Air traffic autonomous operation track generation method and device based on generative AI

Abstract

According to the air traffic autonomous operation track generation method and device based on the generative AI, the operation track in an air traffic autonomous operation scene can be rapidly generated, and sufficient data resources are provided for research and evaluation of air traffic autonomous operation. The method comprises the steps of (1) defining typical autonomous operation scenes, defining six autonomous operation scenes according to autonomous operation demand analysis, and describing and defining features of each scene; (2) constructing a generative AI model, training according to historical data provided by a first-level data center, and storing trained model data; training is carried out according to historical data provided by the first-level data center, and trained model data are stored; and (3) generating track data of each operation scene by using a generative AI model.

Description

Technical Field

[0001] The present invention relates to the technical field of civil aviation general aviation flight, and in particular to a method for generating an air traffic autonomous operation track based on generative AI, and also to a device for generating an air traffic autonomous operation track based on generative AI. Background Art

[0004] Autonomous air traffic operation means that in the future air traffic system, aircraft can autonomously perform navigation tasks without relying on centralized ground control systems. Autonomous air traffic operation achieves intelligent upgrading and transformation of the aviation industry by introducing and utilizing new technologies such as big data, cloud computing, the Internet of Things, artificial intelligence, and blockchain, focusing on promoting information sharing and data exchange to support a more collaborative and integrated aviation ecosystem.

[0005] However, the construction of integrated ground-to-air autonomous operation is still in the stage of conceptual research and key technology research. Therefore, there is a lack of actual operation data in various autonomous operation scenarios, resulting in a lack of effective evaluation methods for the effectiveness of integrated ground-to-air autonomous operation. It is difficult to effectively evaluate the efficiency improvement after achieving autonomous operation and the correctness of the direction of further research. Summary of the invention

[0006] In order to overcome the defects of the prior art, the technical problem to be solved by the present invention is to provide a method for generating air traffic autonomous operation trajectories based on generative AI, which can quickly generate operation trajectories in air traffic autonomous operation scenarios and provide sufficient data resources for the research and evaluation of air traffic autonomous operation.

[0007] The technical solution of the present invention is: the method for generating an autonomous air traffic trajectory based on generative AI comprises the following steps:

[0008] (1) Define typical autonomous operation scenarios. Based on the autonomous operation demand analysis, define six autonomous operation scenarios and describe and define the characteristics of each scenario.

[0009] (2) Build a generative AI model, train it based on the historical data provided by the primary data center, and save the trained model data; build a generative AI model based on the historical data provided by the primary data center, and save the trained model data;

[0010] (3) Generate track data for each operation scenario using a generative AI model;

[0011] The six autonomous operation scenarios in step (1) are described and defined as follows:

[0012] (1.1) Autonomous decision-making operation scenario for approach procedure. Under the concept of autonomous operation, the aircraft can realize the continuous descent operation after autonomous judgment of the continuous descent conditions, flight profile planning, and air-ground negotiation on the airborne side during the approach procedure.

[0013] (1.2) Straightening operation scenarios under flexible use of airspace: In autonomous operation scenarios, the crew can actively select temporary routes and perform straightening flights;

[0014] (1.3) Autonomous rerouting in bad weather: In the autonomous rerouting in bad weather scenario, the aircraft determines the best route to bypass the weather and avoid congested routes based on its business objectives and risk tolerance, and implements the route after automatic negotiation and confirmation with the ground automation system;

[0015] (1.4) Single aircraft TBO required arrival time operation scenario under safety conditions;

[0016] (1.5) In the aircraft emergency response scenario under the influence of severe convection on the route, in the single aircraft TBO required arrival time operation under safe conditions, the aircraft receives the key point control arrival time CTA demand of the ground digital control, and combines the surrounding situation service provided by the ground. On the premise of ensuring the safety of the preceding aircraft and surrounding route aircraft, the aircraft autonomously controls the acceleration and deceleration of the aircraft to realize the required arrival time RTA function of TBO and fly according to the predetermined track;

[0017] (1.6) Autonomous wake turbulence separation operation scenario in high-density scenarios. If the aircraft encounters severe convective weather and cannot avoid or detour, in the autonomous operation scenario, the aircraft makes independent decisions and decides whether to make an emergency descent based on the situation. Through air-ground information sharing and collaborative decision-making, the pilot perceives the operating status of the surrounding airspace, the flight intentions of other flights, the use of each altitude layer, traffic conditions, and potential conflicts, and the pilot independently adjusts and changes the plan.

[0018] The present invention defines typical autonomous operation scenarios, defines six autonomous operation scenarios based on autonomous operation demand analysis, and describes and defines the characteristics of each scenario; constructs a generative AI model, trains according to historical data provided by the primary data center, and saves the trained model data; trains according to historical data provided by the primary data center, and saves the trained model data; uses the generative AI model to generate track data for each operation scenario; therefore, it is possible to quickly generate operation tracks in air traffic autonomous operation scenarios, providing sufficient data resources for the research and evaluation of air traffic autonomous operation.

[0019] A device for generating an autonomous air traffic trajectory based on generative AI is also provided, which comprises:

[0020] Define scenario module, which is configured to define typical autonomous operation scenarios. According to the autonomous operation demand analysis, six autonomous operation scenarios are defined, and the characteristics of each scenario are described and defined;

[0021] Build a model module, which is configured to build a generative AI model, train it based on the historical data provided by the primary data center, and save the trained model data; train it based on the historical data provided by the primary data center, and save the trained model data;

[0022] A data generation module configured to generate track data for each operational scenario using a generative AI model;

[0023] The six autonomous operation scenarios of the scenario module are described and defined as follows:

[0024] (1.1) Autonomous decision-making operation scenario for approach procedure. Under the concept of autonomous operation, the aircraft can realize the continuous descent operation after autonomous judgment of the continuous descent conditions, flight profile planning, and air-ground negotiation on the airborne side during the approach procedure.

[0025] (1.2) Straightening operation scenarios under flexible use of airspace: In autonomous operation scenarios, the crew can actively select temporary routes and perform straightening flights;

[0026] (1.3) Autonomous rerouting in bad weather: In the autonomous rerouting in bad weather scenario, the aircraft determines the best route to bypass the weather and avoid congested routes based on its business objectives and risk tolerance, and implements the route after automatic negotiation and confirmation with the ground automation system;

[0027] (1.4) Single aircraft TBO required arrival time operation scenario under safety conditions;

[0028] (1.5) In the aircraft emergency response scenario under the influence of severe convection on the route, in the single aircraft TBO required arrival time operation under safe conditions, the aircraft receives the key point control arrival time CTA demand of the ground digital control, and combines the surrounding situation service provided by the ground. On the premise of ensuring the safety of the preceding aircraft and surrounding route aircraft, the aircraft autonomously controls the acceleration and deceleration of the aircraft to realize the required arrival time RTA function of TBO and fly according to the predetermined track;

[0029] (1.6) Autonomous wake turbulence separation operation scenario in high-density scenarios. If the aircraft encounters severe convective weather and cannot avoid or detour, in the autonomous operation scenario, the aircraft makes independent decisions and decides whether to make an emergency descent based on the situation. Through air-ground information sharing and collaborative decision-making, the pilot perceives the operating status of the surrounding airspace, the flight intentions of other flights, the use of each altitude layer, traffic conditions, and potential conflicts, and the pilot independently adjusts and changes the plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Shown is an overall flow chart of the method for generating air traffic autonomous operation trajectory based on generative AI according to the present invention.

[0031] Figure 2 Shown is a flowchart of the steps (2.1.2.1) of the method for generating air traffic autonomous operation trajectory based on generative AI according to the present invention.

[0032] Figure 3 Shown is a flowchart of the steps (2.1.2.2) of the method for generating air traffic autonomous operation trajectory based on generative AI according to the present invention.

[0033] Figure 4 Shown is a flow chart of step (3) of the method for generating air traffic autonomous operation trajectory based on generative AI according to the present invention.

[0034] Figure 5 Shown is a flowchart of steps (2.1.2.2.2.3.2) and (2.1.2.2.2.1.2) of the method for generating air traffic autonomous operation trajectory based on generative AI according to the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0036] In order to make the description of the present disclosure more detailed and complete, the following provides an illustrative description of the implementation and specific embodiments of the present invention; however, this is not the only form of implementing or using the specific embodiments of the present invention. The implementation covers the features of multiple specific embodiments and the method steps and sequences used to construct and operate these specific embodiments. However, other specific embodiments can also be used to achieve the same or equivalent functions and step sequences.

[0037] like Figure 1 As shown, this method for generating air traffic autonomous operation trajectory based on generative AI includes the following steps:

[0038] (1) Define typical autonomous operation scenarios. Based on the autonomous operation demand analysis, define six autonomous operation scenarios and describe and define the characteristics of each scenario.

[0039] (2) Build a generative AI model, train it based on the historical data provided by the primary data center, and save the trained model data; build a generative AI model based on the historical data provided by the primary data center, and save the trained model data;

[0040] (3) Generate track data for each operation scenario using a generative AI model;

[0041] The six autonomous operation scenarios in step (1) are described and defined as follows:

[0042] (1.1) Autonomous decision-making operation scenario for approach procedure. Under the concept of autonomous operation, the aircraft can realize the continuous descent operation after autonomous judgment of the continuous descent conditions, flight profile planning, and air-ground negotiation on the airborne side during the approach procedure.

[0043] (1.2) Straightening operation scenarios under flexible use of airspace: In autonomous operation scenarios, the crew can actively select temporary routes and perform straightening flights;

[0044] (1.3) Autonomous rerouting in bad weather: In the autonomous rerouting in bad weather scenario, the aircraft determines the best route to bypass the weather and avoid congested routes based on its business objectives and risk tolerance, and implements the route after automatic negotiation and confirmation with the ground automation system;

[0045] (1.4) Single aircraft TBO required arrival time operation scenario under safety conditions;

[0046] (1.5) In the aircraft emergency response scenario under the influence of severe convection on the route, in the single aircraft TBO required arrival time operation under safe conditions, the aircraft receives the key point control arrival time CTA demand of the ground digital control, and combines the surrounding situation service provided by the ground. On the premise of ensuring the safety of the preceding aircraft and surrounding route aircraft, the aircraft autonomously controls the acceleration and deceleration of the aircraft to realize the required arrival time RTA function of TBO and fly according to the predetermined track;

[0047] (1.6) Autonomous wake turbulence separation operation scenario in high-density scenarios. If the aircraft encounters severe convective weather and cannot avoid or detour, in the autonomous operation scenario, the aircraft makes independent decisions and decides whether to make an emergency descent based on the situation. Through air-ground information sharing and collaborative decision-making, the pilot perceives the operating status of the surrounding airspace, the flight intentions of other flights, the use of each altitude layer, traffic conditions, and potential conflicts, and the pilot independently adjusts and changes the plan.

[0048] The present invention defines typical autonomous operation scenarios, defines six autonomous operation scenarios based on autonomous operation demand analysis, and describes and defines the characteristics of each scenario; constructs a generative AI model, trains according to historical data provided by the primary data center, and saves the trained model data; trains according to historical data provided by the primary data center, and saves the trained model data; uses the generative AI model to generate track data for each operation scenario; therefore, it is possible to quickly generate operation tracks in air traffic autonomous operation scenarios, providing sufficient data resources for the research and evaluation of air traffic autonomous operation.

[0049] Preferably, the step (2) comprises the following sub-steps:

[0050] (2.1) Train the model based on historical data and update the iterative model parameters;

[0051] (2.2) Save the best model for subsequent use;

[0052] (2.3) Maintain the scalability of the model and iteratively upgrade the model as historical data becomes richer and more accumulated.

[0053] Preferably, the step (2.1) comprises the following sub-steps:

[0054] (2.1.1) Establishment of aircraft flight trajectory;

[0055] (2.1.2) Construction of flight trajectory diffusion model;

[0056] (2.1.3) Run scene feature extraction and use the VAE model to map the latent variable space into learnable low-dimensional variables;

[0057] (2.1.4) Conduct training based on historical data provided by the primary data center and save the trained model data;

[0058] Wherein, the step (2.1.1) comprises the following sub-steps:

[0059] (2.1.1.1) The historical data used in this patent are all ADS-B data. Before use, the original data of cat21 needs to be parsed into decimal data for subsequent use;

[0060] (2.1.1.2) The unique aircraft track is determined based on the date, 24-digit address code, and flight tail number;

[0061] (2.1.1.3) Use TRP time to sort ADS-B messages;

[0062] (2.1.1.4) Select the six track-related characteristic information including time, longitude, latitude, speed, altitude and north heading angle from the ADS-B message;

[0063] (2.1.1.5) Use filtering methods to remove abnormal tracks;

[0064] The step (2.1.2) comprises the following sub-steps:

[0065] (2.1.2.1) Define the forward process of the diffusion model and obtain the noise level at each step;

[0066] (2.1.2.2) Define the AirTraj-UNet network structure and learn the track distribution;

[0067] (2.1.2.3) Train the AirTraj-UNet network structure, output the noise level at each step, and optimize the parameters through back propagation and gradient descent techniques according to the difference between the predicted noise level and the actual noise level to adjust the model parameters;

[0068] The step (2.1.3) comprises the following sub-steps:

[0069] (2.1.3.1) Construct a VAE model, including an encoding module and a decoding module;

[0070] (2.1.3.2) Use historical data to train and optimize VAE model parameters;

[0071] (2.1.3.3) Save the best model parameters;

[0072] (2.1.3.4) The encoding part of the VAE model is used to encode the scene features. The encoded embedded information is learned through cross-attention layers and correlation with the track information in each attention module in the AirTraj-UNet structure.

[0073] Preferably, if Figure 2 As shown, the step (2.1.2.1) includes the following sub-steps:

[0074] (2.1.2.1.1) The original data x0 obtained from the training set;

[0075] (2.1.2.1.2) In a series of time steps t, noise is gradually added to the data.

[0076] At each time step t, Gaussian noise is added by the following formula:

[0077]

[0078] Among them, αt is the predetermined decay coefficient at time step t, z t is a random number sampled from a standard normal distribution.

[0079] Machine noise vector;

[0080] (2.1.2.1.3) Repeat step (2.1.2.1.2) until the final time step T is reached. At this time, the data x t Close to pure noise.

[0081] Preferably, if Figure 3 As shown, the step (2.1.2.2) includes the following sub-steps:

[0082] (2.1.2.2.1) The inverse process starts from a data x that is approximately pure noise t We start by reducing the effect of noise in a series of time steps t, using a parameterized model θ to predict the earlier state x at each time step. t-1 The noise component in the noise component or directly predict the noise-free state, each

[0083] One step involves computing a denoised version This is achieved through the following formula:

[0084]

[0085] in, is the estimate of the noise at time step t, α t is the noise adjustment parameter;

[0086] (2.1.2.2.2) Construct the AirTraj-UNet network structure and learn the model parameters θ;

[0087] (2.1.2.2.3) After the last step t=1 is completed, output This is an estimate of the original data used as the resulting track.

[0088] Preferably, the step (2.1.2.2.2) comprises the following sub-steps:

[0089] (2.1.2.2.2.1) Construct a downsampling module. The length of the track is reduced to 1 / 2 of the original length each time it passes through the downsampling module, so that the model can learn the distribution of tracks at multiple scales.

[0090] (2.1.2.2.2.2) Construct a correlation learning module that includes a self-correlation attention mechanism and a cross-attention mechanism;

[0091] (2.1.2.2.2.3) Construct an upsampling module. The length of the track is doubled each time it passes through the upsampling module, so that the track is finally restored to its original length.

[0092] Preferably, the step (2.1.2.2.2.3) comprises the following sub-steps:

[0093] (2.1.2.2.2.3.1) Construct a temporal coding module including a sinusoidal mapping layer, a linear mapping layer, a nonlinear activation layer, and finally a linear mapping layer connected once;

[0094] (2.1.2.2.2.3.2) Construct a layer that includes batch normalization, nonlinear activation layer, and expansion volume

[0095] The residual connection network module of the accumulation layer (such as Figure 5 shown);

[0096] (2.1.2.2.2.3.3) Construct a correlation learning module that includes a self-correlation attention mechanism and a cross-attention mechanism. Use the self-correlation attention mechanism to learn the correlation between the track data itself, and use the cross-attention mechanism to learn the correlation between the track distribution and the scene features.

[0097] (2.1.2.2.2.3.4) The result of time encoding is added to the residual connection network module after linear mapping, and then superimposed multiple times in the order of the residual connection module and the correlation learning module.

[0098] Preferably, the step (2.1.2.2.2.1) comprises the following sub-steps:

[0099] (2.1.2.2.2.1.1) Construct a temporal coding module including a sinusoidal mapping layer, a linear mapping layer, a nonlinear activation layer, and finally a linear mapping layer connected once;

[0100] (2.1.2.2.2.1.2) Construct a layer that includes batch normalization, nonlinear activation layer, and expansion volume

[0101] The residual connection network module of the accumulation layer (such as Figure 5 shown);

[0102] (2.1.2.2.2.1.3) Construct a correlation learning module that includes a self-correlation attention mechanism and a cross-attention mechanism. Use the self-correlation attention mechanism to learn the correlation between the track data itself, and use the cross-attention mechanism to learn the correlation between track distribution and scene features;

[0103] (2.1.2.2.2.1.4) The result of time encoding is added to the residual connection network module after linear mapping, and then superimposed multiple times in the order of residual connection module and correlation learning module.

[0104] Preferably, if Figure 4 As shown, the step (3) includes the following sub-steps:

[0105] (3.1) Use VAE encoding to extract running scene features;

[0106] (3.2) Match the model with the scene;

[0107] (3.3) Generate autonomous air traffic trajectory based on the matching results.

[0108] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes the steps of the above-mentioned embodiment method, and the storage medium can be: ROM / RAM, disk, CD, memory card, etc. Therefore, corresponding to the method of the present invention, the present invention also includes a device for generating an autonomous air traffic trajectory based on generative AI, which is usually represented in the form of functional modules corresponding to the steps of the method. The device includes:

[0109] Define scenario module, which is configured to define typical autonomous operation scenarios. According to the autonomous operation demand analysis, six autonomous operation scenarios are defined, and the characteristics of each scenario are described and defined;

[0110] Build a model module, which is configured to build a generative AI model, train it based on the historical data provided by the primary data center, and save the trained model data; train it based on the historical data provided by the primary data center, and save the trained model data;

[0111] A data generation module configured to generate track data for each operational scenario using a generative AI model;

[0112] The six autonomous operation scenarios of the scenario module are described and defined as follows:

[0113] (1.1) Autonomous decision-making operation scenario for approach procedure. Under the concept of autonomous operation, the aircraft can realize the continuous descent operation after autonomous judgment of the continuous descent conditions, flight profile planning, and air-ground negotiation on the airborne side during the approach procedure.

[0114] (1.2) Straightening operation scenarios under flexible use of airspace: In autonomous operation scenarios, the crew can actively select temporary routes and perform straightening flights;

[0115] (1.3) Autonomous rerouting in bad weather: In the autonomous rerouting in bad weather scenario, the aircraft determines the best route to bypass the weather and avoid congested routes based on its business objectives and risk tolerance, and implements the route after automatic negotiation and confirmation with the ground automation system;

[0116] (1.4) Single aircraft TBO required arrival time operation scenario under safety conditions;

[0117] (1.5) In the aircraft emergency response scenario under the influence of severe convection on the route, in the single aircraft TBO required arrival time operation under safe conditions, the aircraft receives the key point control arrival time CTA demand of the ground digital control, and combines the surrounding situation service provided by the ground. On the premise of ensuring the safety of the preceding aircraft and surrounding route aircraft, the aircraft autonomously controls the acceleration and deceleration of the aircraft to realize the required arrival time RTA function of TBO and fly according to the predetermined track;

[0118] (1.6) Autonomous wake turbulence separation operation scenario in high-density scenarios. If the aircraft encounters severe convective weather and cannot avoid or detour, in the autonomous operation scenario, the aircraft makes independent decisions and decides whether to make an emergency descent based on the situation. Through air-ground information sharing and collaborative decision-making, the pilot perceives the operating status of the surrounding airspace, the flight intentions of other flights, the use of each altitude layer, traffic conditions, and potential conflicts, and the pilot independently adjusts and changes the plan.

[0119] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention are still within the protection scope of the technical solution of the present invention.

Claims

1. A method for generating autonomous air traffic trajectory based on generative AI, characterized by: It includes the following steps: (1) Define typical autonomous operation scenarios. Based on the autonomous operation demand analysis, define six autonomous operation scenarios and describe and define the characteristics of each scenario. (2) Build a generative AI model, train it based on the historical data provided by the primary data center, and save the trained model data; build a generative AI model based on the historical data provided by the primary data center, and save the trained model data; (3) Generate track data for each operation scenario using a generative AI model; The six autonomous operation scenarios in step (1) are described and defined as follows: (1.1) Autonomous decision-making operation scenario for approach procedure. Under the concept of autonomous operation, the aircraft can realize the continuous descent operation after autonomous judgment of the continuous descent conditions, flight profile planning, and air-ground negotiation on the airborne side during the approach procedure. (1.2) Straightening operation scenarios under flexible use of airspace: In autonomous operation scenarios, the crew can actively select temporary routes and perform straightening flights; (1.3) Autonomous rerouting in bad weather: In the autonomous rerouting in bad weather scenario, the aircraft determines the best route to bypass the weather and avoid congested routes based on its business objectives and risk tolerance, and implements the route after automatic negotiation and confirmation with the ground automation system; (1.4) Single aircraft TBO required arrival time operation scenario under safety conditions; (1.5) In the aircraft emergency response scenario under the influence of severe convection on the route, in the single aircraft TBO required arrival time operation under safe conditions, the aircraft receives the key point control arrival time CTA demand of the ground digital control, and combines the surrounding situation service provided by the ground. On the premise of ensuring the safety of the preceding aircraft and surrounding route aircraft, the aircraft autonomously controls the acceleration and deceleration of the aircraft to realize the required arrival time RTA function of TBO and fly according to the predetermined track; (1.6) Autonomous wake turbulence separation operation scenario in high-density scenarios. If the aircraft encounters severe convective weather and cannot avoid or detour, in the autonomous operation scenario, the aircraft makes independent decisions and decides whether to make an emergency descent based on the situation. Through air-ground information sharing and collaborative decision-making, the pilot perceives the operating status of the surrounding airspace, the flight intentions of other flights, the use of each altitude layer, traffic conditions, and potential conflicts, and the pilot independently adjusts and changes the plan.

2. The method for generating air traffic autonomous operation trajectory based on generative AI according to claim 1, characterized in that: The step (2) comprises the following sub-steps: (2.1) Train the model based on historical data and update the iterative model parameters; (2.2) Save the best model for subsequent use; (2.3) Maintain the scalability of the model and iteratively upgrade the model as historical data becomes richer and more accumulated.

3. The method for generating air traffic autonomous operation trajectory based on generative AI according to claim 2, characterized in that: The step (2.1) comprises the following sub-steps: (2.1.1) Establishment of aircraft flight trajectory; (2.1.2) Construction of flight trajectory diffusion model; (2.1.3) Run scene feature extraction and use the VAE model to map the latent variable space into learnable low-dimensional variables; (2.1.4) Conduct training based on historical data provided by the primary data center and save the trained model data; Wherein, the step (2.1.1) comprises the following sub-steps: (2.1.1.1) The historical data used in this patent are all ADS-B data. Before use, the original data of cat21 needs to be parsed into decimal data for subsequent use; (2.1.1.2) The unique aircraft track is determined based on the date, 24-digit address code, and flight tail number; (2.1.1.3) Use TRP time to sort ADS-B messages; (2.1.1.4) Select the six track-related characteristic information including time, longitude, latitude, speed, altitude and north heading angle from the ADS-B message; (2.1.1.5) Use filtering methods to remove abnormal tracks; The step (2.1.2) comprises the following sub-steps: (2.1.2.1) Define the forward process of the diffusion model and obtain the noise level at each step; (2.1.2.2) Define the AirTraj-UNet network structure and learn the track distribution; (2.1.2.3) Train the AirTraj-UNet network structure, output the noise level at each step, and optimize the parameters through back propagation and gradient descent techniques according to the difference between the predicted noise level and the actual noise level to adjust the model parameters; The step (2.1.3) comprises the following sub-steps: (2.1.3.1) Construct a VAE model, including an encoding module and a decoding module; (2.1.3.2) Use historical data to train and optimize VAE model parameters; (2.1.3.3) Save the best model parameters; (2.1.3.4) The encoding part of the VAE model is used to encode the scene features. The encoded embedded information is learned through cross-attention layers and correlation with the track information in each attention module in the AirTraj-UNet structure.

4. The method for generating air traffic autonomous operation trajectory based on generative AI according to claim 3 is characterized in that: The step (2.1.2.1) comprises the following sub-steps: (2.1.2.1.1) The original data x0 obtained from the training set; (2.1.2.1.2) Noise is gradually added to the data in a series of time steps t. For each time step t, Gaussian noise is added by the following formula: Among them, α t is the predetermined decay coefficient at time step t, z t is a random noise vector sampled from a standard normal distribution; (2.1.2.1.3) Repeat step (2.1.2.1.2) until the final time step T is reached. At this time, the data x t Close to pure noise.

5. The method for generating air traffic autonomous operation trajectory based on generative AI according to claim 4, characterized in that: The step (2.1.2.2) comprises the following sub-steps: (2.1.2.2.1) The inverse process starts from a data x that is approximately pure noise t We start by gradually reducing the effect of noise in a series of time steps t, using a parameterized model θ at each time step to predict the earlier state x t-1 The noise component in the image or directly predict the noise-free state, each step involves calculating the denoised version This is achieved through the following formula: in, is the estimate of the noise at time step t, α t is the noise adjustment parameter; (2.1.2.2.2) Construct the AirTraj-UNet network structure and learn the model parameters θ; (2.1.2.2.3) After the last step t=1 is completed, output This is an estimate of the original data used as the resulting track.

6. The method for generating air traffic autonomous operation trajectory based on generative AI according to claim 5, characterized in that: The step (2.1.2.2.2) comprises the following sub-steps: (2.1.2.2.2.1) Construct a downsampling module. The length of the track is reduced to 1 / 2 of the original length each time it passes through the downsampling module, so that the model can learn the distribution of tracks at multiple scales. (2.1.2.2.2.2) Construct a correlation learning module that includes a self-correlation attention mechanism and a cross-attention mechanism; (2.1.2.2.2.3) Construct an upsampling module. The length of the track is doubled each time it passes through the upsampling module, so that the track is finally restored to its original length.

7. The method for generating air traffic autonomous operation trajectory based on generative AI according to claim 6, characterized in that: The step (2.1.2.2.2.3) comprises the following sub-steps: (2.1.2.2.2.3.1) Construct a temporal coding module including a sinusoidal mapping layer, a linear mapping layer, a nonlinear activation layer, and finally a linear mapping layer connected once; (2.1.2.2.2.3.2) Construct a residual connection network module that includes batch normalization, non-linear activation layer, and dilated convolution layer; (2.1.2.2.2.3.3) Construct a correlation learning module that includes a self-correlation attention mechanism and a cross-attention mechanism. Use the self-correlation attention mechanism to learn the correlation between the track data itself, and use the cross-attention mechanism to learn the correlation between the track distribution and the scene features. (2.1.2.2.2.3.4) The result of time encoding is added to the residual connection network module after linear mapping, and then superimposed multiple times in the order of the residual connection module and the correlation learning module.

8. The method for generating air traffic autonomous operation trajectory based on generative AI according to claim 5, characterized in that: The step (2.1.2.2.2.1) comprises the following sub-steps: (2.1.2.2.2.1.1) Construct a temporal coding module including a sinusoidal mapping layer, a linear mapping layer, a nonlinear activation layer, and finally a linear mapping layer connected once; (2.1.2.2.2.1.2) Construct a residual connection network module that includes batch normalization, non-linear activation layer, and dilated convolution layer; (2.1.2.2.2.1.3) Construct a correlation learning module that includes a self-correlation attention mechanism and a cross-attention mechanism. Use the self-correlation attention mechanism to learn the correlation between the track data itself, and use the cross-attention mechanism to learn the correlation between track distribution and scene features; (2.1.2.2.2.1.4) The result of time encoding is added to the residual connection network module after linear mapping, and then superimposed multiple times in the order of residual connection module and correlation learning module.

9. The method for generating air traffic autonomous operation trajectory based on generative AI according to claim 5, characterized in that: The step (3) comprises the following sub-steps: (3.1) Use VAE encoding to extract running scene features; (3.2) Match the model with the scene; (3.3) Generate autonomous air traffic trajectory based on the matching results.

10. An air traffic autonomous operation track generation device based on generative AI, characterized by: It includes: Define scenario module, which is configured to define typical autonomous operation scenarios. According to the autonomous operation demand analysis, six autonomous operation scenarios are defined, and the characteristics of each scenario are described and defined; Build a model module, which is configured to build a generative AI model, train it based on the historical data provided by the primary data center, and save the trained model data; train it based on the historical data provided by the primary data center, and save the trained model data; A data generation module configured to generate track data for each operational scenario using a generative AI model; The six autonomous operation scenarios of the scenario module are described and defined as follows: (1.1) Autonomous decision-making operation scenario for approach procedure. Under the concept of autonomous operation, the aircraft can realize the continuous descent operation after autonomous judgment of the continuous descent conditions, flight profile planning, and air-ground negotiation on the airborne side during the approach procedure. (1.2) Straightening operation scenarios under flexible use of airspace: In autonomous operation scenarios, the crew can actively select temporary routes and perform straightening flights; (1.3) Autonomous rerouting in bad weather: In the autonomous rerouting in bad weather scenario, the aircraft determines the best route to bypass the weather and avoid congested routes based on its business objectives and risk tolerance, and implements the route after automatic negotiation and confirmation with the ground automation system; (1.4) Single aircraft TBO required arrival time operation scenario under safety conditions; (1.5) In the aircraft emergency response scenario under the influence of severe convection on the route, in the single aircraft TBO required arrival time operation under safe conditions, the aircraft receives the key point control arrival time CTA demand of the ground digital control, and combines the surrounding situation service provided by the ground. On the premise of ensuring the safety of the preceding aircraft and surrounding route aircraft, the aircraft autonomously controls the acceleration and deceleration of the aircraft to realize the required arrival time RTA function of TBO and fly according to the predetermined track; (1.6) Autonomous wake turbulence separation operation scenario in high-density scenarios. If the aircraft encounters severe convective weather and cannot avoid or detour, in the autonomous operation scenario, the aircraft makes independent decisions and decides whether to make an emergency descent based on the situation. Through air-ground information sharing and collaborative decision-making, the pilot perceives the operating status of the surrounding airspace, the flight intentions of other flights, the use of each altitude layer, traffic conditions, and potential conflicts, and the pilot independently adjusts and changes the plan.

Images