Space target track positioning method, system, equipment and medium
By using data disassembly and neural network modules in the spatial target track positioning system, combining GRU and FNN networks for training and verification, the problems of complex computing and difficult model tuning in the existing technology are solved, and high accuracy and low accuracy are achieved Cost track calculation.
Patent Information
- Application Number
- CN202411130586.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2024-12-17
AI Technical Summary
The existing spatial target track positioning method requires a lot of calculation and tuning to find a suitable model, and the stress conditions are complex and the calculation results are prone to deviation.
A spatial target track positioning system is proposed, including data dismantling module, neural network module and combination module. By dismantling the original observation data, training the training set is trained using GRU and FNN networks, and matching verification is performed through the verification set, and gradient is performed. Update and backpropagate to ensure the accuracy of the model.
Low-cost and high-accuracy track calculations are realized, reducing dependence on tle data, simplifying the calculation process, and improving the risk resistance and accuracy of the model.
Smart Images
Figure CN119150920A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of space positioning, and in particular relates to a method, system, equipment and medium for positioning a space target orbit. Background Art
[0002] In the technology of space target positioning, the general computing systems are now divided into two categories. One is the Kepler method model, which measures the position and velocity of the satellite and uses Kepler's law to calculate the orbital parameters of the satellite. The other is the high-precision orbit model, which includes numerical integration orbit models such as J2, J4, and HPOP.
[0003] In terms of calculation parameters, both the Kepler method and the high-precision numerical method face two problems in calculating orbital data. The first problem is a large number of input parameters; the second problem is complex calculations. Moreover, the force conditions change in real time, and different forces have different effects at different times, such as orbital attenuation, perturbation, target weight, attitude, force area, atmospheric drag model parameters, gravity model, etc. When one or more of these errors occur, the calculation results will be biased.
[0004] Different neural network models under different parameters describe the trajectory differently. It is necessary to select appropriate network models and parameters, and to find a suitable model and parameters after a lot of calculations and tuning.
[0005] The convergence of the network model is different under different gradients. Different data values and data amounts will affect the convergence of the model. Finding and debugging a suitable gradient strategy can ensure that the model achieves good convergence. Summary of the invention
[0006] The present invention proposes a space target orbit positioning method, system, device and medium to solve the problem in the prior art that a large amount of calculation and tuning is required to find a suitable model.
[0007] To achieve the above object, the present invention proposes the following technical solutions:
[0008] A space target orbit positioning system, comprising a data disassembly module, a neural network module and a combination module;
[0009] The data decomposition module decomposes the original observation data of the target to obtain a training set;
[0010] The neural network module trains the data of the training set to obtain trained data;
[0011] The combination module integrates the trained data to obtain the trajectory of the target.
[0012] Preferably, the data decomposition module also obtains a verification set after decomposing the original observation data of the target.
[0013] Preferably, the system also includes a deduction module, which performs matching verification on the trained data based on the verification set. If the trained data matches correctly, the trained data is retained. If the trained data matches incorrectly, gradient update and back propagation are performed, and the training is restarted by the neural network module.
[0014] Preferably, the matching verification is specifically to set a preset matching degree, and compare the matching degree of the training data matched by the deduction module based on the verification set with the preset matching degree. If the matching degree is greater than the preset matching degree, the training data match is qualified, otherwise it is unqualified.
[0015] Preferably, the neural network module includes a GRU network and an FNN network, and the neural network module trains the data of the training set through the GRU network, the FNN network, or a combination of the GRU network and the FNN network.
[0016] Preferably, the data decomposition module decomposes the original observation data of the target into three single-axis data columns of x, y, and z, and obtains a training set based on the data columns.
[0017] A method for locating a space target orbit comprises the following steps:
[0018] Step 1: Disassemble the original observation data of the target to obtain the training set and the validation set;
[0019] Step 2: Train the training set data to obtain trained data;
[0020] Step 3: Verify the trained data based on the validation set. If the trained data matches the validation set, keep the trained data and proceed to step 4. If the trained data does not match the validation set, return to step 2.
[0021] Step 4: Integrate the trained data and output the coordinate value of the target.
[0022] An electronic device comprising a memory and a processor;
[0023] Memory for storing computer programs;
[0024] The processor is used to execute the computer program, and the computer program, when executed by the processor, implements the steps of a method for orbit positioning of a space target.
[0025] A computer readable storage medium,
[0026] The storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for orbit positioning of a space target.
[0027] The present invention is beneficial in that:
[0028] The present invention proposes a space target orbit positioning system including a data disassembly module and a neural network module. The neural network is used to process the disassembled raw data, and different network models are used to extract rules, thereby avoiding dependence on TLE data. After each training is completed, the verification data is verified, which can realize low-cost and high-accuracy orbit calculation, and is easy to promote and implement. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:
[0030] Figure 1 It is a schematic diagram of a space target orbit positioning system;
[0031] Figure 2 This is a schematic diagram of the GRU network;
[0032] Figure 3 It is a schematic diagram of FNN network;
[0033] Figure 4 Comparison of the deduction results between this system and the SGP4 model;
[0034] Figure 5 A schematic diagram of a method for locating a space target orbit;
[0035] Figure 6 A schematic diagram of an electronic device. DETAILED DESCRIPTION
[0036] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.
[0037] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.
[0038] Embodiment 1:
[0039] See also Figure 1As shown, the present invention provides a space target orbit positioning system. According to Kepler's law, a space object within the earth's orbit performs an elliptical motion around the earth, which can be considered as a time-sequential motion to a certain extent, and the position of the object at the next moment is only affected by the previous moment.
[0040] The force conditions in three-dimensional space vary in a complex manner. A single network model cannot reflect the complete laws of motion, and different models are needed to adapt to the force manifestations at different locations.
[0041] After cleaning the real observation data in the past, data within a certain time range are selected as training data. By disassembling the three-dimensional coordinate axis, different network models are used to extract the rules. After each training, the verification data is verified. In order to avoid the overfitting problem of the training data, only the models that have a good performance on the verification data are retained.
[0042] Specifically, the system includes a data decomposition module, a neural network module, a deduction module and a combination module.
[0043] The data decomposition module decomposes the original observation data to obtain a training set and a validation set.
[0044] The data decomposition module receives the original observation data, which is the original three-dimensional coordinate value in the format of x, y, and z. After decomposition, separate data columns of the three separate coordinate axes x, y, and z are obtained. According to the data columns of the three separate coordinate axes x, y, and z, the training set and verification set of the three separate coordinate axes x, y, and z are obtained.
[0045] The neural network module provides GRU network and FNN network as the computing architecture, and separately calculates and extracts the rules for the three axes of x, y, and z of the observed data. For the continuous coordinates of different axes, the coordinates of three points are taken as a time series data body to deduce the coordinate points of the next moment. According to the different data rules, different network models are used to extract rules to obtain trained data.
[0046] The GRU network includes an update gate and a reset gate. The update gate controls how the hidden state at the previous moment flows into the hidden state at the current moment, and its output is between 0 and 1, where 0 means completely ignoring the past hidden state and 1 means completely retaining the past hidden state; the reset gate determines how the hidden state at the previous moment affects the candidate hidden state at the current moment, and its output determines whether to ignore the information in the past hidden state.
[0047] The candidate hidden state at the current moment is: the new candidate hidden state calculated based on the input data and the hidden state at the previous moment. The hidden state at the current moment is: the final hidden state calculated by combining the update gate and the hidden state at the previous moment. Figure 2 shown.
[0048] The FNN network (feedforward neural network) is a unidirectional multi-layer network structure. Information starts from the input layer and is transmitted layer by layer in one direction, that is, unidirectional transmission, until it ends at the output layer. Feedforward means that the propagation direction is forward. The feedforward neural network consists of three parts: the input layer (the 0th layer), the output layer (the last layer), and the middle part is called the hidden layer. The hidden layer can be one layer or multiple layers. Figure 3 shown.
[0049] Each neuron takes the output of each node in the previous layer as input, obtains the output of this node through linear transformation (combined with bias) and nonlinear function activation, and passes it to the node in the next layer. By continuously integrating the parameters of each neuron, outputting the final result, using its back propagation algorithm to calculate the loss value, and continuously optimizing the neuron parameters, the effect of simulating mechanical changes is achieved.
[0050] Based on the description of a track in three-dimensional space, at least three coordinate points at different times are needed for calculation. The coordinate points at three consecutive times, t-2, t-1, and t0, are used as input, and the coordinate points at time t+1 are used as output. The model is trained and multi-neuron parameters are designed to simulate the forces in different dimensions. The neural network module in the present invention can use a double-layer GRU network for regularity extraction, and an FNN network for fitting the final output. To achieve the simulation of force changes in the form of neurons, thereby performing track coordinate deduction.
[0051] The deduction module verifies the trained data based on the verification set. If the verification result meets the preset standard, the current trained data is saved as the deduction result. If the verification result does not meet the preset standard, gradient update and back propagation are performed, and training is restarted by the neural network module.
[0052] The combination module combines the deduction results of the single axis together and outputs a complete output 3D coordinate value.
[0053] In a specific implementation, the neural network module uses a double-layer GRU+FNN network to extract regularities from Z-axis data, and uses FNN networks to extract regularities from X-axis and Y-axis data separately.
[0054] In a specific embodiment, a matching degree is preset in the deduction module, and the trained data is compared with the corresponding verification set. If the comparison result shows that the matching degree is greater than the preset matching degree, the trained data can be used as the deduction result. Otherwise, gradient update and back propagation are performed, and the neural network module restarts training.
[0055] The system calculation process proposed in this system is simple, no additional auxiliary parameters are required, and the observed data can be directly used.
[0056] Take a certain experimental cabin track as an example, select the real position data,
[0057] The system and SGP4 model are used to predict the orbit of a certain experimental cabin. The comparison results are as follows: Figure 4 As shown, model_distance is the difference between the model deduction result and the real data, and sgp4_distance is the difference between the result calculated by the sgp4 model based on TLE and the real data.
[0058] It can be seen that the deduction results of this system within a certain period of time are closer to the real results than those of the SGP4 model.
[0059] Firstly, the present invention can perform orbit calculation based on direct observation data, thus getting rid of the dependence on TLE data and having better risk resistance.
[0060] The present invention replaces the conventional dynamic model with a neural network, reduces the reliance on professional knowledge, and does not require additional data parameters during calculation, greatly reducing the computational complexity. Under the same observation data, the model system accuracy is better than the current mainstream SGP4 model in orbit calculation accuracy within a certain period of time. The present invention is conducive to expanding the scope of researchers in space physics, providing a favorable guarantee for the research and development of celestial motion; new laws of motion of space objects can be discovered within the scope of mathematics.
[0061] Embodiment 2:
[0062] See also Figure 5 As shown, the present invention provides a space target orbit positioning method, and the space target orbit positioning algorithm method is based on the above-mentioned space target orbit positioning system and specifically includes:
[0063] Step 1: Get the original data set and decompose the original 3D coordinate values in xyz format into separate data columns of three separate coordinate axes: x, y, and z. Decompose into a training set and a validation set.
[0064] Step 2: Use a neural network model to train the coordinate axis data. For example, use an FNN model for the x-axis, an FNN model for the y-axis, and a GRU+FNN model for the z-axis. Train them separately to obtain trained data.
[0065] Step 3: After one round of training, the corresponding post-training data is verified through the validation set. If the post-training data performs well on the validation set, the current model is saved. If the effect is not good or is inferior to the currently saved model, gradient update and back propagation are performed, and training is restarted until a matching performance model is found.
[0066] Step 4, when performing orbital deduction, the observation data is also broken down into three separate coordinate axis data of X, Y, and Z. By applying their respective models for deduction, their respective deduction results will be obtained. The single-axis deduction results are combined together and output as a complete output 3D coordinate value.
[0067] In a specific implementation, the good performance in step 3 specifically means that the matching degree between the trained data and the corresponding validation set exceeds a preset matching degree.
[0068] Embodiment 3:
[0069] See also Figure 6 As shown, the present invention also provides an electronic device 100; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.
[0070] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the method for orbital positioning of a space target described in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 can include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.
[0071] The at least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.
[0072] Embodiment 4:
[0073] If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).
[0074] It is known from common technical knowledge that the present invention can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are included in the present invention.
[0075] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0076] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0077] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0078] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0079] 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 it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A space target orbit positioning system, characterized in that: Including data disassembly module, neural network module and combination module; The data decomposition module decomposes the original observation data of the target to obtain a training set; The neural network module trains the data of the training set to obtain trained data; The combination module integrates the trained data to obtain the trajectory of the target.
2. A space target orbit positioning system as claimed in claim 1, characterized in that: The data decomposition module decomposes the original observation data of the target to obtain a verification set.
3. A space target orbit positioning system as claimed in claim 2, characterized in that: The system also includes a deduction module, which performs matching verification on the trained data based on the verification set. If the trained data matches well, the trained data is retained. If the trained data does not match well, gradient update and back propagation are performed, and the neural network module restarts training.
4. A space target orbit positioning system as claimed in claim 3, characterized in that: The matching verification specifically sets a preset matching degree, and compares the matching degree of the training data matched by the deduction module based on the verification set with the preset matching degree. If the matching degree is greater than the preset matching degree, the training data is qualified, otherwise it is unqualified.
5. A space target orbit positioning system as claimed in claim 1, characterized in that: The neural network module includes a GRU network and an FNN network. The neural network module trains the data of the training set through the GRU network, the FNN network or a combination of the GRU network and the FNN network.
6. A space target orbit positioning system as claimed in claim 1, characterized in that: The data decomposition module decomposes the original observation data of the target into three single-axis data columns of x, y, and z, and obtains a training set based on the data columns.
7. A method for locating a space target orbit, characterized in that: The steps include: Step 1: Disassemble the original observation data of the target to obtain the training set and the validation set; Step 2: Train the training set data to obtain trained data; Step 3: Verify the trained data based on the validation set. If the trained data matches the validation set, keep the trained data and proceed to step 4. If the trained data does not match the validation set, return to step 2. Step 4: Integrate the trained data and output the coordinate value of the target.
8. An electronic device, characterized in that: including memory and processor; Memory for storing computer programs; A processor is used to execute the computer program, and when the computer program is executed by the processor, the steps of a space target orbit positioning method as described in claim 7 are implemented.
9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the space target orbit positioning method as claimed in claim 7 are implemented.