Vehicle matching method and system based on multi-level transport resource pool

Abstract

本申请公开了一种基于多层级运力资源池的车辆匹配方法及系统，涉及物流技术领域，方法通过基础层、能力层、场景层分层组织车辆多维数据并建立联合索引，实现运力资源池的查询与检索；通过熵权法自动计算四维匹配特征的信息熵与差异系数，得到各维度客观权重向量；通过硬约束过滤剔除不满足基础适配条件的车辆，并计算各候选车辆到正负理想解的相对贴近度，实现多维度精准排序；在匹配中断事件触发时从排序序列中自动选取备选车辆替换，保障匹配结果连续性与稳定性；将匹配结果同步推送至装载优化系统、路径优化系统和调度系统，实现多系统联动执行；基于履约完成数据通过随机森林回归模型持续更新各维度客观权重向量，形成数据驱动权重迭代闭环。

Claims

1. A vehicle matching method based on a multi-level transportation capacity resource pool, characterized in that, The vehicle matching method includes: Step S1: Obtain vehicle basic parameters, vehicle performance indicators, and vehicle scenario adaptation tags, and construct a multi-level transportation capacity resource pool according to the basic layer, capability layer, and scenario layer. Establish a multi-dimensional joint index for the multi-level transportation capacity resource pool to obtain an indexed transportation capacity resource pool. Step S2: Extract the matching feature vectors of the current order from the indexed transportation capacity resource pool in four dimensions: basic adaptation, capability compliance, scenario fit, and collaborative efficiency. Calculate the information entropy and difference coefficient of the matching feature vectors using the entropy weight method to obtain the objective weight vectors for each dimension. Step S3: Filter vehicles in the indexed capacity resource pool that do not meet the hard constraints in the basic adaptation dimension to obtain a candidate capacity set, and calculate the relative proximity of each vehicle in the candidate capacity set to the positive ideal solution and the negative ideal solution according to the objective weight vector of each dimension, and sort them in descending order of the relative proximity to obtain a sorted capacity sequence. Step S4: Select the vehicle with the highest relative proximity from the sorted capacity sequence as the matching vehicle, and when the matching interruption event is triggered, select alternative vehicles from the remaining vehicles in the sorted capacity sequence in descending order of relative proximity to replace the matching vehicle, so as to obtain the final matching vehicle. Step S5: Push the vehicle parameters of the finally matched vehicle to the external loading optimization system to generate a loading plan, push them to the external route optimization system to correct the transportation route, and push them to the external dispatching system to generate a transportation task order, thereby obtaining the linkage execution result; Step S6: Obtain the performance completion data from the linkage execution results, input the performance completion data into the random forest regression model to predict and update the objective weight vectors of each dimension, and obtain the updated objective weight vectors for the next round of matching.

2. The vehicle matching method according to claim 1, characterized in that, Step S1: Obtain vehicle basic parameters, vehicle performance indicators, and vehicle scenario adaptation tags, and construct a multi-level transportation capacity resource pool according to the basic layer, capability layer, and scenario layer. Establish a multi-dimensional joint index for the multi-level transportation capacity resource pool to obtain an indexed transportation capacity resource pool, including: Step S1.1: Obtain the vehicle model, rated load, cargo box type, cargo box size, power type, and current vehicle status of each vehicle from the external vehicle management system, and summarize them according to the vehicle identifier to obtain the vehicle basic parameter set; Step S1.2: Using the vehicle identifier in the vehicle basic parameter set as the association key, extract the historical on-time delivery rate and historical cargo damage rate of each vehicle from the pre-acquired historical order database, and extract the vehicle basic parameters of each vehicle from the vehicle basic parameter set, and integrate them to obtain the vehicle-related dataset. Step S1.3: Generate product specification adaptation labels based on cargo box type, loading and unloading point adaptation labels based on power type, and path adaptation labels based on vehicle model in the vehicle basic parameter set, and integrate them into the vehicle association dataset to obtain a full-dimensional vehicle dataset. Step S1.4: Map the full-dimensional vehicle dataset to the basic layer, capability layer, and scenario layer according to attribute categories, and aggregate the three layers with the vehicle identifier as the association key to obtain a multi-level transportation capacity resource pool. Step S1.5: Perform outlier detection on the vehicle data of each layer in the multi-level transportation capacity resource pool, remove vehicle records with missing parameters or indicators exceeding the normal range, and use the same vehicle model mean method to fill in the missing values ​​for the remaining vehicle records to obtain the cleaned multi-level transportation capacity resource pool. Step S1.6: Establish a joint index on the cleaned multi-level transportation capacity resource pool in terms of vehicle type, cargo box type, vehicle status, product specification adaptation, and loading / unloading point adaptation to obtain an indexed transportation capacity resource pool.

3. The vehicle matching method according to claim 1, characterized in that, Step S2: Extract matching feature vectors for the current order from the indexed capacity resource pool across four dimensions: basic compatibility, capability compliance, scenario fit, and collaborative efficiency. Calculate the information entropy and difference coefficient of the matching feature vectors using the entropy weight method to obtain objective weight vectors for each dimension, including: Step S2.1: Obtain the cargo quantity, product type, loading and unloading point configuration, and target transportation route of the current transportation order from the external order system, and integrate them into the order demand data of the current transportation order; Step S2.2: Using the order demand data of the current transportation order as a one-dimensional reference, extract the corresponding tags of each candidate vehicle in the basic layer, capability layer and scenario layer from the indexed transportation capacity resource pool, and construct a four-dimensional original matching feature matrix; Step S2.3: Normalize the features of each dimension in the original four-dimensional matching feature matrix using the range normalization method to obtain the standardized matching feature matrix; Step S2.4: Calculate the eigenvalue weight of each dimension of the standardized matching feature matrix, and calculate the Shannon information entropy of each dimension based on the eigenvalue weight of each dimension. Step S2.5: Subtract the Shannon information entropy of each dimension from 1 to obtain the difference coefficient of each dimension, and normalize the difference coefficient of each dimension to obtain the objective weight vector of each dimension.

4. The vehicle matching method according to claim 1, characterized in that, Step S3: Filter vehicles in the indexed capacity resource pool that do not meet the hard constraints in the basic adaptation dimensions to obtain a candidate capacity set. Calculate the relative proximity of each vehicle in the candidate capacity set to the positive and negative ideal solutions based on the objective weight vectors of each dimension. Sort the vehicles in the candidate capacity set in descending order of relative proximity to obtain a sorted capacity sequence, including: Step S3.1: Determine whether the remaining load of each vehicle in the indexed transportation capacity resource pool meets the current order's cargo volume requirements, whether the cargo box type matches the order's product specification type, and whether the vehicle's current status is idle. Remove vehicles that do not meet at least one of these conditions to obtain a candidate transportation capacity set. Step S3.2: Extract the feature values ​​of each vehicle in the four dimensions of basic adaptation, capability compliance, scenario fit, and collaborative efficiency from the candidate capacity set and perform normalization processing to construct a candidate capacity decision matrix; Step S3.3: Multiply the objective weight vectors of each dimension with the candidate capacity decision matrix to obtain the weighted normalized decision matrix; Step S3.4: Take the maximum value of each dimension from the weighted normalized decision matrix to form a positive ideal solution, and take the minimum value of each dimension to form a negative ideal solution; Step S3.5: Calculate the Euclidean distance from each vehicle in the weighted normalized decision matrix to the positive ideal solution and the negative ideal solution, and calculate the relative proximity of each vehicle based on the Euclidean distance; Step S3.6: Sort the vehicles in the candidate capacity set in descending order according to the relative proximity to obtain the sorted capacity sequence.

5. The vehicle matching method according to claim 1, characterized in that, Step S4: Select the vehicle with the highest relative proximity from the sorted capacity sequence as the matching vehicle, and when a matching interruption event is triggered, select alternative vehicles from the remaining vehicles in the sorted capacity sequence in descending order of relative proximity to replace the matching vehicle, thus obtaining the final matched vehicle, including: Step S4.1: Select the vehicle with the highest relative proximity from the sorted capacity sequence as the initial matching vehicle, and extract the candidate vehicle sequence from the remaining vehicles in the sorted capacity sequence in descending order of relative proximity. Step S4.2: Collect in real time the BeiDou positioning data, vehicle status data, order system occupancy information, and electronic fence constraint change information of the initially matched vehicle, and integrate them into matching monitoring data; Step S4.3: Determine the interruption event of the matching monitoring data. When any of the following is detected: vehicle fault alarm, abnormal vehicle status, temporary high-priority order occupation, or change of loading and unloading point constraints, it is marked as a matching interruption event triggered. Step S4.4: In response to the matching interruption event, check each candidate vehicle in the candidate vehicle sequence in descending order of relative proximity to see if it meets the basic adaptation hard constraints of the current order, and select the first candidate vehicle that meets the hard constraints. Step S4.5: Replace the initial matched vehicle with the first candidate vehicle that meets the hard constraints to obtain the final matched vehicle. If there is no matching interruption event, the initial matched vehicle is used as the final matched vehicle.

6. The vehicle matching method according to claim 1, characterized in that, Step S5 involves pushing the vehicle parameters of the finally matched vehicle to an external loading optimization system to generate a loading plan, simultaneously pushing them to an external route optimization system to correct the transportation route, and simultaneously pushing them to an external dispatching system to generate a transportation task order, thus obtaining the linked execution result, including: Step S5.1: Extract the vehicle type, cargo box size, load parameters, braking performance, and target transportation route information of the finally matched vehicle, and integrate them into a standardized vehicle parameter package; Step S5.2: Push the standardized vehicle parameter package to the external loading optimization system, and receive and confirm the loading scheme generation result; Step S5.3: The standardized vehicle parameter package is synchronously pushed to the external route optimization system, and the transportation route correction results are received and confirmed. Step S5.4: The standardized vehicle parameter package is synchronously pushed to the external dispatch system, and the transportation task order generation result is received and confirmed. Step S5.5: Summarize the loading scheme generation results, the transportation route correction results, and the transportation task order generation results to generate a linkage execution result.

7. The vehicle matching method according to claim 1, characterized in that, Step S6: Obtain performance completion data from the linked execution results, input the performance completion data into the random forest regression model to predict and update the objective weight vectors of each dimension, and obtain the updated objective weight vectors for the next round of matching, including: Step S6.1: Extract the on-time delivery rate, cargo damage control rate, loading plan execution status, route correction times, and transportation task order completion status of the current matching from the linkage execution results, and integrate them into a fulfillment completion dataset; Step S6.2: Sort the indicators in the performance completion dataset according to time series, and use the forward imputation method to fill in the missing indicators to obtain the performance feature vector; Step S6.3: Input the performance feature vector into the pre-trained random forest regression model, and output the predicted weight update amount through the decision tree of the random forest regression model; Step S6.4: The predicted weight update amount is superimposed on the objective weight vector of each dimension to obtain the updated objective weight vector; Step S6.5: Write the updated objective weight vector back to the configuration layer of the indexed capacity resource pool for input of objective weight vectors of each dimension in the next round of vehicle matching.

8. A vehicle matching system based on a multi-level transportation capacity resource pool, wherein the vehicle matching system is applied to the vehicle matching method as described in any one of claims 1 to 7, characterized in that, The vehicle matching system includes: The indexed capacity resource pool construction module is used to obtain vehicle basic parameters, vehicle performance indicators, and vehicle scenario adaptation tags, and to construct a multi-level capacity resource pool in layers according to the basic layer, capability layer, and scenario layer. A multi-dimensional joint index is established on the multi-level capacity resource pool to obtain the indexed capacity resource pool. The objective weight vector calculation module for each dimension is used to extract the matching feature vectors of the current order in four dimensions: basic adaptation, capability compliance, scenario fit, and collaborative efficiency from the indexed transportation capacity resource pool, and to calculate the information entropy and difference coefficient of the matching feature vectors through the entropy weight method to obtain the objective weight vectors for each dimension. The sorted capacity sequence calculation module is used to filter vehicles in the indexed capacity resource pool whose basic adaptation dimensions do not meet the hard constraints to obtain a candidate capacity set, and calculate the relative proximity of each vehicle in the candidate capacity set to the positive ideal solution and the negative ideal solution according to the objective weight vector of each dimension, and sort them in descending order of the relative proximity to obtain the sorted capacity sequence. The final matching vehicle identification module is used to select the vehicle with the highest relative proximity from the sorted capacity sequence as the matching vehicle, and when the matching interruption event is triggered, to select alternative vehicles from the remaining vehicles in the sorted capacity sequence in descending order of relative proximity to replace the matching vehicle, so as to obtain the final matching vehicle. The final matching vehicle linkage execution module is used to push the vehicle parameters of the final matched vehicle to the external loading optimization system to generate a loading plan, simultaneously push them to the external route optimization system to correct the transportation route, and simultaneously push them to the external dispatching system to generate a transportation task order, thereby obtaining the linkage execution result; The objective weight vector iteration module for each dimension is used to obtain performance completion data from the linkage execution results, input the performance completion data into the random forest regression model to predict and update the objective weight vector for each dimension, and obtain the updated objective weight vector for the next round of matching.

9. An electronic device, characterized in that, The method includes a processor and a memory coupled to the processor, the memory storing program instructions that can be executed by the processor; when the processor executes the program instructions stored in the memory, it implements the vehicle matching method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions that, when executed by a processor, enable the vehicle matching method as described in any one of claims 1 to 7.

Images