Dynamic scheduling method based on dynamic priority queue and spatiotemporal conflict prediction

Abstract

The present application discloses a dynamic scheduling method based on dynamic priority queues and spatiotemporal conflict prediction, which relates to the field of conflict prediction and coordinated scheduling. The method includes: obtaining information data of the target airport; determining the time-slotted take-off and landing queue sequence based on the airport flight resource data based on the multi-service station queuing model; predicting potential conflict heat map information based on the radar and airport topology data based on the conflict prediction model; the conflict prediction model is obtained by training the spatiotemporal graph convolutional network obtained by temporal modeling based on the graph convolutional network combined with the gated recurrent unit according to the historical radar and airport topology data; using a mixed integer programming model, determining the scheduling plan based on the take-off and landing queue sequence and the potential conflict heat map information; the scheduling plan is used to schedule flights within the airport. The present application can effectively reduce the risk of conflicts and collisions between flights while improving operational efficiency, thereby optimizing scheduling.

Description

Technical Field

[0001] The present application relates to the field of conflict prediction and coordinated scheduling, and in particular to a dynamic scheduling method based on dynamic priority queues and spatiotemporal conflict prediction. Background Art

[0002] With the continued development of the global economy and the growing demand for air transport, airports, as key hubs for air transportation, face unprecedented challenges in operational efficiency and safety. On the one hand, the ever-increasing number of aircraft takeoffs and landings and the increasing workload have placed higher demands on airport flight handling capacity. How to maximize flight takeoffs and landings within limited airport resources (such as runways, parking spaces, and maintenance stations) to improve airport utilization and passenger transport efficiency has become a pressing issue. In current air traffic control, traditional scheduling methods often struggle to adapt to the complex and ever-changing flight operating environment. The multi-flight, multi-aircraft flight model complicates coordination and management between flights, leading to flight delays and wasted runway resources. On the other hand, flight safety remains the top priority for air transportation. How to effectively reduce the risk of conflicts and collisions between flights while improving operational efficiency and ensuring flight safety is a key issue in air traffic control. Summary of the Invention

[0003] The purpose of this application is to provide a dynamic scheduling method based on dynamic priority queues and spatiotemporal conflict prediction, which can effectively reduce the risk of conflicts and collisions between flights while improving operational efficiency and achieving scheduling optimization.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a dynamic scheduling method based on dynamic priority queues and spatiotemporal conflict prediction, comprising:

[0006] Acquire information data of the target airport; the information data includes: airport flight resource data and radar and airport topology data; the airport flight resource data includes: flight data and airport resource data; the radar and airport topology data includes: radar trajectory data and airport topology map;

[0007] Based on a multi-service station queuing model, a time-slotted takeoff and landing queue sequence is determined based on the airport flight resource data. The multi-service station queuing model treats the airport runways as multiple service stations and is trained based on an improved M / M / c / K queuing model using a hybrid priority algorithm based on historical information data. The improved M / M / c / K queuing model treats takeoff and landing flights as different service queues.

[0008] Based on the conflict prediction model, potential conflict heat map information is predicted based on the radar and airport topology data; the conflict prediction model is obtained by training a spatiotemporal graph convolutional network based on temporal modeling using a graph convolutional network combined with a gated recurrent unit based on historical radar and airport topology data;

[0009] A mixed integer programming model is used to determine a scheduling plan based on the take-off and landing queue sequence and the potential conflict heat map information; the scheduling plan is used to schedule flights within the airport; and the mixed integer programming model is a mathematical model for multi-objective coordinated scheduling optimization.

[0010] In one embodiment, the hybrid priority algorithm determines the priority of each flight by a dynamic priority factor; the mathematical calculation formula corresponding to the hybrid priority algorithm is:

[0011] ;

[0012] in, For flights At the moment priority; is the time sensitivity coefficient; is a parameter; For flights planned time; is the remaining fuel weight coefficient; For flights The remaining fuel amount; is the maximum fuel capacity of the aircraft type; is the emergency identification coefficient; It is an emergency sign. When the flight is in an emergency state, ,otherwise .

[0013] In one embodiment, the method for determining the conflict prediction model specifically includes:

[0014] Based on a random matrix of fixed dimension, the flight position coordinate information in the historical radar trajectory data is mapped into a feature vector;

[0015] Taking flights as nodes, the dynamic edge weights are determined based on the relative positions and speeds between flights;

[0016] Determine the position and speed vector information of the flight according to the dynamic edge weight;

[0017] Convert historical airport topology maps into digital image data;

[0018] A graph convolutional network is combined with a gated recurrent unit for temporal modeling to obtain a spatiotemporal graph convolutional network.

[0019] Inputting the image data and the position and velocity vector information into the spatiotemporal graph convolutional network, and using the optimizer Adam to train the network based on a set learning rate and batch size with the goal of minimizing the loss function, thereby obtaining a trained spatiotemporal graph convolutional network;

[0020] The trained spatiotemporal graph convolutional network is determined as the conflict prediction model.

[0021] In one embodiment, the loss function is determined based on true labels and the conflict probability predicted by the spatiotemporal graph convolutional network; the true labels are obtained by annotating historical radar and airport topology data based on historical conflict event records;

[0022] The expression of the loss function is:

[0023] ;

[0024] in, is the true label; is the predicted conflict probability; C is the total number of potential conflicts; N is the total number of training samples; is the potential conflict sequence number; is the loss function.

[0025] In one embodiment, the mixed integer programming model includes an objective function and constraints;

[0026] The constraints include: minimum safety interval constraint, maintenance continuity constraint and runway capacity limitation constraint;

[0027] The expression of the objective function is:

[0028] ;

[0029] in, is the sum of the squares of the delay times of all flights; For flights Delay time; is the total number of flights; is the sum of the probabilities of all potential conflicts; is the first weight coefficient; It is The probability of potential conflicts; C is the total number of potential conflicts; To maximize the utilization of parking spaces; is the second weight coefficient; is the utilization rate of parking spaces.

[0030] In one embodiment, the expression of the dynamic edge weight is:

[0031] ;

[0032] in, For flights i and flights j The edge weights between them; For flights i The velocity vector of For flights j The velocity vector of For flights i and flights j The norm of the difference between velocity vectors; is the time interval; is the safety distance threshold.

[0033] In a second aspect, the present application provides a dynamic scheduling system based on dynamic priority queues and spatiotemporal conflict prediction, including:

[0034] An information data acquisition module is used to acquire information data of a target airport; the information data includes: airport flight resource data and radar and airport topology data; the airport flight resource data includes: flight data and airport resource data; the radar and airport topology data includes: radar trajectory data and airport topology map;

[0035] A takeoff and landing queue sequence determination module is configured to determine a time-slotted takeoff and landing queue sequence based on the airport flight resource data and a multi-service station queuing model. The multi-service station queuing model uses the airport runways as multiple service stations and is trained using a hybrid priority algorithm based on historical information data based on an improved M / M / c / K queuing model. The improved M / M / c / K queuing model uses takeoff and landing flights as different service queues.

[0036] a potential conflict heat map information determination module, configured to predict potential conflict heat map information based on the radar and airport topology data based on a conflict prediction model; the conflict prediction model is obtained by training a spatiotemporal graph convolutional network based on temporal modeling using a graph convolutional network combined with a gated recurrent unit based on historical radar and airport topology data;

[0037] A scheduling scheme determination module is used to determine a scheduling scheme based on the take-off and landing queue sequence and the potential conflict heat map information using a mixed integer programming model; the scheduling scheme is used to schedule flights within the airport; the mixed integer programming model is a mathematical model for multi-objective coordinated scheduling optimization.

[0038] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned dynamic scheduling method based on dynamic priority queues and spatiotemporal conflict prediction.

[0039] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned dynamic scheduling method based on dynamic priority queues and spatiotemporal conflict prediction.

[0040] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned dynamic scheduling method based on dynamic priority queues and spatiotemporal conflict prediction.

[0041] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0042] The present application provides a dynamic scheduling method based on dynamic priority queues and spatiotemporal conflict prediction, which relates to the field of conflict prediction and coordinated scheduling. The method includes: obtaining information data of the target airport; determining the take-off and landing queue sequence of time slots based on the airport flight resource data based on a multi-service station queuing model; predicting potential conflict heat map information based on radar and airport topology data based on a conflict prediction model; the conflict prediction model is obtained by training a spatiotemporal graph convolutional network obtained by temporal modeling using a graph convolutional network combined with a gated recurrent unit based on historical radar and airport topology data; using a mixed integer programming model, determining a scheduling plan based on the take-off and landing queue sequence and the potential conflict heat map information; the scheduling plan is used to schedule flights within the airport. The present application determines the take-off and landing queue sequence of time slots based on the multi-service station queuing model, which can improve the operating efficiency of each flight at the airport. In addition, the present application uses a conflict prediction model to predict potential conflict heat map information, which can intuitively reflect the possibility of conflict and provide controllers with intuitive conflict risk distribution information. In addition, the use of a mixed integer programming model achieves global optimization of airport operations. When determining the scheduling plan, the mixed integer programming model will adjust flight takeoff and landing times or runway allocations accordingly to avoid conflicts and ensure flight safety. As a result, this application can effectively reduce the risk of conflicts and collisions between flights while improving operational efficiency, achieving optimized scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 A flowchart of a dynamic scheduling method based on dynamic priority queues and spatiotemporal conflict prediction;

[0045] Figure 2 Schematic diagram of the steps to build a dynamic priority queue;

[0046] Figure 3 Schematic diagram of the processing process of the spatiotemporal conflict prediction network;

[0047] Figure 4 Schematic diagram of the multi-objective collaborative scheduling optimization process;

[0048] Figure 5 This is a structural diagram of a dynamic scheduling system based on dynamic priority queues and spatiotemporal conflict prediction;

[0049] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0051] Queuing theory and space-time network theory, both from transportation theory, offer new insights and approaches to solving problems in air traffic control. Queuing theory can be used to optimize flight service sequences, rationally arrange takeoff and landing queues, and improve the efficiency of airport resource utilization. Space-time network theory, by combining flight spatiotemporal information with airport topology, can predict and avoid potential conflicts, ensuring flight safety. However, existing air traffic control methods still have shortcomings in their integrated application of these theories. For example, flight priority determination lacks comprehensive consideration of multiple factors (such as flight delay time, remaining fuel, and emergency status); conflict prediction lacks the ability to handle complex spatiotemporal relationships and multiple conflict patterns; and scheduling optimization struggles to effectively address the combinatorial explosion problem under multidimensional constraints, resulting in low computational efficiency.

[0052] To this end, this application combines transportation theory to propose a dynamic scheduling method for air traffic control based on dynamic priority queues and spatiotemporal conflict prediction. First, a variety of data, including historical flight data, airport resource usage logs, and radar trajectory data, is collected. A dynamic priority queue based on a multi-server queuing model is designed. Using an improved M / M / c / K queuing model, a hybrid priority algorithm is used to calculate priorities, taking into account factors such as aircraft type, remaining fuel, and planned schedule. This allows for a rational scheduling of flight takeoffs and landings, improving airport resource utilization efficiency and ensuring priority processing for low-fuel and emergency flights. Next, a conflict prediction model based on a spatiotemporal graph convolutional network is constructed. Through the collaborative work of multiple modules, including a spatiotemporal embedding layer, relationship graph construction, and a graph convolution module, a heat map of potential conflicts within the next five minutes is predicted, providing controllers with intuitive conflict risk distribution information. Finally, a mixed integer programming model is developed for multi-objective collaborative scheduling optimization, comprehensively considering factors such as flight takeoff and landing times, runway allocation, and maintenance schedules, minimizing flight delays while ensuring flight safety.

[0053] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0054] In an exemplary embodiment, Figure 1 As shown, a dynamic scheduling method based on dynamic priority queue and spatiotemporal conflict prediction is provided, including:

[0055] Step 100: Obtain information data for the target airport. This information data includes: airport flight resource data and radar and airport topology data. Airport flight resource data includes flight data and airport resource data. Flight data includes aircraft type, remaining fuel, and planned schedule. Airport resource data includes runway number and maintenance station capacity. Radar and airport topology data includes radar track data and an airport topology map.

[0056] Step 200: Based on the multi-service station queuing model and airport flight resource data, a time-slotted takeoff and landing queue sequence is determined. The takeoff and landing queue sequence represents the order of takeoff or landing flights. The multi-service station queuing model treats airport runways as multiple service stations and is based on an improved M / M / c / K queuing model, trained using a hybrid priority algorithm based on historical information data. The improved M / M / c / K queuing model treats takeoff and landing flights as separate service queues.

[0057] The hybrid priority algorithm determines the priority of each flight through a dynamic priority factor; the mathematical formula corresponding to the hybrid priority algorithm is:

[0058] ;

[0059] in, For flights At the moment priority; is the time sensitivity coefficient; is a parameter; For flights planned time; is the remaining fuel weight coefficient; For flights The remaining fuel amount; is the maximum fuel capacity of the aircraft type; is the emergency identification coefficient; It is an emergency sign. When the flight is in an emergency state, ,otherwise .

[0060] Step 300: Based on the conflict prediction model, potential conflict heatmap information is predicted based on radar and airport topology data. This potential conflict heatmap information represents the probability distribution of conflicts occurring at various spatiotemporal locations within the airport. The conflict prediction model is trained using historical radar and airport topology data using a spatiotemporal graph convolutional network (GCNN) based on temporal modeling using a graph convolutional network combined with gated recurrent units.

[0061] The method for determining the conflict prediction model specifically includes:

[0062] Based on a random matrix of fixed dimension, the position coordinate information of flights in historical radar trajectory data is mapped into feature vectors.

[0063] Taking flights as nodes, dynamic edge weights are determined based on the relative positions and speeds between flights.

[0064] The expression of dynamic edge weight is:

[0065] .

[0066] in, For flights i and flights j The edge weights between them; For flights i The velocity vector of For flights j The velocity vector of For flights i and flights j The norm of the difference between velocity vectors; is the time interval; is the safety distance threshold.

[0067] The flight's position and velocity vector information is determined based on the dynamic edge weights.

[0068] Convert historical airport topology maps into digital image data.

[0069] The graph convolutional network is combined with the gated recurrent unit for temporal modeling to obtain the spatiotemporal graph convolutional network.

[0070] The image data as well as the position and velocity vector information are input into the spatiotemporal graph convolutional network. The optimizer Adam is used to train the network based on the set learning rate and batch with the goal of minimizing the loss function to obtain the trained spatiotemporal graph convolutional network.

[0071] The trained spatiotemporal graph convolutional network is determined as the conflict prediction model.

[0072] The loss function is determined based on the true label and the conflict probability predicted by the spatiotemporal graph convolutional network; the true label is obtained by annotating historical radar and airport topology data based on historical conflict event records.

[0073] The expression of the loss function is:

[0074] .

[0075] in, is the true label; is the predicted conflict probability; C is the total number of potential conflicts; N is the total number of training samples; is the potential conflict sequence number; is the loss function.

[0076] Step 400: Using a mixed integer programming model, a scheduling solution is determined based on the takeoff and landing queue sequence and the potential conflict heat map information. The scheduling solution is used to schedule flights within the airport; the mixed integer programming model is a mathematical model for multi-objective coordinated scheduling optimization.

[0077] The mixed integer programming model includes an objective function and constraints; the constraints include: minimum safety interval constraint, maintenance continuity constraint and runway capacity limit constraint.

[0078] The expression of the objective function is:

[0079] ;

[0080] in, is the sum of the squares of the delay times of all flights; For flights Delay time; is the total number of flights; is the sum of the probabilities of all potential conflicts; is the first weight coefficient; It is The probability of potential conflicts; C is the total number of potential conflicts; To maximize the utilization of parking spaces; is the second weight coefficient; is the utilization rate of parking spaces.

[0081] In practical applications, the design steps for the method mentioned in this application are as follows:

[0082] Step 0: Data collection requirements.

[0083] Collect historical flight data (such as aircraft type, delay time, fuel consumption, emergency status), airport resource data, and airport resource usage logs to prepare for training dynamic priority queues.

[0084] Collect radar trajectory data (such as timestamp, position, speed), airport topology (such as runway and taxiway coordinates), historical conflict event records (time, location, conflict type), and other data to prepare for training spatiotemporal conflict prediction.

[0085] Step 1: Dynamic priority queue construction, such as Figure 2 shown.

[0086] In actual airport operations, flight situations are complex and diverse. Factors such as aircraft type, remaining fuel, and planned schedule all influence the urgency of takeoff and landing. Furthermore, airport resource data, such as the number of runways and maintenance bay capacity, is limited. To more rationally schedule flight takeoffs and landings and improve the efficiency of airport resource utilization, a priority queue system is needed that comprehensively considers multiple factors and dynamically optimizes the service order of flights. Based on a multi-service station queuing model, the inputs are real-time flight data (including aircraft type, remaining fuel, and planned schedule) and airport resource data (number of runways and maintenance bay capacity). The output is a time-slotted takeoff and landing queue sequence. This system determines the order in which flights should take off or land in each time slot based on priority calculations.

[0087] Specifically, a priority queue system based on a multi-desk queuing model is designed. The airport runways are considered multiple service desks, and takeoff and landing flights are considered separate service queues. The traditional M / M / c / K model treats all flights as a single queue, handled according to the first-come, first-served (FCFS) rule. This fails to distinguish between takeoff and landing priorities and does not allow high-priority flights to preempt resources. To address this, an improved M / M / c / K queuing model is designed, where M represents the exponential distribution of both the arrival and service processes, c represents the number of service desks (i.e., the number of runways), and K represents the system's capacity constraints (e.g., resource constraints such as runways and parking spaces). Takeoff and landing are considered two independent but preemptible service queues.

[0088] To determine the priority of each flight, a hybrid priority algorithm is used, calculated by a dynamic priority factor:

[0089] .

[0090] in, Indicates flight At the moment priority. is the time-sensitive coefficient, and The value increases exponentially with delay time, ensuring that long-delayed flights automatically receive higher priority, thereby balancing on-time performance with emergency response requirements. The specific value can be adjusted based on the actual operation of the airport and its tolerance for delays. It is a positive parameter used to control the decay rate of the influence of time factors on priority. Its value range is [0.01, 0.1] and is set to 0.01 here. It is the present moment. For flights i planned time. The remaining fuel weight coefficient is used to measure the impact of the remaining fuel on the priority. The recommended value is between [0.1, 0.5] and can be set according to different aircraft models and fuel consumption characteristics. For flights The remaining fuel amount. The maximum fuel capacity for this aircraft type. It is the emergency identification coefficient, which is used to indicate whether the flight is in an emergency state. The value is generally 1 (emergency state) or 0 (non-emergency state). It is an emergency sign. When the flight is in an emergency state, ,otherwise .

[0091] When the runway is free, the system will select the highest priority flight at the head of the takeoff and landing queue to perform takeoff and landing operations.

[0092] Step 2: Spatiotemporal conflict prediction network (conflict prediction model), such as Figure 3 shown.

[0093] In a multi-flight, multi-aircraft airport environment, there are numerous potential conflicts between flights, such as overlapping taxi paths and wake turbulence interference. To proactively detect and avoid these conflicts and ensure flight safety, a model that can accurately predict potential conflicts is necessary. A conflict prediction model based on a spatiotemporal graph convolutional network effectively combines spatiotemporal information, such as flight location and velocity, with airport topology to predict future conflicts. Its inputs are radar data (flight location and velocity vectors) and airport topology (taxiway and runway layouts). The output is a heat map of potential conflicts over the next five minutes, visually displaying the probability distribution of potential conflicts at various spatiotemporal locations within the airport.

[0094] Specifically, we design a conflict prediction model based on a spatiotemporal graph convolutional network. This model mainly consists of the following parts:

[0095] 2-1. Spatiotemporal embedding layer: flight coordinate information (representing the longitude, latitude and altitude of the flight respectively) are multiplied by a random matrix W of fixed dimension 3×128 and mapped into a 128-dimensional feature vector.

[0096] 2-2. Relationship graph construction: Using flights as nodes, dynamic edge weights are established based on the relative position and speed relationships between flights. The edge weight calculation formula is:

[0097] .

[0098] Used to measure the difference in their speed. is the time interval, used here to calculate The change in the relative positions of two flights within a certain time period, where the value is 1 minute. is the safety distance threshold. When two flights are The relative distance within a time is less than , the weight of the edge will increase, indicating that there is a higher risk of conflict between them. The value is determined according to different aircraft models and airport environments. Here it is set to 1 kilometer.

[0099] 2-3. Graph Convolutional Module: This module uses a three-layer GCN (Graph Convolutional Network) combined with a GRU (Gated Recurrent Unit) for time series modeling. The GCN captures the spatial relationships between flights, while the GRU processes the changes in flight status over time. After processing by these modules, the model outputs the conflict probability for each spatiotemporal unit, i.e., the probability of a conflict occurring at each possible flight's spatiotemporal location within the next five minutes, forming a potential conflict heatmap.

[0100] 2-4. Loss Function Design: In the aviation control scenario, the core task of the spatiotemporal conflict prediction network is to accurately predict possible future conflicts. This is essentially a binary classification task, that is, to determine whether a conflict will occur within a certain spatiotemporal unit.

[0101] Input: Radar trajectory data (position, velocity): Flight location (latitude, longitude, altitude) and velocity vector information collected in real time by the airport's radar system. Airport topology map: Detailed description of the layout of airport infrastructure such as taxiways and runways. The airport topology map is converted into digital graph data for model processing and analysis. The Adam optimizer (learning rate = 0.001) is used, with a batch size of 32 and 100 training epochs. The loss function formula is designed as follows:

[0102] .

[0103] in, is the true label (0 for no conflict, 1 for conflict). In actual air traffic control data, through detailed analysis and manual annotation of historical radar trajectory data, we determine whether a conflict occurred at a specific time point in each spatiotemporal unit, which serves as the true reference standard for model training.

[0104] is the conflict probability predicted by the model (0-1). The model extracts and analyzes features of the input radar data and airport topology information, and outputs the probability value of a conflict occurring in each space-time unit. This value reflects the model's judgment on the possibility of conflict in that space-time unit. C is the total number of potential conflicts (the number of space-time units divided according to the thermal layer). The spatial range of the airport is divided into multiple space-time units according to certain rules. Each space-time unit corresponds to a potential conflict judgment point. The total number of these space-time units is C. By predicting the conflict for each space-time unit, it is possible to fully cover the areas within the airport where conflicts may occur. N: The total number of training samples. During the training process, the total number of samples used, which includes radar data and corresponding conflict labels in different time periods, different flight flows, and different weather conditions, to ensure that the model can learn conflict patterns in various situations.

[0105] Step 3: Multi-objective collaborative scheduling optimization, such as Figure 4 shown.

[0106] After constructing the flight priority queue and predicting potential conflicts, it is necessary to comprehensively consider multiple factors, such as flight takeoff and landing times, runway allocation, and maintenance schedules, to perform global optimization to maximize airport operational efficiency while satisfying constraints such as safety and resource limitations. The dynamic priority queue constructed in step 1 determines the takeoff and landing order of flights, while the potential conflict heat map predicted in step 2 reflects the conflict risk at different spatiotemporal locations. Combining these two allows scheduling optimization to both consider flight priorities and avoid potential conflicts. Therefore, a mixed integer programming model is designed to perform multi-objective collaborative scheduling optimization.

[0107] The input is the output queues from the first two steps (the time-slotted takeoff and landing queue sequence) and the conflict prediction (a heat map of potential conflicts within the next five minutes). The output is a global optimization plan that includes takeoff and landing times, runway allocation, and maintenance schedules. This plan determines the specific takeoff and landing times for each flight, the runways to be used, and the time and schedule for maintenance.

[0108] Specifically, a mixed integer programming model is established, and the objective function is:

[0109] .

[0110] in, It represents the sum of the squares of the delay time of all flights. By minimizing this term, flight delays can be minimized. It's a flight i The delay time, is the total number of flights. Represents the sum of the probabilities of all potential conflicts, multiplied by the first weight coefficient to balance conflict risks with other objectives, The recommended value range is [0.1, 1], which can be adjusted according to the airport's emphasis on safety. It is The probability of a potential conflict, C is the total number of potential conflicts. Indicates maximizing the utilization of parking spaces, multiplied by the second weight coefficient to balance parking space utilization and other objectives, The recommended value range is [0.1, 1] and can be set according to the tightness of parking space resources. is the utilization rate of parking spaces.

[0111] Constraints include:

[0112] 1. Minimum safety interval: (related to aircraft type combination), i.e. any two flights i and j Time interval between takeoff and landing Must be greater than or equal to the minimum safety distance determined based on their aircraft type combination to ensure flight safety.

[0113] 2. Maintenance continuity:

[0114] , indicating flights k Maintenance time At its arrival time Plus cooling time After that, and at its departure time Subtract preparation time Before that, ensure the continuity and effectiveness of maintenance work.

[0115] 3. Runway capacity limitations: ,in, is an indicator variable, indicating that Runway usage (1 if used, 0 if not used), is a time set, It is the maximum capacity limit of the runway, ensuring that the runway is not overloaded.

[0116] The mixed integer programming problem described above can be solved and optimized using Gurobi or CPLEX solvers, ultimately obtaining the optimal global optimization solution that includes takeoff and landing times, runway allocation, and maintenance scheduling.

[0117] In step 1, a dynamic priority queue was designed. This queue model, based on the multi-desk queue model, employs a modified M / M / c / K queue model, treating airport runways as service desks and takeoff and landing flights as separate service queues. This model comprehensively considers multiple flight attributes (aircraft type, remaining fuel, and planned schedule) as well as airport resource constraints (number of runways and maintenance station capacity). By calculating dynamic priority factors, the takeoff and landing order of flights can be flexibly adjusted based on actual conditions. Furthermore, the time-sensitive coefficient and contingency indicator coefficient enable flexible coordination between long-delayed flights and emergency flights.

[0118] The remaining fuel weight coefficient ensures that flights with low fuel levels are served first, avoiding safety risks caused by fuel depletion, making the airport's flight scheduling more scientific and reasonable, and improving overall operational efficiency.

[0119] In step 2, we designed a spatiotemporal conflict prediction network. This network, based on a spatiotemporal graph convolutional network, uses multiple modules working together to effectively ensure flight safety and improve airport operational efficiency. The spatiotemporal embedding layer maps flight coordinate information into a 128-dimensional feature vector, enabling subsequent models to better process and analyze flight spatiotemporal information. This provides a unified feature representation for conflict prediction and enhances the model's understanding of flight locations and motion.

[0120] The relationship graph construction module establishes dynamic edge weights based on the relative positions and speeds of flights, accurately reflecting the risk of conflict between them. When the relative distance between two flights falls below a safety distance threshold for a short period of time, the edge weight increases, allowing the model to intuitively reflect the increased likelihood of conflict. Ultimately, the model accurately predicts the probability of conflict for each spacetime unit within the next five minutes, creating a potential conflict heat map that provides controllers with intuitive information on the distribution of conflict risks.

[0121] Through step 3, a multi-objective collaborative scheduling optimization was designed. By establishing a mixed integer programming model and comprehensively considering factors such as flight take-off and landing times, runway allocation, and maintenance schedules, the global optimization of airport operations was achieved. To reduce flight delays, by minimizing the sum of the squares of all flight delay times in the objective function, the scheduling plan prioritizes flights with long delays, ensuring that flights take off and land on time, reducing waiting times, and improving the airport's overall operational efficiency. Regarding flight safety, by multiplying the sum of potential conflict probabilities by a weight coefficient, the scheduling plan, while considering flight priorities, fully addresses potential conflict areas and high-risk periods. When the potential conflict probability is high, the scheduling plan will adjust flight take-off and landing times or runway allocations accordingly to avoid conflicts and ensure flight safety.

[0122] In an exemplary embodiment, Figure 5 As shown, a dynamic scheduling system based on dynamic priority queues and spatiotemporal conflict prediction is provided, including:

[0123] The information data acquisition module is used to obtain information data of the target airport. The information data includes: airport flight resource data and radar and airport topology data; airport flight resource data includes: flight data and airport resource data; radar and airport topology data includes: radar trajectory data and airport topology map.

[0124] The takeoff and landing queue sequence determination module is used to determine the time-slotted takeoff and landing queue sequence based on airport flight resource data based on a multi-service desk queuing model. The multi-service desk queuing model treats the airport runways as multiple service desks and is based on an improved M / M / c / K queuing model, trained using a hybrid priority algorithm based on historical information data. The improved M / M / c / K queuing model treats takeoff and landing flights as different service queues.

[0125] A potential conflict heat map information determination module is used to predict potential conflict heat map information based on a conflict prediction model and radar and airport topology data. The conflict prediction model is obtained by training a spatiotemporal graph convolutional network based on temporal modeling using a graph convolutional network combined with a gated recurrent unit based on historical radar and airport topology data.

[0126] The scheduling scheme determination module is used to determine the scheduling scheme based on the take-off and landing queue sequence and potential conflict heat map information using a mixed integer programming model; the scheduling scheme is used to schedule flights within the airport; the mixed integer programming model is a mathematical model for multi-objective coordinated scheduling optimization.

[0127] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a dynamic scheduling method based on a dynamic priority queue and spatiotemporal conflict prediction is implemented.

[0128] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0129] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0130] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0131] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0132] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0133] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A dynamic scheduling method based on dynamic priority queue and spatiotemporal conflict prediction, characterized in that: include: Obtain information data of the target airport; The information data includes: airport flight resource data and radar and airport topology data; the airport flight resource data includes: flight data and airport resource data; the radar and airport topology data includes: radar trajectory data and airport topology map; Based on a multi-service station queuing model, a time-slotted takeoff and landing queue sequence is determined based on the airport flight resource data. The multi-service station queuing model treats the airport runways as multiple service stations and is based on an improved M / M / c / K queuing model, trained using a hybrid priority algorithm based on historical information data. The improved M / M / c / K queuing model treats takeoff and landing flights as different service queues. Based on the conflict prediction model, potential conflict heat map information is predicted based on the radar and airport topology data; the conflict prediction model is obtained by training a spatiotemporal graph convolutional network based on temporal modeling using a graph convolutional network combined with a gated recurrent unit based on historical radar and airport topology data; A mixed integer programming model is used to determine a scheduling plan based on the take-off and landing queue sequence and the potential conflict heat map information; the scheduling plan is used to schedule flights within the airport; and the mixed integer programming model is a mathematical model for multi-objective coordinated scheduling optimization.

2. The dynamic scheduling method based on dynamic priority queue and spatiotemporal conflict prediction according to claim 1, characterized in that: The hybrid priority algorithm determines the priority of each flight through a dynamic priority factor; the mathematical calculation formula corresponding to the hybrid priority algorithm is: ； in, For flights At the moment priority; is the time sensitivity coefficient; is a parameter; For flights planned time; is the remaining fuel weight coefficient; For flights The remaining fuel amount; is the maximum fuel capacity of the aircraft type; is the emergency identification coefficient; It is an emergency sign. When the flight is in an emergency state, ,otherwise .

3. The dynamic scheduling method based on dynamic priority queue and spatiotemporal conflict prediction according to claim 1, characterized in that: The method for determining the conflict prediction model specifically includes: Based on a random matrix of fixed dimension, the flight position coordinate information in the historical radar trajectory data is mapped into a feature vector; Taking flights as nodes, the dynamic edge weights are determined based on the relative positions and speeds between flights; Determine the position and speed vector information of the flight according to the dynamic edge weight; Convert historical airport topology maps into digital image data; A graph convolutional network is combined with a gated recurrent unit for temporal modeling to obtain a spatiotemporal graph convolutional network. Inputting the image data and the position and velocity vector information into the spatiotemporal graph convolutional network, and using the optimizer Adam to train the network based on a set learning rate and batch size with the goal of minimizing the loss function, thereby obtaining a trained spatiotemporal graph convolutional network; The trained spatiotemporal graph convolutional network is determined as the conflict prediction model.

4. The dynamic scheduling method based on dynamic priority queue and spatiotemporal conflict prediction according to claim 3 is characterized in that: The loss function is determined based on the true label and the conflict probability predicted by the spatiotemporal graph convolutional network; the true label is obtained by annotating historical radar and airport topology data based on historical conflict event records; The expression of the loss function is: ； in, is the true label; is the predicted conflict probability; C is the total number of potential conflicts; N is the total number of training samples; is the potential conflict sequence number; is the loss function.

5. The dynamic scheduling method based on dynamic priority queue and spatiotemporal conflict prediction according to claim 1, characterized in that: The mixed integer programming model includes an objective function and constraints; The constraints include: minimum safety interval constraint, maintenance continuity constraint and runway capacity limitation constraint; The expression of the objective function is: ； in, is the sum of the squares of the delay times of all flights; For flights Delay time; is the total number of flights; is the sum of the probabilities of all potential conflicts; is the first weight coefficient; It is The probability of potential conflicts; C is the total number of potential conflicts; To maximize the utilization of parking spaces; is the second weight coefficient; is the utilization rate of parking spaces.

6. The dynamic scheduling method based on dynamic priority queue and spatiotemporal conflict prediction according to claim 3, characterized in that: The expression of the dynamic edge weight is: ； in, For flights i and flights j The edge weights between them; For flights i The velocity vector of For flights j The velocity vector of For flights i and flights j The norm of the difference between velocity vectors; is the time interval; is the safety distance threshold.

Images