Online management method and system for queuing of transport vehicles

Abstract

The invention provides an online management method and system for queuing of transport vehicles. The method comprises the following steps: acquiring a historical customs clearance information set and vehicle cargo information; performing time sequence dependence modeling and customs clearance efficiency analysis on the preprocessed historical customs clearance information set based on a time sequence modeling method to obtain a gateway customs clearance efficiency model; setting a static priority weight set for each queuing vehicle queue, and setting a dynamic priority weight set based on a saturation limit decision algorithm according to a gateway clearance efficiency model and a vehicle cargo information set; entropy weights of the static priority weight set and the dynamic priority weight set are fused into a standard priority weight set, and appointment processing time analysis is performed on each queuing vehicle queue based on the standard priority weight set and a gateway clearance efficiency model to obtain an appointment soft time window set; and based on the appointment soft time window set, extracting the clearance load of each logistics gateway, and according to each clearance load, carrying out business constraint verification and gateway guidance on the transport vehicle.

Description

Technical Field

[0001] This invention relates to the field of logistics scheduling technology, and in particular to an online management method and system for transport vehicle queuing. Background Technology

[0002] In large logistics hubs, cross-border ports, and multi-channel warehousing parks, transport vehicles need to complete inspection, customs clearance, loading and unloading operations in sequence at designated logistics checkpoints. With the increase in cargo volume and the improvement of timeliness requirements, problems such as vehicle queues and congestion at checkpoints, uneven resource utilization, and delays of high-priority goods are becoming increasingly prominent. Achieving intelligent and efficient online queue management has become the key to improving the throughput capacity and service level of logistics hubs.

[0003] However, existing vehicle queue management methods still have significant technical limitations when dealing with dynamic and complex real-world operational scenarios. First, most systems rely on first-come, first-served or static priorities based on fixed rules for scheduling, failing to respond in real-time to fluctuations in checkpoint processing efficiency and dynamic changes in vehicle status. This leads to rigid scheduling strategies and overall low clearance efficiency. Second, the lack of accurate prediction of checkpoint future processing capacity results in severely inaccurate queuing time estimates, leaving vehicle reservation and scheduling without a reliable data foundation and causing frequent plan failures. Finally, existing solutions mostly focus on queue optimization within a single checkpoint, lacking intelligent guidance mechanisms for dynamic load balancing across multiple parallel checkpoints. This fails to prevent uneven distribution of queue resources at the system level, limiting the overall resource utilization ceiling. Summary of the Invention

[0004] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an online management method and system for transport vehicle queuing. It has the advantages of time-series prediction-driven decision-making, multi-level weight dynamic fusion, and global load intelligent guidance. It solves the problems of long average waiting time for vehicles and low overall throughput efficiency of the checkpoint cluster caused by static and rigid scheduling rules, unpredictable processing efficiency, and unbalanced load across multiple checkpoints in the vehicle queuing management scenario at logistics checkpoints.

[0005] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: This invention provides an online management method for transport vehicle queues, comprising the following steps: Obtain historical customs clearance information sets for each logistics checkpoint and vehicle and cargo information sets for the corresponding queuing vehicle queues at each logistics checkpoint. The vehicle and cargo information sets include cargo type, customer level, transportation status, and order data for each queuing vehicle. The historical customs clearance information set is preprocessed with time series data, and the preprocessed historical customs clearance information set is used for time series dependency modeling and customs clearance model training based on the time series modeling method to obtain the customs clearance efficiency model. Based on static business rules, a static priority weight set is set for each queuing vehicle queue according to the vehicle and cargo information set. Based on the saturation limit decision algorithm, a dynamic priority weight set is set for each queuing vehicle queue according to the customs clearance efficiency model and the vehicle and cargo information set. The static priority weight set and the dynamic priority weight set entropy weight are merged into a standard priority weight set, and the reservation processing time of each queuing vehicle queue is analyzed based on the standard priority weight set and the customs clearance efficiency model to obtain a reservation soft time window set. Based on the aforementioned reservation soft time window set, the clearance load of each logistics checkpoint is extracted, and business constraint verification and checkpoint guidance are performed for newly arriving transport vehicles according to each clearance load.

[0006] According to a preferred embodiment of the present invention, the time-series data preprocessing of the historical customs clearance information set includes: The historical customs clearance information set is mapped to fields and processed in a structured manner to obtain a standard customs clearance information set; The standard customs clearance information set is subjected to missing value identification and outlier removal, and the identified missing values ​​and the missing positions generated after outlier removal are smoothly filled to obtain a smooth customs clearance information set. The smooth customs clearance information set is unified on a time axis, and the unified smooth customs clearance information set is arranged in chronological order according to a preset time granularity to obtain a customs clearance data sequence set. Efficiency statistics and feature annotations are performed on each customs clearance data in the customs clearance data sequence set to obtain a customs clearance feature sequence set. The customs clearance feature sequence set is then normalized to obtain a standard customs clearance feature sequence set.

[0007] According to another preferred embodiment of the present invention, the step of performing time-series dependency modeling and customs clearance model training on the preprocessed historical customs clearance information set based on the time-series modeling method to obtain a customs clearance efficiency model includes: Using a preset customs clearance model, static customs clearance feature sequence set, known customs clearance feature sequence set, and observed customs clearance feature sequence set are extracted from the standard customs clearance feature sequence set, respectively. The static customs clearance feature sequence set, the known customs clearance feature sequence set, and the observed customs clearance feature sequence set are weighted and fused using the gated fusion network of the customs clearance model to obtain a fused customs clearance feature sequence set. The fused customs clearance feature sequence set is subjected to multi-layer causal dilated convolution to obtain a local customs clearance time-series feature sequence set; The local customs clearance time-series feature sequence set is subjected to low-rank approximate attention weighting based on the low-rank projection mechanism to obtain the global customs clearance time-series feature sequence set. The static customs clearance feature sequence set is used as a condition and embedded into the global customs clearance time sequence set using a gating mechanism to obtain the standard input feature sequence set. The customs clearance model is trained based on the standard input feature sequence set and the standard customs clearance feature sequence set to obtain the customs clearance efficiency model.

[0008] According to another preferred embodiment of the present invention, the step of applying low-rank approximate attention weighting to the local customs clearance time-series feature sequence set based on a low-rank projection mechanism to obtain a global customs clearance time-series feature sequence set includes: Layer normalization is performed on each local customs clearance time sequence feature sequence in the local customs clearance time sequence set to obtain a standard local customs clearance time sequence feature sequence. Attention feature projection is performed on the standard local customs clearance time-series feature sequence to obtain a query vector sequence, a key vector sequence, and a value vector sequence. The key vector sequence and the value vector sequence are dimensionality-reduced by performing a projection on a preset low-rank projection matrix to obtain a low-rank key vector sequence and a low-rank value vector sequence. The attention weights between the query vector sequence and the low-rank key vector sequence are calculated using a low-rank approximate attention algorithm to obtain a low-rank approximate attention weight matrix. The low-rank value vector sequence is weighted and summed using the low-rank approximate attention weight matrix to obtain the global dependency vector sequence. The global dependency vector sequence is output-projected, and the projected global dependency vector sequence and the local customs clearance time series feature sequence are residually connected to obtain the global customs clearance time series feature sequence. All global customs clearance time series feature sequences are then aggregated into a global customs clearance time series feature sequence set.

[0009] According to another preferred embodiment of the present invention, training the customs clearance model based on the standard input feature sequence set and the standard customs clearance feature sequence set to obtain a customs clearance efficiency model includes: Extract the customs clearance efficiency feature sequence set from the standard customs clearance feature sequence set; The standard input feature sequence set is slid-filtered according to a preset time period window to obtain a periodic input feature sequence set consisting of multiple periodic input feature sequences of the same length; Each clearance efficiency feature located after the last time step of the input feature sequence in each cycle is selected from the clearance efficiency feature sequence set as a lag efficiency feature to obtain a lag efficiency feature set. Using the periodic input feature sequence set as input and the lag efficiency feature set as labels, the customs clearance model is trained under supervision to obtain the customs clearance efficiency model.

[0010] According to another preferred embodiment of the present invention, the step of setting a static priority weight set for each queuing vehicle queue based on the vehicle and cargo information set according to the static business rules includes: Select the transport vehicles in each queue as the target transport vehicles, take the vehicle cargo information corresponding to the target transport vehicles in the vehicle cargo information set as the target vehicle cargo information, and extract the static business weight item set from the target vehicle cargo information. Determine whether an absolute business priority item exists in the set of static business weight items; If so, then set a static priority weight for the target transport vehicle according to the absolute business priority item; If not, then the weight scores of each item in the target transport vehicle are statistically analyzed according to the static business weight itemset to obtain a weight score set, and the static priority weight of the target transport vehicle is set according to the mean of the weight score set.

[0011] According to another preferred embodiment of the present invention, the saturation-constrained decision algorithm sets a dynamic priority weight set for each queuing vehicle queue based on the customs clearance efficiency model and the vehicle and cargo information set, including: Using the customs clearance efficiency model, multi-step rolling efficiency analysis is performed on each queuing vehicle queue based on the vehicle and cargo information set to obtain a queuing clearance efficiency sequence set. Based on the queue clearance efficiency sequence set, the waiting time of each queued vehicle queue is analyzed to obtain a primary waiting time interval sequence set. Extract the transportation status and order data corresponding to each queued vehicle queue from the vehicle and cargo information set to obtain a transportation status sequence set and an order data sequence set. The transportation status sequence set, the order data sequence set, and the primary waiting time interval sequence set are used to construct a vehicle time-sensitive feature sequence set, and priority regression decision is performed on the vehicle time-sensitive feature sequence set to obtain a dynamic priority score set. The dynamic priority score set is subjected to saturation function mapping and change rate threshold limitation, and the dynamic priority score set after change rate limitation is subjected to weight normalization operation to obtain dynamic priority weight set.

[0012] According to another preferred embodiment of the present invention, the step of fusing the static priority weight set and the dynamic priority weight set entropy weight into a standard priority weight set includes: Based on the transport vehicle IDs corresponding to each queuing vehicle queue, the static priority weight set and the dynamic priority weight set are used as column vectors for weight alignment to obtain a weight evaluation matrix. Perform a minimum-maximum normalization operation on the two column vectors of the weight evaluation matrix to obtain a normalized weight matrix; The information entropy of the two column vectors of the normalized weight matrix is ​​calculated to obtain the static weight entropy value and the dynamic weight entropy value. The static difference coefficient and the dynamic difference coefficient are calculated based on the static weight entropy value and the dynamic weight entropy value, respectively. The static difference coefficient and the dynamic difference coefficient are then normalized to obtain the static fusion weight and the dynamic fusion weight. A standard priority weight set is calculated based on the static fusion weight, the dynamic fusion weight, the static priority weight set, and the dynamic priority weight set.

[0013] According to another preferred embodiment of the present invention, the step of performing reservation processing time analysis for each queuing vehicle queue based on the standard priority weight set and the customs clearance efficiency model to obtain a reservation soft time window set includes: Based on the standard priority weight set, each queuing vehicle queue is prioritized to obtain a priority vehicle queue set. Using the customs clearance efficiency model, a multi-step rolling efficiency analysis is performed on the priority vehicle queue set based on the vehicle and cargo information set to obtain a standard customs clearance efficiency sequence set. Based on the standard customs clearance efficiency sequence set, waiting time analysis is performed on the priority vehicle queue set to obtain a standard waiting time interval sequence set; Based on the standard priority weight set, a flexible time window is set for the standard waiting time interval sequence set to obtain a flexible window sequence set. The standard waiting time interval sequence set and the flexible window sequence set are merged and window conflict checks are performed to obtain the reservation soft time window set.

[0014] To achieve at least one of the above-mentioned objectives, the present invention further provides an online management system for transport vehicle queuing, the system comprising an information acquisition module, a customs clearance analysis module, a priority setting module, a reservation analysis module, and a checkpoint guidance module, wherein: The information acquisition module acquires historical customs clearance information sets for each logistics checkpoint and vehicle and cargo information sets for the queuing vehicle queues corresponding to each logistics checkpoint. The vehicle and cargo information sets include cargo type, customer level, transportation status, and order data for each queuing vehicle. The customs clearance analysis module performs time-series data preprocessing on the historical customs clearance information set, and performs time-series dependency modeling and customs clearance model training on the preprocessed historical customs clearance information set based on the time-series modeling method to obtain the customs clearance efficiency model. The priority setting module sets a static priority weight set for each queuing vehicle queue based on the static business rules and the vehicle and cargo information set, and sets a dynamic priority weight set for each queuing vehicle queue based on the saturation limit decision algorithm, the customs clearance efficiency model, and the vehicle and cargo information set. The reservation analysis module merges the static priority weight set and the dynamic priority weight set into a standard priority weight set, and performs reservation processing time analysis for each queuing vehicle queue based on the standard priority weight set and the customs clearance efficiency model to obtain a reservation soft time window set. The checkpoint guidance module extracts the clearance load of each logistics checkpoint based on the reservation soft time window set, and performs business constraint verification and checkpoint guidance for newly arriving transport vehicles according to each clearance load.

[0015] (III) Beneficial Effects Compared with the prior art, the present invention provides an online management method and system for transport vehicle queuing, which has the following beneficial effects: This online management method for transport vehicle queuing improves data quality through time-series data preprocessing, providing stable and standardized input to the model. It achieves intelligent fusion of multi-source heterogeneous information through a gated fusion network, enabling model decisions to consider not only historical data but also business rules and future constraints, significantly increasing the amount of input information. Multi-layer causal dilated convolution expands the receptive field without leaking future information or increasing parameters, efficiently extracting local and intermediate dependency patterns of the sequence. A low-rank attention mechanism solves the computational bottleneck of standard self-attention in long-sequence scenarios, enabling the model to efficiently model global dependencies between all time steps, achieving a balance between local perception and global perspective. Static feature conditional embedding enables personalized prediction, allowing the same model to differentiate the basic processing capabilities of different checkpoints, making prediction results more targeted and improving the accuracy of checkpoint clearance efficiency analysis.

[0016] This online management method for transport vehicle queues unifies business compliance, customer value, and checkpoint capacity into a single priority calculation framework by calculating static and dynamic priority weight sets. Static rules ensure compliance and safety baselines, while dynamic weights reflect real-time congestion and urgency. Furthermore, entropy weight fusion achieves an adaptive balance between the two types of weights, ensuring that the final standard priority is stable and reliable while also possessing real-time adjustment capabilities. This provides a unified, interpretable, and resistant decision-making basis for subsequent soft time window generation and queue scheduling, significantly improving the fairness and efficiency of logistics checkpoint queue management.

[0017] This online management method for transport vehicle queues achieves a crucial transition from priority-based decision-making to executable time-based management by deeply coupling standard priority weights with the customs clearance efficiency model. Through multi-step rolling efficiency analysis and waiting time derivation, the method ensures that the appointment time calculation accurately reflects the temporal changes in the customs clearance capacity, avoiding deviations caused by static queuing or one-time estimations. By introducing a flexible time window mechanism based on priority weights, high-priority vehicles receive more certain processing times, while low-priority vehicles bear more time fluctuations, thus achieving a balance between fairness and efficiency. Through window merging and conflict verification, the method ensures that the appointment results are always limited by the actual processing capacity of the customs clearance, avoiding resource overload and improving the efficiency of queue management. Attached Figure Description

[0018] Figure 1 The diagram shown is a flowchart of an online management method for transport vehicle queues according to the present invention. Detailed Implementation

[0019] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious modifications will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0020] It is understood that the term "a" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0021] Example 1: Please combine Figure 1 This invention discloses an online management method for transport vehicle queues, the method comprising the following steps: Obtain historical customs clearance information sets for each logistics checkpoint and vehicle and cargo information sets for the corresponding queuing vehicle queues at each logistics checkpoint. The vehicle and cargo information sets include cargo type, customer level, transportation status, and order data for each queuing vehicle.

[0022] In large logistics parks, port-type logistics hubs, cross-border logistics centers, or multi-gate distribution scenarios, transport vehicles typically need to complete cargo inspection, security checks, information verification, customs clearance, or business acceptance at multiple logistics checkpoints sequentially. Since these operations are time-consuming, queues of transport vehicles may occur at these checkpoints. Therefore, to improve clearance efficiency, online management of each transport vehicle is necessary. A logistics checkpoint refers to a business processing node that a transport vehicle must pass through during logistics operations to complete specific logistics or regulatory operations. The historical clearance information set records clearance data for each logistics checkpoint, including the checkpoint ID, transport vehicle ID, timestamp of vehicle clearance, cargo type, clearance time, total number of vehicles processed daily or hourly, total cargo volume, checkpoint duty roster, operating status of inspection equipment, and external event data such as holiday information and weather conditions. This historical clearance information set can be obtained through the business systems of each logistics checkpoint.

[0023] In detail, the vehicle and cargo information set includes the vehicle and cargo information corresponding to each transport vehicle in each queue of vehicles. Each vehicle and cargo information includes the cargo type of each cargo in the transport vehicle, the customer level of the customer corresponding to each cargo, the real-time transportation status, and the order data corresponding to the cargo. The cargo type includes general cargo, cold chain cargo, hazardous chemicals, and cross-border e-commerce parcels, which can be obtained directly from the pre-declared electronic waybill or customs declaration form. The customer level includes VIP customers, long-term agreement customers, and ordinary individual customers, which can be obtained through a customer relationship management system (CRM). The transportation status includes the real-time location information of the vehicle and cargo, transportation time, and waiting time, which can be obtained through information uploaded by IoT devices installed on each transport vehicle. The order data includes the order number, order time, and order service commitment for each cargo, such as next-day delivery or two-day delivery.

[0024] The historical customs clearance information set is preprocessed with time series data, and the preprocessed historical customs clearance information set is used for time series dependency modeling and customs clearance model training based on the time series modeling method to obtain the customs clearance efficiency model.

[0025] Specifically, the time-series data preprocessing of the historical customs clearance information set includes: The historical customs clearance information set is mapped to fields and processed in a structured manner to obtain a standard customs clearance information set; The standard customs clearance information set is subjected to missing value identification and outlier removal, and the identified missing values ​​and the missing positions generated after outlier removal are smoothly filled to obtain a smooth customs clearance information set. The smooth customs clearance information set is unified on a time axis, and the unified smooth customs clearance information set is arranged in chronological order according to a preset time granularity to obtain a customs clearance data sequence set. Efficiency statistics and feature annotations are performed on each customs clearance data in the customs clearance data sequence set to obtain a customs clearance feature sequence set. The customs clearance feature sequence set is then normalized to obtain a standard customs clearance feature sequence set.

[0026] The field mapping refers to mapping various types of data in the historical customs clearance information set to a preset unified standard, such as unifying the unit of customs clearance timestamps and mapping cargo types to numerical codes. The structured processing refers to splitting each data type and unifying the data format according to a preset field arrangement structure. Outlier filtering can be performed using the interquartile range method, and smoothing can be performed using the median or mean of the nearest neighbor time domain data. The time axis unification refers to mapping the data corresponding to each data point to a unified time axis. The time granularity refers to the length of the time window, such as one hour or half an hour. The efficiency statistics refer to counting the number of transport vehicles processed within the corresponding time granularity. The feature labeling refers to using the number of transport vehicles processed within the corresponding time granularity as labels, and converting the cargo type quantity and proportion, customs ID, customs duty roster, equipment status, weather conditions, and holidays into feature vectors according to a fixed order and format. The feature vectors are labeled to obtain customs clearance features. The feature normalization method can be the min-max normalization method or the z-score normalization method.

[0027] Specifically, the time-series modeling method performs time-series dependency modeling and clearance model training on the preprocessed historical clearance information set to obtain a customs clearance efficiency model, including: Using a preset customs clearance model, static customs clearance feature sequence set, known customs clearance feature sequence set, and observed customs clearance feature sequence set are extracted from the standard customs clearance feature sequence set, respectively. The static customs clearance feature sequence set, the known customs clearance feature sequence set, and the observed customs clearance feature sequence set are weighted and fused using the gated fusion network of the customs clearance model to obtain a fused customs clearance feature sequence set. The fused customs clearance feature sequence set is subjected to multi-layer causal dilated convolution to obtain a local customs clearance time-series feature sequence set; The local customs clearance time-series feature sequence set is subjected to low-rank approximate attention weighting based on the low-rank projection mechanism to obtain the global customs clearance time-series feature sequence set. The static customs clearance feature sequence set is used as a condition and embedded into the global customs clearance time sequence set using a gating mechanism to obtain the standard input feature sequence set. The customs clearance model is trained based on the standard input feature sequence set and the standard customs clearance feature sequence set to obtain the customs clearance efficiency model.

[0028] The customs clearance model includes a gated fusion network, a multi-layer causal dilated convolutional network, a low-rank approximate attention weighted network, and a gated embedding module. The static customs clearance feature sequence set consists of features in the standard customs clearance feature sequence set that do not change over time, such as features corresponding to the checkpoint ID. The known customs clearance feature sequence set consists of features in the standard customs clearance feature sequence set that can be used as known future features, such as features corresponding to checkpoint duty schedules and holiday situations. The observed customs clearance feature sequence set consists of historical features that can be observed from the standard customs clearance features, such as features corresponding to the number of transport vehicles processed, the number and proportion of cargo types, equipment status, and weather conditions within each time granularity. The weighted fusion using the gated fusion network involves using a gated residual network (GRN) to calculate the feature weights corresponding to the static customs clearance feature sequence set, the known customs clearance feature sequence set, and the observed customs clearance feature sequence set at each time step, performing weighted fusion at each time step, and then aggregating the fused customs clearance features obtained after each time step into a fused customs clearance feature sequence set.

[0029] In detail, the multi-layer causal dilation convolution refers to convolution using multiple residual blocks, where each residual block contains two causal dilation convolution layers, and each causal dilation convolution layer contains multiple causal convolution layers and dilation convolution layers. The data after convolution is passed through a gated activation network. Because convolution is performed through causal convolution and dilation convolution, local temporal features corresponding to the fused customs clearance feature sequence in the short or medium term can be extracted. The gated embedding module's gated embedding mechanism refers to constructing the static customs clearance feature sequence set into a conditional input sequence that is time-aligned with the global customs clearance temporal feature sequence by copying it in the time dimension or by conditional encoding. The conditional input sequence is then embedded into the global customs clearance temporal feature sequence set using a lightweight gated recurrent unit (GRU).

[0030] Specifically, the step of applying low-rank approximate attention weighting to the local customs clearance time-series feature sequence set based on the low-rank projection mechanism to obtain the global customs clearance time-series feature sequence set includes: Layer normalization is performed on each local customs clearance time sequence feature sequence in the local customs clearance time sequence set to obtain a standard local customs clearance time sequence feature sequence. Attention feature projection is performed on the standard local customs clearance time-series feature sequence to obtain a query vector sequence, a key vector sequence, and a value vector sequence. The key vector sequence and the value vector sequence are dimensionality-reduced by performing a projection on a preset low-rank projection matrix to obtain a low-rank key vector sequence and a low-rank value vector sequence. The attention weights between the query vector sequence and the low-rank key vector sequence are calculated using a low-rank approximate attention algorithm to obtain a low-rank approximate attention weight matrix. The low-rank value vector sequence is weighted and summed using the low-rank approximate attention weight matrix to obtain the global dependency vector sequence. The global dependency vector sequence is output-projected, and the projected global dependency vector sequence and the local customs clearance time series feature sequence are residually connected to obtain the global customs clearance time series feature sequence. All global customs clearance time series feature sequences are then aggregated into a global customs clearance time series feature sequence set.

[0031] In one specific embodiment of the present invention, the parameter structure of the low-rank approximate attention weighted network is shown in the table below: batch size: 32, historical time step: 10, local feature dimension: 64, attention projection dimension: 32, low-rank dimension: 8. The layer normalization is used to stabilize the training process and accelerate convergence. The local clearing time-series feature sequence is derived from the result of the previous multi-layer causal dilation convolution, and its shape is (batch size, historical time step, feature dimension). For example, assuming the batch size is 32, the historical time step is 10 (i.e., data from the past 10 hours), and the feature dimension of each time step is 64, then the input shape is (32, 10, 64). Layer normalization performs independent normalization on each time step of each sample, making its mean 0 and variance 1, thereby eliminating the influence of dimensions and making subsequent attention calculations more stable. The attention feature projection refers to generating query vector sequence Q, key vector sequence K, and value vector sequence V by passing the normalized standard local customs clearance time series feature sequence through three independent linear transformation layers respectively. The weight matrices of these three linear layers are all of dimension (feature dimension, projection dimension); assuming the projection dimension is set to 32, after projection, the shapes of Q, K, and V all become (32, 10, 32). The core purpose is to map the original features into three different semantic spaces to calculate the correlation between them subsequently; The dimensionality reduction projection is one of the key improvements of this invention. In a general self-attention mechanism, the computational complexity of calculating attention weights is proportional to the square of the sequence length. To reduce the computational overhead, this invention introduces a learnable low-rank projection matrix P with dimension (projection dimension, low-rank dimension). Assuming the low-rank dimension is set to 8, which is much smaller than the sequence length 10, then multiply the key sequence K and the value sequence V with P respectively to obtain the low-rank key vector sequence (K_low) and the low-rank value vector sequence (V_low), whose shapes are both reduced from (32, 10, 32) to (32, 10, 8). The core purpose is to reduce the computational complexity related to the feature dimension in the attention mechanism from O(T 2 ×d) to O(T 2 ×r) (where r << d), thereby significantly reducing the number of parameters and computational overhead in short and medium sequence scenarios; Calculating the attention weights means calculating the matrix product of the query vector sequence Q and the low-rank key vector sequence K_low to obtain the original attention score matrix with shape (32, 10, 10). Subsequently, apply a normalization function to the last dimension (key dimension) of this score matrix for normalization to obtain the low-rank approximate attention weight matrix with the same shape (32, 10, 10). Each element of the low-rank approximate attention weight matrix represents the correlation strength between a query time step and a key time step. The core purpose is to efficiently calculate the global dependencies between all time steps without calculating the high-cost scaled dot product of standard attention. The weighted summation means that for each sample, multiply the (10, 10) attention weight matrix by the (10, 8) V_low to obtain the global dependency vector sequence with shape (32, 10, 8). The core purpose is to aggregate the information of the query at each time step with the values at all time steps, so that the features at each position contain global context information; The output projection refers to mapping the global dependency vector sequence of shape (32,10,8) back to the original local feature dimension (64) through a linear output projection layer, resulting in a projected sequence of shape (32,10,64). The residual connection refers to adding the local clearing temporal feature sequence of shape (32,10,64) element by element. The core purpose is to integrate information through output projection and adapt it to subsequent networks. The residual connection preserves the original local features, prevents information loss during attention transmission, alleviates the gradient vanishing problem, and makes the network easier to optimize. The final output is the global clearing temporal feature sequence, which has the same shape as the input, (32,10,64).

[0032] Specifically, the step of training the customs clearance model based on the standard input feature sequence set and the standard customs clearance feature sequence set to obtain the customs clearance efficiency model includes: Extract the customs clearance efficiency feature sequence set from the standard customs clearance feature sequence set; The standard input feature sequence set is slid-filtered according to a preset time period window to obtain a periodic input feature sequence set consisting of multiple periodic input feature sequences of the same length; Each clearance efficiency feature located after the last time step of the input feature sequence in each cycle is selected from the clearance efficiency feature sequence set as a lag efficiency feature to obtain a lag efficiency feature set. Using the periodic input feature sequence set as input and the lag efficiency feature set as labels, the customs clearance model is trained under supervision to obtain the customs clearance efficiency model.

[0033] The time period window can be 36 hours. The sliding filter refers to moving the time period window step by step and taking a portion of the standard input feature sequence in the time period window at the corresponding time step as the periodic input feature sequence. The lag efficiency feature can be a feature composed of one or more customs clearance efficiency features. The supervised training refers to using the expected efficiency features corresponding to each periodic input feature sequence output by the customs clearance model, and updating and optimizing the parameters of each network of the customs clearance model based on the loss between the expected efficiency feature set and the lag efficiency feature set, thereby obtaining the customs clearance efficiency model.

[0034] By performing time-series data preprocessing, data quality can be improved, providing stable and standardized inputs for the model. The intelligent fusion of multi-source heterogeneous information is achieved through gated fusion networks, enabling model decisions to be based not only on historical data but also on business rules and future constraints, greatly increasing the amount of information input. Multi-layer causal dilated convolution can expand the receptive field without leaking future information or increasing parameters, efficiently extracting local and intermediate dependency patterns of the sequence. The low-rank attention mechanism solves the computational bottleneck of standard self-attention in long sequence scenarios, enabling the model to efficiently model global dependencies between all time steps, achieving a balance between local perception and global vision. Static feature conditional embedding enables personalized prediction, allowing the same model to distinguish the basic processing capabilities of different checkpoints, making the prediction results more targeted, thereby improving the accuracy of checkpoint clearance efficiency analysis.

[0035] Based on static business rules, a static priority weight set is set for each queuing vehicle queue according to the vehicle and cargo information set. Based on the saturation constraint decision algorithm, a dynamic priority weight set is set for each queuing vehicle queue according to the customs clearance efficiency model and the vehicle and cargo information set.

[0036] The static business rules refer to priority weight rules formulated based on static business constraints that do not change over time. The step of setting a static priority weight set for each queuing vehicle queue based on the vehicle and cargo information set according to the static business rules includes: Select the transport vehicles in each queue as the target transport vehicles, take the vehicle cargo information corresponding to the target transport vehicles in the vehicle cargo information set as the target vehicle cargo information, and extract the static business weight item set from the target vehicle cargo information. Determine whether an absolute business priority item exists in the set of static business weight items; If so, then set a static priority weight for the target transport vehicle according to the absolute business priority item; If not, then the weight scores of each item in the target transport vehicle are statistically analyzed according to the static business weight itemset to obtain a weight score set, and the static priority weight of the target transport vehicle is set according to the mean of the weight score set.

[0037] The static business weight item set includes four business weight items: cargo type, customer level, regulatory requirements, and safety constraints for each cargo in the target vehicle's cargo information. Determining whether an absolute business priority item exists in the static business weight item set refers to identifying cargo with specific regulatory requirements or safety constraints that correspond to absolute priority. Based on the regulatory requirements and safety constraints of the cargo, a static priority weight is set for the transport vehicle. For example, when the vehicle's cargo information indicates that the transported cargo belongs to a category explicitly defined by national or industry regulations that must clear customs within a limited time to avoid significant safety or environmental risks, such as Class 1 explosives or highly toxic chemicals, the static priority weight is set to 100. The weight score statistics refer to calculating the weight item scores corresponding to each weight item of the cargo. For example, when the customer level is VIP, the weight item score is increased by 8. By matching these rules, the accumulated weight item scores are used as the weight score of the cargo.

[0038] Specifically, the saturation-constrained decision-making algorithm sets a dynamic priority weight set for each queuing vehicle queue based on the customs clearance efficiency model and the vehicle and cargo information set, including: Using the customs clearance efficiency model, multi-step rolling efficiency analysis is performed on each queuing vehicle queue based on the vehicle and cargo information set to obtain a queuing clearance efficiency sequence set. Based on the queue clearance efficiency sequence set, the waiting time of each queued vehicle queue is analyzed to obtain a primary waiting time interval sequence set. Extract the transportation status and order data corresponding to each queued vehicle queue from the vehicle and cargo information set to obtain a transportation status sequence set and an order data sequence set. The transportation status sequence set, the order data sequence set, and the primary waiting time interval sequence set are used to construct a vehicle time-sensitive feature sequence set, and priority regression decision is performed on the vehicle time-sensitive feature sequence set to obtain a dynamic priority score set. The dynamic priority score set is subjected to saturation function mapping and change rate threshold limitation, and the dynamic priority score set after change rate limitation is subjected to weight normalization operation to obtain dynamic priority weight set.

[0039] In practical applications, the dynamic priority weight set is updated using a polling mechanism. The multi-step rolling efficiency analysis refers to performing efficiency analysis on a time-period-window basis according to the order of each queued vehicle, continuously iterating and updating the clearance efficiency for future time steps to obtain a queue clearance efficiency sequence set. After obtaining the queue clearance efficiency sequence set, the waiting time interval for each transport vehicle can be calculated using discrete event simulation or cumulative summation methods based on the relationship between clearance efficiency and the number of vehicles in the queue, thus obtaining a primary waiting time interval sequence set. Since the transport status includes information such as waiting time, total transport time, and real-time location information, combined with the order placement time and order service commitment in the order data, it is possible to... The remaining transportation time is determined and compared with the waiting time interval to determine the dynamic urgency of each vehicle. A lightweight decision tree model can be used for priority regression decision-making to generate dynamic priority scores. The saturation function mapping refers to nonlinear compression or threshold constraint on the dynamic priority score set. When the gate is close to saturation or overloaded, high dynamic priority scores are suppressed to obtain a restricted priority score set constrained by the gate's processing capacity, avoiding excessive concentration of scheduling strategies. The change rate threshold limit refers to setting a change rate threshold for each multi-step rolling efficiency analysis and polling, for example, new weight = old weight + sign(Δ)*min(|Δ|,δ), where δ is the threshold, to prevent priority weight oscillation in the short term.

[0040] By calculating static and dynamic priority weight sets, business compliance, customer value, and gateway capacity can be unified into the same priority calculation framework. Static rules ensure bottom-line fairness and security, while dynamic decision-making improves system throughput and timeliness stability. This provides stable, interpretable, and controllable weight inputs for subsequent queuing scheduling and resource allocation, thereby improving the accuracy of queuing management.

[0041] The static priority weight set and the dynamic priority weight set entropy weight are merged into a standard priority weight set. Based on the standard priority weight set and the customs clearance efficiency model, the reservation processing time of each queuing vehicle queue is analyzed to obtain a reservation soft time window set.

[0042] Specifically, the step of fusing the static priority weight set and the dynamic priority weight set entropy weight into a standard priority weight set includes: Based on the transport vehicle IDs corresponding to each queuing vehicle queue, the static priority weight set and the dynamic priority weight set are used as column vectors for weight alignment to obtain a weight evaluation matrix. Perform a minimum-maximum normalization operation on the two column vectors of the weight evaluation matrix to obtain a normalized weight matrix; The information entropy of the two column vectors of the normalized weight matrix is ​​calculated to obtain the static weight entropy value and the dynamic weight entropy value. The static difference coefficient and the dynamic difference coefficient are calculated based on the static weight entropy value and the dynamic weight entropy value, respectively. The static difference coefficient and the dynamic difference coefficient are then normalized to obtain the static fusion weight and the dynamic fusion weight. A standard priority weight set is calculated based on the static fusion weight, the dynamic fusion weight, the static priority weight set, and the dynamic priority weight set.

[0043] The weight alignment refers to forming a row vector by combining the static priority weights and corresponding dynamic priority weights belonging to the same transport vehicle ID. The static difference coefficient is obtained by subtracting the static weight entropy value from a preset constant, and the dynamic difference coefficient is obtained by subtracting the dynamic weight entropy value from a preset constant. The calculation of the standard priority weight set based on the static fusion weight, the dynamic fusion weight, the static priority weight set, and the dynamic priority weight set refers to selecting the target static priority weight and the target dynamic priority weight from the static priority weight set according to each transport vehicle ID, and summing the product of the static fusion weight and the target static priority weight and the product of the dynamic fusion weight and the target dynamic priority weight to obtain the standard priority weight.

[0044] By using static rules to ensure compliance and safety, and dynamic weights to reflect real-time congestion and urgency, and by using entropy weight fusion to achieve an adaptive balance between the two types of weights, the final standard priority is stable and reliable while also having real-time adjustment capabilities. This provides a unified, interpretable, and resistant decision-making basis for subsequent appointment soft time window generation and queue scheduling, significantly improving the fairness and efficiency of queue management at logistics checkpoints.

[0045] Specifically, the step of analyzing the reservation processing time for each queuing vehicle queue based on the standard priority weight set and the customs clearance efficiency model yields a set of reservation soft time windows, including: Based on the standard priority weight set, each queuing vehicle queue is prioritized to obtain a priority vehicle queue set. Using the customs clearance efficiency model, a multi-step rolling efficiency analysis is performed on the priority vehicle queue set based on the vehicle and cargo information set to obtain a standard customs clearance efficiency sequence set. Based on the standard customs clearance efficiency sequence set, waiting time analysis is performed on the priority vehicle queue set to obtain a standard waiting time interval sequence set; Based on the standard priority weight set, a flexible time window is set for the standard waiting time interval sequence set to obtain a flexible window sequence set. The standard waiting time interval sequence set and the flexible window sequence set are merged and window conflict checks are performed to obtain the reservation soft time window set.

[0046] The method for multi-step rolling efficiency analysis is the same as that in the above steps, and will not be repeated here. The method for waiting time analysis is also the same as that in the above steps, and will not be repeated here. The setting of the flexible time window refers to dividing the flexible time based on the interval length of each standard waiting time interval and the corresponding standard priority weight. For example, 5% of the duration of each standard waiting time interval is used as the base flexible time. A narrower time fluctuation range is set for the base flexible time with high standard priority weight, and a wider time fluctuation range is set for the base flexible time with low standard priority weight. By constraining the reduction of the high priority time window and the expansion of the low priority time window to maintain a balance within the same time period, the overall reservation time axis is prevented from drifting systematically. The window conflict verification refers to detecting whether the flexible windows of different vehicles have serious overlap in time, causing the number of reserved vehicles to exceed the physical capacity of the gate at a certain moment. If a conflict is detected, the window of the low priority vehicle is fine-tuned according to the priority order.

[0047] By deeply coupling standard priority weights with the customs clearance efficiency model, a crucial transition from priority decision-making to executable time-based security is achieved. Through multi-step rolling efficiency analysis and waiting time derivation, the appointment time calculation can accurately reflect the temporal changes in the customs clearance capacity, avoiding deviations caused by static queuing or one-time estimation. By introducing a flexible time window mechanism based on priority weights, high-priority vehicles receive more certain processing times, while low-priority vehicles bear more time fluctuations, thus achieving a balance between fairness and efficiency. Through window merging and conflict verification, it can be ensured that the appointment results are always limited by the actual processing capacity of the customs clearance, avoiding resource overload and improving the efficiency of queue management.

[0048] Based on the aforementioned reservation soft time window set, the clearance load of each logistics checkpoint is extracted, and business constraint verification and checkpoint guidance are performed for newly arriving transport vehicles according to each clearance load.

[0049] The clearance load refers to the comprehensive time load required by each logistics checkpoint to complete the clearance processing of all existing queued vehicles under the current reservation soft time window set constraints. It includes at least the remaining processing time corresponding to the latest end time in the reservation soft time window set, and can be further modified by combining the unit time processing capacity and the number of concurrent channels.

[0050] In detail, the process of performing business constraint verification and customs clearance guidance for newly arrived transport vehicles based on their respective clearance loads refers to selecting newly arrived transport vehicles as inbound transport vehicles, obtaining the cargo information of the inbound transport vehicles, extracting business constraints from the cargo information, determining whether the inbound transport vehicles can move through the customs clearance based on the business constraints, and generating guidance suggestions to guide the inbound transport vehicles to the logistics customs clearance with the least clearance load when they can move through the customs clearance, and distributing the guidance suggestions to staff for guidance.

[0051] By adjusting the clearance load based on the latest end time and processing capacity of the reservation soft time window, the system can accurately depict the time saturation of each logistics checkpoint under existing reservation constraints, avoiding misjudgments caused by relying solely on instantaneous queue length. Secondly, by introducing business constraint verification when new vehicles arrive, the checkpoint guidance decision-making can simultaneously meet the requirements of regulations, safety, and business rules, preventing the blind diversion of unqualified vehicles and improving system security and compliance. By generating checkpoint guidance suggestions according to the principle of minimizing clearance load, cross-checkpoint load balancing can be achieved, reducing the risk of local congestion and thus improving the efficiency of queue management.

[0052] Example 2: This invention discloses an online management system for transport vehicle queuing. The system includes an information acquisition module, a customs clearance analysis module, a priority setting module, a reservation analysis module, and a checkpoint guidance module, wherein: The information acquisition module acquires historical customs clearance information sets for each logistics checkpoint and vehicle and cargo information sets for the queuing vehicle queues corresponding to each logistics checkpoint. The vehicle and cargo information sets include cargo type, customer level, transportation status, and order data for each queuing vehicle. The customs clearance analysis module performs time-series data preprocessing on the historical customs clearance information set, and performs time-series dependency modeling and customs clearance model training on the preprocessed historical customs clearance information set based on the time-series modeling method to obtain the customs clearance efficiency model. The priority setting module sets a static priority weight set for each queuing vehicle queue based on the static business rules and the vehicle and cargo information set, and sets a dynamic priority weight set for each queuing vehicle queue based on the saturation limit decision algorithm, the customs clearance efficiency model, and the vehicle and cargo information set. The reservation analysis module merges the static priority weight set and the dynamic priority weight set into a standard priority weight set, and performs reservation processing time analysis for each queuing vehicle queue based on the standard priority weight set and the customs clearance efficiency model to obtain a reservation soft time window set. The checkpoint guidance module extracts the clearance load of each logistics checkpoint based on the reservation soft time window set, and performs business constraint verification and checkpoint guidance for newly arriving transport vehicles according to each clearance load.

[0053] The processes described above with reference to the flowcharts in the embodiments disclosed in this invention can be implemented as computer software programs. The embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wire segments, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless segments, wire segments, optical fibers, RF, etc., or any suitable combination thereof.

[0054] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0055] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the stated principles, the implementation of the present invention may have any variations or modifications.

Claims

1. An online management method for transport vehicle queuing, characterized in that, The method includes: Obtain historical customs clearance information sets for each logistics checkpoint and vehicle and cargo information sets for the corresponding queuing vehicle queues at each logistics checkpoint. The vehicle and cargo information sets include cargo type, customer level, transportation status, and order data for each queuing vehicle. The historical customs clearance information set is preprocessed with time series data, and the preprocessed historical customs clearance information set is used for time series dependency modeling and customs clearance model training based on the time series modeling method to obtain the customs clearance efficiency model. Based on static business rules, a static priority weight set is set for each queuing vehicle queue according to the vehicle and cargo information set. Based on the saturation limit decision algorithm, a dynamic priority weight set is set for each queuing vehicle queue according to the customs clearance efficiency model and the vehicle and cargo information set. The static priority weight set and the dynamic priority weight set entropy weight are merged into a standard priority weight set, and the reservation processing time of each queuing vehicle queue is analyzed based on the standard priority weight set and the customs clearance efficiency model to obtain a reservation soft time window set. Based on the aforementioned reservation soft time window set, the clearance load of each logistics checkpoint is extracted, and business constraint verification and checkpoint guidance are performed for newly arriving transport vehicles according to each clearance load.

2. The online management method for transport vehicle queuing according to claim 1, characterized in that, The time-series data preprocessing of the historical customs clearance information set includes: The historical customs clearance information set is mapped to fields and processed in a structured manner to obtain a standard customs clearance information set; The standard customs clearance information set is subjected to missing value identification and outlier removal, and the identified missing values ​​and the missing positions generated after outlier removal are smoothly filled to obtain a smooth customs clearance information set. The smooth customs clearance information set is unified on a time axis, and the unified smooth customs clearance information set is arranged in chronological order according to a preset time granularity to obtain a customs clearance data sequence set. Efficiency statistics and feature annotations are performed on each customs clearance data in the customs clearance data sequence set to obtain a customs clearance feature sequence set. The customs clearance feature sequence set is then normalized to obtain a standard customs clearance feature sequence set.

3. The online management method for transport vehicle queuing according to claim 2, characterized in that, The time-series modeling method performs time-series dependency modeling and clearance model training on the preprocessed historical clearance information set to obtain a customs clearance efficiency model, including: Using a preset customs clearance model, static customs clearance feature sequence set, known customs clearance feature sequence set, and observed customs clearance feature sequence set are extracted from the standard customs clearance feature sequence set, respectively. The static customs clearance feature sequence set, the known customs clearance feature sequence set, and the observed customs clearance feature sequence set are weighted and fused using the gated fusion network of the customs clearance model to obtain a fused customs clearance feature sequence set. Perform multi-layer causal dilated convolution on the fused customs clearance feature sequence set to obtain a local customs clearance time-series feature sequence set; The local customs clearance time-series feature sequence set is subjected to low-rank approximate attention weighting based on the low-rank projection mechanism to obtain the global customs clearance time-series feature sequence set. The static customs clearance feature sequence set is used as a condition and embedded into the global customs clearance time sequence set using a gating mechanism to obtain the standard input feature sequence set. The customs clearance model is trained based on the standard input feature sequence set and the standard customs clearance feature sequence set to obtain the customs clearance efficiency model.

4. The online management method for transport vehicle queuing according to claim 3, characterized in that, The local customs clearance time-series feature sequence set is subjected to low-rank approximate attention weighting based on a low-rank projection mechanism to obtain a global customs clearance time-series feature sequence set, including: Layer normalization is performed on each local customs clearance time sequence feature sequence in the local customs clearance time sequence set to obtain a standard local customs clearance time sequence feature sequence. Attention feature projection is performed on the standard local customs clearance time-series feature sequence to obtain a query vector sequence, a key vector sequence, and a value vector sequence. The key vector sequence and the value vector sequence are dimensionality-reduced by performing a projection on a preset low-rank projection matrix to obtain a low-rank key vector sequence and a low-rank value vector sequence. The attention weights between the query vector sequence and the low-rank key vector sequence are calculated using a low-rank approximate attention algorithm to obtain a low-rank approximate attention weight matrix. The low-rank value vector sequence is weighted and summed using the low-rank approximate attention weight matrix to obtain the global dependency vector sequence. The global dependency vector sequence is output-projected, and the projected global dependency vector sequence and the local customs clearance time series feature sequence are residually connected to obtain the global customs clearance time series feature sequence. All global customs clearance time series feature sequences are then aggregated into a global customs clearance time series feature sequence set.

5. The online management method for transport vehicle queuing according to claim 3, characterized in that, The step of training the customs clearance model based on the standard input feature sequence set and the standard customs clearance feature sequence set to obtain the customs clearance efficiency model includes: Extract the customs clearance efficiency feature sequence set from the standard customs clearance feature sequence set; The standard input feature sequence set is slid-filtered according to a preset time period window to obtain a periodic input feature sequence set consisting of multiple periodic input feature sequences of the same length; Each clearance efficiency feature located after the last time step of the input feature sequence in each cycle is selected from the clearance efficiency feature sequence set as a lag efficiency feature to obtain a lag efficiency feature set. Using the periodic input feature sequence set as input and the lag efficiency feature set as labels, the customs clearance model is trained under supervision to obtain the customs clearance efficiency model.

6. The online management method for transport vehicle queuing according to claim 1, characterized in that, The static business rules set a static priority weight set for each queuing vehicle queue based on the vehicle and cargo information set, including: Select the transport vehicles in each queue as the target transport vehicles, take the vehicle cargo information corresponding to the target transport vehicles in the vehicle cargo information set as the target vehicle cargo information, and extract the static business weight item set from the target vehicle cargo information. Determine whether an absolute business priority item exists in the set of static business weight items; If so, then set a static priority weight for the target transport vehicle according to the absolute business priority item; If not, then the weight scores of each item in the target transport vehicle are statistically analyzed according to the static business weight itemset to obtain a weight score set, and the static priority weight of the target transport vehicle is set according to the mean of the weight score set.

7. The online management method for transport vehicle queuing according to claim 1, characterized in that, The saturation-constrained decision-making algorithm sets a dynamic priority weight set for each queuing vehicle queue based on the customs clearance efficiency model and the vehicle and cargo information set, including: Using the customs clearance efficiency model, multi-step rolling efficiency analysis is performed on each queuing vehicle queue based on the vehicle and cargo information set to obtain a queuing clearance efficiency sequence set. Based on the queue clearance efficiency sequence set, the waiting time of each queued vehicle queue is analyzed to obtain a primary waiting time interval sequence set. Extract the transportation status and order data corresponding to each queued vehicle queue from the vehicle and cargo information set to obtain a transportation status sequence set and an order data sequence set. The transportation status sequence set, the order data sequence set, and the primary waiting time interval sequence set are used to construct a vehicle time-sensitive feature sequence set, and priority regression decision is performed on the vehicle time-sensitive feature sequence set to obtain a dynamic priority score set. The dynamic priority score set is subjected to saturation function mapping and change rate threshold limitation, and the dynamic priority score set after change rate limitation is subjected to weight normalization operation to obtain dynamic priority weight set.

8. The online management method for transport vehicle queuing according to claim 7, characterized in that, The step of fusing the static priority weight set and the dynamic priority weight set entropy weight into a standard priority weight set includes: Based on the transport vehicle IDs corresponding to each queuing vehicle queue, the static priority weight set and the dynamic priority weight set are used as column vectors for weight alignment to obtain a weight evaluation matrix. Perform a minimum-maximum normalization operation on the two column vectors of the weight evaluation matrix to obtain a normalized weight matrix; The information entropy of the two column vectors of the normalized weight matrix is ​​calculated to obtain the static weight entropy value and the dynamic weight entropy value. The static difference coefficient and the dynamic difference coefficient are calculated based on the static weight entropy value and the dynamic weight entropy value, respectively. The static difference coefficient and the dynamic difference coefficient are then normalized to obtain the static fusion weight and the dynamic fusion weight. A standard priority weight set is calculated based on the static fusion weight, the dynamic fusion weight, the static priority weight set, and the dynamic priority weight set.

9. The online management method for transport vehicle queuing according to claim 1, characterized in that, The method analyzes the reservation processing time for each queuing vehicle queue based on the standard priority weight set and the customs clearance efficiency model, resulting in a reservation soft time window set, including: Based on the standard priority weight set, each queuing vehicle queue is prioritized to obtain a priority vehicle queue set. Using the customs clearance efficiency model, a multi-step rolling efficiency analysis is performed on the priority vehicle queue set based on the vehicle and cargo information set to obtain a standard customs clearance efficiency sequence set. Based on the standard customs clearance efficiency sequence set, waiting time analysis is performed on the priority vehicle queue set to obtain a standard waiting time interval sequence set; Based on the standard priority weight set, a flexible time window is set for the standard waiting time interval sequence set to obtain a flexible window sequence set. The standard waiting time interval sequence set and the flexible window sequence set are merged and window conflict checks are performed to obtain the reservation soft time window set.

10. An online management system for queuing transport vehicles, characterized in that, The system includes an information acquisition module, a customs clearance analysis module, a priority setting module, an appointment analysis module, and a customs guidance module, wherein: The information acquisition module acquires historical customs clearance information sets for each logistics checkpoint and vehicle and cargo information sets for the queuing vehicle queues corresponding to each logistics checkpoint. The vehicle and cargo information sets include cargo type, customer level, transportation status, and order data for each queuing vehicle. The customs clearance analysis module performs time-series data preprocessing on the historical customs clearance information set, and performs time-series dependency modeling and customs clearance model training on the preprocessed historical customs clearance information set based on the time-series modeling method to obtain the customs clearance efficiency model. The priority setting module sets a static priority weight set for each queuing vehicle queue based on the static business rules and the vehicle and cargo information set, and sets a dynamic priority weight set for each queuing vehicle queue based on the saturation limit decision algorithm, the customs clearance efficiency model, and the vehicle and cargo information set. The reservation analysis module merges the static priority weight set and the dynamic priority weight set into a standard priority weight set, and performs reservation processing time analysis for each queuing vehicle queue based on the standard priority weight set and the customs clearance efficiency model to obtain a reservation soft time window set. The checkpoint guidance module extracts the clearance load of each logistics checkpoint based on the reservation soft time window set, and performs business constraint verification and checkpoint guidance for newly arriving transport vehicles according to each clearance load.

Images