Aircraft trajectory prediction method based on neural network

Abstract

The invention provides an aircraft trajectory prediction method based on a neural network, and the method comprises the steps: obtaining the flight data, including flight state data and trajectory influence data, of an aircraft at the current moment in real time; a trajectory prediction model is constructed based on a neural network, and trajectory prediction is performed by using flight data at the current moment to obtain a predicted trajectory at the next moment; performing flight control based on the predicted trajectory; the sample flight data is selected from historical flight data at historical moments based on a flight data threshold. According to the method provided by the invention, through the trajectory prediction model constructed based on the neural network, the predicted trajectory of the next moment is obtained by applying the flight data of the aircraft at the current moment obtained in real time, and flight control is performed on the aircraft based on the predicted trajectory, so that accurate, rapid, flexible and highly practical trajectory prediction is realized; and the requirements of variable flight environments and real-time trajectory prediction are met.

Description

Technical Field

[0001] The present invention relates to the technical field of aircraft control, and particularly to an aircraft trajectory prediction method based on a neural network. Background Art

[0002] Aircraft trajectory prediction is an important topic in the aviation field, which is crucial for aircraft navigation, control, and avoiding mid-air collisions. With the continuous progress of aviation technology, the flight missions of aircraft are becoming more and more complex, and the requirements for the accuracy and real-time performance of trajectory prediction are also getting higher and higher. Traditional aircraft trajectory prediction methods mainly rely on physical models, such as numerical simulation methods based on dynamic equations.

[0003] However, an aircraft is affected by various uncertain factors during actual flight, such as changes in wind speed and direction, and uncertainties in the aircraft's own state. These all make it very difficult to establish an accurate physical model. Moreover, the calculation of physical models is usually relatively complex and requires a large amount of computing resources and time. Therefore, relying on physical models for aircraft trajectory prediction has deficiencies in terms of accuracy and practicality. Summary of the Invention

[0004] The present invention provides an aircraft trajectory prediction method based on a neural network to solve the deficiencies in the prior art that rely on physical models for aircraft trajectory prediction, which have deficiencies in terms of accuracy and practicality.

[0005] The present invention provides an aircraft trajectory prediction method based on a neural network, including: Obtaining in real time the flight data of the aircraft at the current moment, where the flight data includes flight state data and trajectory influence data; Based on the trajectory prediction model, applying the flight data at the current moment for trajectory prediction to obtain the predicted trajectory at the next moment; Based on the predicted trajectory, performing flight control on the aircraft; The trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network; The sample flight data is selected from the historical flight data at historical moments based on a flight data threshold.

[0006] According to the aircraft trajectory prediction method based on a neural network provided by the present invention, the data type of the flight data threshold corresponds one-to-one with the data type of the historical flight data; The construction steps of the sample flight data include: Obtaining the historical flight data at the historical moment; Compare the historical flight data of each data type with the flight data threshold of the corresponding data type respectively; Select the historical flight data whose comparison result exceeds the flight data threshold as the sample flight data.

[0007] According to a flight trajectory prediction method based on a neural network provided by the present invention, the step of selecting the historical flight data whose comparison result exceeds the flight data threshold as the sample flight data includes: Select the historical flight data whose comparison result exceeds the flight data threshold as the initial sample flight data; Add Gaussian noise to the initial sample flight data to obtain the sample flight data.

[0008] According to a flight trajectory prediction method based on a neural network provided by the present invention, the training steps of the trajectory prediction model include: Perform trajectory prediction based on the sample flight data and the initial trajectory prediction model to obtain a sample prediction trajectory; Calculate a prediction loss based on the sample prediction trajectory and the sample label; Based on the prediction loss and the gradient descent algorithm, perform parameter iteration on the initial trajectory prediction model to obtain the trajectory prediction model; The time to which the sample label belongs is later than the time to which the sample flight data belongs.

[0009] According to a flight trajectory prediction method based on a neural network provided by the present invention, the calculating the prediction loss based on the sample prediction trajectory and the sample label includes: Construct a prediction loss function based on a loss function and a regularization term; Based on the prediction loss function, perform calculations using the sample prediction trajectory and the sample label to obtain the prediction loss; The regularization term includes L1 norm regularization and L2 norm regularization.

[0010] According to a flight trajectory prediction method based on a neural network provided by the present invention, the initial trajectory prediction model includes an input network layer, a hidden network layer, and an output network layer. The hidden network layer includes a ReLU activation function, and the output network layer includes a linear activation function; Each network layer includes a plurality of neurons; The performing parameter iteration on the initial trajectory prediction model based on the prediction loss and the gradient descent algorithm to obtain the trajectory prediction model includes: Based on the predicted loss and the gradient descent algorithm, randomly select a target neuron for discarding based on a preset probability, and perform parameter iteration on the initial trajectory prediction model until the number of iterations meets the preset number of iterations to obtain the trajectory prediction model.

[0011] The present invention also provides a flight vehicle trajectory prediction device based on a neural network, including: An acquisition unit, which acquires the flight data of the flight vehicle at the current moment in real time, and the flight data includes flight state data and trajectory influence data; A trajectory prediction unit, which performs trajectory prediction based on the trajectory prediction model and applies the flight data at the current moment to obtain a predicted trajectory for the next moment; A flight control unit, which performs flight control on the flight vehicle based on the predicted trajectory; The trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network; The sample flight data is selected from the historical flight data at historical moments based on a flight data threshold.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the neural network-based flight vehicle trajectory prediction method as described in any one of the above.

[0013] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the neural network-based flight vehicle trajectory prediction method as described in any one of the above.

[0014] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the neural network-based flight vehicle trajectory prediction method as described in any one of the above.

[0015] A neural network-based flight vehicle trajectory prediction method provided by the present invention, through a trajectory prediction model constructed based on a neural network, applies the flight data of the flight vehicle acquired in real time at the current moment to obtain a predicted trajectory for the next moment, and performs flight control on the flight vehicle based on the predicted trajectory, realizing accurate, fast, flexible, and highly practical trajectory prediction, and meeting the requirements of a changing flight environment and real-time trajectory prediction. Description of the Drawings

[0016] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 is a schematic flowchart of a flight vehicle trajectory prediction method based on a neural network provided by the present invention; Figure 2 is a schematic structural diagram of a flight vehicle trajectory prediction device provided by the present invention; Figure 3 is a schematic structural diagram of an electronic device provided by the present invention. Detailed Embodiments

[0018] To make the objectives, technical solutions and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present invention fall within the scope of protection of the present invention.

[0019] It should be noted that with the rapid development of fields such as unmanned aerial vehicles, commercial aviation, and space exploration, higher requirements are put forward for aircraft trajectory prediction technology. These requirements not only include high-precision prediction under known flight conditions, but also involve adaptability and robustness in unknown or changing environments. The existing technologies mainly face the following problems: 1. Strong model dependence: Traditional aircraft trajectory prediction methods highly rely on the accurate modeling of aircraft dynamics, which is often difficult to achieve in practical applications. The dynamic model of an aircraft is complex and affected by various uncertain factors, such as environmental changes and aircraft performance degradation, resulting in difficulty in ensuring the accuracy of the model. 2. High computational cost: Prediction methods based on physical models usually require a large amount of computational resources, which becomes a bottleneck in flight control scenarios that require real-time feedback. Insufficient real-time performance may lead to delays in flight control decisions and increase flight safety risks. 3. Limited generalization ability: Existing methods perform well under specific conditions, but their prediction performance often drops sharply when facing unknown or changing environments. This limits the autonomous flight ability of aircraft in complex and changing environments. 4. Insufficient data utilization: Traditional trajectory prediction methods fail to fully utilize existing flight data. Through data-driven methods, potential patterns in the data can be mined to improve the accuracy and generalization ability of prediction. 5. Poor adaptability: When facing changes in aircraft performance or flight mission adjustments, existing methods are difficult to quickly adapt to new requirements. This requires trajectory prediction methods to have better flexibility and adaptability.

[0020] To address the above problems, the present invention provides a neural network-based aircraft trajectory prediction method to achieve accurate and highly practical trajectory prediction. Figure 1 It is a schematic flowchart of the neural network-based aircraft trajectory prediction method provided by the present invention, as Figure 1 shown. This method includes: Step 110, obtaining in real time the flight data of the aircraft at the current moment, where the flight data includes flight state data and trajectory influence data; Here, the flight state data refers to the motion state data of the aircraft, such as position, speed, acceleration and other data. The trajectory influence data here refers to the data that affects the flight trajectory of the aircraft, which can include internal factor data and external factor data. Among them, the internal factor data such as the control instructions of the aircraft. The control instructions here refer to the instructions used to control the flight trajectory of the aircraft, such as acceleration. In addition, the external factor data such as environmental factors like wind force, temperature, and air pressure.

[0021] Specifically, the flight data at the current moment can be obtained by acquiring the monitoring data collected in real time by the sensors pre-set on the aircraft. For example, through sensors such as GPS, gyroscopes, and accelerometers, the flight state data such as position, speed, acceleration, and attitude angle can be collected in real time. In addition, the trajectory influence data can be obtained by acquiring the control instructions received by the aircraft and the external factor data that may affect the aircraft trajectory through a weather station, radar, or other external data sources, such as wind speed, wind direction, air pressure, temperature, and other data.

[0022] Step 120, based on the trajectory prediction model, apply the flight data at the current moment for trajectory prediction to obtain the predicted trajectory at the next moment; The trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network; the sample flight data is selected from the historical flight data at historical moments based on a flight data threshold.

[0023] Here, the trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network. It should be noted that considering that the aircraft will be affected by various uncertain factors during actual flight, such as changes in wind speed and wind direction, and the uncertainty of the aircraft's own state.

[0024] Therefore, compared with the prior art that establishes an accurate physical model, the trajectory prediction model constructed based on a neural network provided by the embodiments of the present invention for identifying the flight trajectory does not require a complex modeling process, reduces the dependence on the aircraft dynamics model, and can learn the internal laws and characteristics in the sample flight data, thereby having the generalization ability for unknown data.

[0025] In addition, the trajectory prediction model constructed based on a neural network automatically adapts to different flight conditions and external environment changes to achieve accurate trajectory prediction. Moreover, the neural network model is more efficient in calculation, reduces the dependence on high-performance computing resources, lowers the calculation cost, is more convenient for real-time trajectory prediction of the aircraft, is applicable to flight control scenarios for rapid decision-making, and improves the practicality of the trajectory prediction method.

[0026] Specifically, in the sample data acquisition stage of the trajectory prediction model, first, the historical flight data at historical moments can be used as sample data, that is, the sample flight data is obtained. And, the historical flight data at any historical moment can be used as the trajectory label of the sample flight data at the previous moment. Thus, avoiding the annotation of the actual trajectory of the sample flight data greatly improves the training efficiency.

[0027] It should be noted that in order to ensure the prediction effect of the trajectory prediction model, representative historical flight data can be selected from the historical flight data at historical moments as sample flight data. For example, historical flight data with strong wind or large changes in control commands can be selected to improve the training performance of the trajectory prediction model.

[0028] In the training stage of the trajectory prediction model, the sample flight data can be used to iteratively train the initial trajectory prediction model until the number of training times reaches the preset number to obtain the trajectory prediction model. It should be noted that by completing model training in a data-driven manner and constructing the trajectory prediction model, the trajectory prediction model can learn the internal relationships between a large amount of flight data, thereby achieving accurate trajectory prediction.

[0029] In addition, in the actual application stage of the trajectory prediction model, by inputting the flight data at the current moment obtained in real time into the trajectory prediction model and using the mapping relationship learned inside the neural network, the predicted trajectory for the next moment can be quickly calculated.

[0030] Step 130, perform flight control on the aircraft based on the predicted trajectory; Specifically, the control strategy for the aircraft can be adjusted according to the deviation between the predicted trajectory and the desired trajectory to achieve flight control of the aircraft. For example, the acceleration of the aircraft can be adjusted.

[0031] The method provided by the embodiments of the present invention, through the trajectory prediction model constructed based on a neural network, applies the flight data of the aircraft at the current moment obtained in real time to obtain the predicted trajectory for the next moment, and performs flight control on the aircraft based on the predicted trajectory, realizing accurate, fast, flexible and highly practical trajectory prediction, and meeting the requirements of a changing flight environment and real-time trajectory prediction.

[0032] Based on any of the above embodiments, the data type of the flight data threshold corresponds one-to-one with the data type of the historical flight data; The construction steps of the sample flight data include: Obtain the historical flight data at the historical moment; Compare the historical flight data of each data type with the flight data threshold of the corresponding data type; Select the historical flight data with the comparison result exceeding the flight data threshold as the sample flight data.

[0033] Here, the flight data threshold refers to the reference data of flight data for each data type, which can be used to reflect various flight data of the aircraft under normal flight conditions. The data types here include data types such as position, speed, acceleration, attitude angle, wind speed, wind direction, air pressure, temperature, control commands, etc. Thus, the data types included in the flight data threshold correspond one-to-one with the data types of historical flight data, and also include the data thresholds of data such as position, speed, acceleration, attitude angle, wind speed, wind direction, air pressure, temperature, control commands, etc. For example, the wind speed threshold can be 5.5 m / s.

[0034] Specifically, first, historical flight data at a historical moment can be obtained, and the historical flight data can be cleaned to remove outliers and error records. Then, the historical flight data of each data type is respectively compared with the flight data threshold of the corresponding data type. For example, the historical flight data of wind speed is compared with the wind speed flight data threshold, and the historical flight data of speed is compared with the speed flight data threshold. Here, they are not listed one by one. Finally, the historical flight data with the comparison result exceeding the flight data threshold is selected as the sample flight data. For example, when the historical flight data of wind speed is 8 m / s, which exceeds the wind speed flight data threshold, then the historical flight data at this historical moment can be used as the sample flight data.

[0035] It should be noted that selecting the historical flight data with the comparison result exceeding the flight data threshold as the sample flight data can select the sample flight data that has a greater impact on the trajectory, enabling the initial trajectory prediction model to more fully learn the relationship between the sample flight data and the trajectory, improving the accuracy of trajectory prediction, and still being able to achieve accurate trajectory prediction in the face of uncontrollable influencing factors, enhancing the robustness and generalization of the trajectory prediction model.

[0036] Based on any of the above embodiments, the selection of the historical flight data with the comparison result exceeding the flight data threshold as the sample flight data includes: Selecting the historical flight data with the comparison result exceeding the flight data threshold as the initial sample flight data; Adding Gaussian noise to the initial sample flight data to obtain the sample flight data.

[0037] Specifically, historical flight data with a comparison result exceeding the flight data threshold is selected as the initial sample flight data. Then, the initial sample flight data can be normalized to scale all data values to between [-1, 1] or [0, 1]. It should be noted that the initial sample flight data of different data types will have different units or dimensions, that is, the initial sample flight data of different data types have different value ranges in the original values. To avoid larger values masking the importance of relatively important but smaller flight data, the initial sample flight data can be normalized, for example, the maximum-minimum normalization method can be used.

[0038] Furthermore, to enhance the prediction ability of the trajectory prediction model for unknown data and the robustness of the model, Gaussian noise can be added to the normalized initial sample flight data to obtain the final sample flight data.

[0039] Based on any of the above embodiments, the training steps of the trajectory prediction model include: Based on the sample flight data and the initial trajectory prediction model, trajectory prediction is performed to obtain a sample prediction trajectory; Based on the sample prediction trajectory and the sample label, the prediction loss is calculated; Based on the prediction loss and the gradient descent algorithm, parameter iteration is performed on the initial trajectory prediction model to obtain the trajectory prediction model; The belonging time of the sample label is later than the belonging time of the sample flight data.

[0040] Specifically, in the model training stage, first, the sample flight data can be input into the initial trajectory prediction model, and trajectory prediction is performed through the initial trajectory prediction model to output to the sample prediction trajectory. Then, the prediction loss can be calculated by applying a loss function through the sample prediction trajectory and the sample label corresponding to the sample flight data. The loss function here can be the mean square error function or the cross-entropy loss function.

[0041] It should be noted that the belonging time of the sample label here is later than the belonging time of the sample flight data. For example, the sample label corresponding to the sample flight data can be the flight data at the next moment of the belonging time of the sample flight data.

[0042] After calculating the prediction loss of the current iteration round, the initial trajectory prediction model can be iteratively parameterized using the prediction loss. Meanwhile, the gradient descent method can be used as the optimization algorithm, and a suitable optimizer such as Adam or RMSprop can be selected to accelerate the training speed and improve the training efficiency. The gradient descent algorithm here includes but is not limited to stochastic gradient descent, batch gradient descent, or mini-batch gradient descent. Additionally, hyperparameters such as the learning rate and the number of iterations can be set to control the stability and convergence speed of the training process, preventing overfitting. An early stopping strategy can also be used to prevent overfitting. Finally, the final trajectory prediction model is obtained.

[0043] It should be noted that the learning rate is dynamically adjusted according to the error change during the training process to optimize the training process and accelerate the convergence speed. The learning rate decay strategy is adopted to gradually reduce the learning rate as the training progresses to improve the fine-tuning ability of the model.

[0044] The method provided by the embodiments of the present invention designs an adaptive learning algorithm to enable the trajectory prediction model to be quickly updated according to changes in the performance of the aircraft or adjustments to the flight mission. An online learning or incremental learning mechanism is introduced to enable the model to continuously learn from new data to adapt to the changing flight environment.

[0045] Based on any of the above embodiments, calculating the prediction loss based on the sample predicted trajectory and the sample label includes: Constructing a prediction loss function based on the loss function and the regularization term; Calculating the prediction loss based on the prediction loss function by applying the sample predicted trajectory and the sample label; The regularization term includes L1 norm regularization and L2 norm regularization.

[0046] Specifically, the prediction loss function can be constructed by adding a regularization term to the loss function. The regularization term here includes L1 norm regularization and L2 norm regularization. It should be noted that by adding a regularization term to the loss function, the value range of the weights is restricted, the risk of overfitting is reduced, and the prediction performance of the model is ensured.

[0047] Based on any of the above embodiments, the initial trajectory prediction model includes an input network layer, a hidden network layer, and an output network layer. The hidden network layer includes a ReLU activation function, and the output network layer includes a linear activation function; Each network layer includes multiple neurons; Performing parameter iteration on the initial trajectory prediction model based on the prediction loss and the gradient descent algorithm to obtain the trajectory prediction model includes: Based on the predicted loss and the gradient descent algorithm, randomly select target neurons for dropout based on a preset probability, and perform parameter iteration on the initial trajectory prediction model until the number of iterations meets the preset number of iterations to obtain the trajectory prediction model.

[0048] Here, the initial trajectory prediction model includes an input network layer, a hidden network layer, and an output network layer. The hidden network layer includes a ReLU activation function, or Sigmoid and Tanh activation functions, and the output network layer includes a linear activation function. In one embodiment, the initial trajectory prediction model can be designed as a multi-layer feedforward neural network, including 5 neural network layers, 1 input network layer, 3 hidden network layers, and 1 output network layer. Allocate multiple neurons to each network layer. For example, the number of neurons in the input network layer is equal to the total number of input flight data, that is, equal to the total number of flight state data and trajectory influence data. Each hidden network layer has 100 neurons, and the number of neurons in the output network layer is equal to the output dimension of the trajectory prediction.

[0049] In addition, the hidden network layer includes a ReLU activation function, and the output network layer includes a linear activation function.

[0050] Specifically, in the model training stage, based on the predicted loss and the gradient descent algorithm, randomly select target neurons for dropout based on a preset probability, and perform parameter iteration on the initial trajectory prediction model until the number of iterations meets the preset number of iterations to obtain the trajectory prediction model. For example, the Dropout technique can be introduced to randomly dropout some neurons during the training process to improve the generalization ability of the model.

[0051] It should be noted that the total sample flight data can be divided into a training set, a validation set, and a test set. After completing the model training, the model performance can be evaluated on the validation set to adjust the parameters of the trajectory prediction model to optimize the performance. In addition, a final evaluation is performed on the test set to ensure that the trajectory prediction model has good generalization ability. It can be understood that evaluating the performance of the model on independent validation and test sets ensures that the model has good generalization ability. The stability and reliability of the model can be further verified through methods such as cross-validation.

[0052] Thus, the final trajectory prediction model can be integrated into the control system of the aircraft to realize the practical application of trajectory prediction.

[0053] It should also be noted that noise is added to the sample flight data during the training process to enhance the robustness of the model to uncertainty. Apply regularization techniques, such as L1 or L2 regularization, to limit the magnitude of the weights and prevent overfitting. Introduce the Dropout technique to randomly dropout some neurons during the training process to improve the generalization ability of the model.

[0054] The method provided in an embodiment of the present invention aims to achieve high-precision prediction of aircraft trajectories by constructing a multi-layer feedforward neural network and combining technical means such as data preprocessing, model training, and generalization performance improvement. The method provided in an embodiment of the present invention can effectively improve the accuracy and real-time performance of aircraft trajectory prediction, providing strong support for aircraft navigation and control. Furthermore, the trajectory prediction method provided in an embodiment of the present invention is not only applicable to drones but can also be applied to a variety of fields such as commercial aviation and space exploration, and has broad application prospects.

[0055] Based on any of the above embodiments, Figure 2 Schematic diagram of the structure of the aircraft trajectory prediction device based on neural network provided by the present invention, such as Figure 2 As shown, the device includes: An acquisition unit 210 acquires flight data of the aircraft at the current moment in real time, wherein the flight data includes flight status data and trajectory impact data; The trajectory prediction unit 220 performs trajectory prediction based on the trajectory prediction model and applies the flight data at the current moment to obtain a predicted trajectory at the next moment; A flight control unit 230 performs flight control on the aircraft based on the predicted trajectory; The trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network; The sample flight data is selected from historical flight data at historical moments based on a flight data threshold.

[0056] The device provided by the embodiment of the present invention uses a trajectory prediction model constructed based on a neural network to apply the real-time flight data of the aircraft at the current moment to obtain the predicted trajectory at the next moment. Based on the predicted trajectory, the aircraft is controlled in flight, achieving accurate, fast, flexible and practical trajectory prediction, adapting to changing flight environments and the needs of real-time trajectory prediction.

[0057] Based on any of the above embodiments, the data type of the flight data threshold corresponds one-to-one with the data type of the historical flight data; The device also includes a training unit, which is specifically used to: Obtaining historical flight data at the historical moment; Comparing the historical flight data of each data type with the flight data threshold of the corresponding data type; The historical flight data whose comparison result exceeds the flight data threshold is selected as the sample flight data.

[0058] Based on any of the above embodiments, the training unit is further specifically configured to: Select the historical flight data whose comparison result exceeds the flight data threshold as the initial sample flight data; Add Gaussian noise to the initial sample flight data to obtain the sample flight data.

[0059] Based on any of the above embodiments, the training unit is further specifically configured to: Perform trajectory prediction based on the sample flight data and the initial trajectory prediction model to obtain a sample prediction trajectory; Calculate a prediction loss based on the sample prediction trajectory and the sample label; Perform parameter iteration on the initial trajectory prediction model based on the prediction loss and the gradient descent algorithm to obtain the trajectory prediction model; The belonging time of the sample label is later than the belonging time of the sample flight data.

[0060] Based on any of the above embodiments, the training unit is further specifically configured to: Construct a prediction loss function based on the loss function and the regularization term; Calculate the prediction loss based on the prediction loss function by applying the sample prediction trajectory and the sample label; The regularization term includes L1 norm regularization and L2 norm regularization.

[0061] Based on any of the above embodiments, the initial trajectory prediction model includes an input network layer, a hidden network layer, and an output network layer. The hidden network layer includes a ReLU activation function, and the output network layer includes a linear activation function; Each network layer includes multiple neurons; The training unit is further specifically configured to: Perform parameter iteration on the initial trajectory prediction model based on the prediction loss and the gradient descent algorithm, and randomly select target neurons for discard based on a preset probability until the number of iterations meets the preset number of iterations to obtain the trajectory prediction model.

[0062] Figure 3 Illustrates a schematic diagram of the physical structure of an electronic device, such as Figure 3As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340. Among them, the processor 310, the communications interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 can call the logical instructions in the memory 330 to execute a method for predicting the flight trajectory of an aircraft based on a neural network. The method includes: obtaining in real time the flight data of the aircraft at the current moment, where the flight data includes flight state data and trajectory influence data; based on a trajectory prediction model, applying the flight data at the current moment to perform trajectory prediction to obtain a predicted trajectory for the next moment; based on the predicted trajectory, performing flight control on the aircraft; the trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network; the sample flight data is selected from the historical flight data at historical moments based on a flight data threshold.

[0063] In addition, when the logical instructions in the above-mentioned memory 330 can be implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0064] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the neural network-based aircraft trajectory prediction method provided by each of the above methods. The method includes: obtaining in real time the flight data of the aircraft at the current moment, where the flight data includes flight state data and trajectory influence data; based on a trajectory prediction model, applying the flight data at the current moment to perform trajectory prediction to obtain a predicted trajectory for the next moment; based on the predicted trajectory, performing flight control on the aircraft; the trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network; the sample flight data is selected from the historical flight data at historical moments based on a flight data threshold.

[0065] In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it realizes the neural network-based aircraft trajectory prediction method provided by each of the above methods. The method includes: obtaining in real time the flight data of the aircraft at the current moment, where the flight data includes flight state data and trajectory influence data; based on a trajectory prediction model, applying the flight data at the current moment to perform trajectory prediction to obtain a predicted trajectory for the next moment; based on the predicted trajectory, performing flight control on the aircraft; the trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network; the sample flight data is selected from the historical flight data at historical moments based on a flight data threshold.

[0066] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0067] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A flight trajectory prediction method based on a neural network, characterized in that Including: Obtain the flight data of the aircraft at the current moment in real time, where the flight data includes flight status data and trajectory influence data; Based on the trajectory prediction model, apply the flight data at the current moment for trajectory prediction to obtain the predicted trajectory at the next moment; Based on the predicted trajectory, perform flight control on the aircraft; The trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network; The sample flight data is selected from the historical flight data at historical moments based on flight data thresholds; 2. The method for predicting the flight trajectory of an aircraft based on a neural network according to claim 1, wherein The data types of the flight data thresholds correspond one-to-one with the data types of the historical flight data; The construction steps of the sample flight data include: Obtain the historical flight data at the historical moment; Compare the historical flight data of each data type with the flight data threshold of the corresponding data type; Select the historical flight data with the comparison result exceeding the flight data threshold as the sample flight data.

3. The method for predicting the flight trajectory of an aircraft based on a neural network according to claim 2, wherein, The step of selecting the historical flight data with the comparison result exceeding the flight data threshold as the sample flight data includes: Select the historical flight data with the comparison result exceeding the flight data threshold as the initial sample flight data; Add Gaussian noise to the initial sample flight data to obtain the sample flight data.

4. The neural network-based aircraft trajectory prediction method according to any one of claims 1 to 3, characterized in that, The training steps of the trajectory prediction model include: Based on the sample flight data and the initial trajectory prediction model, perform trajectory prediction to obtain a sample predicted trajectory; Calculate the prediction loss based on the sample predicted trajectory and sample labels; Based on the prediction loss and the gradient descent algorithm, perform parameter iteration on the initial trajectory prediction model to obtain the trajectory prediction model; The belonging moment of the sample labels is later than the belonging moment of the sample flight data.

5. The method for predicting the flight trajectory of an aircraft based on a neural network according to claim 4, wherein The step of calculating the prediction loss based on the sample predicted trajectory and sample labels includes: Based on the loss function and regularization term, construct a prediction loss function; Based on the prediction loss function, apply the sample predicted trajectory and sample labels for calculation to obtain the prediction loss; The regularization term includes L1 norm regularization and L2 norm regularization.

6. The method for predicting the flight trajectory of an aircraft based on a neural network according to claim 4, wherein The initial trajectory prediction model includes an input network layer, a hidden network layer, and an output network layer. The hidden network layer includes a ReLU activation function, and the output network layer includes a linear activation function; Each network layer includes multiple neurons; The step of performing parameter iteration on the initial trajectory prediction model based on the prediction loss and the gradient descent algorithm to obtain the trajectory prediction model includes: Based on the prediction loss and the gradient descent algorithm, and randomly select target neurons for dropout based on a preset probability, perform parameter iteration on the initial trajectory prediction model until the number of iterations meets the preset number of iterations to obtain the trajectory prediction model.

7. An aircraft trajectory prediction device based on a neural network, characterized in that, Including: An acquisition unit that obtains the flight data of the aircraft at the current moment in real time, where the flight data includes flight status data and trajectory influence data; A trajectory prediction unit that, based on the trajectory prediction model, applies the flight data at the current moment for trajectory prediction to obtain the predicted trajectory at the next moment; The flight control unit performs flight control on the aircraft based on the predicted trajectory; The trajectory prediction model is obtained by training an initial trajectory prediction model based on sample flight data, and the initial trajectory prediction model is constructed based on a neural network; The sample flight data is selected from historical flight data at historical moments based on a flight data threshold.

8. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the neural network-based aircraft trajectory prediction method according to any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the neural network-based aircraft trajectory prediction method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the neural network-based aircraft trajectory prediction method according to any one of claims 1 to 6.

Images