflict risk coefficient to determine whether the current scheduling task has conflict risk.

[0033] Preferably, the scheduling task conflict risk coefficient generated by the intelligent evaluation of the conflict risk of the current scheduling task by the pre-trained machine learning model is compared and analyzed with the pre-set scheduling task conflict risk coefficient reference threshold value to determine whether the current scheduling task has a conflict risk, and the judgment logic is as follows:

[0034] If the scheduling task conflict risk coefficient is greater than the pre-set scheduling task conflict risk coefficient reference threshold value, it is determined that there is a conflict risk between the scheduling tasks; if the scheduling task conflict risk coefficient is less than or equal to the pre-set scheduling task conflict risk coefficient reference threshold value, it is determined that there is no conflict risk between the scheduling tasks.

[0035] Preferably, when the machine learning model determines that multiple scheduling tasks have a conflict risk, all scheduling instructions are not executed, and a short time delay arbitration window is entered, the urgency, influence range and vehicle resource occupation state of each scheduling task are dynamically evaluated in the arbitration window, and based on the pre-set task priority and scheduling influence factor, real-time conflict resolution and task fusion processing are performed to realize the specific steps of optimal allocation and efficient scheduling of limited resources as follows:

[0036] When multiple scheduling tasks are detected to have a resource conflict, enter the arbitration window, and quantitatively score the priority of all conflict tasks. In order to more comprehensively evaluate the execution urgency and scheduling value of the task, the urgency, remaining schedulable time and influence range of the task in the disaster are considered, a priority scoring function is constructed, and the formula is as follows:

[0037]

[0038] , wherein P q is the task priority score, U q is the task urgency coefficient, D q is the task remaining schedulable time, I q is the task influence range index, λ is the time compression adjustment factor, α1, α2 and α3 are weight coefficients of the task urgency coefficient U q , the task remaining schedulable time D q and the task influence range index I q , and α1+α2+α3=1.

[0039] After the priority calculation is completed, the competition degree between tasks in the resource layer is evaluated, the conflict risk when the same resource is called by multiple tasks at the same time is focused on, a resource conflict intensity model between tasks is constructed through the resource usage flag and the time overlap ratio, and the formula is as follows:

[0040]

[0041] , wherein C qp is a resource conflict intensity, representing the degree of resource-level conflict between the scheduling task q and the scheduling task p in the scheduling system, h is the total number of resources used by the scheduling task, and qk and pk are binary variables, representing whether the resource k is used by the scheduling task q or the scheduling task p, qk is the actual occupation time interval of the scheduling task q for the resource k, pk is the actual occupation time interval of the scheduling task p for the resource k, respectively, and

[0042] After the quantification of the task priority and the resource conflict intensity, the scheduling tasks are fused and scored to comprehensively judge the rationality of their execution and the cost of resource scheduling. A scheduling fusion scoring function is introduced, and the formula is as follows:

[0043]

[0044] , wherein S q represents the final scheduling score of the task q, qp is a conflict influence factor.

[0045] In the above technical solution, the technical effects and advantages provided by the present application are as follows:

[0046] The present application can effectively identify and resolve the resource conflict problem between scheduling instructions in a multi-department concurrent scheduling environment by unified access, standardized processing and feature modeling of multi-source scheduling instructions, combined with intelligent identification and dynamic arbitration mechanism of conflict risks by a machine learning model, thereby avoiding task failure and resource waste caused by repeated allocation or scheduling interference. At the same time, by introducing a short-latency arbitration window, flexible control and priority allocation of scheduling instructions are realized, so that limited emergency vehicle resources can be dynamically reorganized and optimally allocated according to the task urgency and influence range, thereby significantly improving the continuity of material distribution, the stability of scheduling and the overall collaborative efficiency in the emergency response process. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0048] Figure 1 The method flowchart of the emergency material distribution vehicle scheduling method based on collaborative guarantee provided by the present application. DETAILED DESCRIPTION

[0049] Example implementations are now described with reference to the drawings. Example implementations can, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the inventive gist to those skilled in the art.

[0050] The application provides an emergency material distribution vehicle scheduling method based on cooperative guarantee as shown in the formula (I): Figure 1 The application provides an emergency material distribution vehicle scheduling method based on cooperative guarantee as shown in the formula (I):

[0051] The instructions from different platforms are uniformly accessed into the centralized management scheduling system, and the scheduling requests and resource information of each department are collected and aggregated.

[0052] The instructions from different platforms are uniformly accessed into the centralized management scheduling system, which means that the scheduling instructions issued by multiple functional agencies (such as emergency management bureaus, transportation departments, medical institutions, and storage centers) through their respective business systems or platforms are all gathered into a unified scheduling system for centralized management and processing. Since the platforms and instruction formats used by each department are often different, if not uniformly accessed, it is easy to cause information fragmentation, missing instructions, or conflict difficult to identify. Through uniform access, centralized collection and comprehensive control of various scheduling requests and resource states (such as vehicle location, material type, task progress, etc.) can be achieved, ensuring that the scheduling system has a complete information foundation when facing multi-source tasks, which helps subsequent standardized processing, conflict detection, and intelligent allocation, thereby improving the coordination and overall scheduling efficiency of emergency response.

[0053] The instructions from different platforms are uniformly accessed into the centralized management scheduling system, which includes the following key steps: first, an interface module supporting multi-data source access is built, allowing each functional agency to send scheduling instructions to the unified platform through API calling, data subscription, or message bus; second, corresponding conversion rules and field mapping relationships are established for the instruction formats of each agency, converting the instruction content (such as task number, scheduling object, destination, time requirement, etc.) in different systems into a unified standard data model; then, all the accessed and standardized converted scheduling instructions are uniformly imported into the core instruction center of the scheduling system and associated with the vehicle, material, and personnel resource databases, forming a complete task-resource association view. Through this process, the structured integration and centralized management of cross-department scheduling information are achieved, providing a reliable data foundation for subsequent scheduling analysis, task coordination, and intelligent optimization of the system.

[0054] The obtained scheduling request and resource information are standardized and preprocessed, including field unification, format conversion, abnormal value elimination and missing value completion, and all the processed effective scheduling information is stored in a unified data set according to a predetermined data structure to support subsequent feature extraction and intelligent analysis.

[0055] Due to the differences in scheduling instruction formats from different agencies, including data structure, field naming, coding method, etc., it is easy to cause matching errors or information missing in the subsequent analysis process. Therefore, it is necessary to define a unified field mapping and analysis rule for the instruction structure of each agency in advance, and convert the scheduling information into a general data format of the system; for example, unified vehicle ID, task priority field, demand material type code, etc. Through standardized analysis, it can be ensured that the subsequent processing links can understand and compare the scheduling requirements of each agency at the same semantic level.

[0056] After completing the standardized analysis, the parsed scheduling information needs to be further preprocessed, including deduplication, error value correction, data filling, timestamp alignment, etc. Through preprocessing, some poor quality or obviously incorrect scheduling records can be filtered out, and missing fields can be inferred and supplemented, making the data more complete and reliable. Subsequently, all effective scheduling information is stored in a data set according to the predetermined structure, providing a high-quality and consistent basic data source for subsequent feature extraction.

[0057] Key indicators reflecting potential resource conflicts between scheduling tasks are extracted from the constructed data set through feature engineering technology, and the extracted key indicators are analyzed in multiple dimensions to quantify the potential conflict state in the current scheduling task.

[0058] Key indicators reflecting potential resource conflicts between scheduling tasks are extracted from the constructed data set through feature engineering technology, including the probability of different agencies issuing the same scheduling target for the same resource category and the clarity of the boundary division between scheduling tasks of multiple functional agencies. The probability of different agencies issuing the same scheduling target for the same resource category and the clarity of the boundary division between scheduling tasks of multiple functional agencies are analyzed under the detection window to generate organization command convergence reference value and scheduling domain boundary ambiguity reference value respectively, and the organization command convergence reference value and the scheduling domain boundary ambiguity reference value are used to quantify the potential conflict state in the current scheduling task.

[0059] The greater the probability that the same institution sends the same or highly similar scheduling targets to the same resource category, the more potential resource conflicts there are between scheduling tasks. This phenomenon reflects the tendency of multiple independent instruction sources to simultaneously call for the same type of critical resources (such as the same type of vehicle, supplies, path, or destination) without sufficient coordination, resulting in an increase in the concentration of resource requests at the system level. If the resource supply capacity is insufficient or the scheduling response mechanism is lagging, resource competition, task preemption, and repeated scheduling will occur, leading to a series of scheduling interference phenomena such as repeated assignment of vehicles, task execution overlap, and path conflicts. In particular in emergency response scenarios, resources are limited and time is tight, and such "instruction convergence" can easily cause an increase in the load of the scheduling system and imbalance in deployment. Therefore, by analyzing the convergence degree of scheduling instructions between different institutions in terms of target resources, time, and regional dimensions, potential scheduling conflict risks can be effectively warned, and key judgment basis can be provided for subsequent conflict resolution and task fusion.

[0060] The specific steps for analyzing the probability of different institutions sending the same scheduling target to the same resource category in the detection window to generate the organization command convergence reference value are as follows:

[0061] First, the scheduling targets of different institutions are grouped by resource category (such as ambulances, cold chain vehicles, unmanned transportation equipment, etc.). For each resource category, a set of scheduling targets for each institution is constructed, where the target can be a specific task point, route, or supply delivery point. Then, the intersection strength ratio index of all institution scheduling target sets is generated, representing the degree of overlap between scheduling targets of multiple institutions in the same resource category. The generation formula is as follows:

[0062]

[0063] In the formula, ISR is the intersection strength ratio index, which measures the actual overlap between scheduling targets of multiple institutions in resource category R, G i and G j are the sets of scheduling targets sent by the i th and j th scheduling institutions, respectively, within the detection window for the resource category, and n is the total number of institutions participating in scheduling.

[0064] The above steps identify the degree of overlap between scheduling targets of different institutions in the same resource category, thereby quantifying the consistency strength of multi-source instructions. By calculating the intersection strength ratio between scheduling target sets, a key structured index basis is provided for subsequent judgment of potential scheduling conflicts.

[0065] To further enhance the sensitivity of the model to the target consistency trend, a discrete structure convergence function is introduced to measure the tension relationship between the distribution of the overall scheduling target and the concentration of instructions, and the final organization command convergence reference value is constructed, as follows:

[0066]

[0067] , wherein OCC-RV is an organization command convergence reference value, U R is the unique dispatch target number, representing the total number of unique targets after merging and deduplication of all institutional dispatch target sets under a resource category.

[0068] Unique dispatch target number U R is used to measure the overall distribution range and dispersion degree of dispatch task targets of different institutions within a detection window under a certain resource category. It represents the total number of target points (such as destinations, material receiving points, service areas, etc.) after deduplication of all institutional dispatch instructions. The key significance of this index is that when U R is small, it indicates that each institution concentrates on dispatching a small number of targets, with a high risk of concentration and task overlap, and a higher possibility of scheduling conflicts. When U R is large, it indicates that the dispatch task targets are widely distributed and the use of resources is scattered, and even if there is local target overlap, it is more likely to be accidental overlap rather than systematic conflict. Therefore, U R as a regulating factor in the organization command convergence index, can effectively suppress the "conflict risk amplification" caused by too few targets, or prevent the phenomenon of "misjudgment as conflict" when there are too many targets, thereby enhancing the judgment accuracy and stability of the scheduling system in the conflict identification process.

[0069] The above steps couple the degree of overlap of multiple institutional dispatch targets with the complexity of overall target distribution, thereby generating an organization command convergence reference value that accurately reflects the concentration trend of dispatch instructions. This reference value not only identifies whether the dispatch targets are consistent, but also automatically adjusts the sensitivity according to the diversity of targets, avoiding false conflict risks due to the large number of targets.

[0070] The larger the organization command convergence reference value generated by analyzing the probability of different institutions issuing the same dispatch target for the same resource category within the detection window, the higher the degree of overlap of dispatch instructions in time, space, or task content by different institutions. This highly convergent scheduling behavior means that multiple tasks are likely to compete for the same batch of resources, significantly increasing the resource conflict risk between dispatch tasks. Conversely, when the organization command convergence reference value is low, it indicates that the dispatch instructions have large differences in resource type, dispatch target, time window, etc., and the intersection of resource use between tasks is small, with a lower conflict probability.

[0071] The unclear boundary between multiple functional agencies' dispatch tasks is indeed one of the important manifestations of potential resource conflicts between dispatch tasks. The unclear dispatch boundary mainly manifests as follows: different agencies lack clear definition in terms of dispatch goals, resource usage range, and time and space coverage area, leading to problems such as overlapping instructions, repeated resource calls, or overlapping responsibilities during task execution. For example, if the medical department and the traffic department simultaneously dispatch the same batch of vehicles for different tasks without sharing their respective dispatch intentions and resource allocation information, they may issue conflicting instructions to the same vehicle, causing execution confusion or task failure. Such ambiguous boundaries often mean that the system lacks “instruction attribution judgment ability” and “resource uniqueness occupation mechanism” when receiving dispatch instructions, thus laying the foundation for potential dispatch conflicts. In particular, in high-intensity emergency response scenarios, unclear boundaries will directly affect the accuracy and timeliness of resource allocation, and are an important signal for identifying the risk of dispatch task conflicts.

[0072] The specific steps of analyzing the degree of clearness of the boundary between multiple functional agencies' dispatch tasks in the detection window to generate a dispatch domain boundary ambiguity reference value are as follows:

[0073] The dispatch tasks published by multiple functional agencies in a certain detection window are constructed into a task dispatch domain overlap graph (TOG), where each node represents a dispatch task, and the edge represents the existence of a repeated relationship between the target areas of the dispatch tasks. Then, the intersection index of each dispatch task is calculated to measure the degree of repetition of the task with other tasks, and the expression is as follows:

[0074] The formula is as follows:

[0075]

[0076] CI q is the intersection index of dispatch task q, indicating the strength of resource overlap between dispatch task q and other tasks in the current detection window, R q and R p are the resource sets of dispatch task q and dispatch task p, respectively, indicating the set of resource objects called by each dispatch task, which can include vehicle number, driver ID, warehouse number, and road segment number, etc. δ(g q , g p ) is a functional agency difference function used to determine whether two tasks come from different functional agencies, g q and g p represent the agency identifiers of dispatch task q and dispatch task p, respectively; Γ(q) is the set of tasks that have resource intersection with task q.

[0077] In the construction of the scheduling task cross-mapping, the node represents each independent scheduling task in the system, usually initiated by different functional agencies, containing task objectives, resource requirements, scheduling time windows, service areas, etc. Each node as a basic unit in the graph carries the "entity information" of the task itself, such as a task of transporting medical supplies is a node, and another task of assigning the same type of vehicle to transport patients in the disaster area is also a node. By representing tasks as nodes, the distribution and scale of scheduling activities can be clearly identified in the graph structure.

[0078] The edge represents the existence of some "relationship link" between two scheduling tasks, which specifically refers to resource overlap or target area overlap in this scenario. If two tasks use the same batch of vehicles, schedule the same warehouse, or have an intersection of service areas, an edge will be established between the corresponding nodes of the two tasks. The existence of the edge means that these tasks may interfere with each other or compete for resources during execution, which is an important signal of potential conflict. By constructing these edges, isolated task nodes can be connected into a complex relationship network, facilitating subsequent calculation of task cross and scheduling conflict degree.

[0079] This step can more sensitively mine the weak signals of cross-agency resource conflicts by identifying the resource overlap degree between multiple functional agencies and filtering the "natural overlap" of scheduling within the same agency through the heterogeneous agency factor, forming a structural cross index at the task level. This provides a "node perspective" for the subsequent global determination of boundary ambiguity.

[0080] After obtaining the cross index of all tasks, the boundary ambiguity of the scheduling domain is further extracted at the global level to generate a scheduling domain boundary ambiguity reference value, which quantifies the "instruction overlap confusion degree" of the overall scheduling domain, and the formula is as follows:

[0081]

[0082] In the formula, DDBF-RV is the scheduling domain boundary ambiguity reference value, Φ is the set of cross indexes of all tasks, Φ = {CI q} = {CI1, CI2, …, CI m}, m is the total number of all scheduling tasks published by multiple functional agencies within the current detection window, max(Φ) is the maximum value of the cross index of all scheduling tasks, which finds the most serious resource conflict task, n is the total number of agencies participating in scheduling, and D split is the scheduling responsibility domain fragmentation, which represents the number of tasks from different functional agencies scheduling the same resource category (such as the same type of vehicle, warehouse, and transit route).

[0083] By integrating the maximum cross index between tasks and the fragmentation degree of multi-agency resource scheduling, the ambiguity degree of scheduling task boundaries in the entire scheduling system is quantified comprehensively. It can accurately identify high-risk conflict areas in the scheduling domain caused by overlapping responsibilities or chaotic resource allocation between agencies, thereby providing effective early warning basis and optimization scheduling strategy support for the system.

[0084] The greater the scheduling domain boundary ambiguity reference value generated after analyzing the boundary division degree between multiple functional agency scheduling tasks under the detection window, the higher the degree of responsibility overlap or resource overlap between tasks of different agencies, the poorer the independence between scheduling tasks, and the more likely the same resource is repeatedly called or the multi-party scheduling target conflicts, thereby causing potential resource conflicts. Conversely, when the reference value is small, the scheduling task boundary is clear, the responsibility is clear, the tasks do not interfere with each other, and the scheduling resource allocation is also more orderly, so it can be considered that there is no significant resource conflict risk between tasks.

[0085] The quantitatively processed scheduling conflict key indicators are input into the pre-trained machine learning model as feature vectors, and the conflict risk of the current scheduling task is intelligently evaluated by the machine learning model to determine whether the current scheduling task has a conflict risk.

[0086] The quantitatively processed organization command convergence reference value and scheduling domain boundary ambiguity reference value are input into the pre-trained machine learning model as feature vectors, and the scheduling task conflict risk coefficient is generated by the machine learning model. The conflict risk of the current scheduling task is intelligently evaluated by the scheduling task conflict risk coefficient to determine whether the current scheduling task has a conflict risk.

[0087] The "pre-trained machine learning model" means that the model has been fully trained and optimized based on historical data before the current scheduling system is formally run, so that it has the ability to accurately judge and predict the input features. In this system, the training phase usually uses a large number of past scheduling instructions, resource states, and conflict result samples to construct a training set, and through supervised learning, the task features and the actual conflict results are one-to-one corresponding, input into the machine learning algorithm for learning. During the training process, the model automatically identifies the complex relationship between features and conflicts and forms internal parameter structures (such as decision tree structure, weight matrix, neural network connection weight, etc.), thereby possessing the ability of "analogical reasoning" and risk assessment for new scheduling tasks. The training process also includes cross-validation, model evaluation and optimization steps to ensure that the model has good generalization ability and can adapt to scheduling conflict identification in different scenarios in the future.

[0088] The pre-training completion is crucial in that it enables the system to quickly and automatically assess the conflict risk intelligently without complex data modeling or manual rule judgment when facing new scheduling tasks accessed in real time. In the system, the organization command convergence index and the scheduling domain boundary ambiguity index are two important input features that reflect the potential conflict risks of scheduling instructions at the organization level and the spatial management level. When these indicators are input into the trained model, the model will make a quantitative judgment on the conflict risk that the current task may cause based on the knowledge learned by itself, and output a scheduling conflict risk coefficient. This risk coefficient can be used as a core decision-making basis before scheduling execution, for judging whether to enter the arbitration window, whether to need task fusion or priority adjustment, etc. Compared with the traditional rule engine based on artificial threshold setting, the pre-trained model has higher adaptability and judgment accuracy, especially in the complex multi-variable interaction and conflict feature difficult to express explicitly in the emergency scheduling scene, which can significantly improve the intelligent level and actual scheduling efficiency of the system.

[0089] The machine learning model is not limited here, and any machine learning model that can realize the comprehensive analysis of the organization command convergence reference value OCC-RV and the scheduling domain boundary ambiguity reference value DDBF-RV to generate the scheduling task conflict risk coefficient DCRC can be used. To realize the technical scheme of the present application, the present application provides a specific implementation manner;

[0090] The scheduling task conflict risk coefficient DCRC generation formula is as follows: DCRC = δ1·OCC-RV + δ2·DDBF-RV, wherein δ1 and δ2 are preset proportion coefficients of the organization command convergence reference value OCC-RV and the scheduling domain boundary ambiguity reference value DDBF-RV, and δ1 and δ2 are both greater than 0.

[0091] The preset proportion coefficient (i.e. δ1 and δ2 in the formula) refers to a set of weight parameters artificially or systemically set in advance when the system assesses the scheduling task conflict risk, which is used to reflect the relative influence degree of different reference values such as the organization command convergence reference value OCC-RV and the scheduling domain boundary ambiguity reference value DDBF-RV on the overall conflict risk coefficient DCRC.

[0092] Specifically, δ1 represents the influence weight of the organization command convergence degree on the conflict risk in the overall risk assessment; and δ2 represents the influence weight of the scheduling domain boundary ambiguity degree on the conflict risk. By assigning different proportional coefficients to the two key indicators, the sensitivity of the system to different conflict sources can be flexibly adjusted to adapt to the scheduling strategy in different scenarios. For example, in large-scale scheduling involving multiple agencies, more attention may be paid to the weight of OCC-RV (δ1 is larger); and in cross-regional joint scheduling, more attention may be paid to the risk brought by the ambiguity of the regional boundary (δ2 is larger). Therefore, the preset proportional coefficient not only reflects the understanding and preference of the model designer for the risk source, but also provides an adjustable intelligent evaluation strategy for the scheduling system.

[0093] According to the scheduling task conflict risk coefficient, the greater the organization command convergence reference value generated by analyzing the probability of different agencies issuing the same scheduling target for the same resource category in the detection window, the greater the scheduling domain boundary ambiguity reference value generated by analyzing the clear degree of the boundary between the scheduling tasks of multiple functional agencies in the detection window. Then, the greater the scheduling task conflict risk coefficient generated by the intelligent evaluation of the conflict risk of the current scheduling task by the pre-trained machine learning model, the greater the probability of the conflict risk of the current scheduling task, and vice versa.

[0094] The scheduling task conflict risk coefficient generated by the intelligent evaluation of the conflict risk of the current scheduling task by the pre-trained machine learning model is compared and analyzed with the pre-set scheduling task conflict risk coefficient reference threshold value to determine whether the current scheduling task has a conflict risk. The judgment logic is as follows:

[0095] If the scheduling task conflict risk coefficient is greater than the pre-set scheduling task conflict risk coefficient reference threshold value, it is determined that there is a conflict risk between the scheduling tasks; if the scheduling task conflict risk coefficient is less than or equal to the pre-set scheduling task conflict risk coefficient reference threshold value, it is determined that there is no conflict risk between the scheduling tasks.

[0096] When the machine learning model determines that multiple scheduling tasks have a conflict risk, all scheduling instructions are not executed, and the arbitration window with a short time delay is entered. In the arbitration window, the urgency, influence range and occupation state of the vehicle resources of each scheduling task are dynamically evaluated, and based on the pre-set task priority and scheduling influence factor, real-time conflict resolution and task fusion processing are performed to realize optimal allocation and efficient scheduling of limited resources;

[0097] When the machine learning model determines that there is a risk of resource conflict among multiple scheduling tasks, the system does not immediately execute the scheduling instructions, but enters a short time delay "arbitration window". The purpose is to provide a high real-time decision buffer for the scheduling system to avoid problems such as repeated occupation of vehicle resources, mutual coverage of instructions, and conflict of transportation paths caused by concurrent conflicts. In the arbitration window, the system dynamically evaluates the urgency of each scheduling instruction (such as task response time limit, disaster area urgent level), the influence range (such as the number of people affected, the radiation radius of materials), and the current vehicle resource occupation state (such as whether it can be allocated, whether it is executing a task, whether it is in a maintenance cooling period). These evaluation results will be fused and analyzed with the preset task priority level, scheduling influence factor, etc., to calculate the comprehensive scheduling priority index of each task. Based on the index, the system will perform conflict resolution operations, such as selecting the optimal task execution, reasonably merging tasks (path fusion, batch distribution), or delaying the allocation of resources for low-priority tasks. The core role of this step is to ensure that in the multi-source concurrent and resource-intensive emergency scenario, the scheduling system can make non-blind decisions and not miss high-priority tasks, truly implementing "on-demand allocation and efficient execution" of limited resources, and significantly improving the intelligent level and response efficiency of the entire emergency logistics system.

[0098] When the machine learning model determines that multiple scheduling tasks exist conflict risk, do not execute all scheduling instructions, enter a short time delay arbitration window, dynamically evaluate the urgency of each scheduling task, the influence range and the occupation state of vehicle resources, and based on the preset task priority and scheduling influence factor, real-time conflict resolution and task fusion processing is performed to realize the specific steps of optimal allocation and efficient scheduling of limited resources as follows:

[0099] When multiple scheduling tasks are detected to have resource conflicts, enter the arbitration window, and quantitatively score the priority of all conflict tasks. In order to more comprehensively evaluate the execution urgency and scheduling value of the task, the emergency degree, the remaining schedulable time and its influence range in the disaster are considered, and a priority scoring function is constructed, the formula is as follows:

[0100]

[0101] , in the formula, P q is the task priority score, which represents the final priority score of task q, which is the core reference value for comparing the priority of multiple conflict tasks, U q is the task urgency coefficient, which is used to measure the execution urgency of the task, such as whether it relates to life safety, public health, infrastructure, etc., D q is the remaining schedulable time of the task, which represents the remaining time between the current system time and the task must be completed time, I qis the task impact range index, representing the number of affected people, geographical range, and material coverage, λ is the time compression adjustment factor, a nonlinear adjustment parameter to enhance the sensitivity to tasks with near deadline, α1, α2 and α3 are weight coefficients of the task urgency coefficient U q , the remaining schedulable time D q and the task impact range index I q , and α1+α2+α3=1.

[0102] The above steps are used to quantify the scheduling urgency and importance of conflicting tasks, helping the system to identify the tasks that need to be responded to most urgently. By introducing a multi-dimensional index to build a priority scoring model, the urgency, remaining time and impact range of the task are quantified, providing a decision basis for subsequent conflict handling.

[0103] After the priority calculation is completed, the competition degree between tasks in the resource layer is evaluated, and the conflict risk when the same resource is called by multiple tasks at the same time is focused on. By using the resource usage flag and the time overlap ratio, a resource conflict intensity model between tasks is constructed, as follows:

[0104]

[0105] , where C qp is the resource conflict intensity, representing the resource-level conflict degree between the scheduling task q and the scheduling task p in the scheduling system, h is the total number of resources used by the scheduling task, δ qk and δ pk are binary variables representing whether the resource k is used by the scheduling task q or the scheduling task p, T qk is the actual occupation time interval of the resource k by the scheduling task q, T pk is the actual occupation time interval of the resource k by the scheduling task p, and γ is the conflict penalty index, used to amplify the conflict risk caused by the greater overlap time.

[0106] The above steps are used to calculate the overlap degree of each scheduling task in resource usage, revealing the potential scheduling interference between tasks due to sharing resources. By constructing a resource conflict tensor, the system can quantify the scheduling conflict risk in the resource dimension, laying a foundation for reasonable allocation of resources.

[0107] After the task priority and resource conflict intensity are quantified, the scheduling tasks are fused to score to comprehensively judge the rationality of their execution and the cost of resource scheduling. A scheduling fusion scoring function is introduced, as follows:

[0108]

[0109] , where S qFinal scheduling score of task q, β qp is a conflict impact factor, which is set according to the level difference, urgency difference and other factors between tasks, and is used to adjust different conflicts to tasks.

[0110] The above steps are used to integrate the task priority and the resource conflict strength to calculate the final scheduling execution score of each task. The system can determine the execution order of the tasks, the fusion possibility or the delay processing strategy, so as to realize the optimal scheduling decision in the conflict environment.

[0111] The present application can effectively identify and resolve the resource conflict problem between scheduling instructions in a multi-department concurrent scheduling environment by unified access, standardized processing and feature modeling of multi-source scheduling instructions, combined with intelligent identification and dynamic arbitration mechanism of conflict risk by a machine learning model, avoid task failure and resource waste caused by repeated allocation or scheduling interference; at the same time, by introducing a short delay arbitration window to realize flexible control and priority allocation of scheduling instructions, limited emergency vehicle resources can be dynamically reorganized and optimally allocated according to the task urgency and influence range, so as to significantly improve the material distribution continuity, scheduling stability and overall collaborative efficiency in the emergency response process.

[0112] The above formulas are dimensionless values, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the latest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0113] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above drawings and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

[0114] It should be noted that in this text, if there are relationship terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or device including the element.

[0115] It should be understood that the size of the sequence number of the above processes does not mean the order of execution in various embodiments of the present application, and the execution order of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0116] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0117] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0118] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment according to actual needs.

[0119] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically, or two or more units can be integrated into one unit.

[0120] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of changes or replacements within the technical scope disclosed in the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0121] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without deviating from the spirit and scope of the present application. Therefore, the above figures and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

Claims

1. A method for dispatching emergency material delivery vehicles based on collaborative support, characterized in that, Includes the following steps: Instructions from different platforms are uniformly integrated into a centralized management scheduling system, which collects and aggregates scheduling requests and resource information from various departments. The acquired scheduling requests and resource information are standardized and preprocessed, and all processed valid scheduling information is stored in a unified data set according to a predetermined data structure to support subsequent feature extraction and intelligent analysis. Feature engineering techniques are used to extract key indicators reflecting potential resource conflicts between scheduling tasks from the constructed dataset, and multi-dimensional comprehensive analysis is performed on the extracted key indicators to quantify the potential conflict status in the current scheduling tasks. The quantified key indicators of scheduling conflicts are used as feature vectors and input into a pre-trained machine learning model. The machine learning model is then used to intelligently assess the conflict risk of the current scheduling task and determine whether there is a conflict risk in the current scheduling task. When the machine learning model determines that multiple scheduling tasks are at risk of conflict, it does not execute all scheduling instructions and enters a short-delay arbitration window. Within the arbitration window, it dynamically evaluates the urgency, scope of impact, and vehicle resource occupancy status of each scheduling task. Based on the preset task priority and scheduling impact factors, it performs real-time conflict resolution and task fusion processing to achieve optimal allocation and efficient scheduling of limited resources. Feature engineering techniques are used to extract key indicators reflecting potential resource conflicts between scheduling tasks from the constructed dataset. The extracted key indicators include the probability that different organizations issue the same scheduling target for the same resource category and the clarity of the boundary division between scheduling tasks of multiple functional organizations. The probability that different organizations issue the same scheduling target for the same resource category and the clarity of the boundary division between scheduling tasks of multiple functional organizations are analyzed under the detection window to generate organizational command convergence reference values ​​and scheduling domain boundary fuzzy reference values. The potential conflict status in the current scheduling task is quantified by organizational command convergence reference values ​​and scheduling domain boundary fuzzy reference values. The specific steps for analyzing the clarity of boundary delineation between scheduling tasks of multiple functional agencies under a detection window to generate fuzzy reference values ​​for scheduling domain boundaries are as follows: The scheduling tasks issued by multiple functional departments under a certain detection window are constructed into a task scheduling domain cross-graph, where each node represents a scheduling task, and edges indicate the repetition of target regions between scheduling tasks. Then, the cross-reference index of each scheduling task is calculated, and the calculation expression is as follows: The formula is as follows: , It is a scheduling task Cross-linking index, and These are the scheduling tasks and scheduling tasks A collection of resources It is a functional department difference function used to determine whether two tasks come from different functional departments. and They represent the scheduling tasks respectively. and scheduling tasks The logo of the organization to which it belongs; Is related to the task A set of tasks that have overlapping resources; After obtaining the cross-reactivity index of all tasks, the degree of boundary ambiguity of the scheduling domain is further extracted at the global level to generate a reference value for the boundary ambiguity of the scheduling domain, so as to quantify the "instruction overlap disorder" of the overall scheduling domain. The generation formula is as follows: In the formula, It is a fuzzy reference value for the scheduling domain boundary. It is the set of cross-functional indices for all tasks. , This refers to the total number of all scheduling tasks issued by multiple functional departments within the current detection window. It takes the maximum value among the cross-reactivity indices of all scheduled tasks to identify the task with the most severe resource conflict. It is the total number of organizations participating in the scheduling. It is the scheduling responsibility domain split degree, which represents the number of tasks of the same resource category that are simultaneously scheduled from different functional departments.

2. The emergency material delivery vehicle dispatching method based on collaborative support according to claim 1, characterized in that, Unifying instructions from different platforms into a centralized management scheduling system involves the following key steps: First, we built an interface module that supports access to multiple data sources, enabling various functional departments to send scheduling instructions to the unified platform through API calls, data subscriptions, and message bus methods. Secondly, establish corresponding conversion rules and field mapping relationships for the instruction formats of various institutions, and convert the instruction content in different systems into a unified standard data model; Then, all the dispatch instructions that have been connected and standardized are uniformly merged into the core command center of the dispatch system and linked with the vehicle, material, and personnel resource databases to form a complete task-resource association view.

3. The emergency material delivery vehicle dispatching method based on collaborative support according to claim 1, characterized in that, The quantized organizational command convergence reference value and the fuzzy reference value of the scheduling domain boundary are used as feature vectors and input into a pre-trained machine learning model. The machine learning model generates a scheduling task conflict risk coefficient, and the conflict risk of the current scheduling task is intelligently evaluated based on the scheduling task conflict risk coefficient to determine whether the current scheduling task has a conflict risk.

4. The emergency material delivery vehicle dispatching method based on collaborative support according to claim 3, characterized in that, The conflict risk coefficient of the current scheduled task, generated by the intelligent assessment of the conflict risk of the current scheduled task using a pre-trained machine learning model, is compared and analyzed with a pre-set reference threshold for the conflict risk coefficient of the current scheduled task to determine whether there is a conflict risk in the current scheduled task. The judgment logic is as follows: If the scheduling task conflict risk coefficient is greater than the preset scheduling task conflict risk coefficient reference threshold, it is determined that there is a conflict risk between the scheduling tasks; if the scheduling task conflict risk coefficient is less than or equal to the preset scheduling task conflict risk coefficient reference threshold, it is determined that there is no conflict risk between the scheduling tasks.

5. The emergency material delivery vehicle dispatching method based on collaborative support according to claim 4, characterized in that, When a machine learning model determines that multiple scheduling tasks have a risk of conflict, it does not execute all scheduling instructions and enters a short-delay arbitration window. Within the arbitration window, it dynamically evaluates the urgency, scope of impact, and vehicle resource occupancy status of each scheduling task. Based on preset task priorities and scheduling impact factors, it performs real-time conflict resolution and task fusion processing to achieve optimal allocation and efficient scheduling of limited resources. The specific steps are as follows: When multiple scheduling tasks are detected to have resource conflicts, the arbitration window is activated to quantify the priority of all conflicting tasks. To more comprehensively assess the urgency and scheduling value of tasks, a priority scoring function is constructed, taking into account the urgency of the tasks, the remaining schedulable time, and the scope of their impact in a disaster. The formula is as follows: In the formula, It is a task priority rating. This is the task urgency coefficient, used to measure the urgency of task execution. It is the remaining schedulable time of the task. It is the task's impact range index. It is a time compression adjustment factor. , and These are the mission urgency levels. Remaining schedulable time for the task and the scope of the task's impact index The weighting coefficients, ; After priority calculation, the degree of competition between tasks at the resource level is assessed, focusing on the conflict risk when the same resource is called by multiple tasks simultaneously. A resource conflict intensity model between tasks is constructed by combining resource usage indicators with time overlap ratio, as shown in the following formula: In the formula, It represents the intensity of resource conflicts, indicating the scheduling task. and scheduling tasks The degree of resource-level conflicts existing in the scheduling system This is the total number of resources used by the scheduled task. and It is a binary variable representing resources. Is the task scheduled? Or scheduling tasks use, It is a scheduling task Resources The actual time interval occupied, They represent the scheduling tasks respectively. Resources The actual time interval occupied, It is a conflict punishment index; After quantifying task priority and resource conflict intensity, each scheduled task is given a fusion score to comprehensively assess its execution rationality and resource scheduling cost. A scheduling fusion scoring function is introduced, as shown in the following formula: In the formula, Indicates task The final scheduling score, It is a factor influencing conflict.

Images