A multi-queue parallel logistics reservation queuing system and method

Abstract

The present application provides a multi-queue parallel logistics reservation queuing system and method, which relates to the technical field of logistics management. The method includes aligning real-time data such as vehicle arrival times, parking lot capacities, loading and unloading equipment status, external traffic, and environmental information collected at multiple queuing queues and parking lots in a logistics park to generate a multi-dimensional data set. Subsequently, sparse reconstruction technology is used to extract the target data and abnormal indication data required for queuing scheduling from the multi-dimensional data set to detect potential problems such as faults, congestion, and vehicle reservation mismatches. And a multi-queue parallel queuing scheduling scheme is generated based on the target data and vehicle reservation information and sent to vehicles and operators through a network or a mobile terminal. When an abnormal situation is detected, the system dynamically adjusts the queuing scheme and performs cross-queue collaborative scheduling when necessary to improve vehicle turnover efficiency and park operation capacity.

Description

Technical Field

[0001] The present application relates to the technical field of logistics management, and in particular, to a logistics reservation queuing system and method with multi-queue parallelism. Background Art

[0002] With the continuous growth of the business volume in the logistics industry, logistics parks often need to manage multiple queuing queues simultaneously to efficiently dispatch vehicles and reduce congestion caused by queuing. However, in traditional queuing management, a single queue or a multi-queue strategy based on manual experience is usually adopted, making it difficult to timely obtain and comprehensively analyze multi-dimensional data such as vehicle arrival time, queue load, parking lot capacity, and loading and unloading equipment status, and unable to quickly respond to emergencies (such as equipment failures, sudden increases in vehicle flow, external congestion, etc.). This management method lacking dynamic scheduling capabilities and cross-queue coordination means easily leads to disorderly flow or repeated queuing of vehicles among different queues, increasing the management cost of the park and the waiting time of vehicles, and also reducing the overall operation efficiency and service level. Therefore, how to flexibly utilize real-time data and achieve cross-queue linkage in large-scale, multi-queue parallel scenarios has become an urgent problem to be solved in the current logistics queuing scheduling field. Summary of the Invention

[0003] In view of the deficiencies of the prior art, the present application provides a logistics reservation queuing system and method with multi-queue parallelism.

[0004] In a first aspect, the present application provides a logistics reservation queuing method with multi-queue parallelism, including:

[0005] Obtaining real-time operation data of multiple queuing queues and a parking lot in a logistics park, aligning the real-time operation data to form a multi-dimensional data set, where the real-time operation data includes vehicle arrival time, queue load, parking lot capacity, loading and unloading equipment status, and external traffic and environmental information;

[0006] Performing sparse reconstruction processing on the multi-dimensional data set to obtain target data for queuing scheduling and abnormal indication data for dynamically adjusting the queuing scheduling;

[0007] Based on the target data and combined with the vehicle reservation information of each queue, generating a queuing scheduling plan with multi-queue parallelism, where the queuing scheduling plan includes the queuing order of vehicles and parking lot arrangements;

[0008] Issuing the queuing scheduling plan to vehicles and operation terminals;

[0009] Based on the abnormal indication data, determining the abnormal situations that occur in the real-time update of the real-time operation data, and dynamically adjusting the queuing scheduling plan according to the abnormal situations to generate a new queuing scheduling plan.

[0010] As an alternative implementation, the generation of the queuing scheduling scheme with multi-queue parallelism includes:

[0011] Integrate the reservation period, vehicle arrival time, estimated waiting duration of each vehicle, and the remaining capacity of the parking lot, and determine the available time period and queue load of each loading and unloading device;

[0012] Queue and sort the vehicles and allocate parking lots according to the integrated data and preset priority rules, and output the queuing scheduling scheme with multi-queue parallelism.

[0013] As an alternative implementation, the generation of the new queuing scheduling scheme includes:

[0014] Monitor the abnormal indication data, and determine the vehicle status, parking lot capacity change, and loading and unloading equipment failure that cause the abnormality, and generate a monitoring result;

[0015] Based on the monitoring result, re-evaluate the queues and vehicle operation sequences affected by the abnormality, and make real-time corrections to the vehicle queuing sequence, parking lot arrangement, and loading and unloading equipment allocation, generate a new queuing scheduling scheme and send it to the corresponding vehicles and operation terminals.

[0016] As an alternative implementation, the generation of the new queuing scheduling scheme further includes:

[0017] Merge and analyze the vehicle reservation status, external traffic conditions, and loading and unloading equipment availability corresponding to multiple abnormal indication data within the same time period or adjacent time periods, and determine whether there is a large-scale or cross-queue linkage abnormality;

[0018] For the determined linkage abnormality situation, execute the multi-queue collaborative scheduling strategy, allocate spare parking spaces or additional loading and unloading equipment resources to the affected queues, and synchronously update the queuing sequence and operation time of each vehicle.

[0019] As an alternative implementation, the sparse reconstruction process for the multi-dimensional dataset includes:

[0020] Convert the multi-dimensional dataset aligned according to the time, queue, and device dimensions into a target data matrix;

[0021] Decompose the target data matrix to obtain a low-rank matrix and a sparse matrix, where the low-rank matrix represents the target data structure after interpolation and denoising; the sparse matrix is used to identify abnormal data corresponding to faults, congestion, and extreme values;

[0022] Use the low-rank matrix as the target data and the sparse matrix as the abnormal indication data.

[0023] As an alternative implementation, the forming of the multi-dimensional dataset includes:

[0024] Based on the vehicle reservation information and the real-time operation data, align the reservation time period, reservation queue, reservation handling equipment, and vehicle type corresponding to each vehicle with the actual arrival time and queuing waiting duration, and generate a reservation mapping matrix;

[0025] The sparse reconstruction processing of the multi-dimensional dataset further includes:

[0026] Based on the sparse matrix, determine the entries marked as abnormal, and through association with the reservation mapping matrix, cluster the feature vectors corresponding to the entries marked as abnormal to generate a first sub-sparse matrix and a second sub-sparse matrix;

[0027] Among them, the first sub-sparse matrix is used to characterize the abnormalities caused by vehicle reservation information; the second sub-sparse matrix is used to characterize the abnormalities caused by non-vehicle reservation information;

[0028] The elements in the first sub-sparse matrix and the second sub-sparse matrix are distinguished by the reservation deviation degree.

[0029] As an alternative implementation, the clustering of the feature vectors corresponding to the entries marked as abnormal includes:

[0030] Based on the sparse matrix, extract all the entries marked as abnormal therein, and add a multi-dimensional feature vector reflecting the business status to each abnormal record;

[0031] Obtain the operation demand information of the logistics park;

[0032] Based on the operation demand information, configure weight coefficients for each dimension in the multi-dimensional feature vector;

[0033] Compare the multi-dimensional feature vectors corresponding to each abnormal record, and use a preset clustering algorithm for grouping to generate a clustering result;

[0034] Based on the clustering result, generate the first sub-sparse matrix and the second sub-sparse matrix;

[0035] Among them, the multi-dimensional feature vector includes: the specific time period when the abnormality occurs and whether the adjacent time period is the peak operation period of the park;

[0036] The queue number corresponding to the abnormality, the handling equipment ID, and the parking lot area ID;

[0037] The reservation deviation degree;

[0038] Weather conditions, traffic congestion index, holiday information.

[0039] As an alternative implementation, the sparse reconstruction process for the cube further includes:

[0040] Construct a queue coupling matrix according to the business characteristics of multi-queue parallel queuing to characterize the resource sharing relationship and / or job collaboration relationship between multiple queues;

[0041] During the decomposition process of the target data matrix, introduce a parallel constraint term corresponding to the queue coupling matrix, and combine the parallel constraint term with the sparse matrix for iterative solution to identify anomalies that occur synchronously across queues or parallelism anomalies caused by shared resource failures;

[0042] Among them, in response to the parallel association degree between any two queues in the queue coupling matrix being greater than a first preset threshold, and the value of the corresponding abnormal entry in the sparse matrix exceeding a second preset threshold within the same or adjacent time periods, it is determined that the anomaly corresponding to this entry is a multi-queue parallel anomaly; the multi-queue parallel anomaly is used to perform linkage adjustment of queuing scheduling across queues.

[0043] In a second aspect, the present application provides a logistics reservation queuing system for multi-queue parallelism, including:

[0044] An acquisition unit that obtains real-time operation data of multiple queuing queues and parking lots in a logistics park, aligns the real-time operation data to form a multi-dimensional data set, and the real-time operation data includes vehicle arrival time, queue load, parking lot capacity, loading and unloading equipment status, and external traffic and environmental information;

[0045] A processing unit that performs sparse reconstruction processing on the multi-dimensional data set to obtain target data for queuing scheduling and anomaly indication data for dynamically adjusting the queuing scheduling;

[0046] A scheduling unit that, based on the target data and in combination with the vehicle reservation information of each queue, generates a queuing scheduling plan for multi-queue parallelism, and the queuing scheduling plan includes the queuing order of vehicles and parking lot arrangements; the queuing scheduling plan is sent to vehicles and operation terminals;

[0047] An adjustment unit that, based on the anomaly indication data, determines the abnormal conditions that occur in the real-time update of the real-time operation data, and dynamically adjusts the queuing scheduling plan according to the abnormal conditions to generate a new queuing scheduling plan.

[0048] Compared with the prior art, by aligning and sparsely reconstructing multi-source real-time data, the present application can not only effectively denoise and interpolate missing information, but also output in real time a sparse matrix that can indicate anomalies such as faults, congestion, or reservation mismatches. On this basis, by combining vehicle reservation information with the load status of each queue, a more accurate queuing scheduling scheme is generated, and dynamic adjustments are made to abnormal situations. Through this scheme, the logistics park can significantly improve the vehicle queuing efficiency and the coherence of operation connection, and shorten the detention time of vehicles during peak hours. At the same time, for sudden cross-queue linkage anomalies, multi-queue collaborative scheduling can also be achieved, so as to maximize the overall operation efficiency of the park while avoiding resource waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 FIG. is a flowchart of a logistics reservation queuing method with multi-queue parallelism provided by an embodiment of the present application;

[0050] Figure 2 FIG. is a flowchart of a method for generating a new queuing scheduling scheme provided by an embodiment of the present application;

[0051] Figure 3 FIG. is a schematic diagram of a logistics reservation queuing system with multi-queue parallelism provided by an embodiment of the present application;

[0052] Figure 4 FIG. is a schematic diagram of processing a multi-dimensional data set provided by an embodiment of the present application.

[0053] Reference numerals: 10, acquisition unit; 20, processing unit; 30, scheduling unit; 40, adjustment unit. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0055] See Figure 1 As shown, an embodiment of the present application provides a logistics reservation queuing method with multi-queue parallelism. The method includes steps S101 to S104, where:

[0056] S101: Obtain the real-time operation data of multiple queuing queues and parking lots in the logistics park, and align the real-time operation data to form a multi-dimensional data set. The real-time operation data includes vehicle arrival time, queue load, parking lot capacity, loading and unloading equipment status, and external traffic and environmental information;

[0057] S102: Perform sparse reconstruction processing on the multi-dimensional data set to obtain target data for queuing scheduling and anomaly indication data for dynamically adjusting the queuing scheduling;

[0058] S103: Generate a queuing scheduling plan with multi-queue parallelism based on the target data and in combination with the vehicle reservation information of each queue. The queuing scheduling plan includes the queuing order of vehicles and parking lot arrangements; send the queuing scheduling plan to the vehicles and operation terminals;

[0059] S104: Based on the abnormal indication data, determine the abnormal conditions that occur during the real-time update of the real-time operation data, and dynamically adjust the queuing scheduling plan according to the abnormal conditions to generate a new queuing scheduling plan.

[0060] This application first obtains real-time operation data such as vehicle arrival time, queue load, parking lot capacity, loading and unloading equipment status, and external traffic and environmental information from multiple queuing queues and parking lots in the logistics park. By aligning these data in terms of time and business dimensions, a multi-dimensional data set for queuing scheduling analysis is formed; the sparse reconstruction technology is introduced to analyze the multi-dimensional data set, so as to extract the target data representing the vehicle queuing status and site load, and capture the abnormal indication data that can indicate abnormal changes during real-time operation. The sparse reconstruction process can utilize elements such as historical queuing data, loading and unloading operation efficiency, and external traffic conditions, and combine algorithms such as compressive sensing or matrix completion to reduce the dimension and complete the original data, thereby saving computing resources while ensuring the analysis accuracy.

[0061] After obtaining the target data, in combination with the vehicle reservation information of each queue in the logistics park, according to conditions such as vehicle arrival time, queuing duration, remaining available capacity of the parking lot, and available time periods of the loading and unloading equipment, comprehensively calculate the queuing order and allocate parking lot positions for the corresponding vehicles to form a queuing scheduling plan with multi-queue parallelism; this plan gives priority to queue load balance and the operation connection requirements of each vehicle, and can consider the influence of external traffic and environmental factors within a reasonable range to improve the overall queuing efficiency and vehicle turnover rate.

[0062] Subsequently, send the queuing scheduling plan to the vehicle drivers and operation terminals through the network or mobile terminal, so that the vehicles can operate according to the queue order and parking arrangements of the plan. During the actual operation process, use the abnormal indication data to identify abnormal conditions such as new vehicle arrival volume, changes in parking lot capacity, loading and unloading equipment failures, or sudden traffic events in real time. Once it is determined that these abnormal conditions will impact or affect the original queuing scheduling plan, re-evaluate the load of each queue, equipment availability, and vehicle waiting duration according to the real-time operation data, dynamically adjust the queuing order and parking lot allocation strategy, generate a new queuing scheduling plan and send it in a timely manner to ensure the continuous and efficient progress of the multi-queue parallel logistics reservation queuing operation and cope with various emergencies.

[0063] Regarding the above S101:

[0064] In specific implementation, a number of data collection devices and information interaction interfaces are deployed within the logistics park, which are used to collect in real time information such as the vehicle arrival time, the number of queuing vehicles, the license plate recognition result, the remaining parking spaces in the parking lot, the traffic flow in and out, and the status of loading and unloading equipment for each queuing queue. At the same time, external traffic flow and environmental monitoring data, such as road congestion level, weather conditions, and air quality, can also be obtained based on the traffic management system or third-party services.

[0065] The above data collection can be carried out according to a unified sampling period or event-triggered mode, and data cleaning and screening rules are configured during the data collection process to filter abnormal or duplicate data, so as to reduce the impact of inaccurate information on subsequent analysis.

[0066] After the data collection is completed, the multi-source real-time operation data is imported into the queuing scheduling platform and aligned.

[0067] Specifically, in the time dimension, data from different sources and possibly out of sync can be mapped to the same time axis according to a unified timestamp or fixed sampling period; for missing or incomplete time period data, methods such as linear interpolation or moving average can be used for compensation.

[0068] In the business identification dimension, based on the queuing queue number, parking lot number, or loading and unloading equipment identification where the vehicle is located, the corresponding data items are integrated under a unified logical entity. For example, for data collected for the same vehicle at different time periods, they can be associated and merged through the license plate number or reservation number to ensure that the queuing and parking dynamics of the vehicle can be accurately located and traced during subsequent analysis.

[0069] Through the above data cleaning, compensation, and alignment operations, a multi-dimensional data set containing elements such as vehicle arrival information, queue load, parking lot capacity, loading and unloading equipment status, and external traffic and environmental information is finally formed, providing a data basis for subsequent sparse reconstruction analysis and queuing scheduling scheme generation. This multi-dimensional data set maintains consistency and integrity at both the time and business levels, thus being able to more effectively reflect the state changes of each queue in the logistics park at different moments.

[0070] Regarding the above S102:

[0071] The main purpose of this step is: in the case where there may be missing, duplicate, or noise interference in the collected multi-dimensional data, to extract the core data (i.e., target data) reflecting vehicle queuing and site load through sparse reconstruction technology, and identify abnormal change signals (i.e., abnormal indication data), providing a reliable reference for the generation and dynamic adjustment of the queuing scheduling scheme in the subsequent steps.

[0072] Among them, sparse reconstruction is based on the intrinsic low-rank or sparse characteristics of multi-dimensional data. By compressing or decomposing the data space under an appropriate mathematical model, it can effectively eliminate or compensate for noise and missing information.

[0073] Specifically, by connecting elements such as different queuing queues, time slices, parking lot capacities, and loading and unloading equipment statuses in a logistics park into a high-dimensional matrix or tensor, certain sparse patterns or low-rank modes can be discovered in this structure. For example, the change in the number of vehicles queuing in adjacent time periods is usually relatively smooth or shows periodicity, and there is also a correlation in the queue lengths of different queues during peak hours. Using sparse reconstruction algorithms can effectively capture these internal correlations. On the one hand, it can represent the main components of the data with fewer feature coefficients, and on the other hand, it can reasonably interpolate or infer missing values. At the same time, potential emergencies can be identified by measuring the reconstruction residuals and their abnormal degrees.

[0074] In specific implementation, before performing sparse reconstruction on the multi-dimensional data set, it is first necessary to further preprocess the multi-source data obtained in S101 and construct a suitable structural representation. According to the needs of the queuing scheduling platform, data such as the number of vehicles queuing, vehicle arrival time, parking lot capacity, and operating status of loading and unloading equipment can be regarded as matrices or tensors of different dimensions and spliced together.

[0075] For example, "time-queue-parking lot capacity-loading and unloading equipment status" can be regarded as a four-dimensional tensor, and a unified time step is divided in the time dimension (such as one time slice every 10 minutes or every 30 minutes), the specific queue number is identified in the queue dimension, and the available number of parking spaces or the available time period of the equipment is recorded in the capacity dimension and the loading and unloading equipment dimension. For external traffic and environmental information (such as traffic congestion index, weather conditions, etc.), it can be added as an exogenous variable to this tensor or an association matrix is established with it to ensure that external influencing factors can be comprehensively considered during subsequent reconstruction.

[0076] After constructing the tensor or matrix, a suitable sparse reconstruction model can be selected for solution. Common techniques include:

[0077] Compressed Sensing: After vectorizing the high-dimensional data, design a sparse basis or dictionary, perform sparse representation on the data, and use a small number of observation points for recovery;

[0078] Matrix Completion: For a matrix with missing values, with the goal of minimizing the nuclear norm, use the low-rank property of the matrix for interpolation and reconstruction;

[0079] Tensor Factorization: When the data exists in the form of a tensor, methods such as Canonical Decomposition (CANDECOMP), Parallel Factor Analysis (PARAFAC), and Tucker decomposition can be used to iteratively solve its core tensor and factor matrices to correct missing values and outliers.

[0080] In actual implementation, according to the data scale and computing resources, optimization methods (such as Alternating Direction Method of Multipliers ADMM, Stochastic Gradient Descent SGD, or Conjugate Gradient Method CG) can be used to solve the above model. For outliers or extreme points in the data, a robust term (such as norm or Huber loss, etc.) can be introduced into the objective function to reduce the impact of abnormal data on the reconstruction result, thereby further improving the stability and accuracy of the algorithm.

[0081] After completing the sparse reconstruction of multi-dimensional data, a difference or residual distribution between the denoised reconstructed data and the actual collected data will be obtained. According to the statistical analysis of this residual distribution in dimensions such as time and queue, the part that exceeds the preset threshold or significantly deviates from the normal pattern can be marked as abnormal indication data.

[0082] For example, when in a certain queue and time period, the gap between the reconstructed value and the measured value of the vehicle queue length is too large and this gap cannot be explained by normal fluctuations or exogenous variables, abnormal indication data can be generated to remind the subsequent steps to pay attention to or give priority to processing this queue. At the same time, for abnormalities in the parking lot capacity or the status of loading and unloading equipment, the same residual analysis method can be used to detect and alarm in time, thereby providing a basis for dynamic adjustment of queue scheduling.

[0083] In this way, through sparse reconstruction, the most business-related core information can be mined from multi-source data with noise and missing values, significantly reducing the estimation error of important indicators such as queue length and site capacity, and providing more reliable data support for queue scheduling. With the help of residual analysis and threshold judgment, sudden situations such as a large-scale influx of vehicles, equipment failures, or external extreme weather impacts can be quickly warned, converted into abnormal indication data, and support the system for real-time monitoring and response. Compared with directly performing complex modeling on all original multi-dimensional data, using sparse reconstruction can effectively compress the data dimension and reduce the processing overhead of high-dimensional noise, so that the entire queue scheduling process can still maintain good real-time performance in large-scale application scenarios.

[0084] As an alternative implementation manner, the generation of a queuing scheduling scheme with multi-queue parallelism includes:

[0085] Integrate the reservation period, vehicle arrival time, estimated waiting duration for each vehicle, and the remaining capacity of the parking lot, and determine the available periods and queue loads of each loading and unloading device;

[0086] Queue and sort the vehicles and allocate parking lots according to the integrated data and preset priority rules, and output a queuing scheduling scheme with multi-queue parallelism.

[0087] As an optional implementation manner, the generating a new queuing scheduling scheme includes:

[0088] Monitor the abnormal indication data, and determine the vehicle status, parking lot capacity change, and loading and unloading device failure that cause the abnormality, and generate a monitoring result;

[0089] Based on the monitoring result, re-evaluate the queues and vehicle operation sequences affected by the abnormality, and perform real-time correction on the vehicle queuing sequence, parking lot arrangement, and loading and unloading device allocation, generate a new queuing scheduling scheme and send it to the corresponding vehicles and operation terminals.

[0090] Please refer to Figure 2 , Figure 2 which is a flowchart of a method for generating a new queuing scheduling scheme provided by an embodiment of the present application, including steps S201~S202, where:

[0091] S201: Combine and analyze the vehicle reservation status, external traffic conditions, and loading and unloading device availability corresponding to multiple abnormal indication data within the same time period or adjacent time periods, and determine whether there is a large-scale or cross-queue linkage abnormality;

[0092] S202: For the determined linkage abnormality situation, execute a multi-queue collaborative scheduling strategy, allocate spare parking spaces or additional loading and unloading device resources to the affected queues, and synchronously update the queuing sequence and operation time of each vehicle.

[0093] In specific implementation, the queuing scheduling platform first reads the reservation information and actual operation data of each vehicle from the aligned multi-dimensional dataset, and processes them according to the following process:

[0094] First, the system will match different types of vehicles with the remaining capacity of the parking lot according to the vehicle reservation period, the expected arrival time of the vehicle, the vehicle type, and the external traffic environment. During this process, the available periods and queue load distributions of each loading and unloading device will be taken into account. If some devices are in a maintenance or high-load state, other devices will be preferentially allocated to avoid excessive congestion in local queues. Subsequently, the system will comprehensively consider the expected waiting duration of each vehicle and the queue length of its queue, and optimize and adjust the vehicle sequence. This optimization process can be based on pre-configured priority rules, such as vehicles in urgent need of departure first, perishable goods first, or vehicles with a relatively long actual waiting duration being appropriately advanced, etc.

[0095] Once the above data integration and priority rule application are completed, the queuing scheduling platform will output the result of "vehicle queuing sequence and parking lot arrangement", form a queuing scheduling plan with multiple queues in parallel, and send it to the driver and operating personnel through the network or terminal device. This plan will clearly specify information such as which queue the vehicle should enter, when to perform loading and unloading operations, and the corresponding parking space number, etc., so as to achieve efficient and orderly logistics queuing operations in the scenario of multiple queues in parallel.

[0096] During the normal queuing execution, the system continuously monitors the abnormal indication data output by sparse reconstruction, and locates and tracks abnormal situations that may affect queuing efficiency or pose safety risks. Specifically, the sparse matrix at the current moment can be periodically read in the background process , and the distribution and amplitude of abnormal elements are analyzed:

[0097] When it is detected that some vehicles have been detained for too long during the waiting process, or the remaining parking spaces in the parking lot are rapidly decreasing, or a fault warning occurs for the loading and unloading equipment, the system will associate such abnormal information with the vehicle, equipment, parking lot number, etc., to form a monitoring result. For the queues and vehicles affected by the abnormality in the monitoring result, the queuing scheduling platform will re-evaluate their waiting duration, equipment availability, and operation priority.

[0098] On this basis, the system will dynamically correct the queuing sequence, parking lot allocation, and loading and unloading equipment adjustment of these vehicles. For example, if the parking lot capacity is significantly reduced, some vehicles can be transferred to other reserved parking spaces; if a loading and unloading device stops due to a fault, the relevant vehicles will be preferentially transferred to the queue where the normal device is located; if it is found that a certain vehicle deviates significantly from the reserved arrival period, and this deviation affects the operation connection of other vehicles, the system will also adjust its queuing position accordingly. After the above corrections are completed, the system immediately generates a new queuing scheduling plan and sends this plan to the driver and on-site operators through the mobile terminal or scheduling terminal to respond to the identified abnormal situation in the shortest time.

[0099] When the system monitors the abnormal indication data, it is possible that multiple abnormalities break out simultaneously within the same time period or adjacent time periods. To further reduce the linkage risk, based on the original monitoring and scheduling, this embodiment adds the detection logic for large-scale or cross-queue linkage abnormalities:

[0100] First, the system will comprehensively analyze all newly generated abnormal data within a certain time window (such as 15 minutes or 30 minutes), and determine whether these abnormalities break out concentratedly in the same device, the same area, or the same time period by comprehensively comparing the vehicle reservation status, external traffic information, availability of loading and unloading equipment, etc. If it is confirmed that it belongs to a linkage abnormality, such as multiple devices almost simultaneously failing within the same time period, or the abnormal congestion state of a certain queue affecting other related queues, the system will trigger a multi-queue collaborative scheduling strategy.

[0101] In the multi-queue collaborative scheduling, the system uniformly evaluates the operating loads of the affected queues, and preferentially allocates available resources (such as spare parking spaces, spare loading and unloading equipment, or other vacant sites) to the queues at the peak of the abnormality. If the external traffic conditions deteriorate, such as the main road being unexpectedly closed, resulting in a large number of vehicles being unable to enter the site on time, the system will also reserve a certain buffer capacity in other queues accordingly, and synchronously update the queuing order and operation time of the vehicles. In this way, the entire queuing process is no longer a fine-tuning for a single point or a single queue, but through the linkage processing of the operation information of multiple queues, the overall queuing efficiency and the stability of vehicle turnover are maximally maintained.

[0102] After the scheduling platform completes this multi-queue collaborative scheduling, it will immediately push the updated queuing instructions to the terminal devices of relevant vehicles and on-site staff to ensure the rapid implementation of the scheduling plan for dealing with linkage abnormalities. For the abnormal indication data continuously collected subsequently, the system still continuously conducts dynamic tracking and evaluation. If a new linkage risk appears again, the corresponding collaborative strategy can be triggered again.

[0103] Please refer to Figure 4 , Figure 4 which is a schematic diagram of processing a multi-dimensional data set provided by an embodiment of this application.

[0104] As an alternative implementation, the sparse reconstruction processing of the multi-dimensional data set includes:

[0105] Converting the multi-dimensional data set aligned according to the time, queue, and device dimensions into a target data matrix;

[0106] Decomposing the target data matrix to obtain a low-rank matrix and a sparse matrix, where the low-rank matrix represents the target data structure after interpolation and denoising; the sparse matrix is used to identify abnormal data corresponding to faults, congestion, and extreme values;

[0107] Take the low-rank matrix as the target data and the sparse matrix as the anomaly indication data.

[0108] By introducing low-rank and sparse decomposition techniques in the sparse reconstruction stage, this application realizes the precise separation of noise and outliers, improves the detection and recognition capabilities for faults, congestion, and extreme values, and thus provides more targeted and robust support for queuing scheduling.

[0109] In the specific implementation, first, based on the alignment operation in S101, a multi-dimensional data set containing data elements such as vehicle arrival time, queue load, parking lot capacity, loading and unloading equipment status, and external traffic and environmental information is uniformly sorted in multiple dimensions such as time, queue, and equipment; subsequently, one or more target data matrices are formed according to the established time step and service identifier. To facilitate the decomposition process, it can be executed as follows:

[0110] In the aligned multi-dimensional data, each time slice may contain multiple queue information, corresponding equipment usage, and external traffic data, etc.; to balance the operation efficiency and data structure characteristics, it can be mapped into a two-dimensional matrix in the way of time × service metrics. For example, assuming there are m time slices and n service metrics (including queue length, parking lot capacity, loading and unloading equipment working status, etc.), then a target data matrix with the size of can be constructed. For missing or incomplete entries, appropriate placeholders or interpolations can be temporarily stored in this matrix to ensure the integrity of the matrix dimensions.

[0111] After obtaining the target data matrix , introduce low-rank-sparse decomposition (such as Robust PCA, etc.) to decompose it, which can be specifically expressed as:

[0112] ;

[0113] where is the low-rank matrix, is the sparse matrix.

[0114] The low-rank matrix reflects the target data structure after interpolation and denoising, and depicts the core change trend between vehicle queuing and parking lot capacity and other elements with fewer feature dimensions;

[0115] The sparse matrix is used to identify those abnormal data points that rarely appear in the target data matrix but have a high degree of mutation or deviation, and can indicate abnormal situations such as faults, congestion, and extreme values.

[0116] During the actual operation process, the decomposition process can be expressed as the following optimization problem, aiming to find by minimizing and these two terms and :

[0117] ;

[0118] where, indicates that the objective to be minimized is the matrix and , that is, to find these two matrices such that the value of the objective function is minimized under the satisfaction of the constraint conditions. is the constraint condition, requiring that the low-rank matrix and the sparse matrix add up to the original matrix . is the nuclear norm (i.e., the sum of singular values), used to ensure that has the low-rank property. The nuclear norm is the sum of all singular values of the matrix (singular values are a generalization similar to the eigenvalues of the matrix). Minimizing with the nuclear norm can effectively approximate the rank of the matrix, making most of the singular values of the matrix zero or tending to zero, ensuring that the matrix has a low rank (that is, the matrix structure is as simple and compact as possible, capturing the overall trend and structural information of the data); represents the norm of the sparse matrix , used to highlight and retain a finite number of outliers. The norm of the sparse matrix is defined as the sum of the absolute values of all elements in the matrix. By minimizing the norm of the sparse matrix , it can be ensured that most of the elements in the matrix tend to zero, only retaining a small number of non-zero values (outliers or noise), thereby realizing the detection and extraction of data anomalies or emergencies; is the balance coefficient, used to control the trade-off between the low-rank and sparse terms. Common solution methods include the alternating direction method of multipliers (ADMM), the augmented Lagrangian multiplier method, etc.; during the solution process, if it is found that some entries in the matrix are missing or have abnormal weights, further adaptive adjustment can be made to ensure the true effectiveness of the low-rank part and the high confidence of the sparse part.

[0119] After the above decomposition is completed, the matrix can be used as the denoised target data structure for load analysis and job time prediction in the subsequent queuing scheduling process, that is, as the target data for queuing scheduling; while the matrix The non-zero elements in often indicate obvious abnormal situations occurring in a certain time slice, queue, and device dimension, such as a sudden increase in the number of vehicle arrivals (which may lead to traffic congestion), the parking lot capacity exceeding the normal value range (which may indicate errors or omissions in the management system records), malfunctions of loading and unloading equipment, or strong impacts of extreme weather on logistics operations. Therefore, in this application, is regarded as abnormal indication data, which is convenient for subsequent steps to identify and respond to the above-mentioned emergencies.

[0120] Furthermore, the queuing scheduling platform can directly estimate the queue length, allocate the parking lot capacity, and schedule the loading and unloading equipment based on the target data structure obtained from so as to generate a more accurate queuing scheduling plan.

[0121] Once the amplitude or density of certain values in exceed the preset threshold, it means that the corresponding time slice or queue is in an abnormal state. The system can actively send out warning messages or automatically execute abnormal handling strategies, such as temporarily opening a standby parking lot, dispatching additional loading and unloading equipment, or recalculating the queuing order of vehicles.

[0122] After continuously obtaining new multi-dimensional data and appending it to , the decomposition process can be repeated regularly or in real time to update and to ensure the timeliness of monitoring abnormal situations and make the queuing scheduling always follow the changes of actual business.

[0123] By introducing low-rank matrix and sparse matrix decomposition, this specific implementation can more accurately distinguish normal business changes from potential faults or extreme values in a multi-source data environment that may contain noise and local missing data;

[0124] It can be seen that this implementation effectively separates the main business information and abnormal information by converting the aligned multi-dimensional data set into a target data matrix and performing low-rank matrix and sparse matrix decomposition, and solves the problem of inaccurate scheduling caused by noise interference, improving the system's recognition and processing capabilities for vehicle congestion, equipment failures, and extreme values.

[0125] As an optional implementation, the reservation information of vehicles in each queue is further incorporated into the sparse matrix to further complete more refined abnormal recognition and processing, so as to solve the technical problem that it is impossible to distinguish different types of problems such as mismatched reservation information, equipment failures, or external traffic anomalies when only relying on historical or external data. This solution can help the system more accurately locate the causes of abnormalities, improve the refined management level of queuing scheduling, and provide more targeted basis for subsequent dynamic adjustments.

[0126] In the aforementioned low-rank sparse decomposition process, the obtained sparse matrix is mainly used to identify those abnormal data points that deviate significantly from the target data structure. However, the abnormal deviation may be caused by various reasons: for example, the vehicle arrival batches far exceed the expectations, equipment failures cause queuing congestion, or the reserved time slots of some vehicles do not match the actual arrival times, etc.

[0127] If all the above reasons are indiscriminately incorporated into the same sparse matrix , although abnormalities can be detected as a whole, it is difficult to timely and accurately identify whether the error is caused by the mismatch of vehicle reservations or the abnormalities caused by external emergencies. Therefore, on the basis of retaining the original functions (detecting abnormalities and indicating data deviation situations), combined with the reservation information of each queue of vehicles, is further decomposed or labeled, so as to subdivide the abnormalities into:

[0128] Abnormalities caused by the mismatch of reservation information (for example, the actual arrival time of the vehicle deviates significantly from the reserved time);

[0129] Abnormalities caused by non-reservation factors such as equipment failures or external traffic congestion.

[0130] In specific implementation, when forming the aforementioned multi-dimensional data set or target data matrix, the reserved time slot, reserved queue, reserved loading and unloading equipment, vehicle type, etc. corresponding to each vehicle can be additionally aligned with fields such as the actual arrival time and queuing waiting duration to form a reservation mapping matrix or embed it into the same tensor.

[0131] Exemplarily, a reservation mapping matrix with a size of can be constructed, where M corresponds to the time slice or arrival time number, N represents the features related to vehicle reservations (such as reserved time period, reserved queue ID, reserved equipment ID, etc.), and the reservation mapping matrix can be associated with or under the same time index.

[0132] After decomposing the target data matrix to obtain , can be further subdivided, including:

[0133] For the entries marked as abnormal in , by associating with the reservation mapping matrix , the feature vectors or records of these abnormal entries are reclustered or decomposed again. Let:

[0134] ;

[0135] Among them, is used to characterize the anomalies caused by reservation mismatches (such as the vehicle arriving late, arriving early, or reservation information being lost, etc.), while is used to characterize the anomalies caused by other factors (such as equipment failures, external traffic delays, environmental emergencies, etc.).

[0136] In actual implementation, a threshold can be set according to the reservation deviation degree (such as the difference between the "actual arrival time - reserved arrival time" or other statistical indicators). If it exceeds a certain range, it can be preliminarily determined that the anomaly is mainly caused by reservation mismatch, and thus classified into ; otherwise, classified into .

[0137] For the anomaly entries that are still uncertain, they can be dynamically updated or label-corrected after the system collects more real-time data.

[0138] Furthermore, add constraint terms or penalty terms related to reservation deviation in the original optimization model. For example, in order to distinguish vehicle reservation factors from equipment or external factors, a specific sparse penalty coefficient can be assigned to reservation deviation in the optimization objective ; when there are obvious reservation deviations for some data points, automatically impose a higher sparse cost on the corresponding positions of so that it is easier to be extracted during matrix factorization in. This can achieve a rough division of reservation anomalies and other anomalies in one decomposition, reducing the difficulty of subsequent secondary processing.

[0139] When the system detects a significant deviation between the actual arrival time of the vehicle and the reservation information through , it can send a reservation anomaly warning to the queuing and scheduling platform; the system can select different scheduling strategies according to this type of anomaly, such as notifying the vehicle to enter early or postpone queuing, in order to minimize the impact on the overall scheduling efficiency.

[0140] For corresponding anomalies, such as equipment failures, road congestion, or extreme weather, etc., corresponding maintenance processes or emergency plans can be triggered, such as reallocating loading and unloading equipment, calling a standby parking lot, or coordinating with the local traffic department, in order to reduce the queuing delay caused by non-reservation factors.

[0141] After continuously obtaining new vehicle arrival records and reservation mappings, can be updated periodically or in real time and , incorporating the latest reservation data and measured data into the secondary decomposition or additional constraint model; for the already marked , Continue to make corrections to ensure that the system's recognition of different types of exceptions always remains accurate and timely.

[0142] In this way, while retaining the primary sparse decomposition function (identifying abnormal deviations), introducing vehicle reservation information to perform re-decomposition or constraint processing can clearly distinguish predictable but unmatched exceptions such as reservation defaults and early arrivals from truly uncontrollable equipment or external shock-type exceptions. Targeted scheduling or reminders are carried out for exceptions that do not match the reservation, preventing the system from misjudging ordinary deviations caused by reservation information as major faults or emergencies; thus, more flexible vehicle scheduling and resource allocation can be carried out at the queuing scheduling level, reducing unnecessary overall rearrangements or emergency measures and saving the operating costs of the logistics park.

[0143] Combined with the reservation mapping matrix , the reservation fulfillment status of different vehicles can be dynamically monitored, and additional control can be exerted on vehicles that deviate from the reservation information for a long time or frequently; at the same time, when major changes occur in the external environment (such as severe congestion), the system can also quickly identify the change and execute linkage emergency response.

[0144] It should be noted that the secondary decomposition or additional constraints in this solution are not limited to a single algorithm and can be adapted to different sparse decomposition or tensor decomposition models; in different business scenarios, the reservation mapping can also be replaced with other scenario-specific prior information (such as warehouse operation plans, batch order time windows, etc.) to achieve a wider range of applications.

[0145] As an optional implementation method, in order to more precisely and business-targetedly identify and process the entries marked as abnormal in the sparse matrix , on the basis of the original two types of exceptions, namely "vehicle reservation information" and "equipment or external factors", multi-dimensional business characteristics can be further combined and a unique clustering method can be designed. Through this clustering, various potential abnormal causes can be effectively distinguished, avoiding problems of over- or under-application in subsequent scheduling strategies and emergency measures, and improving the refined management level of multi-queue parallel queuing.

[0146] In specific implementation, first extract all the entries marked as abnormal from the sparse matrix , and add a multi-dimensional feature vector reflecting the business status to each abnormal record. This feature can include the time dimension (such as the specific period when the exception occurs and whether the adjacent period is the peak operation period of the park), queue and equipment characteristics (such as the queue number corresponding to the exception, the ID of the loading and unloading equipment, the ID of the parking lot area, etc.), the deviation degree between vehicle reservation and actual arrival (such as the absolute value of the difference between the reserved arrival time and the actual arrival time or the ratio of the vehicle waiting time to the theoretical time), and the external environment conditions (such as weather conditions, traffic congestion index, holiday information, etc.).

[0147] Since there are usually operating shifts or peak hours in a logistics park, when sorting out the characteristics of these abnormal entries, information such as the time axis, queue number, or geographical area can be preferentially referred to for pre-grouping to reduce confusion between different operation batches or completely different devices.

[0148] After obtaining the above multi-dimensional characteristics, a set of weighted similarity or distance metrics is defined to reflect the importance of each dimension.

[0149] For example, different weight coefficients can be configured for the time dimension, queue / device dimension, appointment deviation dimension, and external environment dimension according to the actual operation requirements of the logistics park. When the system needs to give priority to abnormal situations caused by equipment failures, a higher weight can be given to the queue / device dimension; if the impact of appointment defaults or large-scale early arrivals of vehicles on the park is more serious, a higher weight can be given to the appointment deviation dimension. Then, according to the set similarity (or distance) calculation method, the feature vectors of all abnormal entries are compared with each other, and a clustering algorithm suitable for large-scale, multi-dimensional data (such as hierarchical clustering, spectral clustering, or DBSCAN, etc.) is used to group them.

[0150] After obtaining the clustering results, comprehensive analysis can be performed on the abnormal entries in the same cluster based on their clustering centers or clustering characteristics.

[0151] For example, if a large number of abnormalities in a cluster occur at the same time period, on the same device, and the external congestion index is similar, it is more likely to be caused by local reasons such as equipment failures or road closures; if the abnormalities in another cluster all have significant appointment deviation characteristics and are distributed on different queues and devices, it indicates that such abnormalities are mainly due to problems such as vehicles not arriving on time and poor communication between the appointment system and drivers.

[0152] Based on this, the queuing scheduling platform can execute more targeted handling measures for different abnormal clusters. On the one hand, for the equipment failure cluster, the equipment inspection or repair process can be immediately started; on the other hand, for vehicles with serious appointment deviations, the queuing strategy can be adjusted, or corresponding thresholds can be set in the system to take restrictive measures against frequently defaulting vehicles. If the clustering results are still uncertain about the causes of some abnormal entries, they can be classified or corrected after more sampling data of time slices are collected later.

[0153] Through the above unique clustering method, this application uses a sparse matrix Based on the preliminary anomaly marking, a more refined distinction of anomaly types is achieved. Compared with only distinguishing between the two major categories of "reservation information mismatch" and "external failure", the weighted similarity and hierarchical clustering designed according to multiple business dimensions here can quickly discover the commonalities in the causes of anomalies during peak hours, at specific site equipment, or in special traffic environments. This can not only arrange the vehicle queuing order and resource allocation more reasonably, but also avoid resource waste or scheduling chaos caused by uniformly adopting the same emergency measures. Finally, by clustering and subdividing the anomaly entries and linking them with the queuing scheduling strategy, the overall efficiency and resilience of the multi-queue parallel queuing in the logistics park can be further improved.

[0154] Exemplarily, in a certain logistics park, queuing chaos often occurs during peak hours due to vehicles arriving early or late, and operation blockages are caused by loading and unloading equipment failures or external road congestion. The park management deployed data collection devices and information interaction interfaces within the logistics park to collect core operation data such as vehicle arrival times, queuing quantities, remaining parking spaces in the parking lot, and loading and unloading equipment status of each queue in real time. At the same time, it accessed the traffic management system to obtain external environment data such as congestion indices, weather conditions, and surrounding road construction information. After aligning these multi-source data according to dimensions such as time, queue number, and loading and unloading equipment, a target data matrix containing several dimensions is formed. , and uses the method of low-rank and sparse decomposition to be divided into a low-rank matrix and a sparse matrix used to identify anomalies .

[0155] During a typical early morning peak hour, the decomposition results show that the sparse matrix has a large number of non-zero anomaly entries in some time slices and specific queues, and its distribution also shows a certain degree of aggregation. To deeply analyze the causes of anomalies, the system extracts multiple feature vectors such as the deviation between vehicle arrival and reservation, queue number, loading and unloading equipment ID, weather, and congestion index for these anomaly entries in the sparse matrix , and performs hierarchical clustering under the weighted similarity model. After preliminary clustering, it is found that there is a large cluster of anomaly vehicles whose reservation arrival times differ significantly from the actual arrival times, and most of these vehicles did not inform the park in advance; further verification reveals that some carriers changed their schedules due to temporarily receiving other transportation tasks and did not update the reservation information in time, and were finally identified as anomalies of the "high deviation from reservation" type by the system.

[0156] For these vehicles, a strategy of batch release and dynamic adjustment can be adopted at the queuing scheduling end, which not only avoids affecting other vehicles that arrive on time, but also reduces the waste of parking resources caused by large-scale mismatches.

[0157] In another cluster of anomalies during the same morning shift, it was observed that they were mainly concentrated on the same loading and unloading equipment, and there were no obvious abnormalities in the external road conditions and weather. After on-site investigation, it was confirmed that the equipment had a sensor failure, which led to a significant reduction in the vehicle operation efficiency and indirectly caused a significant prolongation of the queuing time. For such "equipment failure" anomalies, the emergency response plan can be immediately activated, technicians can be arranged to repair the faulty equipment, and some of the vehicles originally planned to operate on this equipment can be allocated to the standby loading and unloading points, significantly alleviating the queuing congestion caused by the failure. Since this measure focuses on specific equipment and time periods under the guidance of the clustering results, it avoids blindly expanding the emergency response across the entire park and also shortens the fault handling time.

[0158] During subsequent operations, as new vehicle arrival data and reservation information continue to be incorporated, the system will regularly update the target data matrix and perform the above clustering analysis again after decomposing to obtain the sparse matrix If certain vehicles or loading and unloading equipment repeatedly appear in the anomaly entries in different time periods, they can be marked by the system as continuously abnormal or objects that need to be monitored key, for further follow-up investigation. By means of this process that is executed cyclically, the park management party continuously improves the dynamic understanding of vehicle reservation habits, equipment health status, and the impact of the external environment, can not only issue warnings and response measures in a timely manner, but also accumulate and analyze the anomaly patterns in the long term, so as to continuously optimize the queuing management strategy and equipment operation and maintenance plan, and ultimately achieve the improvement of the efficiency and stability of multi-queue parallel queuing operations.

[0159] As an alternative implementation method, when there may be coupling relationships such as resource sharing, operation linkage, or synchronous blocking between multiple queues, how to better capture and utilize the associations between these queues to solve the technical problem that it is difficult to detect cross-queue anomalies or optimize the overall queuing efficiency when only considering each queue as an independent dimension. This application further discloses an optional implementation method that can both maintain the core mechanism of sparse reconstruction and fully reflect the dynamic coupling and collaborative characteristics of each queue in the multi-queue parallel scenario, achieving more accurate anomaly detection and better queuing scheduling decisions.

[0160] In existing solutions, sparse decomposition usually treats the load or vehicle arrival information of each queue as independent columns or tensor dimensions. Although it can identify anomalies of a single queue on the timeline, when bottlenecks occur in cross-queue resources (such as shared loading and unloading equipment, shared buffers, or parking lot partitions), or when different queues experience synchronous delays during peak hours, simply splicing multiple queue data together often fails to highlight the coupling structure brought about by "parallelism", and it is also difficult to timely discover parallel anomalies that occur in multiple queues at the same time. To this end, on the basis of the original low-rank-sparse decomposition, a multi-queue parallel mapping or queue coupling matrix specifically used to characterize the parallel correlation of multiple queues can be introduced, and corresponding constraints or regularization terms can be added during the sparse decomposition process to model the linkage characteristics of multiple queues.

[0161] In specific implementation, the improvement of incorporating queue parallelism into sparse matrices is mainly reflected in the following aspects:

[0162] First, in order to quantify the parallel association between queues, a queue coupling matrix or parallel association matrix can be constructed in data preprocessing, denoted as G. G can be a matrix of size Q×Q, where Q represents the number of queues, and each element G(i,j) in G represents the degree of parallelism, resource sharing intensity, or job mutual exclusion / cooperation relationship between queue i and queue j.

[0163] For example:

[0164] When two queues often share the same batch of loading and unloading equipment or parking areas during peak hours, G(i,j) can be set to a larger positive value;

[0165] If some queues are mutually exclusive (e.g. a device can only be used by one of the queues at a time), a weight or symbol indicating exclusivity can be given to G(i,j) in the model;

[0166] If there is almost no job association between the two queues, G(i,j) can be close to 0.

[0167] Next, we form the target data matrix (or tensor) When the basic dimensions consistent with the existing implementation methods are retained, such as "time-queue-device status-external environment", etc., the above parallel association information is embedded in the regular term of the decomposition process in some way, or with the sparse matrix Used together to detect whether there is a cross-queue synchronization anomaly. For example, in the original decomposition target On the basis of G, add a parallel constraint or parallel regularization related to G, denoted as This parallel constraint aims to reflect the following engineering expectation: if the parallel correlation degree of two queues is relatively high (G(i,j) is large), then the possibility of anomalies occurring in the same or similar time slices is relatively greater or more common; or, once a loading and unloading device of a certain queue fails, resulting in a sharp increase in queuing, then another queue with a high correlation degree with it may also simultaneously experience an increase in queuing pressure, etc.

[0168] Exemplarily, can be designed as a form of penalty for cross-queue difference or cross-queue similarity to constrain the abnormal patterns presented simultaneously in highly correlated queues. For example:

[0169] ;

[0170] Among them, can indicate the component difference between queue i and queue j within the same time slice (or adjacent time slices) in the sparse matrix (i.e., the anomaly identifier); if G(i,j) is large, but it is found that the abnormal patterns distributed in a large number of time slices of these two queues are extremely different, it may indicate an asynchronous anomaly that does not meet expectations, or the anomaly is only locally affected, and key inspections need to be carried out within the algorithm; vice versa.

[0171] can be a certain loss function, such as the second norm, norm or Huber function, is the balance coefficient for parallel coupling, used to make a trade-off between the computational amount and the strength of the parallel constraint.

[0172] After the decomposition of the parallel constraint is completed, the new sparse matrix can not only capture the abnormal points of each queue itself, but also highlight parallel anomalies or synchronization failures at the cross-queue level.

[0173] For example, when certain highly correlated queues simultaneously show a sudden increase in the queuing volume within the same time window and the external congestion index does not increase significantly, it can be judged that this may be a parallel anomaly caused by a shared resource failure or a malfunction of the coordination mechanism, so as to distinguish it from the anomalies that only occur in individual queues; it can also identify why there is no synchronous anomaly between queues that should influence each other, and then prompt that a specific queue may have independently adopted an emergency plan or there may be a potential scheduling conflict.

[0174] In a specific implementation, the parallel constraint sparse decomposition can be performed using a block alternating optimization or a multi-level iterative strategy. First, perform several rounds of iteration on the basic term without the regularization constraint term to obtain the initial and ; then, under the action of the parallel constraint, Correct or reconstruct the abnormal distribution across queues to make it more consistent with the queue coupling information given by G.

[0175] For possible missing data or local extreme values, existing matrix filling methods or robust terms can be continued to perform interpolation and denoising. When performing optimization and solution, it is necessary to consider It may take effect simultaneously in both the time and queue dimensions. Therefore, it is possible to perform piecewise smoothing in the time dimension and calculate the cross-queue similarity sum in the queue dimension to ensure the convergence and computational efficiency of the algorithm.

[0176] In the multi-queue parallel scenario of this application, through the additional defined parallel association matrix G and the corresponding regular constraint terms , the decomposition result of the sparse matrix S has a stronger cross-queue coupling awareness: it can not only identify abnormalities within a single queue but also actively detect parallel abnormalities that multiple queues may be affected by simultaneously or near simultaneously during the same period. For these parallel abnormalities, the system can trigger corresponding linkage measures preferentially during the queuing and scheduling process, such as jointly scheduling multiple queues, temporarily reorganizing the resources of parallel queues, or adjusting the usage rights of shared buffers, thereby avoiding the overall efficiency decline caused by isolated handling of single queues. In addition, if the actual coupling degree of some queues in the G matrix is very weak, their parallel impact is relatively small, thus ensuring that the normal handling of those independent queues by the system is not interfered with.

[0177] Exemplarily, if a device or a parking lot area is shared by multiple queues, its failure often manifests as a linkage among multiple queues. This parallel constraint scheme can automatically identify the parallel abnormal pattern and prompt the operation and maintenance personnel for more accurate fault location. During peak hours, if several associated queues almost simultaneously show abnormal queuing lengths without obvious external traffic reasons, it indicates that there may be internal organizational or equipment scheduling failures, guiding the system to quickly issue a linkage warning and adjust the queuing strategies of multiple queues. Through the queue coupling matrix G, local single-queue abnormalities can be distinguished from global abnormalities that affect multiple coupled queues, and then the priorities and processing scopes can be determined during scheduling. With the help of parallel abnormal information, targeted linkage emergency measures (such as switching to the backup channel of a specific shared resource) can be executed to prevent large-scale and indiscriminate rescheduling once a queue fails, saving time and equipment occupancy costs.

[0178] In this way, based on the improved sparse decomposition idea of multi-queue parallel association, it is possible to fully integrate the description of the parallel relationship by the "queue coupling matrix" within the framework of the traditional "low-rank and sparse" model, and then provide a more delicate and accurate decision-making basis for anomaly detection and dynamic scheduling in the multi-queue parallel logistics reservation queuing scenario. This solution is not only applicable to parallel queues with tightly shared or mutually exclusive resources in practice, but can also be extended to the multi-queue collaborative operation environment with different levels of coupling relationships, bringing higher robustness and overall efficiency to large-scale, multi-dimensional, and parallel modern logistics scheduling systems.

[0179] Based on the same inventive concept, an embodiment of the present application also provides a multi-queue parallel logistics reservation queuing system corresponding to a multi-queue parallel logistics reservation queuing method. Since the principle of problem-solving in the system in the embodiment of the present application is similar to that of the above-mentioned multi-queue parallel logistics reservation queuing method in the embodiment of the present application, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be elaborated.

[0180] Refer to Figure 3 As shown in the figure, an embodiment of the present application provides a multi-queue parallel logistics reservation queuing system, and the system includes:

[0181] The acquisition unit 10 obtains the real-time operation data of multiple queuing queues and parking lots in the logistics park, aligns the real-time operation data to form a multi-dimensional data set, and the real-time operation data includes vehicle arrival time, queue load, parking lot capacity, loading and unloading equipment status, and external traffic and environmental information;

[0182] The processing unit 20 performs sparse reconstruction processing on the multi-dimensional data set to obtain target data for queuing scheduling and anomaly indication data for dynamically adjusting the queuing scheduling;

[0183] The scheduling unit 30 generates a multi-queue parallel queuing scheduling plan based on the target data and in combination with the vehicle reservation information of each queue. The queuing scheduling plan includes the queuing order of vehicles and parking lot arrangements; and issues the queuing scheduling plan to vehicles and operation terminals;

[0184] The adjustment unit 40 determines the abnormal situation that appears in the real-time operation data during real-time update based on the anomaly indication data, and dynamically adjusts the queuing scheduling plan according to the abnormal situation to generate a new queuing scheduling plan.

[0185] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present invention can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

Claims

1. A multi-queue parallel logistics reservation queuing method, characterized in that Including: Obtain the real-time operation data of multiple queuing queues and parking lots in the logistics park, align the real-time operation data to form a multi-dimensional data set, and the real-time operation data includes vehicle arrival time, queue load, parking lot capacity, loading and unloading equipment status, and external traffic and environmental information; Perform sparse reconstruction processing on the multi-dimensional data set to obtain target data for queuing scheduling and abnormal indication data for dynamically adjusting the queuing scheduling; Based on the target data and combined with the vehicle reservation information of each queue, generate a queuing scheduling scheme with multi-queue parallelism, and the queuing scheduling scheme includes the queuing order of vehicles and parking lot arrangements; send the queuing scheduling scheme to vehicles and operation terminals; Based on the abnormal indication data, determine the abnormal conditions that occur in the real-time update of the real-time operation data, and dynamically adjust the queuing scheduling scheme according to the abnormal conditions to generate a new queuing scheduling scheme; Among them, the sparse reconstruction processing of the multi-dimensional data set includes: Convert the multi-dimensional data set aligned according to time, queue, and equipment dimensions into a target data matrix; Decompose the target data matrix to obtain a low-rank matrix and a sparse matrix, where the low-rank matrix represents the target data structure after interpolation and denoising; the sparse matrix is used to identify abnormal data corresponding to faults, congestion, and extreme values; Use the low-rank matrix as the target data and the sparse matrix as the abnormal indication data; The sparse reconstruction processing of the multi-dimensional data set further includes: According to the business characteristics of multi-queue parallel queuing, construct a queue coupling matrix to characterize the resource sharing relationship and / or job collaboration relationship between multiple queues; During the decomposition process of the target data matrix, introduce a parallel constraint term corresponding to the queue coupling matrix, and combine the parallel constraint term with the sparse matrix for iterative solution to identify abnormal conditions that occur synchronously across queues or parallelism abnormalities caused by shared resource failures; Among them, in response to the parallel association degree between any two queues in the queue coupling matrix being greater than a first preset threshold, and the value of the corresponding abnormal entry in the sparse matrix exceeding a second preset threshold within the same or adjacent time period, it is determined that the abnormality corresponding to this entry is a multi-queue parallel abnormality; the multi-queue parallel abnormality is used to perform cross-queue queuing scheduling linkage adjustment.

2. The multi-queue parallel logistics reservation queuing method according to claim 1, wherein The generation of the queuing scheduling scheme with multi-queue parallelism includes: Integrate the reservation period, vehicle arrival time, estimated waiting time, and remaining capacity of the parking lot for each vehicle, and determine the available time period of each loading and unloading equipment and the queue load; Queue and sort the vehicles and allocate parking lots according to the integrated data and preset priority rules, and output a queuing scheduling scheme with multi-queue parallelism.

3. A multi-queue parallel logistics reservation queuing method according to claim 1, characterized in that, The generation of the new queuing scheduling scheme includes: Monitor the abnormal indication data, and determine the vehicle status, parking lot capacity change, and loading and unloading equipment failure that cause the abnormality, and generate a monitoring result; Based on the monitoring results, re-evaluate the queues affected by anomalies and the vehicle operation sequence, make real-time corrections to the vehicle queuing sequence, parking lot arrangement, and loading and unloading equipment allocation, generate a new queuing and scheduling plan, and send it to the corresponding vehicles and operation terminals.

4. A method for logistics reservation queuing with multi-queue parallelism according to claim 3, characterized in that The generation of the new queuing and scheduling plan further includes: Combined analysis of the vehicle reservation status, external traffic conditions, and the availability of loading and unloading equipment corresponding to multiple anomaly indication data within the same time period or adjacent time periods to determine whether large-scale or cross-queue linkage anomalies occur; For the determined linkage anomalies, execute a multi-queue collaborative scheduling strategy, allocate spare parking spaces or additional loading and unloading equipment resources to the affected queues, and synchronously update the queuing sequence and operation time of each vehicle.

5. A multi-queue parallel logistics reservation queuing method according to claim 1, characterized in that, The formation of the multi-dimensional data set includes: Based on the vehicle reservation information and the real-time operation data, align the reserved time period, reserved queue, reserved loading and unloading equipment, and vehicle type corresponding to each vehicle with the actual arrival time and queuing waiting duration to generate a reservation mapping matrix; The sparse reconstruction process of the multi-dimensional data set further includes: Based on the sparse matrix, determine the entries marked as anomalies, and through association with the reservation mapping matrix, cluster the eigenvectors corresponding to the entries marked as anomalies to generate a first sub-sparse matrix and a second sub-sparse matrix; Among them, the first sub-sparse matrix is used to characterize the anomalies caused by vehicle reservation information; the second sub-sparse matrix is used to characterize the anomalies caused by non-vehicle reservation information; The elements in the first sub-sparse matrix and the second sub-sparse matrix are distinguished by the reservation deviation degree.

6. A method for logistics reservation queuing with multi-queue parallelism according to claim 5, characterized in that The clustering of the eigenvectors corresponding to the entries marked as anomalies includes: Based on the sparse matrix, extract all the entries marked as anomalies, and add a multi-dimensional eigenvector reflecting the business status to each anomaly record; Obtain the operation demand information of the logistics park; Based on the operation demand information, configure weight coefficients for each dimension in the multi-dimensional eigenvector; Compare the multi-dimensional eigenvectors corresponding to each anomaly record, and use a preset clustering algorithm for grouping to generate a clustering result; Based on the clustering result, generate the first sub-sparse matrix and the second sub-sparse matrix; Among them, the multi-dimensional eigenvector includes: the specific time period when the anomaly occurs and whether the adjacent time period is the peak operation period of the park; The queue number, loading and unloading equipment ID, and parking lot area ID corresponding to the anomaly; Reservation deviation degree; Weather conditions, traffic congestion index, holiday information.

7. A logistics reservation queuing system with multi-queue parallelism, characterized in that, It includes: A collection unit that obtains the real-time operation data of multiple queuing queues and parking lots in the logistics park, aligns the real-time operation data to form a multi-dimensional data set, and the real-time operation data includes vehicle arrival time, queue load, parking lot capacity, loading and unloading equipment status, and external traffic and environmental information; A processing unit that performs sparse reconstruction processing on the multi-dimensional data set to obtain target data for queuing and scheduling and anomaly indication data for dynamically adjusting the queuing and scheduling; A scheduling unit, based on the target data and in combination with the vehicle reservation information of each queue, generates a queuing scheduling scheme with multi-queue parallelism. The queuing scheduling scheme includes the queuing order of vehicles and parking lot arrangements; and issues the queuing scheduling scheme to vehicles and operation terminals. An adjustment unit, based on the exception indication data, determines the exception situation that occurs during the real-time update of the real-time operation data, and dynamically adjusts the queuing scheduling scheme according to the exception situation to generate a new queuing scheduling scheme. Among them, the sparse reconstruction process for the multi-dimensional data set includes: Converting the multi-dimensional data set aligned according to time, queue, and device dimensions into a target data matrix. Decomposing the target data matrix to obtain a low-rank matrix and a sparse matrix, where the low-rank matrix represents the target data structure after interpolation and denoising; the sparse matrix is used to identify abnormal data corresponding to faults, congestion, and extreme values. Taking the low-rank matrix as the target data and the sparse matrix as the exception indication data. The sparse reconstruction process for the multi-dimensional data set further includes: According to the business characteristics of multi-queue parallel queuing, constructing a queue coupling matrix to characterize the resource sharing relationship and / or job collaboration relationship between multiple queues. During the decomposition process of the target data matrix, introducing a parallel constraint term corresponding to the queue coupling matrix, and combining the parallel constraint term with the sparse matrix for iterative solution to identify exceptions that occur synchronously across queues or parallelism exceptions caused by shared resource failures. Among them, in response to the parallel association degree between any two queues in the queue coupling matrix being greater than a first preset threshold, and the value of the corresponding abnormal entry in the sparse matrix exceeding a second preset threshold within the same or adjacent time periods, it is determined that the exception corresponding to this entry is a multi-queue parallel exception; the multi-queue parallel exception is used to perform cross-queue queuing scheduling linkage adjustment.

Images