Training method of gate assignment model and gate assignment method

Abstract

The application belongs to the technical field of intelligent scheduling, and particularly relates to a training method of a parking stand allocation model and a parking stand allocation method. The training method of the parking stand allocation model determines the similarity of parking stand allocation problems of a first airport and a second airport; determines a model migration strategy based on the similarity of the parking stand allocation problems; trains a first parking stand allocation model of the first airport according to the model migration strategy and historical parking stand allocation data of the second airport, and obtains a second parking stand allocation model suitable for the second airport. The method enables the allocation model to automatically adapt to airport operation changes, reduces the complexity of the construction of the allocation model, and improves the universality of the allocation model. The application also provides a parking stand allocation method. The method dynamically responds to the parking stand allocation problem of the airport based on the parking stand allocation model trained by the training method of the parking stand allocation model, and realizes efficient allocation of parking stand resources.

Description

Technical Field

[0001] This application belongs to the field of intelligent scheduling technology, specifically involving a training method and a parking space allocation method for a parking space allocation model. Background Technology

[0002] Airport parking space allocation is a critical aspect of aviation operations management, involving the rational allocation of arriving and departing flights to airport parking spaces. Effective parking space allocation can improve airport operational efficiency, reduce flight delays, optimize ground services, and enhance the overall passenger experience.

[0003] In existing technologies, mathematical optimization models, heuristic algorithms, and artificial intelligence techniques are commonly used to solve the airport parking space allocation problem. However, existing parking space allocation models need to be built independently for each airport, making the development process cumbersome and resulting in the complexity of traditional parking space allocation models. Summary of the Invention

[0004] This application provides a training method and a parking stand allocation method for a parking stand allocation model, which solves the technical problem that the process of building a parking stand allocation model independently for each airport is too cumbersome and easily leads to the complexity of model construction in traditional parking stand allocation models.

[0005] Firstly, this application provides a training method for a parking space allocation model, comprising:

[0006] Determine the similarity of the parking space allocation problem between the first and second airports;

[0007] Based on the similarity of the parking space allocation problem, a model transfer strategy is determined;

[0008] According to the model migration strategy and the historical parking stand allocation data of the second airport, the first parking stand allocation model of the first airport is trained to obtain a second parking stand allocation model applicable to the second airport. The first parking stand allocation model is a model that has been trained. The input of the first parking stand allocation model is the parking stand environmental state information, and the output is the parking stand allocation action.

[0009] In one possible implementation, determining the similarity of the parking space allocation problem between the first and second airports includes:

[0010] Acquire multimodal data related to the parking stand allocation problem for the first airport and the second airport respectively, wherein the multimodal data includes multiple of the following types of data: airport images, airport layout topology maps, parking stand related facility data, flight dynamic data, environmental data, and parking stand allocation optimization objectives;

[0011] The similarity of the parking space allocation problem is determined based on the multimodal data of the first airport and the second airport respectively.

[0012] In one possible implementation, determining the similarity of the parking space allocation problem based on the multimodal data of the first airport and the second airport respectively includes:

[0013] The multimodal data of the first airport and the second airport are input into a large model. The initial features of each data item in the multimodal data are extracted through the large model, and the initial features of each data item are mapped to a unified feature space to obtain the feature vectors of each data item corresponding to the first airport and the second airport. Attention weights between different data items of the first airport and the second airport are determined.

[0014] Based on the feature vectors of the data corresponding to the first airport and the second airport respectively, and the attention weights between different data items of the first airport and the second airport, the multimodal fusion features of the first airport and the second airport are determined respectively.

[0015] The similarity of the parking space allocation problem is determined based on the multimodal fusion features of the first airport and the second airport respectively.

[0016] In one possible implementation, determining the similarity of the parking space allocation problem based on the multimodal data of the first airport and the second airport respectively includes:

[0017] Determine the similarity of each data item in the multimodal data of the first airport and the second airport respectively;

[0018] The similarity of the parking space allocation problem is determined based on the weight of each data item according to the similarity of each data item.

[0019] In one possible implementation, determining the model transfer strategy based on the similarity of the parking space allocation problem includes:

[0020] If the similarity of the parking space allocation problem is greater than or equal to a first threshold, the model migration strategy is determined to be state alignment and model output layer adjustment.

[0021] The step of training the first parking space allocation model for the first airport according to the model migration strategy and the historical parking space allocation data of the second airport to obtain a second parking space allocation model suitable for the second airport includes:

[0022] When the model migration strategy is state alignment and adjustment of the model output layer, the parameters of the feature extraction layer and intermediate hidden layer of the first stop position allocation model are frozen.

[0023] The state space applicable to the second airport is mapped to the state space of the first parking space allocation model. The historical parking space allocation data of the second airport is used to train the first parking space allocation model to adjust the output layer parameters of the first parking space allocation model, thereby obtaining the second parking space allocation model.

[0024] In one possible implementation, determining the model transfer strategy based on the similarity of the parking space allocation problem includes:

[0025] If the similarity of the parking space allocation problem is greater than or equal to the second threshold and less than the first threshold, the model migration strategy is determined to be state alignment and model hierarchical adjustment.

[0026] The step of training the first parking space allocation model for the first airport according to the model migration strategy and the historical parking space allocation data of the second airport to obtain a second parking space allocation model suitable for the second airport includes:

[0027] When the model transfer strategy is state alignment and model hierarchical adjustment, the state space applicable to the second airport is mapped to the state space of the first parking space allocation model. The historical parking space allocation data of the second airport is used to train the first parking space allocation model in stages to obtain the second parking space allocation model.

[0028] In the first stage, the parameters of the feature extraction layer and the intermediate hidden layer of the first parking space allocation model are frozen, and the output layer parameters of the first parking space allocation model are adjusted.

[0029] In the second stage, the parameters of the feature extraction layer of the first parking space allocation model are frozen, and the parameters of the intermediate hidden layer and the output layer of the first parking space allocation model are adjusted.

[0030] In the third stage, the parameters of each layer of the first parking space allocation model are adjusted.

[0031] In one possible implementation, determining the model transfer strategy based on the similarity of the parking space allocation problem includes:

[0032] If the similarity of the parking space allocation problem is greater than or equal to the third threshold and less than the second threshold, the model migration strategy is determined to be adding a new model branch, and the new model branch and the original model branch are adjusted hierarchically.

[0033] The step of training the first parking space allocation model for the first airport according to the model migration strategy and the historical parking space allocation data of the second airport to obtain a second parking space allocation model suitable for the second airport includes:

[0034] When the model migration strategy is to add a new model branch and adjust the new model branch and the original model branch in a hierarchical manner, a special model branch is added on the basis of the original model branch of the first parking position allocation model. The special model branch is used to handle the state space and action space that are different between the first airport and the first airport.

[0035] The first parking space allocation model with added dedicated model branches is trained in stages to obtain the second parking space allocation model.

[0036] In the first stage, the original model branch of the first parking position allocation model is frozen, and the parameters of the dedicated model branch are adjusted.

[0037] In the second stage, the feature extraction layer of the original model branch of the first parking position allocation model is frozen, the parameters of the intermediate hidden layer and output layer of the original model branch are adjusted, and the parameters of the dedicated model branch are adjusted.

[0038] In the third stage, the parameters of the original model branch and the special model branch are adjusted.

[0039] Secondly, this application provides a parking space allocation method, including:

[0040] Obtain the parking stand environmental status information of the second airport, which includes parking stand information, taxiway information, runway information, and unassigned flight status information;

[0041] The parking stand environmental status information of the second airport is input into the second parking stand allocation model to obtain the parking stand allocation action output by the second parking stand allocation model. The second parking stand allocation model is trained using the training method of the parking stand allocation model as described in the first aspect and various possible implementations of the first aspect.

[0042] Thirdly, this application provides a training device for a parking space allocation model, comprising:

[0043] The determination module is used to determine the similarity of the parking space allocation problem between the first and second airports.

[0044] The determining module is also used to determine a model transfer strategy based on the similarity of the parking space allocation problem.

[0045] The processing module is used to train the first parking space allocation model of the first airport according to the model migration strategy and the historical parking space allocation data of the second airport, so as to obtain a second parking space allocation model applicable to the second airport. The first parking space allocation model is a model that has been trained. The input of the first parking space allocation model is the parking space environmental state information, and the output is the parking space allocation action.

[0046] Fourthly, this application provides a parking space allocation device, comprising:

[0047] The acquisition module is used to acquire the parking stand environmental status information of the second airport. The parking stand environmental status information includes parking stand information, taxiway information, runway information, and unassigned flight status information.

[0048] The processing module is used to input the parking stand environmental status information of the second airport into the second parking stand allocation model to obtain the parking stand allocation action output by the second parking stand allocation model. The second parking stand allocation model is trained using the training method of the parking stand allocation model as described in the first aspect and various possible implementations of the first aspect.

[0049] Fifthly, this application provides an electronic device, comprising:

[0050] A processor, and a memory communicatively connected to the processor;

[0051] The memory stores computer-executed instructions;

[0052] The processor executes computer execution instructions stored in the memory to implement the training method for the parking space allocation model as described in the first aspect and various possible implementations of the first aspect, or the parking space allocation method as described in the second aspect and various possible implementations of the second aspect.

[0053] Sixthly, this application provides a computer storage medium storing computer execution instructions, which are executed by a processor to implement the training method for the parking space allocation model as described in the first aspect and various possible implementations thereof, or the parking space allocation method as described in the second aspect and various possible implementations thereof.

[0054] Seventhly, this application provides a computer program product, including a computer program that, when executed by a processor, implements a training method for a parking space allocation model as described in the first aspect and various possible implementations thereof, or a parking space allocation method as described in the second aspect and various possible implementations thereof.

[0055] The training method for the parking stand allocation model provided in this application determines the similarity between the parking stand allocation problems of a first airport and a second airport; based on the similarity, a model transfer strategy is determined; and the first parking stand allocation model for the first airport is trained according to the model transfer strategy and historical parking stand allocation data of the second airport to obtain a second parking stand allocation model suitable for the second airport. This method significantly reduces the model development and maintenance costs while maintaining the personalized adaptability of the allocation model, enabling the allocation model to automatically adapt to changes in airport operations, reducing the complexity of building allocation models across different airports, and improving the versatility of the allocation model. This application also provides a parking stand allocation method, which uses the parking stand allocation model trained by the same method to dynamically respond to the parking stand allocation problem of the corresponding airport, further optimizing and improving the accuracy and effectiveness of parking stand allocation, ensuring that the allocation strategy is always synchronized with the airport's operational status. Attached Figure Description

[0056] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0057] Figure 1 This is a schematic diagram of a scenario for the parking space allocation method provided in this application;

[0058] Figure 2 This is a flowchart illustrating the training method of the parking space allocation model provided in this application. Figure 1 ;

[0059] Figure 3 This is a flowchart illustrating the training method of the parking space allocation model provided in this application. Figure 2 ;

[0060] Figure 4 This is a flowchart illustrating the parking space allocation method provided in this application;

[0061] Figure 5 This is a schematic diagram of the structure of the training device for the parking space allocation model provided in this application;

[0062] Figure 6 This is a schematic diagram of the parking space allocation device provided in this application;

[0063] Figure 7 This is a schematic diagram of the structure of the electronic device provided in this application.

[0064] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0066] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented, for example, in orders other than those illustrated or described herein.

[0067] In this application, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0068] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with relevant laws, regulations and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0069] First, the terms used in this application will be explained.

[0070] Transformer is a deep learning architecture based on self-attention mechanism, used to process sequential data (such as text, time series, etc.). Its core idea is to dynamically calculate the correlation weights between different parts of the input data through the attention mechanism, replacing traditional recurrent neural networks or convolutional neural networks.

[0071] Jaccard similarity: Used to measure the similarity between two sets, defined as the size of the intersection of the two sets divided by the size of the union.

[0072] Cosine similarity: Cosine similarity measures the degree of similarity between two vectors by calculating the cosine of the angle between them.

[0073] Euclidean distance: Euclidean distance is a measure of the straight-line distance between two points in Euclidean space.

[0074] Weighted Euclidean distance: Weighted Euclidean distance is an extension of Euclidean distance, which allows different weights to be assigned to each dimension.

[0075] Structural similarity: This is a metric that measures the similarity between two images, taking into account brightness, contrast, and structural information; the value ranges from -1 to 1, with a higher value indicating a higher similarity.

[0076] Proportional similarity: This is usually used to measure whether the proportional relationship between two values ​​or vectors is similar; it can be obtained by calculating the proportion of corresponding elements in two vectors and comparing these proportions.

[0077] Model fine-tuning is a commonly used technique in deep learning, mainly used to optimize the performance of pre-trained models on specific tasks. The core idea of ​​fine-tuning is to use a model pre-trained on a large-scale dataset as a foundation, and then adjust the model's parameters by further training it on a dataset specific to the task, thereby improving its performance on that task.

[0078] Airport parking space allocation is a critical aspect of aviation operations management, involving the rational allocation of arriving and departing flights to airport parking spaces. Effective parking space allocation can improve airport operational efficiency, reduce flight delays, optimize ground services, and enhance the overall passenger experience.

[0079] In existing technologies, airport parking space allocation is a complex and dynamic optimization problem involving multiple constraints and objectives. Its complexity is reflected in the need to consider various factors such as flight time, aircraft type, parking space restrictions, airline preferences, and safety requirements.

[0080] In existing technologies, mathematical optimization models, heuristic algorithms, and artificial intelligence techniques are commonly used to solve the airport parking space allocation problem. However, existing parking space allocation models need to be built independently for each airport, making the development process cumbersome and resulting in the complexity of traditional parking space allocation models.

[0081] To address the aforementioned issues, this application provides a training method and a parking space allocation method for a parking space allocation model.

[0082] First, the implementation scenarios involved in this application will be explained.

[0083] Figure 1 This is a schematic diagram illustrating a scenario of the parking space allocation method provided in this application. For example... Figure 1As shown, multiple aircraft 101 are connected to the airport ground management system 102, and the multiple aircraft 101 include, but are not limited to, aircraft that are in flight, aircraft that are parked at parking positions, and aircraft that are waiting to take off. The airport ground management system 102 is equipped with a parking position dynamic control module. The parking position dynamic control module can be trained on the parking position allocation model required by the current airport, and the parking position allocation model can be used to allocate parking positions for multiple aircraft in the airport in real time.

[0084] For example, multiple aircraft 101 include: aircraft A, aircraft B, and aircraft C, and aircraft A, aircraft B, and aircraft C can all communicate with the airport ground management system 102. The parking stand dynamic control module in the airport ground management system 102, based on the real-time dynamic data information of aircraft A, aircraft B, and aircraft C, specifically including: time until takeoff, whether it is occupying the runway, whether it is occupying the taxiway, and whether it has parked in a parking stand, inputs the obtained information into the parking stand dynamic control module, and the parking stand dynamic control module outputs parking stand allocation information for aircraft A, aircraft B, and aircraft C. For example, aircraft A drives from parking stand 1-1 into taxiway 1, aircraft B drives from taxiway 2 into parking stand 2-1, and aircraft C drives from taxiway 3 into runway 2.

[0085] This application provides a training method for an aircraft parking space allocation model. Based on an existing parking space allocation model, it determines the similarity of the parking space allocation problem between the airport for which the current allocation model needs to be built and the airport with the existing parking space allocation model. Based on different levels of similarity, corresponding model transfer strategies are determined. The existing parking space allocation model is then transferred according to the model transfer strategies corresponding to different levels of similarity, thereby achieving rapid model transfer. This method optimizes the allocation model construction strategy, improves the model construction efficiency, and reduces the complexity of allocation model construction.

[0086] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0087] Figure 2 This is a flowchart illustrating the training method of the parking space allocation model provided in this application. Figure 1 The implementing entity in this embodiment can be, for example, an airport ground management system equipped with a dynamic parking space control module. Figure 2 As shown, the training method for the parking space allocation model provided in this embodiment includes:

[0088] S201: Determine the similarity of the parking space allocation problem between the first and second airports.

[0089] The similarity metric describes the degree of similarity between the first and second airports in terms of parking space allocation, including their characteristics, challenges, and solutions.

[0090] Understandably, the airport parking stand allocation problem refers to the rational allocation of limited parking stand resources to various flights during airport operations, based on factors such as arrival and departure times, aircraft type, airline, and special needs (such as jet bridge requirements), to ensure safe and efficient flight operations. While the parking stand allocation problem varies across airports due to differences in airport size, layout, and operational rules, there are similarities in core issues, objectives, constraints, and data processing. Specifically, the parking stand allocation problem at different airports requires allocating parking positions to different flights within limited parking stand resources, comprehensively considering core factors such as flight schedules, aircraft type, and airline needs. For example, both hub airports and regional airports need to resolve parking stand conflicts during peak flight periods. The parking stand allocation problem at different airports aims to improve operational efficiency (such as reducing taxiing time), ensure safety (avoiding aircraft conflicts), enhance service quality (such as prioritizing jet bridge allocation), and reduce costs (optimizing resource utilization).

[0091] The parking stand allocation problems of the first and second airports are determined separately, and data processing is performed on the parking stand allocation problems of the first and second airports respectively. The corresponding problem information is extracted according to different dimensions, and the similarity of the corresponding dimensions is calculated based on the problem information of multiple dimensions.

[0092] For example, the first airport could be airport A, and the second airport could be airport B. The parking stand allocation problem for airport A includes: airport A is a large hub airport with high flight density, a complex route network, and multiple types of parking stands (near stands, distant stands, combined stands, etc.), requiring improved parking stand utilization, reduced aircraft taxiing time, ensured safe spacing, optimized passenger experience (e.g., jet bridge allocation), and improved transfer efficiency. The parking stand allocation problem for airport B includes: airport B is a regional airport with relatively fewer flights, but needs to balance domestic and international flight demands, improving parking stand utilization, reducing aircraft taxiing time, ensuring safe spacing, and optimizing passenger experience (e.g., jet bridge allocation), and improving transfer efficiency. To ensure safe spacing and optimize passenger experience (such as jet bridge allocation), and to guarantee normal flight operations, data processing was performed on the parking stand allocation issues at Airport A and Airport B respectively. Relevant problem information was extracted according to different dimensions, and the corresponding similarity was calculated based on these dimensions. These dimensions include, but are not limited to: number and type of parking stands, flight volume and type, aircraft type distribution, aircraft type compatibility requirements, airline-specific requirements, and jet bridge allocation priority. Specifically, the similarity for different dimensions includes: "Number and type of parking stands: Airport A has 200 (50 jet bridges), Airport B has 120..." "30 jet bridges (airport A and airport B), the similarity of the number of parking stands (e.g., cosine similarity) is calculated to be 0.2". "Flight volume and type: Airport A has 1500 flights (50% international), Airport B has 800 flights (30% international), the similarity of flight volume and type (e.g., weighted Euclidean distance) is calculated to be 0.25". "Aircraft type compatibility requirements: Airport A requires Category E wide-body aircraft, Airport B requires Category E wide-body aircraft parking stands, therefore the parking stand constraints are completely identical, and the corresponding similarity is calculated to be 10.15". "Safety interval rules: Airport A requires a 5-minute interval between consecutive flights, Airport B requires a 5-minute interval between consecutive flights, the constraints are completely identical." The similarity between airports A and B is calculated to be 0.81, based on the similarity calculated from multiple dimensions. Specifically, the similarity between airport A and airport B for parking space allocation is 0.1: "Airport-dedicated parking spaces: Airport A has 10 dedicated parking spaces, and Airport B has 5 dedicated parking spaces." The similarity between airport A and airport B for parking space allocation is 0.1, with the calculation result being 10.15. This application does not impose any special restrictions on the parking space allocation problem itself, nor does it impose any special restrictions on the calculation method for the similarity of parking space allocation problems between different airports.

[0093] S202: Determine the model transfer strategy based on the similarity of parking space allocation problems.

[0094] S203: Based on the model transfer strategy and the historical parking space allocation data of the second airport, train the first parking space allocation model of the first airport to obtain the second parking space allocation model applicable to the second airport.

[0095] Among them, the first parking space allocation model is a model that has been trained; the input of the first parking space allocation model is the parking space environmental state information, and the output is the parking space allocation action.

[0096] Understandably, model transfer strategies typically refer to the methods used in machine learning and deep learning to apply a pre-trained model to another related but different task or dataset. Model transfer strategies are suitable for situations where the target task has a small amount of data, high annotation costs, or excessive costs of training the model from scratch. By utilizing the knowledge of the pre-trained model, the performance of the new task can be effectively improved, while reducing training time and data requirements.

[0097] In this embodiment, due to significant structural and operational differences in the parking space allocation problem among different airports, and these differences are reflected in several key factors, such as flight density, number of parking spaces, airline operating models, and the existence of dedicated parking spaces, the differences in these key factors lead to a certain degree of inconsistency in the modeling methods, constraints, and optimization objectives of the parking space allocation problem among different airports. If a model trained at one airport is directly transferred to another airport, it may lead to problems such as poor adaptability, inaccurate predictions, or decreased scheduling efficiency. Therefore, the similarity of the parking space allocation problem can be divided into different similarity levels from high to low, and different similarity levels adopt corresponding model transfer strategies.

[0098] Based on the similarity of the parking stand allocation problem, the similarity level corresponding to the currently acquired similarity is determined. Then, according to the model transfer strategy corresponding to the similarity level, combined with the historical parking stand allocation data of the second airport, the first parking stand allocation model of the first airport is trained to obtain the second parking stand allocation model applicable to the second airport.

[0099] For example, the similarity levels determined based on the similarity of the parking space allocation problem include: high similarity (similarity greater than or equal to 0.8), moderate similarity (similarity greater than or equal to 0.5 but less than 0.8), and low similarity (similarity less than 0.5). If airport A already has a corresponding parking space allocation model, and a parking space allocation model for airport B needs to be constructed, if the similarity between the parking space allocation problems of airport A and airport B is 0.81 (0.81 > 0.8), it indicates a high degree of similarity between the parking space allocation problems of airport A and airport B, corresponding to a high similarity level. Similar optimization algorithms or system architectures can be used. In this case, the parking space data information of airport B can be input into the parking space allocation model of airport A, and the parking space allocation model corresponding to airport A can be fine-tuned using the parking space data information of airport B. The fine-tuned model can then be used as the parking space allocation model for airport B. If the similarity between the parking space allocation problems of airport A and airport B is 0.6 (0.5 < 0.6), then... If the similarity is less than 0.8, it indicates that the similarity between the parking space allocation problems of airport A and airport B is moderate, corresponding to a medium similarity level. Targeted adjustments to some constraints or objective functions are needed. In this case, analyze the differences between airport A and airport B in terms of constraints (such as airline-specific parking space rules) and objective functions (such as jet bridge allocation priority), and adjust the constraint parameters or weights in the algorithm accordingly. Transfer the core algorithm framework (such as rule-based allocation logic), but retrain or adjust some modules (such as dynamically adjusting rules). Then, train the adjusted parking space allocation model for airport A using parking space data from airport B, and use the trained model as the parking space allocation model for airport B. If the similarity between the parking space allocation problems of airport A and airport B is 0.3 (0.3 < 0.5), it indicates that the similarity between the parking space allocation problems of airport A and airport B is low, corresponding to a low similarity level. In this case, the similarity between the two airports is too low to perform model transfer, and an independent allocation strategy needs to be designed.

[0100] In this embodiment, regarding the similarity of parking space allocation models between different airports, when the similarity is low, model transfer based on the existing parking space allocation model is not possible. In other words, the training method for the parking space allocation model provided in this application is suitable for cases with high or medium similarity, but not for cases with excessively low similarity. For example, excessively low similarity can be defined as a similarity of less than 0.5. Furthermore, the similarity value corresponding to excessively low similarity can be dynamically set for different usage scenarios. This application does not impose any special restrictions on the specific similarity value corresponding to excessively low similarity.

[0101] The training method for the parking space allocation model provided in this embodiment determines the similarity between the parking space allocation problems of the first airport and the second airport; based on the similarity, a model transfer strategy is determined; and the first parking space allocation model for the first airport is trained according to the model transfer strategy and the historical parking space allocation data of the second airport to obtain a second parking space allocation model suitable for the second airport. This method improves the efficiency of building the allocation model, enables the allocation model to automatically adapt to changes in airport operations, and reduces the complexity of building allocation models between different airports.

[0102] Figure 3 This is a flowchart illustrating the training method of the parking space allocation model provided in this application. Figure 2 .like Figure 3 As shown, in this embodiment... Figure 2 Based on the embodiments, the training method of the parking space allocation model is described in detail. The training method of the parking space allocation model shown in this embodiment includes:

[0103] S301: Obtain multimodal data related to the parking stand allocation problem for both the first and second airports.

[0104] S302: Determine the similarity of the parking space allocation problem based on the multimodal data of the first and second airports respectively.

[0105] Multimodal data is used to indicate a collection of information about data of multiple types or sources.

[0106] Understandably, multimodal data includes airport-related data of various types or dimensions. Each type or dimension of data represents a specific "modality," that is, data of different dimensions, i.e., data of different modalities. Therefore, based on multimodal data, it is possible to analyze and process the parking space allocation problem of the first and second airports from different dimensions.

[0107] In some embodiments, determining the similarity of the parking space allocation problem between the corresponding airports based on the multimodal data of the first airport and the second airport includes: inputting the multimodal data of the first airport and the second airport into a large model, extracting the initial features of each data item in the multimodal data through the large model, mapping the initial features of each data item to a unified feature space to obtain the feature vectors of each data item corresponding to the first airport and the second airport, and determining the attention weights between different data items of the first airport and the second airport; determining the multimodal fusion features of the first airport and the second airport based on the feature vectors of each data item corresponding to the first airport and the second airport, and the attention weights between different data items of the first airport and the second airport; and determining the similarity of the parking space allocation problem based on the multimodal fusion features of the first airport and the second airport.

[0108] The initial features are used to indicate the basic features extracted from the parking space allocation problem, and the attention weights are used to measure the contribution of different modalities of data on the same parking space allocation problem to the final similarity calculation. The large model can be, for example, a multimodal pre-trained model based on the Transformer architecture.

[0109] In this embodiment, the large model refers to a deep learning model with strong generalization ability trained with massive data and huge computing power, such as GPT (Generative Pre-trained Transformer), BERT (Bidirectional Encoder Representations from Transformers), Transformer, etc. In the scenario of airport parking space allocation, the large model has multimodal processing capabilities and can simultaneously parse heterogeneous data such as numerical data (flight volume), text (rules), spatiotemporal data (flight dynamics), and images (airport layout map). It can also map different multimodal data corresponding to different airports to a unified feature space, that is, convert various types of data from different airports into comparable standardized vectors (such as 1024-dimensional feature vectors). In addition, the large model also has a self-attention mechanism, which can automatically identify the correlation between key features such as "aircraft type distribution - parking space size" and dynamically adjust the corresponding attention weights.

[0110] Multimodal data related to the parking space allocation problem for both airports (Airport 1 and Airport 2) are acquired. A large model set up in the airport ground management system is invoked, and the multimodal data of both airports are input into the large model. Initial features of different dimensions of data from different airports are extracted through the large model, and these initial features are mapped to a unified feature space to obtain feature vectors for each data item corresponding to airports 1 and 2. Based on the self-attention mechanism in the large model, attention weights between different modalities within the same airport are calculated to reflect the relative importance of each modality feature in the task. Based on the attention weights between different data items of airport 1, the feature vectors of each data item corresponding to airport 1 are fused to obtain the multimodal fused features of airport 1. Based on the attention weights between different data items of airport 2, the feature vectors of each data item corresponding to airport 2 are fused to obtain the multimodal fused features of airport 2. Finally, based on the multimodal fused features of airports 1 and 2, the similarity between their corresponding multimodal fused features is calculated, that is, the similarity of the parking space allocation problem between airports 1 and 2 is determined.

[0111] For example, the currently available multimodal data for airports A and B includes, but is not limited to: aircraft type compatibility requirements, flight operation data (including flight volume and type), parking space resource allocation information (including parking space quantity and type), meteorological data, and historical scheduling records. For airports A and B, the corresponding multimodal data is input into a multimodal pre-trained model based on the Transformer architecture. This model extracts features from the multimodal data of the corresponding airports to obtain initial features (e.g., aircraft type compatibility ratio, flight volume, number of parking spaces, etc.). Then, the initial features of each data point are mapped to a unified feature space, resulting in feature vectors for airports A and B respectively. Through a self-attention mechanism, the large model can dynamically focus on key data (such as peak flight times and parking space compatibility) to optimize resource allocation. Therefore, by utilizing the self-attention mechanism and the multimodal data of the Transformer architecture... The pre-trained model calculates attention weights between different modalities within the same airport. Specifically, flight density data for airport A may receive higher weights because it has a greater impact on parking space allocation; meteorological data may receive lower weights because it has a smaller direct impact on parking space allocation. Based on the attention weights of various data points within the corresponding airports, the feature vectors of various modalities for airports A and B are fused to obtain the multimodal fusion features for airports A and B. For example, the multimodal fusion features for airport A may focus more on flight density and aircraft type compatibility, while the multimodal fusion features for airport B may focus more on parking space type and historical scheduling records. Based on the multimodal fusion features of airport A and airport B, the similarity between the two airports is calculated. For example, cosine similarity and Euclidean distance can be used to measure the similarity between the multimodal fusion features of the two airports.

[0112] In some embodiments, determining the similarity of the parking space allocation problem between the respective airports based on the multimodal data of the first airport and the second airport further includes: determining the similarity of each data item in the multimodal data of the first airport and the second airport; and determining the similarity of the parking space allocation problem according to the weight of each data item in the similarity of each data item.

[0113] Understandably, during the construction of the parking stand allocation model for the second airport, corresponding weight information is pre-set for data of different modalities. This application does not impose special restrictions on the pre-setting method of weight information for data of different modalities.

[0114] For example, the dimensions corresponding to the currently acquired multimodal data include, but are not limited to: aircraft type compatibility requirements, flight operation data, parking space resource configuration information, meteorological data, and historical scheduling records, with corresponding preset weights of 0.2, 0.3, 0.25, 0.1, and 0.15, respectively. For data of different modalities, the similarity of the same modal data corresponding to different airports is calculated. Specifically, Jaccard similarity or cosine similarity can be used to calculate the overlap of aircraft type lists and the difference in the proportion of jet bridge parking spaces; Euclidean distance or cosine similarity can be used to calculate the difference in flight volume and flight type proportions; and structural similarity or proportional similarity can be used to calculate the number of parking spaces. Differences in quantity and type; differences in meteorological data can be calculated using Euclidean distance or proportional similarity; differences in historical scheduling records can be calculated using proportional similarity; the similarities of different modal data between airport A and airport B calculated in this instance are as follows: aircraft type adaptation requirements -0.7, flight operation data -0.928, parking space resource allocation information -0.867, meteorological data -0.708, and historical scheduling records -0.84; combined with the preset weights of different modal data, the similarity of the parking space allocation problem between airport A and airport B can be calculated to be 0.83195, that is, the similarity between airport A and airport B in the parking space allocation problem is approximately 83.2%.

[0115] S303: Determine the model transfer strategy based on the similarity of the parking space allocation problem.

[0116] S304: Based on the model transfer strategy and the historical parking space allocation data of the second airport, train the first parking space allocation model of the first airport to obtain the second parking space allocation model applicable to the second airport.

[0117] Given that there is an existing parking stand allocation model for the first airport and a corresponding parking stand allocation model for the second airport, the model transfer strategy for constructing the parking stand allocation model for the second airport is determined based on the similarity between the parking stand allocation problems of the first and second airports. In accordance with this model transfer strategy and combined with the historical parking stand allocation data of the second airport, the first parking stand allocation model for the first airport is retrained to obtain a second parking stand allocation model suitable for the second airport.

[0118] In some embodiments, if the similarity of the parking space allocation problem is greater than or equal to a first threshold, the model transfer strategy is determined to be state alignment and model output layer adjustment; if the model transfer strategy is state alignment and model output layer adjustment, the parameters of the feature extraction layer and intermediate hidden layer of the first parking space allocation model are frozen; the state space applicable to the second airport is mapped to the state space of the first parking space allocation model, and the historical parking space allocation data of the second airport is used to train the first parking space allocation model to adjust the output layer parameters of the first parking space allocation model, thereby obtaining the second parking space allocation model.

[0119] The output layer is used to output the parking stand allocation action, the feature extraction layer is used to extract the basic features in the parking stand allocation problem, the intermediate hidden layer is used to determine the parking stand allocation action, and the historical parking stand allocation data includes: the historical parking stand environmental status information and historical parking stand allocation actions of the second airport; the first threshold can be, for example, 0.8.

[0120] Understandably, the state space is used to reflect the operational status of the airport, and it includes the input information required by the first parking stand allocation model. In other words, the data information included in this state space is consistent with the parking stand environmental state information corresponding to the first airport. Since the second parking stand allocation model is obtained by model transfer based on the first parking stand allocation model, and the first parking stand allocation model is designed based on the state space of the first airport, if the historical parking stand allocation data of the second airport is directly input into the first parking stand allocation model, the first parking stand allocation model cannot correctly understand or process the historical parking stand allocation data of the second airport because the definition, structure, or dimension of the state space may be different. Therefore, it is necessary to convert the parking stand environmental state information of the second airport that is currently acquired into the parking stand environmental state information of the first airport, that is, to map the state space applicable to the second airport to the state space of the first parking stand allocation model.

[0121] For example, if the similarity between the parking space allocation problems of airport A and airport B is 0.81 (0.81 > 0.8), it indicates that the parking space allocation problems of airport A and airport B have a high degree of overlap in the state space. In this case, the model transfer strategy is to align the states of airport A and airport B and adjust only the output layer of the first parking space allocation model. With the model transfer strategy being state alignment and adjusting the model output layer, the parameters of the feature extraction layer and intermediate hidden layer of the first parking space allocation model are frozen. The state space applicable to the second airport is mapped to the state space of the first parking space allocation model. The historical parking space allocation data of the second airport is used to train the first parking space allocation model to adjust the output layer parameters of the first parking space allocation model, and the trained model is determined as the second parking space allocation model. Specifically, when the two airports have similar flight structures and resource configurations, model fine-tuning can be directly adopted to make small adjustments to the output layer parameters of the first parking space allocation model of airport A to quickly adapt to the data distribution of airport B.

[0122] In some embodiments, if the similarity of the parking space allocation problem is greater than or equal to a second threshold and less than a first threshold, the model transfer strategy is determined to be state alignment and model hierarchical adjustment. If the model transfer strategy is state alignment and model hierarchical adjustment, the state space applicable to the second airport is mapped to the state space of the first parking space allocation model. The first parking space allocation model is trained in stages using the historical parking space allocation data of the second airport to obtain the second parking space allocation model.

[0123] The second threshold can be, for example, 0.6.

[0124] In this embodiment, model hierarchical adjustment refers to adjusting the parameters of different layers of the first parking space allocation model to adapt to the dataset of the second airport. When the similarity of the parking space allocation problem is greater than or equal to a second threshold and less than a first threshold, the phased training of the first parking space allocation model specifically includes: a first stage, a second stage, and a third stage. Specifically, in the first stage, the parameters of the feature extraction layer and intermediate hidden layers of the first parking space allocation model are frozen, and the parameters of the output layer of the first parking space allocation model are adjusted. In the second stage, the parameters of the feature extraction layer of the first parking space allocation model are frozen, and the parameters of the intermediate hidden layers and output layers of the first parking space allocation model are adjusted. In the third stage, the parameters of each layer of the first parking space allocation model are adjusted.

[0125] For example, if the similarity between the parking space allocation problems of airport A and airport B is 0.7, where 0.6 < 0.7 < 0.8, it indicates that the parking space allocation problems of airport A and airport B partially overlap in state space. They exhibit certain similarities in key features of parking space allocation (such as the number of parking spaces, flight arrival or departure time distribution, and aircraft type), but also show significant differences. In this case, the model transfer strategy is determined to be state alignment between airport A and airport B, and hierarchical adjustment of the first parking space allocation model. With state alignment and hierarchical model adjustment as the model transfer strategy, the state space applicable to the second airport is mapped to the state space of the first parking space allocation model. Historical parking space allocation data from the second airport is used to adjust the first parking space allocation model. The model is trained in stages, with parameters gradually adjusted to adapt to the needs of Airport B. Specifically, the staged training includes: In the first stage, the parameters of the feature extraction layer and intermediate hidden layers of the first parking space allocation model are frozen, and only the output layer parameters are adjusted to enable it to make decisions based on the actual needs of the second airport; In the second stage, the parameters of the feature extraction layer of the first parking space allocation model are frozen, and the parameters of the intermediate hidden layers and output layers are adjusted to better capture the specific features and rules of the second airport; In the third stage, the parameters of all layers are unfrozen, global adjustments are made, and the model is fully trained using historical parking space allocation data of the second airport to optimize its performance on the parking space allocation problem of Airport B, thus obtaining the second parking space allocation model.

[0126] In some embodiments, if the similarity of the parking space allocation problem is greater than or equal to a third threshold and less than a second threshold, the model transfer strategy is determined to be adding a new model branch, and the new model branch and the original model branch are adjusted hierarchically. If the model transfer strategy is adding a new model branch and the new model branch and the original model branch are adjusted hierarchically, a dedicated model branch is added to the original model branch of the first parking space allocation model. The first parking space allocation model after adding the dedicated model branch is trained in stages to obtain the second parking space allocation model.

[0127] The dedicated model branch is used to handle the state space and action space that differ between the first airport and the first airport. The third threshold can be, for example, 0.45.

[0128] Understandably, adding a new model branch refers to adding one or more sub-models (i.e., branches) on top of the model assigned to the first stop position. This branch shares some or all of the inputs with the original model, but has independent parameters and structure, and is used to handle features or rules specific to the target task. By adding a new model branch, the model can learn and adapt to new task requirements while retaining its original knowledge.

[0129] In this embodiment of the application, when the similarity of the parking space allocation problem is greater than or equal to a third threshold and less than a second threshold, the phased training of the first parking space allocation model specifically includes: a first-stage training, a second-stage training, and a third-stage training; in the first stage, the original model branches of the first parking space allocation model are frozen, and the parameters of the dedicated model branches are adjusted; in the second stage, the feature extraction layer of the original model branches of the first parking space allocation model is frozen, the parameters of the intermediate hidden layer and the output layer of the original model branches are adjusted, and the parameters of the dedicated model branches are adjusted; in the third stage, the parameters of the original model branches and the dedicated model branches are adjusted.

[0130] For example, if the similarity between the parking space allocation problems of airport A and airport B is 0.5, and 0.45 < 0.7 < 0.6, it indicates that the parking space allocation problems of airport A and airport B have a small overlap in state space, but there are significant differences in key features of parking space allocation (such as the number of parking spaces, flight arrival or departure time distribution, aircraft type, etc.). In this case, the model transfer strategy is to add a new model branch and adjust the new model branch and the original model branch in a layered manner. With the model transfer strategy being to add a new model branch and adjust the new model branch and the original model branch in a layered manner, a dedicated model branch is added based on the original model branch of the first parking space allocation model, and the parameters of the dedicated model branch are initialized. Typically, some parameters of the original model branch can be copied or pre-trained weights can be used. Initialization is performed to accelerate the training process. The first parking bay allocation model, after adding a dedicated model branch, is trained in stages to obtain the second parking bay allocation model. Specifically, in the first stage, the original model branches of the first parking bay allocation model are frozen, the parameters of the dedicated model branch are adjusted, and the parameters of the feature extraction layer, intermediate hidden layer, and output layer of the newly added model branch are adjusted based on the historical parking bay allocation data of the second airport. In the second stage, the feature extraction layer of the original model branch of the first parking bay allocation model is frozen, the parameters of the intermediate hidden layer and output layer of the original model branch are adjusted, and the parameters of the dedicated model branch are also adjusted. In the third stage, all layers of the newly added model branch and the original model branch are globally optimized, and the model is fully trained using the historical parking bay allocation data of the second airport to optimize model performance.

[0131] The training method for the parking stand allocation model provided in this embodiment acquires multimodal data related to the parking stand allocation problem for both the first and second airports. Based on this data, the similarity of the parking stand allocation problems is determined. A model transfer strategy is then determined based on this similarity. The first parking stand allocation model for the first airport is trained using the determined transfer strategy and the historical parking stand allocation data of the second airport, resulting in a second parking stand allocation model suitable for the second airport. This method significantly reduces model development and maintenance costs while maintaining the personalized adaptability of the allocation model. It enables the allocation model to automatically adapt to changes in airport operations, improves the efficiency of model construction, reduces the complexity of building allocation models across different airports, enhances the versatility of the allocation model, and improves its efficiency in solving dynamic parking stand allocation problems.

[0132] The training method for the parking space allocation model provided in the above embodiments allows for the dynamic allocation of parking spaces at airports, thereby improving airport operational efficiency and service quality.

[0133] Based on the above embodiments, the parking space allocation model is applied to the parking space allocation method. Figure 4 This is a flowchart illustrating the parking space allocation method provided in this application. The implementing entity in this embodiment can, for example, be an airport ground management system equipped with a dynamic parking space control module. Figure 4 As shown, the training method for the parking space allocation model provided in this embodiment includes:

[0134] S401: Obtain the environmental status information of the parking positions at the second airport.

[0135] Among them, the parking stand environmental status information is a dynamic data set used to describe the real-time status of all parking stands in the airport.

[0136] Understandably, the airport ground management system can collect real-time data on the status of all parking positions in the secondary airport, thereby obtaining the occupancy status of the corresponding parking positions, such as: idle, occupied, or under maintenance. Furthermore, the airport ground management system can also obtain real-time dynamic flight information for the current airport. This application does not impose any special restrictions on the data information that the airport ground management system can obtain.

[0137] For example, the current acquired parking stand environmental status information for Airport B includes: "Stand 1-1: Aircraft A is occupied, estimated departure time 14:30", "Stand 1-2: Available", "Stand 1-3: Aircraft B is occupied, estimated departure time 15:15, ground service (refueling) in progress", "Stand 2-1: Under maintenance, jet bridge equipment malfunction, expected to be restored at 16:00", "Stand 2-2: Available", "Stand 2-3: Aircraft C is occupied, estimated departure time 16:00, boarding is underway at the jet bridge", "Stand 3-1: Available", "Stand 3-2: Aircraft D is occupied, estimated departure time 17:00, remote stand, shuttle bus is in place", "Stand 3-3: Available".

[0138] S402: Input the parking stand environmental status information of the second airport into the second parking stand allocation model to obtain the parking stand allocation action output by the second parking stand allocation model.

[0139] Understandably, the parking stand allocation action is a specific operational instruction generated by the parking stand allocation model based on the parking stand environmental status information. Based on the parking stand allocation action output by this second parking stand allocation model, the spatiotemporal relationship between aircraft, parking stands, and ground services is precisely arranged to avoid flight arrival and departure conflicts and improve airport ground operation efficiency.

[0140] For example, the parking stand environmental status information of the second airport is input into the second parking stand allocation model to obtain the parking stand allocation actions output by the second parking stand allocation model. Specifically, the parking stand allocation actions obtained in this instance include: "14:10: Guide aircraft A from stand 1-1 to taxiway 1; 14:20: Taxiway 1 is open, aircraft A enters runway 1 to queue"; "15:00: Refueling completed, ground crew notified to evacuate equipment; 15:05: Guide aircraft B from stand 1-3 to taxiway 2; 15:10: Taxiway 2 is open, aircraft B enters runway 2 to queue"; "15:45: Boarding completed, jet bridge evacuation notified"; 15:50: Guide aircraft C from stand 2-3 to taxiway 3; 15:55: Taxiway 3 is open, aircraft C enters runway 3 to queue".

[0141] The parking stand allocation method provided in this embodiment obtains the parking stand environmental status information of the second airport, inputs this information into a second parking stand allocation model, and obtains the parking stand allocation action output by the second parking stand allocation model. This method can respond to flight dynamic changes in real time, achieving efficient and dynamic allocation of parking stand resources.

[0142] Figure 5 This is a schematic diagram of the training device for the parking space allocation model provided in this application. (See diagram below.) Figure 5As shown, this application provides a training apparatus for a parking space allocation model. The training apparatus 50 for the parking space allocation model includes:

[0143] Module 501 is used to determine the similarity of the parking space allocation problem between the first airport and the second airport.

[0144] The determining module 501 is further configured to determine a model migration strategy based on the similarity of the parking space allocation problem.

[0145] The processing module 502 is used to train the first parking space allocation model of the first airport according to the model migration strategy and the historical parking space allocation data of the second airport, so as to obtain a second parking space allocation model applicable to the second airport. The first parking space allocation model is a model that has been trained. The input of the first parking space allocation model is the parking space environmental state information, and the output is the parking space allocation action.

[0146] Optionally, the training device for the parking space allocation model further includes: an acquisition module 503.

[0147] The acquisition module 503 is used to acquire multimodal data related to the parking space allocation problem for the first airport and the second airport respectively. The multimodal data includes multiple data from the following categories: airport images, airport layout topology maps, parking space related facility data, flight dynamic data, environmental data, and parking space allocation optimization targets.

[0148] The determining module 501 is further configured to determine the similarity of the parking space allocation problem based on the multimodal data of the first airport and the second airport respectively.

[0149] Optionally, the processing module 502 is further configured to input the multimodal data of the first airport and the second airport into a large model, extract the initial features of each data item in the multimodal data through the large model, map the initial features of each data item to a unified feature space, obtain the feature vectors of each data item corresponding to the first airport and the second airport, and determine the attention weights between different data items of the first airport and the second airport.

[0150] The determining module 501 is further configured to determine the multimodal fusion features of the first airport and the second airport based on the feature vectors of the data corresponding to the first airport and the second airport, and the attention weights between the different data items of the first airport and the second airport.

[0151] The determining module 501 is further configured to determine the similarity of the parking space allocation problem based on the multimodal fusion features of the first airport and the second airport respectively.

[0152] Optionally, the determining module 501 is further configured to determine the similarity of each data item in the multimodal data of the first airport and the second airport respectively.

[0153] The determining module 501 is further configured to determine the similarity of the parking space allocation problem based on the weight of each data item in terms of the similarity of each data item.

[0154] Optionally, the determining module 501 is further configured to determine the model migration strategy as state alignment and model output layer adjustment if the similarity of the parking space allocation problem is greater than or equal to a first threshold.

[0155] The processing module 502 is further configured to freeze the parameters of the feature extraction layer and the intermediate hidden layer of the first stop position allocation model when the model migration strategy is state alignment and adjustment of the model output layer.

[0156] The processing module 502 is further configured to map the state space applicable to the second airport to the state space of the first parking space allocation model, and use the historical parking space allocation data of the second airport to train the first parking space allocation model to adjust the output layer parameters of the first parking space allocation model, thereby obtaining the second parking space allocation model.

[0157] Optionally, the determining module 501 is further configured to determine the model migration strategy as state alignment and model hierarchical adjustment when the similarity of the parking space allocation problem is greater than or equal to a second threshold and less than a first threshold.

[0158] The processing module 502 is further configured to, when the model migration strategy is state alignment and model hierarchical adjustment, map the state space applicable to the second airport to the state space of the first parking space allocation model, and use the historical parking space allocation data of the second airport to train the first parking space allocation model in stages to obtain the second parking space allocation model.

[0159] In the first stage, the parameters of the feature extraction layer and the intermediate hidden layer of the first parking space allocation model are frozen, and the output layer parameters of the first parking space allocation model are adjusted.

[0160] In the second stage, the parameters of the feature extraction layer of the first parking space allocation model are frozen, and the parameters of the intermediate hidden layer and the output layer of the first parking space allocation model are adjusted.

[0161] In the third stage, the parameters of each layer of the first parking space allocation model are adjusted.

[0162] Optionally, the determining module 501 is further configured to determine the model migration strategy as adding a new model branch when the similarity of the parking space allocation problem is greater than or equal to a third threshold and less than a second threshold, and to adjust the new model branch and the original model branch in a hierarchical manner.

[0163] The processing module 502 is further configured to add a dedicated model branch on the basis of the original model branch of the first parking space allocation model when the model migration strategy is to add a new model branch and adjust the new model branch and the original model branch in a hierarchical manner. The dedicated model branch is used to process the state space and action space that are different between the first airport and the first airport.

[0164] The processing module 502 is also used to perform phased training on the first parking position allocation model after adding a dedicated model branch to obtain the second parking position allocation model.

[0165] In the first stage, the original model branch of the first parking position allocation model is frozen, and the parameters of the dedicated model branch are adjusted.

[0166] In the second stage, the feature extraction layer of the original model branch of the first parking position allocation model is frozen, the parameters of the intermediate hidden layer and output layer of the original model branch are adjusted, and the parameters of the dedicated model branch are adjusted.

[0167] In the third stage, the parameters of the original model branch and the special model branch are adjusted.

[0168] Figure 6 This is a schematic diagram of the parking space allocation device provided in this application. Figure 6 As shown, this application provides a parking space allocation device, the parking space allocation device 60 including:

[0169] The acquisition module 601 is used to acquire the parking stand environmental status information of the second airport. The parking stand environmental status information includes parking stand information, taxiway information, runway information, and unassigned flight status information.

[0170] The processing module 602 is used to input the parking stand environmental status information of the second airport into the second parking stand allocation model to obtain the parking stand allocation action output by the second parking stand allocation model. The second parking stand allocation model is trained using the training method of the parking stand allocation model in the above embodiment.

[0171] Figure 7 A schematic diagram of the structure of the electronic device provided in this application. Figure 7 As shown, this application provides an electronic device 70, which includes: a receiver 701, a transmitter 702, a processor 703, and a memory 704.

[0172] Receiver 701 is used to receive instructions and data;

[0173] Transmitter 702 is used to send commands and data;

[0174] Memory 704 is used to store instructions executed by the computer;

[0175] The processor 703 is used to execute computer execution instructions stored in the memory 704 to implement the various steps performed by the training method or the stop position allocation method of the stop position allocation model in the above embodiments. For details, please refer to the relevant descriptions in the foregoing embodiments of the training method or the stop position allocation method of the stop position allocation model.

[0176] Optionally, the memory 704 can be either standalone or integrated with the processor 703.

[0177] When the memory 704 is set up independently, the electronic device also includes a bus for connecting the memory 704 and the processor 703.

[0178] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the training of a stop position allocation model or a stop position allocation method as described above by the electronic device.

[0179] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the training of the stop position allocation model or the stop position allocation method described in any of the foregoing embodiments.

[0180] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0181] The technical solutions of this application have been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it is readily understood by those skilled in the art that the scope of protection of this application is obviously not limited to these specific embodiments. The above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A training method for a parking space allocation model, characterized in that, include: Determine the similarity of the parking space allocation problem between the first and second airports; Based on the similarity of the parking space allocation problem, a model transfer strategy is determined; According to the model migration strategy and the historical parking stand allocation data of the second airport, the first parking stand allocation model of the first airport is trained to obtain a second parking stand allocation model applicable to the second airport. The first parking stand allocation model is a model that has been trained. The input of the first parking stand allocation model is the parking stand environmental state information, and the output is the parking stand allocation action. The process of determining a model transfer strategy based on the similarity of the parking space allocation problem includes: If the similarity of the parking space allocation problem is greater than or equal to the third threshold and less than the second threshold, the model migration strategy is determined to be adding a new model branch, and the new model branch and the original model branch are adjusted hierarchically. When the model migration strategy is to add a new model branch and adjust the new model branch and the original model branch in a hierarchical manner, a special model branch is added on the basis of the original model branch of the first parking position allocation model. The special model branch is used to handle the state space and action space that are different between the first airport and the first airport. The first parking space allocation model with added dedicated model branches is trained in stages to obtain the second parking space allocation model. In the first stage, the original model branch of the first parking position allocation model is frozen, and the parameters of the dedicated model branch are adjusted. In the second stage, the feature extraction layer of the original model branch of the first parking position allocation model is frozen, the parameters of the intermediate hidden layer and output layer of the original model branch are adjusted, and the parameters of the dedicated model branch are adjusted. In the third stage, the parameters of the original model branch and the special model branch are adjusted.

2. The method according to claim 1, characterized in that, The determination of the similarity of the parking space allocation problem between the first airport and the second airport includes: Acquire multimodal data related to the parking stand allocation problem for the first airport and the second airport respectively, wherein the multimodal data includes multiple of the following types of data: airport images, airport layout topology maps, parking stand related facility data, flight dynamic data, environmental data, and parking stand allocation optimization objectives; The similarity of the parking space allocation problem is determined based on the multimodal data of the first airport and the second airport respectively.

3. The method according to claim 2, characterized in that, The step of determining the similarity of the parking space allocation problem based on the multimodal data of the first airport and the second airport includes: The multimodal data of the first airport and the second airport are input into a large model. The initial features of each data item in the multimodal data are extracted through the large model, and the initial features of each data item are mapped to a unified feature space to obtain the feature vectors of each data item corresponding to the first airport and the second airport. Attention weights between different data items of the first airport and the second airport are determined. Based on the feature vectors of the data corresponding to the first airport and the second airport respectively, and the attention weights between different data items of the first airport and the second airport, the multimodal fusion features of the first airport and the second airport are determined respectively. The similarity of the parking space allocation problem is determined based on the multimodal fusion features of the first airport and the second airport respectively.

4. The method according to claim 2, characterized in that, The step of determining the similarity of the parking space allocation problem based on the multimodal data of the first airport and the second airport includes: Determine the similarity of each data item in the multimodal data of the first airport and the second airport respectively; The similarity of the parking space allocation problem is determined based on the weight of each data item according to the similarity of each data item.

5. The method according to any one of claims 1-4, characterized in that, The process of determining a model transfer strategy based on the similarity of the parking space allocation problem includes: If the similarity of the parking space allocation problem is greater than or equal to a first threshold, the model migration strategy is determined to be state alignment and model output layer adjustment. The step of training the first parking space allocation model for the first airport according to the model migration strategy and the historical parking space allocation data of the second airport to obtain a second parking space allocation model suitable for the second airport includes: When the model migration strategy is state alignment and adjustment of the model output layer, the parameters of the feature extraction layer and intermediate hidden layer of the first stop position allocation model are frozen. The state space applicable to the second airport is mapped to the state space of the first parking space allocation model. The historical parking space allocation data of the second airport is used to train the first parking space allocation model to adjust the output layer parameters of the first parking space allocation model, thereby obtaining the second parking space allocation model.

6. The method according to any one of claims 1-4, characterized in that, The process of determining a model transfer strategy based on the similarity of the parking space allocation problem includes: If the similarity of the parking space allocation problem is greater than or equal to the second threshold and less than the first threshold, the model migration strategy is determined to be state alignment and model hierarchical adjustment. The step of training the first parking space allocation model for the first airport according to the model migration strategy and the historical parking space allocation data of the second airport to obtain a second parking space allocation model suitable for the second airport includes: When the model transfer strategy is state alignment and model hierarchical adjustment, the state space applicable to the second airport is mapped to the state space of the first parking space allocation model. The historical parking space allocation data of the second airport is used to train the first parking space allocation model in stages to obtain the second parking space allocation model. In the first stage, the parameters of the feature extraction layer and the intermediate hidden layer of the first parking space allocation model are frozen, and the output layer parameters of the first parking space allocation model are adjusted. In the second stage, the parameters of the feature extraction layer of the first parking space allocation model are frozen, and the parameters of the intermediate hidden layer and the output layer of the first parking space allocation model are adjusted. In the third stage, the parameters of each layer of the first parking space allocation model are adjusted.

7. A method for allocating parking spaces, characterized in that, include: Obtain the parking stand environmental status information of the second airport, which includes parking stand information, taxiway information, runway information, and unassigned flight status information; The parking stand environmental status information of the second airport is input into the second parking stand allocation model to obtain the parking stand allocation action output by the second parking stand allocation model, wherein the second parking stand allocation model is trained using the method described in any one of claims 1-6.

8. A training device for a parking space allocation model, characterized in that, include: The determination module is used to determine the similarity of the parking space allocation problem between the first airport and the second airport; The determining module is also used to determine a model transfer strategy based on the similarity of the parking space allocation problem; The processing module is used to train the first parking space allocation model of the first airport according to the model migration strategy and the historical parking space allocation data of the second airport to obtain a second parking space allocation model applicable to the second airport. The first parking space allocation model is a model that has been trained. The input of the first parking space allocation model is parking space environmental status information, and the output is parking space allocation action. The process of determining a model transfer strategy based on the similarity of the parking space allocation problem includes: The determining module is further configured to determine the model migration strategy as adding a new model branch when the similarity of the parking space allocation problem is greater than or equal to a third threshold and less than a second threshold, and to make hierarchical adjustments to the new model branch and the original model branch. The processing module is further configured to add a dedicated model branch on the basis of the original model branch of the first parking space allocation model when the model migration strategy is to add a new model branch and adjust the new model branch and the original model branch in a hierarchical manner. The dedicated model branch is used to process the state space and action space that are different between the first airport and the first airport. The processing module is also used to perform phased training on the first parking position allocation model after adding a dedicated model branch to obtain the second parking position allocation model. In the first stage, the original model branch of the first parking position allocation model is frozen, and the parameters of the dedicated model branch are adjusted. In the second stage, the feature extraction layer of the original model branch of the first parking position allocation model is frozen, the parameters of the intermediate hidden layer and output layer of the original model branch are adjusted, and the parameters of the dedicated model branch are adjusted. In the third stage, the parameters of the original model branch and the special model branch are adjusted.

9. A parking space allocation device, characterized in that, include: The acquisition module is used to acquire the parking stand environmental status information of the second airport. The parking stand environmental status information includes parking stand information, taxiway information, runway information, and unassigned flight status information. The processing module is used to input the parking stand environmental status information of the second airport into the second parking stand allocation model to obtain the parking stand allocation action output by the second parking stand allocation model, wherein the second parking stand allocation model is trained using the method described in any one of claims 1-6.

10. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-7.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

12. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1-7.

Images