Training method and device of jamming bomb scattering trajectory estimation model, equipment and medium

Through the training method of interference bomb dispersion trajectory estimation model of the combination of neural networks and physical control equations, the problem of interference bomb identification and performance evaluation in complex multi-physics environments in the existing technology is solved, and high-dimensional, multi-scale, non-linearity is achieved Accurate estimation and adaptability of characteristics.

CN119004960APending Publication Date: 2024-11-22NORTHWESTERN POLYTECHNICAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411020253.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The prior art is difficult to adapt to the rapid identification and effectiveness evaluation of interference bombs in complex multi-physics environments, and the conventional multi-physics model calculation has limited ability to fit all environmental goals and interference physics.

Method used

The model training method of interference bomb dispersion trajectory estimation based on neural network is adopted. The model is pre-trained by simulation data, and the loss function is constructed by combining field test data and physical control equations, transfer learning and model correction are carried out, and physical consistency terms are integrated to adjust the model. parameters to improve the accuracy and reliability of the model.

Benefits of technology

It improves the accuracy and reliability of the estimation of the dispersal trajectory of interference bombs, can represent the high-dimensional, multi-scale and nonlinear characteristics of the data at multiple levels, adapt to complex actual environments, reduce the risk of overfitting, and enhance the cross-domain migration capabilities of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004960A_ABST
    Figure CN119004960A_ABST
Patent Text Reader

Abstract

The invention provides a training method and device for a jamming bomb scattering trajectory estimation model, equipment and a medium. The training method comprises the following steps: pre-training an initialized jamming bomb scattering trajectory estimation model by using analog simulation data to obtain a pre-trained jamming bomb scattering trajectory estimation model; obtaining an input value of the pre-trained jamming bomb scattering trajectory estimation model based on an external field test condition; inputting the input value into the pre-trained jamming bomb scattering trajectory estimation model, so that the pre-trained jamming bomb scattering trajectory estimation model outputs a model prediction value; and adjusting model parameters of the pre-trained jamming bomb scattering trajectory estimation model according to the model predicted value, an external field test data sample and a loss function formed by a physical consistency item and a data fitting item constructed by a physical control equation to obtain a trained jamming bomb scattering trajectory estimation model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of guided missile simulation, and in particular to a training method, device, equipment and medium for a jammer dispersion trajectory estimation model. Background Art

[0002] In actual physical processes, the environment in which the target (such as an aircraft) is located will undergo multiple state changes, such as altitude, pressure, density, and Mach number. These changes cause the target and the jamming system (such as the jammer) itself to suffer from severe radiation, motion, and physical field interference. The variability of the above states leads to the multi-configuration problem, that is, a variety of different working states and environmental conditions need to be considered. This brings challenges to the design and effectiveness evaluation of jammers.

[0003] At present, limited by the cost of simulating the physical field of each element and acquiring data for identification under various practical conditions (complex geometric shapes, high Mach numbers), conventional solutions use traditional multi-physics model calculation and identification methods. This method is designed for specific typical working conditions based on response surface and principal component analysis. Most of them are shallow models that can only fit a small subset of the original function space, and have limited fitting capabilities for the entire complex function space. The multi-element and multi-configuration field information of jammers has complex high-dimensional and strong nonlinear characteristics, making it difficult for conventional multi-physics model calculation and identification methods to adapt to the rapid identification needs of full-environment targets and interference physical fields in actual mission scenarios. Summary of the invention

[0004] The present invention discloses a training method, device, equipment and medium for a jammer dispersion trajectory estimation model; the method is capable of mining and learning deep hidden information from big data, and multi-level representation of the high-dimensional, multi-scale and nonlinear characteristics of data, thereby improving the accuracy of jammer dispersion trajectory estimation.

[0005] The technical solution of the present disclosure is achieved as follows:

[0006] In a first aspect, the present disclosure provides a training method for a jammer dispersion trajectory estimation model, the training method comprising:

[0007] Pre-training the initialized jammer dispersion trajectory estimation model using simulation data to obtain a pre-trained jammer dispersion trajectory estimation model;

[0008] Acquiring input values ​​of the pre-trained jammer dispersion trajectory estimation model based on field test conditions;

[0009] Inputting the input value into the pre-trained jammer dispersion trajectory estimation model so that the pre-trained jammer dispersion trajectory estimation model outputs a model prediction value;

[0010] The model parameters of the pre-trained jammer dispersion trajectory estimation model are adjusted according to the model prediction value, the field test data sample, and the loss function formed by the physical consistency term and the data fitting term constructed by the physical control equation to obtain the trained jammer dispersion trajectory estimation model.

[0011] In a second aspect, the present disclosure provides a training device for a jammer dispersion trajectory estimation model, the training device comprising: a pre-training part, an acquisition part, a jammer dispersion trajectory estimation model and a training part; wherein,

[0012] The pre-training part is configured to pre-train the initialized jammer dispersion trajectory estimation model using simulation data to obtain a pre-trained jammer dispersion trajectory estimation model;

[0013] The acquisition part is configured to acquire the input value of the pre-trained jammer dispersion trajectory estimation model based on field test conditions;

[0014] The jammer dispersion trajectory estimation model is configured to receive the input value and output a model prediction value;

[0015] The training part is configured to adjust the model parameters of the pre-trained jammer dispersion trajectory estimation model according to the model prediction value, field test data samples, and a loss function formed by physical consistency terms and data fitting terms constructed by physical control equations to obtain a trained jammer dispersion trajectory estimation model.

[0016] In a third aspect, the present disclosure provides a computing device, comprising: a processor and a memory; the processor is used to execute instructions stored in the memory to implement the training method of the jamming bomb dispersion trajectory estimation model as described in the first aspect.

[0017] In a fourth aspect, the present disclosure provides a computer-readable storage medium storing at least one instruction, wherein the at least one instruction is used to be executed by a processor to implement the training method of the jamming projectile dispersion trajectory estimation model as described in the first aspect.

[0018] The present disclosure provides a training method, device, equipment and medium for a jammer dispersion trajectory estimation model; for the pre-trained jammer dispersion trajectory estimation model, an input value is obtained using field test conditions, so that the jammer dispersion trajectory estimation model can start transfer learning from a state close to the actual situation; and the physical consistency term is integrated into the loss function used to adjust the model parameters, so that the trained model has both statistical learning advantages and conforms to physical laws, thereby improving the accuracy and reliability of the model. The scheme provided by the present disclosure can represent the high-dimensional, multi-scale and nonlinear characteristics of data at multiple levels, thereby improving the accuracy of jammer dispersion trajectory estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flowchart of a method for training a jamming projectile dispersion trajectory estimation model provided in the present invention.

[0020] Figure 2 A schematic diagram of the composition of a jamming bomb dispersion trajectory estimation model provided in the present invention.

[0021] Figure 3 A schematic diagram of the flow of the training jammer dispersion trajectory estimation model provided in the present invention.

[0022] Figure 4 A schematic diagram of the composition of a training device for a jamming projectile dispersion trajectory estimation model provided in the present invention.

[0023] Figure 5 A schematic diagram of the structure of a computing device provided by the present disclosure. DETAILED DESCRIPTION

[0024] The technical solutions in the present disclosure will be described clearly and completely below in conjunction with the accompanying drawings in the present disclosure.

[0025] In view of the technical characteristics of the multi-factor and multi-configuration field information of jammers, which have complex high-dimensional and strong nonlinear characteristics, this disclosure is based on the idea of ​​data-driven, and uses artificial intelligence technology, such as the outstanding ability of neural networks to represent complex and strong nonlinear systems, to directly mine and learn deep hidden information from data, so as to be able to represent the nonlinear characteristics of data at multiple levels to solve high-dimensional, multi-scale and nonlinear problems. Based on this, see Figure 1 , which shows a training method for a jamming bomb dispersion trajectory estimation model provided by the present disclosure, and the training method includes steps S101 to S104.

[0026] In step S101, the initialized jammer dispersion trajectory estimation model is pre-trained using simulation data to obtain a pre-trained jammer dispersion trajectory estimation model.

[0027] In this disclosure, Figure 2 As shown, the jammer scatter trajectory estimation model includes an encoder 21 and a decoder 22, wherein the encoder 21 is constructed based on a graph convolutional neural network (GNN), and the decoder 22 is constructed based on a deconvolutional neural network with an upsampling function. In detail, the jammer unit is used as the input element (also referred to as a node in the graph), and the field feature information of different nodes in different ranges is gradually extracted through the graph convolution kernel of the GNN to achieve feature learning and dimensionality reduction of the physical field, and output the feature representation after dimensionality reduction, which can capture the key information of the input element. The decoder gradually reconstructs or generates the physical field information required to estimate the scatter trajectory, such as the velocity field and pressure field of the jammer, according to the feature representation output by the encoder.

[0028] In some examples, the jammer unit is used to describe different physical or dynamic characteristics of the jammer, such as the physical characteristics of the jammer, such as mass, size and shape; the motion information of the jammer, such as position, speed or acceleration; the environmental factors of the jammer's flight, such as atmospheric conditions, pressure and density; the radiation characteristics of the jammer, such as radiation intensity, wavelength and radiation distribution; the combat and parameters of the jammer, such as altitude and flight Mach number; the material properties of the jammer and the dynamic behavior of the jammer, such as rotation, vibration, etc. GNN learns the information of these jammer units to learn how this information affects the behavior of the jammer and predict its dispersion trajectory.

[0029] In some examples, when constructing Figure 2 After the jammer dispersion trajectory estimation model shown in the figure is initialized and the weights and biases of the model are initialized, it is first necessary to make the initialized jammer dispersion trajectory estimation model learn the general physical field behavior and characteristics of the jammer. In the present disclosure, the physical behavior of the jammer is simulated by running numerical simulation software on a computer to generate simulation data, such as flight characteristics, radiation patterns, etc. under different environmental conditions and operating parameters. These large amounts of data obtained through simulation can form a standardized data set for training the initialized jammer dispersion trajectory estimation model.

[0030] It should be noted that after completing the pre-training process described in the above example, the jammer dispersion trajectory estimation model has learned the general physical field behavior and characteristics of the jammer, but its adaptability to the complex actual battlefield environment is still low, resulting in the pre-trained jammer dispersion trajectory estimation model being unable to accurately predict or estimate the dispersion trajectory of the jammer in a complex actual environment. Based on this, the present disclosure uses subsequent steps S102 to S104 to further adjust and optimize the model parameters based on the pre-trained model using field test data, that is, to complete the transfer learning and model correction process, so that the model finally obtained by training can better adapt to the complex actual combat environment, thereby improving the accuracy and reliability of the prediction.

[0031] In step S102, the input value of the pre-trained jammer dispersion trajectory estimation model is obtained based on the field test conditions.

[0032] In the present disclosure, the process of transfer learning and model correction of the pre-trained jammer dispersion trajectory estimation model is to hope that the model can be transferred from the pre-trained state to a state that is more in line with the actual environment scene. Therefore, in the process of transfer learning and model correction of the model, the actual battlefield situation can be introduced for correction of the input value of the model. The present disclosure obtains the input value of the pre-trained jammer dispersion trajectory estimation model based on field test conditions.

[0033] In step S103, the input value is input into the pre-trained jammer dispersion trajectory estimation model, so that the jammer dispersion trajectory estimation model outputs a model prediction value.

[0034] In the present disclosure, Figure 2 In the model composition example shown, the input value is first input to the encoder 21 built based on the graph convolutional neural network (GNN), and the encoder 21 extracts the field feature information in different ranges to achieve feature learning and dimensionality reduction of the physical field, and then outputs the feature representation after dimensionality reduction. The decoder 22 is used to gradually reconstruct or generate the physical field information required for estimating the scattering trajectory, such as the velocity field and pressure field of the jammer, according to the feature representation output by the encoder 21, and then obtain the model prediction value of the model output.

[0035] In some examples, the model prediction value may include predicted observation values ​​used to characterize the motion behavior of the jammer, such as position, velocity, acceleration, etc.; it may also include predicted physical quantities used to characterize physical field data, such as the environmental state around the jammer, such as velocity field, pressure field, and temperature field.

[0036] In step S104, the model parameters of the pre-trained jammer dispersion trajectory estimation model are adjusted according to the model prediction value, the field test data sample, and the loss function formed by the physical consistency term and the data fitting term constructed by the physical control equation to obtain the trained jammer dispersion trajectory estimation model.

[0037] In the present disclosure, in combination with the predicted observation values ​​and predicted physical quantities included in the above-mentioned model prediction values, the present disclosure, in addition to the observation deviation represented by the data fitting term, also introduces the physical quantity deviation that can be expressed by the physical control equation when constructing the loss function for adjusting the model parameters, that is, the physical consistency term in step S104. By combining the physical consistency term and the data fitting term, a loss function is obtained. By adjusting the model parameters of the jamming projectile dispersion trajectory estimation model with the goal of minimizing the loss function, a trained jamming projectile dispersion trajectory estimation model is finally obtained.

[0038] It should be noted that in the loss function of the present disclosure, due to the integration of the physical consistency term, the trained model can have both statistical learning advantages and conform to physical laws, thereby improving the accuracy and reliability of the model. In detail, the trained model, due to the constraints of the physical control equation, can not only ensure that the prediction results of the model are consistent with the physical phenomenon, but also can be better generalized to new, unlearned data, making the prediction results of the model easier to interpret, increasing the transparency and credibility of the model, and being able to remain robust to small perturbations or changes in the input data. In addition, the physical control equation provides additional structured information, which helps to train the model when the amount of data is limited, reduces the dependence on a large amount of training data and the number of iterations required, thereby improving the training efficiency. In addition, combining the physical control equation with the loss function can help prevent the model from being overly sensitive to noise data or outliers during the training process, thereby reducing the risk of overfitting and improving the cross-domain migration ability of the model. Therefore, the solution provided by the present disclosure can mine and learn deep hidden information from big data, and multi-level representation of the high-dimensional, multi-scale and nonlinear characteristics of data, thereby improving the accuracy and reliability of the estimation of the dispersion trajectory of jamming bombs.

[0039] for Figure 1 In some possible implementations of the technical solution shown, the step S102 of obtaining the input value of the pre-trained jammer dispersion trajectory estimation model based on the field test conditions includes:

[0040] The initial state of the jammer is obtained according to the field test conditions, and the initial state of the jammer is converted into an input value of the pre-trained jammer dispersion trajectory estimation model.

[0041] for Figure 1The technical solution shown in the figure may be implemented in some possible ways, such as Figure 3 As shown, the model parameters of the pre-trained jammer dispersion trajectory estimation model are adjusted according to the model prediction value, the field test data sample, and the loss function formed by the physical consistency term and the data fitting term constructed by the physical control equation to obtain the trained jammer dispersion trajectory estimation model, including:

[0042] S301: Selecting data samples from field test data;

[0043] S302: Obtaining data fitting items according to data samples and model prediction values;

[0044] S303: Obtaining a physical consistency term based on a physical control equation according to a model prediction value;

[0045] S304: constructing a loss function for characterizing the deviation of the model prediction value according to the physical consistency term and the data fitting term;

[0046] S305: With the goal of minimizing the loss function, the model parameters of the pre-trained jammer scatter trajectory estimation model are adjusted to obtain the trained jammer scatter trajectory estimation model.

[0047] For the above implementation, specifically, in the field test data, samples corresponding to the time of the model prediction value are selected as data samples, and the data in the sample used to characterize the motion behavior of the jammer are used as actual observation values, such as position, speed, acceleration, etc. In the present disclosure, in addition to constructing a loss function based on the deviation of the observed value, the present disclosure also introduces the deviation of the physical field data into the loss function, that is, the physical consistency term.

[0048] Based on this, in some examples, obtaining the data fitting item according to the data sample and the model prediction value includes:

[0049] The data fitting term is constructed based on the deviation between the actual observation value in the data sample and the observation prediction value in the model prediction value.

[0050] For the above example, specifically, set L data represents the data fitting term, then we have:

[0051]

[0052] Where N is the number of observations, y i represents the i-th actual observation value, Represents the i-th predicted observation value

[0053] In some examples, obtaining the physical consistency term based on a physical control equation according to the model prediction value includes:

[0054] Based on the model prediction value, using the physical control equation to obtain an evaluation index for measuring the degree of satisfaction of the physical properties;

[0055] The evaluation index for measuring the degree of satisfaction of the physical property is determined as the physical consistency item.

[0056] For the above examples, it should be noted that the present disclosure can establish physical control equations according to the jamming bomb unit, such as the Navier-Stokes (NS) equations of fluid mechanics, the combustion equations and the heat transfer equations. For the NS equations, the present disclosure can use the Smoothed Particle Hydrodynamics (SPH) method for fluid simulation. In detail, the SPH method is used to discretize the NS equations of the fluid, that is, the fluid is regarded as consisting of a large number of discrete particles, each particle carries certain physical properties (such as mass, velocity, pressure, viscosity coefficient, etc.), and the behavior of the fluid is simulated by the interaction between the particles. Based on the above physical control equations, the relevant physical quantities constrained by the physical control equations can be obtained. Specifically, according to the data in the model prediction value, the physical properties satisfied by the physical quantity are predicted by the above physical control equations, thereby obtaining an evaluation index for measuring the degree of satisfaction of the physical properties. That is to say, in the present disclosure, the physical consistency term L physics can be represented as a function of the model prediction value, namely This function can represent the residual of the predicted physical quantity calculated by the physical control equation to measure the degree of satisfaction of the physical property. For example, for the continuity condition in fluid mechanics, after substituting the model prediction value into the discrete continuity equation, the residual term obtained is the indicator of the degree of satisfaction of the physical property.

[0057] Based on the above two examples, in some examples, the loss function provided by the present disclosure is shown as follows:

[0058] L=λL dara +(1-λ)L physics

[0059] Wherein, λ is the weight coefficient, and 0<λ<1.

[0060] For the above loss function L, it not only enables the model to learn how to fit the observed data, but also learn how to conform to the laws of physics, thereby improving the prediction accuracy and physical rationality of the model.

[0061] It should be noted that after obtaining the loss function L, the gradient of the loss function with respect to the model parameters is calculated by the back propagation algorithm. Subsequently, the model parameters and weights are updated according to an optimization algorithm, such as the Adam algorithm, and the iteration is repeated until the predetermined number of iterations or the loss function L is minimized, thereby obtaining the trained jammer dispersion trajectory estimation model.

[0062] Based on the same inventive concept as the above technical solution, see Figure 4 , which shows a training device 40 for a jamming projectile dispersion trajectory estimation model provided by the present disclosure, the training device 40 comprises: a pre-training part 401, an acquisition part 402, a jamming projectile dispersion trajectory estimation model 403 and a training part 404; wherein,

[0063] The pre-training part 404 is configured to pre-train the initialized jammer dispersion trajectory estimation model 403 using simulation data to obtain the pre-trained jammer dispersion trajectory estimation model 403;

[0064] The acquisition part 402 is configured to acquire the input value of the pre-trained jammer dispersion trajectory estimation model 403 based on the field test conditions;

[0065] The jammer dispersion trajectory estimation model 403 is configured to receive the input value and output a model prediction value;

[0066] The training part 404 is configured to adjust the model parameters of the pre-trained jammer dispersion trajectory estimation model 403 according to the model prediction value, the field test data sample and the loss function formed by the physical consistency term and the data fitting term constructed by the physical control equation to obtain the trained jammer dispersion trajectory estimation model 403.

[0067] In some examples, the obtaining portion 402 is configured to:

[0068] The initial state of the jammer is obtained according to the field test conditions, and the initial state of the jammer is converted into an input value of the pre-trained jammer dispersion trajectory estimation model 403.

[0069] In some examples, the training portion 404 is configured to:

[0070] Selecting a data sample from the field test data;

[0071] Acquire the data fitting item according to the data sample and the model prediction value;

[0072] According to the model prediction value, obtaining the physical consistency term based on the physical control equation;

[0073] Constructing a loss function for characterizing the deviation of the model prediction value according to the physical consistency term and the data fitting term;

[0074] With the goal of minimizing the loss function, the model parameters of the pre-trained jammer scatter trajectory estimation model 403 are adjusted to obtain the trained jammer scatter trajectory estimation model 403.

[0075] In some examples, the training portion 404 is configured to:

[0076] The data fitting term is constructed based on the deviation between the actual observation value in the data sample and the observation prediction value in the model prediction value.

[0077] In some examples, the training portion 404 is configured to:

[0078] Based on the model prediction value, using the physical control equation to obtain an evaluation index for measuring the degree of satisfaction of the physical properties;

[0079] The evaluation index for measuring the degree of satisfaction of the physical property is determined as the physical consistency item.

[0080] In some examples, the loss function is as follows:

[0081] L=λL data +(1-λ)L physics

[0082] Where λ is the weight coefficient, and 0<λ<1; L data represents the data fitting term, L physics represents a physical consistency term; N represents the number of observations, y i represents the i-th actual observation value, represents the i-th predicted observation value; It represents the residual of the predicted physical quantity calculated by the physical control equation, that is, the physical consistency term.

[0083] In some examples, the jammer dispersion trajectory estimation model 403 includes: an encoder constructed based on a graph convolutional neural network and a decoder constructed based on a deconvolutional neural network.

[0084] Please refer to Figure 5, which shows a structural block diagram of a computing device provided by an exemplary embodiment of the present disclosure. In some examples, the computing device 50 can be at least one of a smart phone, a smart watch, a desktop computer, a laptop, a virtual reality terminal, an augmented reality terminal, a wireless terminal, and a laptop portable computer. The computing device 50 has a communication function and can access a wired network or a wireless network. The computing device 50 can generally refer to one of a plurality of terminals, and those skilled in the art can know that the number of the above terminals can be more or less. In some examples, the computing device 50 can receive data based on the wired network or wireless network to which it is connected. It can be understood that the computing device 50 undertakes the calculation and processing work of the technical solution of the present disclosure, and the present disclosure does not limit this.

[0085] like Figure 5 As shown, the computing device in the present disclosure may include one or more of the following components: a processor 510 and a memory 520 .

[0086] Optionally, the processor 510 uses various interfaces and lines to connect various parts of the entire computing device, and executes various functions of the computing device and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 520, and calling data stored in the memory 520. Optionally, the processor 510 can be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), and programmable logic array (PLA). The processor 510 can integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), a neural network processor (NPU), and a baseband chip. Among them, the CPU mainly processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing the content that needs to be displayed on the touch display; the NPU is used to implement artificial intelligence (AI) functions; and the baseband chip is used to process wireless communications. It is understandable that the above baseband chip may not be integrated into the processor 510, but may be implemented by a separate chip.

[0087] The memory 520 may include a random access memory (RAM) or a read-only memory (ROM). Optionally, the memory 520 includes a non-transitory computer-readable storage medium. The memory 520 may be used to store instructions, programs, codes, code sets, or instruction sets. The memory 520 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above various method embodiments, etc.; the data storage area may store data created according to the use of the computing device, etc.

[0088] In addition, those skilled in the art will appreciate that the structure of the computing device shown in the above figures does not constitute a limitation on the computing device, and the computing device may include more or fewer components than shown, or combine certain components, or arrange the components differently. For example, the computing device also includes a display screen, a camera assembly, a microphone, a speaker, a radio frequency circuit, an input unit, a sensor (such as an acceleration sensor, an angular velocity sensor, a light sensor, etc.), an audio circuit, a WiFi module, a power supply, a Bluetooth module, and other components, which will not be described in detail here.

[0089] The present disclosure also provides a computer-readable storage medium storing at least one instruction, wherein the at least one instruction is used to be executed by a processor to implement the training method of the jamming projectile dispersion trajectory estimation model as described in the above embodiments.

[0090] The present disclosure also provides a computer program product, which includes computer instructions, which are stored in a computer-readable storage medium; a processor of a computing device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computing device executes to implement the training method of the jamming bomb dispersion trajectory estimation model described in the above-mentioned embodiments.

[0091] Those skilled in the art should be aware that in one or more of the above examples, the functions described in the present disclosure can be implemented using hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any media that facilitates the transmission of computer programs from one place to another. Storage media can be any available media that can be accessed by a general or special-purpose computer.

[0092] It should be noted that the technical solutions described in the present disclosure can be combined arbitrarily without conflict.

[0093] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A training method for a jammer dispersion trajectory estimation model, characterized in that: The training method comprises: Pre-training the initialized jammer dispersion trajectory estimation model using simulation data to obtain a pre-trained jammer dispersion trajectory estimation model; Acquiring input values ​​of the pre-trained jammer dispersion trajectory estimation model based on field test conditions; Inputting the input value into the pre-trained jammer dispersion trajectory estimation model so that the pre-trained jammer dispersion trajectory estimation model outputs a model prediction value; The model parameters of the pre-trained jammer dispersion trajectory estimation model are adjusted according to the model prediction value, the field test data sample, and the loss function formed by the physical consistency term and the data fitting term constructed by the physical control equation to obtain the trained jammer dispersion trajectory estimation model.

2. The training method according to claim 1, characterized in that: The step of obtaining the input value of the jammer dispersion trajectory estimation model based on the field test conditions includes: The initial state of the jammer is acquired according to the field test conditions, and the initial state of the jammer is converted into an input value of the pre-trained jammer scatter trajectory estimation model.

3. The training method according to claim 1, characterized in that: The method further comprises: adjusting the model parameters of the pre-trained jammer dispersion trajectory estimation model according to the model prediction value, the field test data sample, and the loss function formed by the physical consistency term and the data fitting term constructed by the physical control equation to obtain the trained jammer dispersion trajectory estimation model, including: Selecting a data sample from the field test data; Acquire the data fitting item according to the data sample and the model prediction value; According to the model prediction value, obtaining the physical consistency term based on the physical control equation; Constructing a loss function for characterizing the deviation of the model prediction value according to the physical consistency term and the data fitting term; With the goal of minimizing the loss function, the model parameters of the pre-trained jammer scatter trajectory estimation model are adjusted to obtain the trained jammer scatter trajectory estimation model.

4. The training method according to claim 3, characterized in that: The obtaining the data fitting item according to the data sample and the model prediction value includes: The data fitting term is constructed based on the deviation between the actual observation value in the data sample and the observation prediction value in the model prediction value.

5. The training method according to claim 3, characterized in that: The obtaining the physical consistency term based on the physical control equation according to the model prediction value includes: Based on the model prediction value, using the physical control equation to obtain an evaluation index for measuring the degree of satisfaction of the physical properties; The evaluation index for measuring the degree of satisfaction of the physical property is determined as the physical consistency item.

6. The method according to claim 3, characterized in that: The loss function is shown as follows: L=λL data +(1-λ)L physics Where λ is the weight coefficient, and 0<λ<1; L data represents the data fitting term, L physics represents a physical consistency term; N represents the number of observations, y i represents the i-th actual observation value, represents the i-th predicted observation value; L physics It represents the residual of the predicted physical quantity calculated by the physical control equation, that is, the physical consistency term.

7. The training method according to claim 1, characterized in that: The jammer scatter trajectory estimation model includes: an encoder constructed based on a graph convolutional neural network and a decoder constructed based on a deconvolutional neural network.

8. A training device for a jamming projectile dispersion trajectory estimation model, characterized in that: The training device comprises: a pre-training part, an acquisition part, a jammer dispersion trajectory estimation model and a training part; wherein, The pre-training part is configured to pre-train the initialized jammer dispersion trajectory estimation model using simulation data to obtain a pre-trained jammer dispersion trajectory estimation model; The acquisition part is configured to acquire the input value of the pre-trained jammer dispersion trajectory estimation model based on field test conditions; The jammer dispersion trajectory estimation model is configured to receive the input value and output a model prediction value; The training part is configured to adjust the model parameters of the pre-trained jammer dispersion trajectory estimation model according to the model prediction value, the field test data sample, and the loss function formed by the physical consistency term and the data fitting term constructed by the physical control equation to obtain the trained jammer dispersion trajectory estimation model.

9. A computing device, characterized in that The computing device comprises: a processor and a memory; the processor is used to execute instructions stored in the memory to implement the training method of the jamming projectile dispersion trajectory estimation model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, and the at least one instruction is used to be executed by a processor to implement the training method of the jamming projectile dispersion trajectory estimation model as described in any one of claims 1 to 7.