Space-based angle-measurement-only target orbit determination method for compensating orbit position error

Abstract

The invention discloses a space-based angle-measurement-only target orbit determination method for compensating orbit position errors, and belongs to the technical field of satellite application measurement. The method comprises the following steps: establishing an error model of an empirical orbit dynamics model error introduction measurement error according to a real orbit dynamics model of a target; training the initial neural network according to a preset training sample to obtain an orbit position error estimation model; wherein the training sample is determined through historical orbit data and an empirical orbit dynamics model; and according to the orbit position error estimation model, compensating the error model, and processing the initial state of a to-be-measured target by using a deviation compensation least square method to obtain an optimal estimation value of the orbit of the to-be-measured target. The method can effectively improve the determination precision of the target track.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite application measurement, and particularly relates to a space-based angle-only target orbit determination method for compensating orbit position errors. Background Art

[0002] The space-based observation platform is equipped with an optical camera, and by utilizing the characteristics of small volume and low power consumption and long detection distance of the optical camera, it can realize the long-distance discovery, positioning and tracking of space targets, and improve the perception ability of the space situation. When the distance between the observation platform and the space target is far, the observation platform can only obtain the relative angle measurement data of the space target, and it is necessary to study the space target orbit determination method under the relative angle measurement data.

[0003] In the related art, when estimating the target orbit, it is necessary to accumulate measurement data for a period of time to make the target orbit determination result converge. Random errors and systematic errors will occur in this process, and the systematic error often introduces a constant deviation into the orbit determination result, resulting in a decrease in the accuracy of the space target orbit determination.

[0004] Based on this, there is an urgent need for a space-based angle-only target orbit determination method for compensating orbit position errors to solve the above technical problems. Summary of the Invention

[0005] The present invention provides a space-based angle-only target orbit determination method for compensating orbit position errors, which can solve the problem of compensating the orbit position estimation error under long-term cumulative observation in the related art. The technical solution is as follows:

[0006] On the one hand, a space-based angle-only target orbit determination method for compensating orbit position errors is provided, and the method includes:

[0007] According to the true orbit dynamics model of the target, establish an error model in which the empirical orbit dynamics model error introduces measurement errors;

[0008] Train the initial neural network according to the preset training samples to obtain an orbit position error estimation model; wherein the training samples are determined by historical orbit data and the empirical orbit dynamics model;

[0009] According to the orbit position error estimation model, compensate the error model, and use the deviation compensation least squares method to process the initial state of the target to be measured to obtain the optimal estimated value of the orbit of the target to be measured.

[0010] On the other hand, a space-based angle-only target orbit determination device for compensating orbit position errors is provided, and the device includes:

[0011] A modeling module, configured to establish an error model in which the error introduced by the empirical orbit dynamics model into the measurement error is based on the true orbit dynamics model of the target;

[0012] A training module, configured to train an initial neural network according to preset training samples to obtain an orbit position error estimation model; wherein the training samples are determined by historical orbit data and an empirical orbit dynamics model;

[0013] A processing module, configured to compensate the error model according to the orbit position error estimation model, and process the initial state of the target to be measured by using the deviation compensation least squares method to obtain the optimal estimated value of the orbit of the target to be measured.

[0014] On the other hand, a computer device is provided, where the computer device includes a memory and a processor. The memory is used to store a computer program, and the processor is used to execute the computer program stored on the memory to implement the steps of the above-mentioned space-based angle-only measurement target orbit determination method for compensating orbit position errors.

[0015] On the other hand, a computer-readable storage medium is provided, where a computer program is stored in the storage medium, and when the computer program is executed by a processor, the steps of the above-mentioned space-based angle-only measurement target orbit determination method for compensating orbit position errors are implemented.

[0016] On the other hand, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned space-based angle-only measurement target orbit determination method for compensating orbit position errors are implemented.

[0017] The technical solution provided by the present invention can at least bring the following beneficial effects: First, an error model between the empirical orbit dynamics model error and the measurement error is established, and an initial error estimation network based on deep learning is designed; then the initial error estimation network is trained by preset training samples to obtain an orbit position error estimation model. Finally, a hybrid dynamics model combining the orbit position error estimation model and the empirical orbit dynamics model is proposed, and a deviation compensation least squares orbit determination method is designed to determine the optimal estimated value of the target orbit. Aiming at the problem that there is an error between the target empirical orbit dynamics model and the target true motion model, which "pulls off" the estimated value of the target initial state, this method proposes to calculate the target orbit by using a hybrid dynamics model of "empirical orbit dynamics model + neural network error estimation", solves the problem of compensating the orbit position estimation error under long-term cumulative observation when determining the target orbit by using relative angle measurement data, and improves the accuracy of target orbit determination. Description of the Drawings

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0019] Figure 1 It is a flowchart of a space-based angle-only target orbit determination method for compensating orbit position error provided by an embodiment of the present invention;

[0020] Figure 2 It is a schematic diagram of error-offset orbit estimation provided by an embodiment of the present invention;

[0021] Figure 3 It is a schematic diagram of an initial neural network structure provided by an embodiment of the present invention;

[0022] Figure 4 It is a flowchart of a least squares orbit determination method for deviation compensation provided by an embodiment of the present invention;

[0023] Figure 5 It is a schematic diagram of the composition of a digital simulation environment provided by an embodiment of the present invention;

[0024] Figure 6 It is a structural diagram of a space-based angle-only target orbit determination device for compensating orbit position error provided by an embodiment of the present invention;

[0025] Figure 7 It is a hardware architecture diagram of a computer device provided by an embodiment of the present invention. Detailed implementation manners

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0027] As mentioned above, when using relative angle measurement data to determine the orbit position of a target, it is necessary to accumulate measurement data for a period of time to make the target orbit determination result converge. Random errors and systematic errors will occur in this process, and it is difficult to establish an accurate mathematical model for systematic errors.

[0028] Based on this, the concept of the present invention is to replace the empirical orbit dynamics model with a hybrid dynamics model of "empirical orbit dynamics model + neural network error compensation model" to achieve system error compensation, thereby improving the accuracy of space target orbit determination.

[0029] The following describes the specific implementation manners of the above concept.

[0030] Please refer to Figure 1 , a space-based angle-only target orbit determination method for compensating orbit position error provided by an embodiment of the present invention, the method includes:

[0031] Step 100, according to the true orbit dynamics model of the target, establish an error model in which the error of the empirical orbit dynamics model introduces measurement error;

[0032] Step 102, train an initial neural network according to a preset training sample to obtain an orbit position error estimation model; wherein the training sample is determined by historical orbit data and an empirical orbit dynamics model;

[0033] Step 104, according to the orbit position error estimation model, compensate the error model, and process the initial state of the target to be measured by using the deviation compensation least squares method to obtain the optimal estimated value of the orbit of the target to be measured.

[0034] In the embodiment of the present invention, first, an error model of the empirical orbit dynamics model error and the measurement error is established, and an initial error estimation network based on deep learning is designed; then, the initial error estimation network is trained by a preset training sample to obtain an orbit position error estimation model. Finally, a hybrid dynamics model combining the orbit position error estimation model and the empirical orbit dynamics model is proposed, and a deviation compensation least squares orbit determination method is designed to determine the optimal estimated value of the target orbit. This method aims at the problem that there is an error between the target empirical orbit dynamics model and the target true motion model, which "pulls off" the estimated value of the target initial state, and proposes to calculate the target orbit by using a hybrid dynamics model of "empirical orbit dynamics model + neural network error estimation", improving the accuracy of target orbit determination.

[0035] The following describes Figure 1 the execution manners of each step shown in

[0036] First, for step 100, according to the true orbit dynamics model of the target, establish an error model in which the error of the empirical orbit dynamics model introduces measurement error.

[0037] When determining the true state x0 of the target orbit by using the relative angle measurement data of the target by a space-based observation platform, it is necessary to use the angle measurement data within the observation time period of the target from t0 to t N During the observation time period from t0 to tN The internal target has a true orbital motion, and correspondingly, there is true angular measurement data of the target. For the optimal state estimate value of the initial true state x0 By extrapolating through the empirical orbital dynamics model and combining with the measurement model, the estimated measurement value corresponding to the observation time can be calculated. The error between the empirical orbital dynamics model and the true orbital motion of the target gradually increases with the observation time. At this time, a constant bias will appear in the estimated measurement value, thus "pulling" the optimal state estimate value As Figure 2 shown

[0038] In the embodiment of the present invention, the establishment process of the error model includes: establishing the true orbital dynamics model of the target according to the first acceleration determined by the target through the empirical orbital dynamics model and the second acceleration determined by the empirical orbital dynamics model and the true orbital motion of the target; respectively establishing the first equation and the second equation of the target at the target time according to the state transition matrices of the true orbital dynamics model and the empirical orbital dynamics model and the true angular measurement data and the measurement equation; performing an approximate substitution process on the difference between the first equation and the second equation and substituting it into the first equation to establish the error model

[0039] Specifically, assume that the true orbital dynamics model of the target is:

[0040]

[0041] In the formula, r is the orbital position vector of the target in the inertial system, and the two dots represent its second derivative; a0 is the acceleration that can be accurately modeled and is used in the empirical orbital dynamics model; a μ is the unknown non-modelable acceleration, which is the error between the empirical orbital dynamics model and the true orbital motion of the target

[0042] The state transition matrix corresponding to this model satisfies the following differential equation:

[0043]

[0044] In the formula, I represents the identity matrix, 0 represents the 0 matrix, t k-1 ≤t≤t k-1 +τ, and the state transition matrix satisfies where Φ(t k ,t k-1 ) is the state transition matrix corresponding to a0, and Φ μ (t k ,t k-1 ) is the state transition matrix corresponding to a μ

[0045] At the target time t k ​The true state is x k , including the position r k and the velocity v k , denoted as x k = [r k ; v k , obtaining the first equation between the initial state deviation and the measurement deviation:

[0046]

[0047] In the formula, z k is the true measured angle value, h(x k ) is the measured angle value calculated according to the state x k , is the Jacobian matrix of the measurement equation calculated according to the state x k , is the shorthand of the state transition matrix from time t0 to time t k , is the shorthand of is the random error vector that ensures the equation holds, Δx0 is the initial state deviation, and the state x k is a six-dimensional vector, including the target position and the target velocity.

[0048] Similarly, for an estimated value of the initial state The state at time t k extrapolated using the empirical orbit dynamics model is including the position and the velocity denoted as Similarly, the second equation between the initial state deviation and the measurement deviation can be obtained:

[0049]

[0050] In the formula, is the estimated measurement value calculated according to the state , H k is the Jacobian matrix of the measurement equation calculated according to the state , Φ is the shorthand of the state transition matrix from time t0 to time t k , ε k , t0), and ε k is the random error vector that ensures the equation holds.

[0051] Subtract the first equation from the second equation and substitute the relationship to obtain:

[0052]

[0053] Among them, Φ μ is the state transition matrix from time t0 to time t kMoment state transition matrix Φ μ (t k , t0) shorthand

[0054] Since Generally, Δx k is a small quantity, so there is Then the above equation can be simplified to:

[0055]

[0056] Substituting this equation into the first equation gives:

[0057]

[0058] Denote s k = H k Φ(I - Φ μ )Δx0, where s k is the constant deviation introduced by the deviation between the empirical orbit dynamics model and the target's true orbit motion in the measurement, that is, the error model, which will introduce a constant deviation in the orbit determination result. When N groups of measurement values are accumulated within the time period from t0 to t N , the orbit state estimation equation containing the error model can be obtained, where the forms of vectors Y, H, S, and V are defined respectively.

[0059]

[0060] Because S is unknown and cannot be accurately modeled, in this embodiment, the historical orbit motion data of the target will be used later, and a neural network will be used to approximate S.

[0061] Then, for step 102, the initial neural network is trained according to the preset training samples to obtain an orbit position error estimation model.

[0062] In the embodiment of the present invention, the orbit position error estimation model is trained in the following manner: An initial neural network with the orbit state estimation and epoch time as the input and the orbit position error as the output is established based on deep learning; among them, the input layer of the initial neural network is the linear transfer purelin function, and the activation functions of the hidden layers are the linear transfer purelin function and the linear transfer satlin function;

[0063] According to the number of the training samples and the output of the initial neural network, a training loss function L is established:

[0064]

[0065] where M is the number of training samples, Δr i and They are the expected output and the actual output of the neural network in sequence;

[0066] Input the training samples into the initial neural network according to the pre-divided training set, validation set, and test set for training to obtain the orbit position error estimation model.

[0067] Specifically, as Figure 3 shown, Figure 3 is the structure diagram of the initial neural network provided in this embodiment, including an input layer, two hidden layers, and an output layer. The input layer has 7 neurons, each hidden layer has 14 neurons, and the output layer has 3 neurons. The activation function of the input layer and one of the hidden layers is selected as the linear transfer purelin function:

[0068] purelin(x) = x

[0069] The activation function of the other hidden layer is selected as the linear transfer satlin function with saturation characteristics:

[0070]

[0071] The 7 inputs of the neural network are the epoch time t k and the state estimation at the corresponding time The 3 outputs of the neural network are the position deviation Δr k .

[0072] In the embodiment of the present invention, the training samples are determined by historical orbit data and an empirical orbit dynamics model, including: determining multiple initial states and multiple initial epochs of the target according to the historical orbit data; wherein, the first interval of each initial epoch is at least 24 hours; according to the empirical orbit dynamics model, calculating the orbit state estimation of each initial state within a first interval at a preset second interval; calculating the position error of the target according to the orbit state estimation and the historical orbit true state, and taking the orbit state estimation and the corresponding position error as a set of data of the training samples.

[0073] For example, first, 20 initial states x 0,n (n = 1,..., 20) of the target are given, and at an interval of 24 hours, 15 initial epochs t 0,m (m = 1,…, 15) are given, from 2021 - 08 - 10 04:00:00.00 (UTC) to 2021 - 08 - 24 04:00:00.00 (UTC); then, for each initial state x 0,n and the initial epoch time t 0,m , according to the empirical orbit dynamics model, calculate the orbit state within a 24 - hour time period at an interval of 3 minutes And calculate the position deviation Δr according to the motion state x of the target true historical orbit k k where any initial state x 0,n and any initial epoch t 0,m generate a set of samples of the orbit state and position deviation sequence. One set of samples contains 480 samples, so a total of 20×15×480 samples are generated.

[0074] Furthermore, divide the determined samples into a training set, a validation set and a test set. Set any set of samples with the initial epoch of 2021-08-24 04:00:00.00 (UTC) as the test set, and randomly divide the remaining samples into a training set and a validation set according to 9:1. Then train the initial neural network according to the loss function in the above process to obtain an orbit position error estimation model, which can estimate the vector parameter S calculated in the above process.

[0075] For step 104, according to the orbit position error estimation model, compensate the error model, and use the deviation compensation least squares method to process the initial state of the target to be measured, so as to obtain the optimal estimated value of the orbit of the target to be measured.

[0076] In the embodiment of the present invention, the optimal orbit estimated value is determined through the following process: given a state estimated value x0, calculate the state of the target to be measured by using the empirical orbit dynamics model to obtain the state estimation at the target time; input the state estimation at the target time into the orbit position error estimation model, and output the position error at the target time; input the position error at the target time into the error model, and output a constant deviation; calculate the optimal orbit estimated value according to the vector parameter and the constant deviation

[0077]

[0078] Figure 4 Specifically, by combining the error model and the orbit position error estimation model calculated in the above process, a hybrid dynamics model of "empirical orbit dynamics + neural network error compensation" is obtained, and a deviation compensation least squares orbit determination method as shown in is proposed. By given the initial state of the target and the initial epoch t0, calculate the state at time t according to the empirical orbit dynamics model k and further calculate the vector parameters Y and H.

[0079] Furthermore, use the trained neural network to estimate the position deviation Δr at time t k k ​​​, and then calculate s according to the measurement equation k , and substitute it into the above formula to calculate the orbital estimated value of the target. Finally, when the calculation result meets the preset judgment, determine it as the optimal orbital estimated value.

[0080] The following uses an embodiment to verify the feasibility of the above method:

[0081] Use the digital simulation method to verify the above steps, and use the digital simulation environment as shown in Figure 5 to verify the effectiveness of the present invention. The digital simulation environment includes a system parameter initialization module, an empirical orbit dynamics model, a line-of-sight measurement simulation module, an error estimation module, a space target orbit determination module, a historical motion data excitation module, and an estimation performance analysis module. The system parameter initialization module sets the key system parameters and the orbit parameters of the observation platform and the target; the empirical orbit dynamics model is used to recursively calculate the orbits of the target and the observation platform; the line-of-sight measurement simulation module is used to simulate the line-of-sight measurement of the observation platform to the target; the error estimation module uses a trained neural network to estimate the error of the empirical orbit dynamics model; the space target orbit determination module completes the target orbit estimation according to the orbit determination method in step 5; the historical motion data excitation module is used to provide the true on-orbit motion data of the target, complete the neural network training and be used for the algorithm performance evaluation; the estimation performance analysis module is used to evaluate the performance of the method of the present invention.

[0082] The orbit parameters of the observation platform are as shown in Table 1 below. Table 1 is the orbit parameters of the observation platform;

[0083] Table 1

[0084]

[0085] The target orbit parameters are shown in Table 2. Table 2 is the target orbit parameters;

[0086] Table 2

[0087]

[0088] The starting time is 2021-08-26 00:00:00.00 (UTC), the sampling period is 60 s, and the total observation duration is 20 h. According to the mean value M of the absolute value of the three-axis position estimation error during 200 Monte Carlo shooting simulations e to evaluate the orbit estimation performance, which is defined as follows:

[0089]

[0090] In the formula, r0 = [x, y, z] is the true position of the target, is the estimated value of the position of the i-th shooting simulation.

[0091] By comparing the performance of the traditional least - squares method and the method of the present invention in estimating the target position, the absolute values of the three - axis position estimation errors of the two methods are shown in Table 3. Table 3 is a comparison of the target three - axis position estimation errors of the two methods. It can be seen that the method of the present invention is effective, better compensates for the empirical orbit dynamics model error, and improves the target orbit determination accuracy.

[0092] Table 3

[0093]

[0094] Please refer to Figure 6 , an embodiment of the present invention provides a space - based angle - only target orbit determination device for compensating orbit position errors. The device includes:

[0095] A modeling module 600, configured to establish an error model for introducing measurement errors by an empirical orbit dynamics model error according to the true orbit dynamics model of the target;

[0096] A training module 602, configured to train an initial neural network according to a preset training sample to obtain an orbit position error estimation model; wherein the training sample is determined by historical orbit data and an empirical orbit dynamics model;

[0097] A processing module 604, configured to compensate the error model according to the orbit position error estimation model, and process the initial state of the target to be measured by using the bias - compensation least - squares method to obtain the optimal estimated value of the orbit of the target to be measured.

[0098] In an embodiment of the present invention, when the modeling module 600 executes to establish an error model for introducing measurement errors by an empirical orbit dynamics model error according to the true orbit dynamics model of the target to be measured, it is specifically configured to perform the following operations:

[0099] Establish the true orbit dynamics model of the target according to the first acceleration determined by the empirical orbit dynamics model of the target and the second acceleration determined by the empirical orbit dynamics model and the true orbit motion of the target; respectively establish the first equation and the second equation of the target at the target moment according to the state transition matrices of the true orbit dynamics model and the empirical orbit dynamics model and the true angle - measurement data and the measurement equation; perform an approximate substitution process on the difference between the first equation and the second equation and substitute it into the first equation to establish the error model.

[0100] In an embodiment of the present invention, the error model is established by the following formula:

[0101]

[0102] In the formula, Y, H, S, and V are all vector parameters of the error model; is the error between the true angle measurement value and the estimated angle measurement value; N is the number of groups of measurement data within the target time period; s is the constant deviation introduced into the measurement due to the deviation between the empirical orbit dynamics model and the target true orbit motion; H N is the Jacobian matrix of the measurement equation determined according to the trajectory state; Φ is the state transition matrix from t0 to t N at the moment; Δx0 is the initial state deviation; ε N is the random error vector to ensure the equality holds.

[0103] In the embodiment of the present invention, when the training module 602 executes the training of the initial neural network according to the preset training samples to obtain the orbit position error estimation model, it is specifically used to perform the following operations:

[0104] Based on deep learning, an initial neural network is established with the orbit state estimation and epoch time as the input and the orbit position error as the output; wherein, the input layer of the initial neural network is the linear transfer purelin function, and the activation functions of the hidden layers are the linear transfer purelin function and the linear transfer satlin function;

[0105] According to the number of the training samples and the output of the initial neural network, a training loss function L is established:

[0106]

[0107] wherein, M is the number of training samples, Δr i and are the expected output and the actual output of the neural network in sequence;

[0108] The training samples are input into the initial neural network according to the pre-divided training set, validation set and test set for training to obtain the orbit position error estimation model.

[0109] In the embodiment of the present invention, the training samples are determined by historical orbit data and an empirical orbit dynamics model, including: determining a plurality of initial states and a plurality of initial epochs of the target according to the historical orbit data; wherein, the first interval of each initial epoch is at least 24 hours; according to the empirical orbit dynamics model, calculating the orbit state estimation of each initial state within a first interval at a preset second interval; calculating the position error of the target according to the orbit state estimation and the historical orbit true state, and taking the orbit state estimation and the corresponding position error as a set of data of the training samples.

[0110] In an embodiment of the present invention, when the processing module 604 executes the operation of compensating the error model according to the orbital position error estimation model and processing the initial state of the target to be measured by using the least squares method of deviation compensation to obtain the optimal estimated value of the orbit of the target to be measured, it is specifically used to perform the following operations: calculating the state of the target to be measured by using the empirical orbit dynamics model according to the preset state estimated value to obtain the state estimation at the target moment; inputting the state estimation at the target moment into the orbital position error estimation model to output the position error at the target moment; inputting the position error at the target moment into the error model to output the constant deviation; calculating the optimal estimated value of the orbit according to the vector parameter and the constant deviation

[0111]

[0112] It should be noted that: the space-based angle-only target orbit determination device for compensating the orbital position error provided in the above embodiment is only illustrated by dividing the above functional modules. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the space-based angle-only target orbit determination device for compensating the orbital position error provided in the above embodiment and the embodiment of the space-based angle-only target orbit determination method for compensating the orbital position error belong to the same concept. For the specific implementation process, please refer to the method embodiment, which will not be elaborated here

[0113] An embodiment of the present application also provides a computer device. Please refer to Figure 7 . The computer device includes a processor and a memory. At least one instruction, at least one program, a code set or an instruction set is stored in the memory. The at least one instruction, at least one program, the code set or the instruction set is loaded and executed by the processor to implement the space-based angle-only target orbit determination method for compensating the orbital position error provided in the above method embodiments

[0114] An embodiment of the present application also provides a computer-readable storage medium. At least one instruction, at least one program, a code set or an instruction set is stored on the computer-readable storage medium. The at least one instruction, at least one program, the code set or the instruction set is loaded and executed by the processor to implement the space-based angle-only target orbit determination method for compensating the orbital position error provided in the above method embodiments

[0115] An embodiment of the present application further provides a computer program product, which includes a computer program. The processor of the computer device reads the computer program from a computer-readable storage medium, and the processor executes the computer program, so that the computer device executes the space-based angle-only target orbit determination method for compensating for orbit position errors described in any one of the above embodiments.

[0116] For convenience of description, when describing the above system or device, it is divided into various modules or units according to functions for separate description. Of course, when implementing the present application, the functions of each unit can be implemented in one or more software and / or hardware.

[0117] From the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present application.

[0118] Finally, it should also be noted that in this document, relational terms such as first, second, third, and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0119] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for determining a space-based target orbit by only measuring angles and compensating for orbit position errors, characterized in that: The method comprises: According to the real orbital dynamics model of the target, an error model for introducing measurement error into the empirical orbital dynamics model error is established; The initial neural network is trained according to a preset training sample to obtain an orbit position error estimation model; wherein the training sample is determined by historical orbit data and an empirical orbit dynamics model; The error model is compensated according to the track position error estimation model, and the initial state of the target to be measured is processed using the deviation compensation least square method to obtain the optimal estimation value of the track of the target to be measured.

2. The method according to claim 1, characterized in that The method of establishing an error model for introducing measurement error into the empirical orbital dynamics model error according to the real orbital dynamics model of the target to be measured comprises: Establishing a real orbital dynamics model of the target according to a first acceleration of the target determined by the empirical orbital dynamics model and a second acceleration determined by the empirical orbital dynamics model and the real orbital motion of the target; Establishing a first equation and a second equation of the target at a target time according to the state transfer matrix of the real orbit dynamics model and the empirical orbit dynamics model, as well as the real angle measurement data and the measurement equation; The error model is established by performing approximate substitution processing on the difference between the first equation and the second equation and substituting the difference into the first equation.

3. The method according to claim 2, characterized in that The error model is established by the following formula: Wherein, Y, H, S, and V are all vector parameters of the error model; is the error between the actual angle measurement value and the estimated angle measurement value; N is the number of measurement data sets in the target time period; s is the constant deviation introduced into the measurement by the deviation between the empirical orbit dynamics model and the target's actual orbit motion; H N The Jacobian matrix of the measurement equation is determined according to the trajectory state; Φ is from t0 to t N The state transfer matrix at the moment; Δx0 is the initial state deviation; ε N is the random error vector that ensures the equality.

4. The method according to claim 3, characterized in that The initial neural network is trained according to the preset training samples to obtain the track position error estimation model, including: Based on deep learning, an initial neural network is established, whose input is orbit state estimation and epoch time, and whose output is orbit position error; wherein the input layer of the initial neural network is a linear transmission purelin function, and the hidden layer activation function is a linear transmission purelin function and a linear transmission satlin function; According to the number of training samples and the output of the initial neural network, a training loss function L is established: Where M is the number of training samples, Δr i and The expected output and actual output of the neural network respectively; The training samples are input into the initial neural network for training according to the pre-divided training set, validation set and test set to obtain the track position error estimation model.

5. The method according to claim 4, characterized in that The training samples are determined by using historical orbit data and an empirical orbit dynamics model, including: Determining a plurality of initial states and a plurality of initial epochs of the target based on historical orbit data; wherein a first interval of each initial epoch is at least 24 hours; calculating an orbital state estimate for each initial state within a first interval at a preset second interval based on an empirical orbital dynamics model; The position error of the target is calculated based on the track state estimate and the historical track real state, and the track state estimate and the corresponding position error are used as a set of data of the training sample.

6. The method according to claim 4, characterized in that The method of compensating the error model according to the track position error estimation model, processing the initial state of the target to be measured by using the deviation compensation least square method, and obtaining the optimal estimation value of the track of the target to be measured includes: According to the preset state estimation value, the state of the target to be measured is calculated using the empirical orbit dynamics model to obtain the state estimation at the target time; Inputting the state estimation of the target time into the orbit position error estimation model, and outputting the position error of the target time; Inputting the position error at the target time into the error model and outputting a constant deviation; The optimal estimated value of the orbit is calculated based on the vector parameters and the constant deviation.

7. A space-based target orbit determination device that only measures angles and compensates for orbit position errors, characterized in that: The device comprises: A modeling module, used to establish an error model for introducing measurement errors into the empirical orbital dynamics model according to the real orbital dynamics model of the target; A training module, used to train the initial neural network according to a preset training sample to obtain an orbit position error estimation model; wherein the training sample is determined by historical orbit data and an empirical orbit dynamics model; The processing module is used to compensate the error model according to the orbit position error estimation model, and use the deviation compensation least square method to process the initial state of the target to be measured to obtain the optimal estimation value of the orbit of the target to be measured.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of any one of the methods described in claims 1-6.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The method comprises a computer program, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images