Methods, apparatus, equipment, media and products for analyzing the error of two-line track elements
By introducing a sliding window and a prediction-correction path tracking method into TLE data, a residual objective function and homotopy equation are constructed, which solves the problem of traditional differential correction being sensitive to initial values, realizes robust improvement and error quantification of TLE data, and improves the accuracy of orbit prediction.
Patent Information
- Application Number
- CN202611610120.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-19
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional differential correction methods are highly sensitive to the initial values of the two-line orbital elements (TLE), and are prone to getting trapped in local optima or diverging, which affects the accuracy of orbit prediction.
By employing a sliding window sampling and prediction-correction path tracing method, and constructing a residual objective function and homotopy equation, the variables to be optimized are iteratively optimized to achieve robust improvement of TLE data.
It overcomes the shortcomings of traditional methods that are sensitive to initial values, and can improve single-epoch root number and error quantification by relying solely on historical TLE data, thereby improving the accuracy of orbit prediction.
Smart Images

Figure CN123115574A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of space target monitoring and orbital mechanics technology, and in particular to a method, apparatus, equipment, medium and products for analyzing the error of two orbital elements. Background Technology
[0002] Two-line orbital elements (TLE) is currently the most widely available, open-source, regularly updated orbital dataset covering the largest number of space targets. TLE data is used to obtain the predicted orbital status of space targets with low computational cost and is widely applied in fields such as space target collision risk assessment, reentry prediction, and orbital evolution analysis.
[0003] However, TLE data itself has inherent limitations: (1) The generation process of TLE depends on observation data, and its observation error, model error, and truncation error are not clearly quantified, and the TLE document does not provide any error information; (2) The nominal accuracy of TLE varies significantly with target type, observation geometry, and epoch differences, and TLEs of the same target at different epochs may contain inconsistent systematic errors; (3) Since TLE uses a simplified SGP4 model, model errors accumulate over time during orbit prediction, further affecting prediction accuracy. The above problems severely restrict the direct use of TLE in high-precision space missions.
[0004] For determining the Time-Like Error (TLE), traditional differential correction techniques optimize locally by minimizing the residuals. However, these techniques are highly sensitive to the initial root numbers and are prone to getting trapped in local optima or diverging. Therefore, there is an urgent need for a two-line orbital element error analysis method that is insensitive to initial conditions and can converge stably. Summary of the Invention
[0005] This invention provides a method, apparatus, equipment, medium, and product for analyzing the error of two-line track elements. It solves the problem that traditional differential correction methods are highly sensitive to initial element counts and prone to getting trapped in local optima or diverging. The technical solution is as follows: On the one hand, a method for analyzing the error of two-line track elements is provided, the method comprising: Obtain the original TLE sequence to be analyzed, and use a sliding window to perform sliding sampling with a step size of 1; For each sliding window of the sampling, execute: Using the SGP4 orbit prediction model, the first position vector and first velocity vector of each subsequent observation epoch within the sliding window are predicted based on the variables to be optimized in the first TLE data within the sliding window. The second position vector and second velocity vector of each observation epoch are then constructed by directly converting the original TLE data of each subsequent observation epoch within the sliding window. A residual objective function is then constructed based on the residual objective function, and a homotopy equation is constructed accordingly. The variables to be optimized include at least the orbital inclination, right ascension of the ascending node, orbital eccentricity, argument of perigee, mean perigee angle, and average rate of motion of the orbit. The homotopy equation is solved iteratively using the prediction-correction path tracking method. The original TLE data of subsequent observation epochs within the sliding window are used to continuously optimize the variable to be optimized in the first observation epoch within the sliding window, so as to obtain the optimal variable after improvement in the first observation epoch. After improving the original TLE sequence, the position vectors are directly converted based on the original TLE sequence and the improved TLE sequence respectively, in order to perform statistical analysis of the position error distribution.
[0006] On the other hand, a two-line track element error analysis device is provided, the device comprising: The sliding unit is used to acquire the original TLE sequence to be analyzed, and the sliding window is used to perform sliding sampling with a step size of 1. A defined unit is used to perform the following for each sampling sliding window: using the SGP4 orbit prediction model, based on the variables to be optimized in the first TLE data within the sliding window, predict the first position vector and the first velocity vector for each subsequent observation epoch within the sliding window; combine this with the second position vector and the second velocity vector for each observation epoch directly converted from the original TLE data of each subsequent observation epoch within the sliding window; construct a residual objective function; and construct a homotopy equation based on the residual objective function; wherein the variables to be optimized include at least the orbital inclination, right ascension of the ascending node, orbital eccentricity, argument of perigee, mean perigee angle, and average rate of motion of the orbit. An improvement unit is used to iteratively solve the homotopy equation using a prediction-correction path tracking method, so as to continuously optimize the variable to be optimized in the first observation epoch of the sliding window using the original TLE data of subsequent observation epochs within the sliding window, and obtain the improved optimal variable for the first observation epoch. The analysis unit is used to directly convert the position vector based on the original TLE sequence and the improved TLE sequence after the improvement of the original TLE sequence is completed, so as to perform statistical analysis of the position error distribution.
[0007] On the other hand, a computer device is provided, the computer device including a memory and a processor, the memory for storing a computer program, and the processor for executing the computer program stored in the memory to implement the steps of the method described above.
[0008] On the other hand, a computer-readable storage medium is provided, wherein a computer program is stored therein, and when the computer program is executed by a processor, it implements the steps of the method described above.
[0009] On the other hand, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described above.
[0010] The technical solution provided by this invention can bring at least the following beneficial effects: A homotopy path tracking framework is introduced to make the iterative process of traditional differential correction continuous, thereby achieving robust improvement on TLE data and constructing error samples to complete statistical characteristic analysis. It overcomes the defect of traditional differential correction being sensitive to initial values and can achieve single-epoch root improvement and error quantification by relying solely on historical TLE. Attached Figure Description
[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0012] Figure 1 This is a flowchart of a two-line track element error analysis method provided by an embodiment of the present invention; Figure 2 This is a normal probability diagram of the improved TLE sequence R-direction position error provided in an embodiment of the present invention; Figure 3 This is a normal probability diagram of the T-direction position error of the improved TLE sequence provided in an embodiment of the present invention; Figure 4 This is a normal probability diagram of the N-direction position error of the improved TLE sequence provided in an embodiment of the present invention; Figure 5 This is a statistical result diagram of forecast error provided by an embodiment of the present invention; Figure 6 This is a structural diagram of a spatial pointing two-line orbital element error analysis device provided in an embodiment of the present invention; Figure 7 This is a hardware architecture diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation
[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0014] The specific implementation of the above concept is described below.
[0015] Please refer to Figure 1 The present invention provides a method for analyzing the error of two-line track elements, the method comprising: Step 100: Obtain the original TLE sequence to be analyzed, and perform sliding sampling with a step size of 1 using a sliding window; Step 102: For each sampling sliding window, the following steps are performed: Using the SGP4 orbit prediction model, the first position vector and the first velocity vector of each subsequent observation epoch within the sliding window are predicted based on the variables to be optimized in the first TLE data within the sliding window. The second position vector and the second velocity vector of each observation epoch are then constructed by directly converting the original TLE data of each subsequent observation epoch within the sliding window. A residual objective function is then constructed based on the residual objective function. The variables to be optimized include at least the orbital inclination, right ascension of the ascending node, orbital eccentricity, argument of perigee, mean perigee angle, and average rate of motion of the orbit. Step 104: The homotopy equation is solved iteratively using the prediction-correction path tracking method. The original TLE data of subsequent observation epochs within the sliding window are used to continuously optimize the variable to be optimized in the first observation epoch within the sliding window, so as to obtain the optimal variable after improvement in the first observation epoch. Step 106: After improving the original TLE sequence, the position vectors are directly converted based on the original TLE sequence and the improved TLE sequence respectively, in order to perform statistical analysis of the position error distribution.
[0016] In this embodiment of the invention, a homotopy path tracking framework is introduced to make the iterative process of traditional differential correction continuous, thereby achieving robust improvement of TLE data and constructing error samples to complete statistical characteristic analysis; it overcomes the defect of traditional differential correction being sensitive to initial values, and can achieve single-epoch root number improvement and error quantification by relying only on historical TLE.
[0017] The following description Figure 1 The execution method of each step is shown.
[0018] For step 100: Download the target TLE historical data from the public source, and remove redundant TLEs and BLEs with epoch intervals less than half an orbital period. Parameter (B) The ballistic coefficient (a fitting parameter used to describe the effect of atmospheric drag in TLE data) is less than zero, indicating anomalies in TLEs. The original TLE sequence to be analyzed is obtained. A sliding window length N is set. The window slides with a step size of 1. The first TLE data point within each window is the target for improvement, and the remaining TLEs are the observed samples. By sliding sequentially with a step size of 1, improvements are achieved for each TLE data point.
[0019] Regarding step 102: In some implementations, a 6-dimensional optimization variable is defined for the TLE data. ,in For track inclination, Right ascension of the ascending node For orbital eccentricity, For the perigee argument, For the near point angle, The average rate of motion of the orbital motion.
[0020] Using the 6-dimensional variable to be optimized from the first TLE within the window, the SGP4 orbital prediction model is used to calculate the first position vector and the first velocity vector for each subsequent observation epoch (epochs from the 2nd to the Nth TLE within the window). The TLE of subsequent observation epochs (the 2nd to Nth TLE epochs in the window) are directly converted to obtain the second position vector and the second velocity vector. Using the 2nd to Nth TLE data points in the window as observation samples, construct the objective function for the forecast residual: In the formula, Let x be the residual objective function, x be the vector of variables to be optimized from the first TLE data within the sliding window, and N be the length of the sliding window. and To use the SGP4 orbit prediction model, based on the variables to be optimized in the first TLE data within this sliding window, predict the first position vector and the first velocity vector for the j-th subsequent observation epoch within this sliding window. and The second position vector and the second velocity vector are obtained by directly converting the raw TLE data of the j-th observation epoch. The homotopy equation is: In the formula, It is a homotopy equation. Let t be the gradient and t be the homotopy parameter. This is the original vector composed of the original variables to be optimized in the first TLE data within the sliding window.
[0021] It is understandable that in addition to the 6-dimensional variables to be optimized, there is other data in the TLE data. The SGP4 orbit prediction model needs to use the variables to be optimized in the first TLE data for prediction. Therefore, the path tracking method is mainly used to iteratively optimize the variables to be optimized in the first TLE data, thereby achieving the goal of improving the TLE data.
[0022] Regarding step 104: In some implementations, step 104 may include the following steps S1-S7: S1, take the original variable vector to be optimized in the first observation epoch within the sliding window as the starting point of the path, calculate the initial gradient vector of the starting point of the path using Richardson extrapolation, set the initial value of the homotopy parameter to 0, and set the initial homotopy step size and the initial BFGS inverse matrix.
[0023] In this step, the initial value of the homotopy parameter is set. The starting point of the path is the original variable vector to be optimized at the first observation epoch within the sliding window. Calculate the initial gradient vector. The initial gradient vector is solved using Richardson extrapolation (a numerical method that eliminates low-order terms of truncation error by calculating differences at different gradient lengths and performing linear extrapolation on the results, thus achieving higher gradient accuracy than single-difference). The BFGS inverse matrix is initialized as an identity matrix. At the same time, save the current point as an old point: , Set the initial homotopy step size. And specify the allowable range of homotopy step size. Set the homotopy residual tolerance. .
[0024] In some implementations, Richardson extrapolation computes the gradient vector as follows: For each dimension of the variable at the current path point, the gradient corresponding to the current dimension is calculated using the following formula, so that a gradient vector can be formed using the gradients corresponding to each dimension: in, In the formula, Let be the gradient of the i-th variable, and h be the numerical difference step size. Unit coordinate vector, Add the i-th dimension to the current path point x The objective function of the residual calculated after perturbation. and It is an intermediate variable and has no real meaning.
[0025] It is understandable that when the initial gradient vector is calculated using the Richardson extrapolation method described above, it is calculated based on the original vector of variables to be optimized. Add a perturbation to each dimension, for example, in the formula The calculation is performed using the SGP4 orbit prediction model based on the original vector of variables to be optimized from the first TLE data within the sliding window. Add perturbation to the i-th dimension Then, predict the first position vector and the first velocity vector for each subsequent observation epoch within the sliding window. Use the second position vector and the second velocity vector for each observation epoch, obtained directly from the original TLE data of each subsequent observation epoch within the sliding window, as the true values to calculate the residual objective function. The value of . , and The calculation is similar. Therefore, the gradient of each dimension of the initial gradient vector can be calculated, i.e. This is how the initial gradient vector is obtained. .
[0026] S2 is executed for each iteration; the current path point and current homotopy parameter of the iteration are obtained, the derivative of the homotopy equation with respect to t is obtained to obtain the tangential direction expression of the implicit function, and the inverse of the Hessian matrix in the tangential direction expression is replaced by the inverse of the current BFGS matrix to obtain the prediction direction, so as to predict the new path point and new homotopy parameter based on the prediction direction and the current homotopy step size respectively; the path point is the current variable vector to be optimized in the first observation epoch within the sliding window.
[0027] In some implementations, new waypoints are predicted in the following way: Get the current path point in this iteration and current homotopy parameters Hotopism equation Implicit functions are defined Differentiating the current homotopy equation with respect to t, we obtain the expression for the tangential direction of the implicit function: In the formula, For gradient, The objective function is the residual corresponding to the current path point. The starting point of the path; Due to the inverse of the Hessian matrix Since it cannot be obtained directly, the inverse matrix of the current BFGS is used. The inverse of the Hessian matrix in the expression for the tangential direction The predicted direction is obtained: In the formula, This is the initial gradient vector; Predicting new waypoints New homotopy parameters : In the formula, This is the current homotopy step size.
[0028] S3. The Trust-Region-Dogleg algorithm is used to iteratively correct the new path points to obtain the correction points.
[0029] In some implementations, the Trust-Region-Dogleg algorithm is used to iteratively correct new path points to obtain correction points, including: The homotopy parameter is fixed. It is necessary to start from the new path point and correct back to the homotopy surface. Define auxiliary functions above. : In the formula, For gradient, Let the residual objective function be... For the new homotopy parameters, This is the initial gradient vector; The Trust-Region-Dogleg algorithm is used to process the equations. Perform iterative solutions to obtain new path points. Iterate through the initial value to obtain the correction point. .
[0030] In this step, the equation is solved. This is equivalent to finding the corresponding point on the homotopy surface. The Trust-Region-Dogleg algorithm is used for iterative solution. Iterate using the initial value to obtain the correction point for that iteration in the first TLE data within the sliding window. Calibration point It also provides a 6-dimensional optimization variable vector. The Trust-Region-Dogleg algorithm is a nonlinear optimization algorithm that combines the advantages of the steepest descent method and Newton's method. It exhibits good robustness in iteratively solving problems when the initial guess deviates significantly from the true value.
[0031] S4. The gradient vector of the correction point is calculated using Richardson extrapolation, and the homotopy residual is evaluated. The decision on whether to accept the correction point as the next path point is based on whether the homotopy residual is less than or equal to the set homotopy residual tolerance.
[0032] In this step, the gradient vector of the correction point is calculated using the Richardson extrapolation method in the same way as the initial gradient vector is calculated in step S1.
[0033] It is understandable that when the gradient vector of the correction point is calculated using the Richardson extrapolation method described above, it is calculated based on the correction point. Add a perturbation to each dimension, for example, in the formula The calculation is for the correction point Add perturbation to the i-th dimension Using the SGP4 orbit prediction model, the correction point with perturbation added is based on the i-th dimension corresponding to the first TLE data within the sliding window. Predict the first position vector and first velocity vector for each subsequent observation epoch within the sliding window. Then, use the second position vector and second velocity vector for each observation epoch, obtained directly from the original TLE data of each subsequent observation epoch within the sliding window, as the true values to calculate the residual objective function. The value of . , and The calculation is similar. Therefore, the gradient of each dimension of the gradient vector at the correction point can be calculated, i.e. This is used to obtain the gradient vector of the correction point. .
[0034] Next, evaluate the homotopy residuals: If homotopic residuals This indicates that the correction point is sufficiently close to the homotopy surface, so the point is accepted, and the update operation of S5 is then performed.
[0035] S5, if accepted, the correction point is used as the next path point, the homotopy parameter, gradient vector and homotopy step size corresponding to the next path point are updated, and the step size vector and gradient difference vector are calculated to update the BFGS inverse matrix, so as to perform the next iteration based on the updated path point.
[0036] First, update the path status, making , Current gradient The correction point is used as the next path point, and the new homotopy parameter is updated to the homotopy parameter corresponding to the next path point. The gradient vector of the correction point is then used. This serves as the gradient vector updated for the next path point.
[0037] The method for updating the homotopy step size during acceptance is as follows: In the formula, For the current homotopy step size, This is the upper limit of the allowed range of homotopy step size. Based on the current formula, the step size is increased to track longer path segments and speed up the iteration.
[0038] Next, the BFGS inverse matrix is updated. The step size vector is calculated. and gradient difference vector That's understandable. and This is the old point information saved from the previous iteration.
[0039] Curvature test conditions: In the formula, Represents transpose. This represents the 2-norm.
[0040] If satisfied, then execute the BFGS inverse matrix update formula: in, , It is an intermediate variable with no real meaning.
[0041] If the curvature condition is not met, skip the BFGS update and maintain... .
[0042] Finally, save the repair information and old data. , This is used for updating the inverse matrix in the next iteration of BFGS.
[0043] It is understandable that after the update is completed, the correction point will be used as the next path point, and the process will jump to step S2 for the next iteration.
[0044] S6: If not accepted, the path point is not updated, but the homotopy step size is reduced for the next iteration.
[0045] like If the current correction point is rejected, then the homotopy step size is considered to be corrected. If the value is too large, the prediction will deviate too far. In this case, the path points are not updated, but the homotopy step size is reduced and the process jumps back to step S2.
[0046] The method to reduce the homotopy step size is as follows: By halving the homotopy step size, the tracking path is narrowed, allowing for a fresh prediction iteration, thus avoiding issues caused by the homotopy step size. Too large an amount will cause the prediction to deviate too far.
[0047] S7, until the homotopy parameter reaches 1, complete the path tracking, and use the path point accepted in the last iteration as the improved optimal variable vector of the first observation epoch within the sliding window.
[0048] In this step, if Then the step size will be truncated to Ensure that the homotopy parameter reaches exactly t=1 in the last iteration. The main loop continues to execute steps S2 to S6 until the homotopy parameter t reaches 1.0. At this point, path tracing is complete, and the path point accepted in the last iteration is... That is, the improved optimal variable vector for the first observation epoch within the sliding window. If the homotopy step size falls below the lower limit after multiple reductions. The algorithm determines that the path is broken, issues a warning, and terminates early. At this point, it returns the last valid path point as the optimal variable vector. .
[0049] In summary, this prediction-correction path tracking method can overcome the shortcomings of traditional differential corrections that are sensitive to initial values, and can achieve single-epoch root improvement and error quantification by relying solely on historical TLE.
[0050] Regarding step 106: In this step, the original variables to be optimized are sequentially improved for the first epoch within the sliding window to obtain the improved TLE sequence. The position vector directly converted from the original TLE is subtracted from the position vector directly converted from the improved TLE to obtain the position error in the RTN coordinate system for that epoch (R: radial, along the geocentric radius vector of the space target; T: lateral, along the velocity direction; N: normal, perpendicular to the orbital plane). The sample mean and standard deviation of all position errors are calculated, and a normal probability plot is plotted to perform a normality test, completing the statistical analysis of the position error distribution. The improved TLE sequence and error statistical characteristics are output for subsequent applications.
[0051] To verify the effectiveness of the present invention, error analysis was performed on the TLE data of the COSMOS 2251 disintegration fragments.
[0052] Fragments from the COSMOS 2251 satellite (SSN 34323) were selected, and TLE sequences from March 2009 to March 2011 were downloaded. After preprocessing, 980 valid TLEs were obtained. With a sliding window of N=6, 975 TLE data points were obtained that require improvement. Parameter initialization: , , Homotopy residual threshold Numerical difference step size The average iteration path takes approximately 50 steps, with no breakage or exit. After improvement, error statistical analysis was performed in the RTN coordinate system, as shown below. Figure 2 , 3 The normal probability plots of the improved TLE sequence's position errors in the R, T, and N directions, shown in Figure 4, reveal that the position error distributions in these directions approximately conform to a normal distribution: the mean position error in the R direction is 62.15m, with a standard deviation of 128.09m; the mean position error in the T direction is -68.43m, with a standard deviation of 1438.83m; and the mean position error in the N direction is -152.13m, with a standard deviation of 137.52m. The standard deviation of the T-direction error is significantly higher than that in the R and N directions, consistent with the trajectory determination error distribution pattern in trajectory determination theory.
[0053] Using all improved TLE data and their corresponding original data as benchmarks, the SGP4 orbital prediction model was extrapolated to the epochs of the subsequent 11 TLE data points. Position prediction errors for both the original and improved TLE data were calculated using the TLE ephemeris as a reference. The resulting prediction errors were then binned at equal time intervals according to the prediction time, and the mean and standard deviation of the prediction errors for each bin were statistically analyzed. The statistical results are shown in the attached figure. Figure 5 As shown, using the improved TLE data for orbit prediction can significantly reduce prediction errors, proving the effectiveness of the algorithm disclosed in this invention.
[0054] Please refer to Figure 6 This invention provides a two-line track element error analysis device, which includes: The sliding unit 601 is used to acquire the original TLE sequence to be analyzed and to perform sliding sampling with a step size of 1 using a sliding window; Unit 602 is defined to perform the following for each sampling sliding window: using the SGP4 orbit prediction model, based on the variables to be optimized in the first TLE data within the sliding window, predict the first position vector and the first velocity vector for each subsequent observation epoch within the sliding window; combine this with the second position vector and the second velocity vector for each observation epoch directly converted from the original TLE data within the sliding window; construct a residual objective function; and construct a homotopy equation based on the residual objective function; wherein the variables to be optimized include at least the orbital inclination, right ascension of the ascending node, orbital eccentricity, argument of perigee, mean perigee angle, and average rate of motion of the orbit. Improvement unit 603 is used to iteratively solve the homotopy equation using the prediction-correction path tracking method, so as to continuously optimize the variable to be optimized in the first observation epoch of the sliding window using the original TLE data of subsequent observation epochs in the sliding window, and obtain the improved optimal variable of the first observation epoch. The analysis unit 604 is used to directly convert the position vector based on the original TLE sequence and the improved TLE sequence after the improvement of the original TLE sequence is completed, so as to perform statistical analysis of the position error distribution.
[0055] It should be noted that the two-line track element error analysis device provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the above device embodiments and method embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.
[0056] Embodiments of this application also provide a computer device, please refer to... Figure 7 The computer device includes a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the at least one instruction, at least one program, code set or instruction set being loaded and executed by the processor to implement the two-line orbital element error analysis method provided in the above method embodiments.
[0057] Embodiments of this application also provide a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the two-line orbital element error analysis method provided in the above-described method embodiments.
[0058] Embodiments of this application also provide a computer program product, which includes a computer program. A processor of a computer device reads the computer program from a computer-readable storage medium and executes the computer program, causing the computer device to perform any of the two-line track element error analysis methods described in the above embodiments.
[0059] For ease of description, the above systems or devices are described separately as various modules or units based on their functions. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware components.
[0060] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, 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 storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.
[0061] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0062] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.
Claims
1. A method for analyzing the error of two-line track elements, characterized in that, include: Obtain the original TLE sequence to be analyzed, and use a sliding window to perform sliding sampling with a step size of 1; For each sliding window of the sampling, execute: Using the SGP4 orbit prediction model, the first position vector and first velocity vector of each subsequent observation epoch within the sliding window are predicted based on the variables to be optimized in the first TLE data within the sliding window. The second position vector and second velocity vector of each observation epoch are then constructed by directly converting the original TLE data of each subsequent observation epoch within the sliding window. A residual objective function is then constructed based on the residual objective function, and a homotopy equation is constructed accordingly. The variables to be optimized include at least the orbital inclination, right ascension of the ascending node, orbital eccentricity, argument of perigee, mean perigee angle, and average rate of motion of the orbit. The homotopy equation is solved iteratively using the prediction-correction path tracking method. The original TLE data of subsequent observation epochs within the sliding window are used to continuously optimize the variable to be optimized in the first observation epoch within the sliding window, so as to obtain the optimal variable after improvement in the first observation epoch. After improving the original TLE sequence, the position vectors are directly converted based on the original TLE sequence and the improved TLE sequence respectively, in order to perform statistical analysis of the position error distribution.
2. The method as described in claim 1, characterized in that, The objective function for the residual is: In the formula, Let x be the residual objective function, x be the vector of variables to be optimized from the first TLE data within the sliding window, and N be the length of the sliding window. and To use the SGP4 orbit prediction model, based on the variables to be optimized in the first TLE data within this sliding window, predict the first position vector and the first velocity vector for the j-th subsequent observation epoch within this sliding window. and The second position vector and the second velocity vector are obtained by directly converting the original TLE data of the j-th observation epoch. The homotopy equation is: In the formula, It is a homotopy equation. Let t be the gradient and t be the homotopy parameter. This is the original vector composed of the original variables to be optimized in the first TLE data within the sliding window.
3. The method as described in claim 1, characterized in that, The method employs a prediction-correction path tracking approach to iteratively solve the homotopy equation. This involves using the raw TLE data from subsequent observation epochs within the sliding window to continuously optimize the variable to be optimized for the first observation epoch within that window, resulting in the improved optimal variable for the first observation epoch. This optimal variable includes: The original variable vector to be optimized in the first observation epoch within the sliding window is taken as the starting point of the path. The initial gradient vector of the starting point of the path is calculated using the Richardson extrapolation method. The initial value of the homotopy parameter is set to 0, and the initial homotopy step size and the initial BFGS inverse matrix are set. Execute for each iteration; Obtain the current path point and current homotopy parameters for this iteration, differentiate the homotopy equation with respect to t to obtain the tangential direction expression of the implicit function, and use the current BFGS inverse matrix to replace the inverse of the Hessian matrix in the tangential direction expression to obtain the predicted direction. Based on the predicted direction and the current homotopy step size, predict the new path point and new homotopy parameters respectively; the path point is the current vector of variables to be optimized in the first observation epoch within the sliding window. The Trust-Region-Dogleg algorithm is used to iteratively correct the new path points to obtain the correction points. The gradient vector of the correction point is calculated using Richardson extrapolation, and the homotopy residual is evaluated. The decision on whether to accept the correction point as the next path point is based on whether the homotopy residual is less than or equal to the set homotopy residual tolerance. If accepted, the correction point is used as the next path point. The homotopy parameter, gradient vector and homotopy step size corresponding to the next path point are updated. The step size vector and gradient difference vector are calculated to update the BFGS inverse matrix, so as to perform the next iteration based on the updated path point. If the path point is not accepted, the homotopy step size is reduced to proceed to the next iteration instead of updating the path point. The path tracking is completed when the homotopy parameter reaches 1, and the path point accepted in the last iteration is used as the improved optimal variable vector for the first observation epoch within the sliding window.
4. The method as described in claim 3, characterized in that, Richardson extrapolation calculates the gradient vector as follows: For each dimension of the variable at the current path point, the gradient corresponding to the current dimension is calculated using the following formula, so that a gradient vector can be formed using the gradients corresponding to each dimension: in, In the formula, Let be the gradient of the i-th variable, and h be the numerical difference step size. Unit coordinate vector, Add to the i-th dimension of the current path point x The objective function of the residual calculated after perturbation. and It is an intermediate variable and has no real meaning.
5. The method as described in claim 3, characterized in that, The new waypoints are predicted in the following way: Get the current path point in this iteration and current homotopy parameters Hotopism equation Implicit functions are defined Differentiating the current homotopy equation with respect to t, we obtain the expression for the tangential direction of the implicit function: In the formula, For gradient, The objective function is the residual corresponding to the current path point. The starting point of the path; Using the current BFGS inverse matrix The inverse of the Hessian matrix in the tangential direction expression is used instead. The predicted direction is obtained: In the formula, The initial gradient vector; Predicting new waypoints New homotopy parameters : In the formula, This is the current homotopy step size.
6. The method as described in claim 3, characterized in that, The Trust-Region-Dogleg algorithm is used to iteratively correct the new path points to obtain correction points, including: Define auxiliary functions : In the formula, For gradient, Let the residual objective function be... For the new homotopy parameters, The initial gradient vector; The Trust-Region-Dogleg algorithm is used to process the equations. Perform iterative solutions to obtain new path points. Iterate through the initial value to obtain the correction point. .
7. A two-track element error analysis device, used to implement the method described in any one of claims 1-6, characterized in that, The device includes: The sliding unit is used to acquire the original TLE sequence to be analyzed, and the sliding window is used to perform sliding sampling with a step size of 1. A defined unit is used to perform the following for each sampling sliding window: using the SGP4 orbit prediction model, based on the variables to be optimized in the first TLE data within the sliding window, predict the first position vector and the first velocity vector for each subsequent observation epoch within the sliding window; combine this with the second position vector and the second velocity vector for each observation epoch directly converted from the original TLE data of each subsequent observation epoch within the sliding window; construct a residual objective function; and construct a homotopy equation based on the residual objective function; wherein the variables to be optimized include at least the orbital inclination, right ascension of the ascending node, orbital eccentricity, argument of perigee, mean perigee angle, and average rate of motion of the orbit. An improvement unit is used to iteratively solve the homotopy equation using a prediction-correction path tracking method, so as to continuously optimize the variable to be optimized in the first observation epoch of the sliding window using the original TLE data of subsequent observation epochs within the sliding window, and obtain the improved optimal variable for the first observation epoch. The analysis unit is used to directly convert the position vector based on the original TLE sequence and the improved TLE sequence after the improvement of the original TLE sequence is completed, so as to perform statistical analysis of the position error distribution.
8. A computer device, characterized in that, The computer device includes a memory and a processor. The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to implement the steps of the method according to any one of claims 1-6.
9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the method described in any one of claims 1-6.
10. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1-6.
