Agile satellite attitude fusion method and system considering multi-source reference differences

Abstract

The present invention discloses an agile satellite attitude fusion method and system taking into account multi-source benchmark differences, belonging to the field of remote sensing satellite ground data processing, the method comprising: calculating thermal environment factors based on satellite measurement data and satellite sensor installation parameters; extracting spatiotemporal low-frequency errors based on observation data of star cameras and star sensors; normalizing the thermal environment factors, using a multi-layer perceptron network to construct a relationship model between spatiotemporal low-frequency errors and thermal environment factors, and obtaining the attitude data of compensated star sensors and star cameras; establishing a multi-source absolute attitude sensor data fusion model based on the attitude data of compensated star sensors and star cameras; using a bidirectional dynamic unscented Kalman filter to perform absolute and relative attitude data fusion filtering on the processed gyro data and the fusion results of star sensors and star cameras to obtain high-precision attitude data results. The present invention improves the attitude determination accuracy of agile satellites and meets the needs of high-precision remote sensing tasks.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing satellite ground data processing, and in particular relates to an agile satellite attitude fusion method and system taking into account multi-source reference differences. Background Art

[0002] Traditional optical satellites mostly adopt passive subsatellite imaging mode, which can only obtain optical image data of the area near the subsatellite point trajectory. High-resolution agile optical satellites can use the rapid maneuverability in pitch and roll directions to adjust the ground pointing of the optical camera to any target area of ​​interest, breaking the limitation that traditional optical satellites can only obtain subsatellite point data, greatly improving the observation efficiency of optical satellites. However, imaging at different orbital positions, different times, and different roll pitch angles causes large differences in the heated area, heated area, and thermal radiation energy of agile optical satellites, resulting in thermal deformation errors of multi-source attitude sensors such as star sensors, star cameras, fiber optic gyroscopes, and laser gyroscopes. The deformation is complex and difficult to model and compensate. At the same time, in the agile imaging mode, there are certain differences in the measurement accuracy of multi-source attitude sensors. Directly performing indifferent attitude fusion processing on the data of star sensors, star cameras, fiber optic gyroscopes, and laser gyroscopes without considering the accuracy weights of various attitude data will result in low fusion accuracy and fail to meet the high-precision processing requirements of agile satellites. Therefore, it is necessary to compensate for the spatiotemporal thermal deformation errors of agile satellites and select the optimal attitude fusion solution based on the real-time error conditions of multi-source data. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes an agile satellite attitude fusion method and system taking into account multi-source reference differences to solve the problems existing in the above-mentioned prior art.

[0004] To achieve the above objectives, the present invention provides an agile satellite attitude fusion method taking into account multi-source reference differences, including.

[0005] The thermal environment factor is calculated based on satellite measurement data and satellite sensor installation parameters; wherein the satellite measurement data includes: agile satellite attitude measurement data and orbit measurement data.

[0006] Extracting spatiotemporal low-frequency errors based on observational data from star cameras and star sensors.

[0007] The thermal environment factors are normalized, and a multi-layer perceptron network is used to build a relationship model between spatiotemporal low-frequency errors and thermal environment factors to obtain the compensated attitude data of the star sensor and star camera.

[0008] Based on the compensated attitude data of the star sensor and the star camera, a multi-source absolute attitude sensor data fusion model is established.

[0009] A bidirectional dynamic unscented Kalman filter is used to perform absolute and relative attitude data fusion filtering on the processed gyro data and the fusion results of the star sensor and star camera to obtain high-precision attitude data results.

[0010] Optionally, the thermal environment factors include: satellite attitude in the orbital coordinate system, satellite geodetic coordinates, solar altitude angle and azimuth angle, and solar pointing parameters.

[0011] Optionally, the process of acquiring the satellite attitude in the orbital coordinate system includes:

[0012] The position and speed of the satellite in earth-fixed coordinates at the current moment are obtained based on the orbit measurement data.

[0013] Based on the earth's rotation parameters, the position and velocity of the satellite in the earth-fixed coordinate system are transformed into the inertial coordinate system to obtain the attitude rotation matrix from the satellite orbital coordinate system to the inertial coordinate system.

[0014] The rotation matrix from the satellite body to the orbital coordinate system is obtained based on the attitude rotation matrix from the satellite orbital coordinate system to the inertial coordinate system.

[0015] The yaw angle, pitch angle and roll angle of the satellite at the current moment are calculated based on the rotation matrix of the satellite body to the orbital coordinate system.

[0016] The satellite attitude in the orbital coordinate system is characterized based on the yaw angle, pitch angle and roll angle of the satellite at the current moment.

[0017] Optionally, the process of extracting spatiotemporal low-frequency errors based on observation data of a star camera and a star sensor includes:

[0018] A multi-star sensor attitude fusion equation is constructed based on the observation data of the star sensor, and the star sensor attitude parameters are obtained by solving the multi-star sensor attitude fusion equation.

[0019] A multi-star camera attitude fusion equation is constructed based on the observation data of the star camera, and the star camera attitude parameters are obtained by solving the multi-star camera attitude fusion equation.

[0020] An attitude parameter difference is obtained based on the star sensor attitude parameter and the star camera attitude parameter.

[0021] The random error filtering process is performed on the attitude parameter differences to extract the spatiotemporal low-frequency errors.

[0022] Optionally, the spatiotemporal low-frequency error includes: a yaw angle spatiotemporal low-frequency error, a pitch angle spatiotemporal low-frequency error and a roll angle spatiotemporal low-frequency error.

[0023] Among them, the expression of the spatiotemporal low-frequency error of the yaw angle.

[0024] .

[0025] In the formula, is the extracted spatiotemporal low-frequency error of the yaw angle, are the distance weight factor and the error similarity factor, respectively. Represents the index of the current spatiotemporal low-frequency error data, is the size of the window for data filtering, ss represents the star sensor, sc represents the star camera, represents the yaw angle parameter difference, The index representing the current yaw angle parameter difference.

[0026] Optionally, the expression of the multi-source absolute attitude sensor data fusion model is:

[0027] .

[0028] In the formula, is the rotation matrix from the inertial coordinate system to the satellite body coordinate system, are the three Euler angles corresponding to the rotation matrix from the inertial coordinate system to the satellite body coordinate system, is the 3*3 identity matrix, They are the directions of the X-axis, Y-axis, and Z-axis of the satellite body in the inertial system determined by the star sensor measurement data. They are the directions of the X-axis, Y-axis, and Z-axis of the satellite body in the inertial system determined by the star camera after time and space error compensation.

[0029] Optionally, the process of obtaining processed gyro data includes:

[0030] Get raw high-frequency gesture data.

[0031] The original high-frequency attitude data is sequentially subjected to frequency reduction processing and cross-validation to obtain processed gyroscope data.

[0032] To achieve the above object, the present invention also provides an agile satellite attitude fusion system that takes into account multi-source reference differences, the system comprising.

[0033] The thermal deformation factor calculation module is used to calculate the satellite attitude, satellite geodetic coordinates, solar altitude angle and azimuth angle and solar pointing parameters in the orbital coordinate system corresponding to the attitude data measurement time according to the orbit measurement data, attitude measurement data, installation parameters and ephemeris file of the agile satellite, and output them as thermal deformation factors.

[0034] The spatiotemporal error extraction module is used to use the original observation data of a single star sensor and a star camera to calculate the attitude parameters of the multi-star sensor and the multi-star camera according to the multi-star sensor attitude determination model and the multi-star camera attitude determination model, and calculate the difference between the two types of attitude parameters, perform filtering processing, and obtain the spatiotemporal error between the two types of attitude sensors.

[0035] The spatiotemporal error modeling and compensation module is used to train the model according to the set multi-layer perceptron network, use the normalized thermal deformation factor and the spatiotemporal error extracted by denoising to obtain the spatiotemporal error model, and perform spatiotemporal error compensation on the measured data to obtain the posture parameters after benchmark unification.

[0036] The multi-source data fusion processing module is used to fuse the attitude parameters determined by the multi-star sensors and multi-star cameras after time and space error compensation to obtain high-precision absolute attitude parameters, and use a bidirectional dynamic unscented Kalman filter to perform absolute and relative attitude data fusion filtering processing to obtain high-precision and high-frequency attitude data results.

[0037] Compared with the prior art, the present invention has the following advantages and technical effects.

[0038] The present invention proposes an agile satellite attitude fusion method that takes into account multi-source benchmark differences. By accurately calculating thermal environment factors and extracting spatiotemporal low-frequency errors, the thermal deformation error in satellite attitude measurement is effectively compensated. The method uses a multi-layer perceptron network to establish a relationship model between errors and thermal environment factors, and realizes accurate compensation of star sensor and star camera attitude data. Furthermore, by constructing a multi-source attitude sensor data fusion model and combining a bidirectional dynamic unscented Kalman filter, high-precision fusion of star sensor, star camera and gyroscope data is achieved. This method significantly improves the attitude determination accuracy of agile satellites, enhances the robustness of the system, meets the needs of high-precision remote sensing tasks, and has important practical application value and broad market prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings constituting a part of the present application are used to provide a further understanding of the present application, and the illustrative embodiments and descriptions of the present application are used to explain the present application, and do not constitute an improper limitation on the present application.

[0040] Figure 1 The present invention is a flowchart of an agile satellite attitude fusion method taking into account multi-source reference differences according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0042] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0043] Embodiment 1.

[0044] like Figure 1 As shown, in this embodiment, an agile satellite attitude fusion method taking into account multi-source benchmark differences is provided, including: calculating thermal environment factors based on satellite measurement data and satellite sensor installation parameters; extracting spatiotemporal low-frequency errors based on observation data of star cameras and star sensors; normalizing the thermal environment factors, and using a multi-layer perceptron network to construct a relationship model between spatiotemporal low-frequency errors and thermal environment factors to obtain compensated attitude data of star sensors and star cameras; establishing a multi-source absolute attitude sensor data fusion model based on the compensated attitude data of star sensors and star cameras; and using a bidirectional dynamic unscented Kalman filter to perform absolute and relative attitude data fusion filtering on the processed gyro data and the fusion results of the star sensor and star camera to obtain high-precision attitude data results.

[0045] Specifically, the following steps are included: Step 1: Based on the agile satellite attitude measurement data, orbit measurement data and the installation parameters of the sensor in the satellite body coordinate system, various factors related to the satellite thermal deformation error are calculated.

[0046] Furthermore, based on the satellite attitude and orbit measurement data, combined with the sensor installation parameters, the satellite position and velocity in different coordinate systems are calculated; the attitude rotation matrix from the satellite orbit coordinate system to the inertial coordinate system is established, and the rotation matrix and attitude Euler angles from the satellite body coordinate system to the inertial coordinate system are calculated. The satellite's geodetic longitude and latitude are calculated, and thus the solar altitude and azimuth are calculated. Based on the rotation matrix between different coordinate systems, the direction of the sun in different coordinate systems is determined.

[0047] Furthermore, the process of obtaining the satellite attitude in the orbital coordinate system includes: obtaining the position and velocity of the satellite in the earth-fixed coordinate system at the current moment based on the orbital measurement data; converting the position and velocity of the satellite in the earth-fixed coordinate system to the inertial coordinate system based on the earth's rotation parameters to obtain the attitude rotation matrix from the satellite orbital coordinate system to the inertial coordinate system; obtaining the rotation matrix from the satellite body to the orbital coordinate system based on the attitude rotation matrix from the satellite orbital coordinate system to the inertial coordinate system; calculating the yaw angle, pitch angle and roll angle of the satellite at the current moment based on the rotation matrix from the satellite body to the orbital coordinate system; and obtaining the satellite attitude in the orbital coordinate system based on the yaw angle, pitch angle and roll angle of the satellite at the current moment.

[0048] As a specific implementation of this embodiment, for measuring time According to the satellite's orbit sensor measurement data, the satellite's position in the ground-fixed coordinates at that moment can be obtained. and speed , and according to the earth's rotation parameters, the satellite position and velocity information in the earth-fixed coordinate system can be converted into the inertial coordinate system, which can be expressed as , , thereby establishing the attitude rotation matrix from the satellite orbit coordinate system to the inertial coordinate system , specifically expressed as follows.

[0049]

[0050] According to the satellite attitude sensor measurement data, the rotation matrix from the satellite body coordinate system to the inertial coordinate system can be obtained According to the matrix multiplication operation, the rotation matrix of the satellite body to the orbital coordinate system can be obtained , , and further calculate the attitude Euler angle value of the satellite body to the orbital coordinate system, as shown in the following formula.

[0051]

[0052] in, , , They represent the satellite at the measurement time The yaw angle, pitch angle, and roll angle under express Middle Row, No. The numeric value of the column.

[0053] For measuring time , based on the satellite's position in Earth-fixed coordinates , the geodetic longitude corresponding to the satellite can be calculated , geodetic latitude The location information is as shown below.

[0054]

[0055] in, is the length of the semi-major axis of the Earth ellipsoid, is the length of the semi-minor axis of the Earth ellipsoid, is the first eccentricity of the Earth ellipsoid, is the second eccentricity of the Earth ellipsoid, .

[0056] Further, calculate the solar altitude angle and solar azimuth , as shown in the following formula.

[0057]

[0058] in, is the solar altitude angle, is the solar declination, is the solar hour angle.

[0059] Calculate the orientation of the sun in the Earth-fixed coordinate system, the satellite coordinate system, and the attitude sensor coordinate system as shown in the following formula.

[0060]

[0061] in, It is the spatial orientation of the sun in the inertial system; is the direction of the sun in the Earth-fixed coordinate system; is the direction of the sun in the satellite's coordinate system; is the direction of the sun in the attitude sensor coordinate system; is the rotation matrix from the inertial system to the Earth-fixed coordinate system, obtained through the Earth's rotation parameters; It is the rotation matrix from the inertial system to the satellite body coordinate system, which is obtained by measuring the attitude sensor; It is the installation parameter from the satellite body to the attitude sensor, which is obtained through installation parameter calibration.

[0062] Therefore, for the measurement time , and obtained various factors related to the satellite thermal environment changes at that moment: satellite attitude in orbital coordinate system , Satellite geodetic coordinates , solar altitude angle and azimuth , Sun pointing parameters .

[0063] Step 2: Based on the observation data of the star camera and star sensor, dual-vector attitude determination is performed, and the differences between the two types of sensors are compared. De-noising is then used to extract the thermal deformation errors between the two types of sensors.

[0064] Furthermore, Gaussian filtering is used to filter out random errors, reduce the impact of noise, and extract spatiotemporal low-frequency errors.

[0065] Furthermore, the process of extracting spatiotemporal low-frequency errors based on the observation data of the star camera and the star sensor includes: constructing a multi-star sensor attitude fusion equation based on the observation data of the star sensor, and solving the multi-star sensor attitude fusion equation to obtain the star sensor attitude parameters; constructing a multi-star camera attitude fusion equation based on the observation data of the star camera, and solving the multi-star camera attitude fusion equation to obtain the star camera attitude parameters; obtaining attitude parameter differences based on the star sensor attitude parameters and the star camera attitude parameters; performing random error filtering processing on the attitude parameter differences to extract spatiotemporal low-frequency errors.

[0066] As a specific implementation of this embodiment, the optical axis measurement data of the star sensor A is used. , optical axis measurement data of star sensor B 、Optical axis installation parameters of star sensor A 、Optical axis installation parameters of star sensor B , construct the multi-star sensor attitude fusion equation, and solve the attitude parameters of the star sensor AB solution , to eliminate the interference of poor transverse axis measurement accuracy of the sensor, as shown in the following formula.

[0067]

[0068] in, , .

[0069] Similarly, the optical axis measurement data of star camera A, the optical axis measurement data of star camera B, the optical axis installation parameters of star camera A, and the optical axis installation parameters of star camera B are used to construct the multi-star camera attitude fusion equation and solve the attitude parameters of star camera AB. .

[0070] Comparison of the differences in attitude parameters calculated by the two attitude sensors , which includes the spatiotemporal low-frequency error caused by satellite thermal deformation and the random error of attitude measurement. The random error is further filtered out to extract the spatiotemporal low-frequency error, as shown in the following formula.

[0071]

[0072] in, is the temporal and spatial low-frequency error of the extracted attitude yaw angle, are the distance weight factor and the error similarity factor, respectively. Indicates the index of the current data. is the size of the window for data filtering. ss represents the star sensor, sc represents the star camera, represents the yaw angle parameter difference, The index representing the current yaw angle parameter difference.

[0073] Similarly, the above operations are performed on the attitude parameter differences of the pitch angle and the roll angle to achieve the low-frequency error of the pitch angle in time and space , low-frequency error of the side swing angle in time and space Extraction.

[0074] Step 3, based on the multi-layer perceptron network, with the normalized thermal environment factor as input and the spatiotemporal low-frequency error between the star sensor and the star camera as output, a relationship model between the spatiotemporal low-frequency error and the thermal environment factor is established, and the multi-source attitude sensor benchmark is compensated and unified.

[0075] Furthermore, a multi-layer perceptron network is used to establish a relationship model between spatiotemporal low-frequency errors and thermal environment factors, linking thermal environment factors with spatiotemporal low-frequency errors for error compensation.

[0076] As a specific implementation of this embodiment, in order to ensure the accuracy of the model, the thermal environment factor needs to be normalized, wherein the normalization of the solar altitude angle parameter is shown in the following formula.

[0077]

[0078] in, For the The original solar altitude angle value of the data, For the The normalized solar altitude angle value of the data is is the total number of data, , It is the maximum and minimum value of the solar altitude angle in all data.

[0079] The normalized thermal environment factors are further constructed into a feature matrix , the spatiotemporal low-frequency error values ​​are constructed as the output matrix , as shown below.

[0080]

[0081] A multi-layer perceptron network with one hidden layer for regression tasks was constructed, in which the input layer had 16 features, the fully connected layer was designed to have 200 neurons, and the ReLU function was used as the activation layer, and the output layer had 3 neurons. In order to ensure the reliability of model fitting, the original data was processed in blocks according to the imaging time period, and each block of data was randomly sampled. 60% of the data in the block was used for training, 20% for verification, and 20% for testing. The training data set, verification data set, and test data set of the entire network were constructed, and model training processing was performed to obtain the thermal deformation relationship model of the attitude sensor. .

[0082] For any measurement data of the star camera, the thermal environment transformation factor corresponding to the measurement data can be calculated based on relevant data, and then brought into the thermal deformation relationship model of the attitude sensor to obtain the current error compensation value, thereby compensating the star camera attitude and unifying the benchmarks of the two types of attitude sensors.

[0083] Step 4: Based on the random error values ​​measured by star cameras and star sensors, a data fusion model of multiple star cameras and multiple star sensors is established to further improve the accuracy of attitude solution.

[0084] Furthermore, according to the rotation matrix from the inertial coordinate system to the satellite body coordinate system, the directions of the X-axis, Y-axis and Z-axis of the satellite body in the inertial system determined by the star sensor measurement data and the directions of the X-axis, Y-axis and Z-axis of the satellite body in the inertial system determined by the star camera after time-space error compensation, a multi-source attitude sensor data fusion model is constructed to improve the accuracy of attitude solution.

[0085] As a specific implementation of this embodiment, the satellite attitude determined by the star sensor measurement data is , after compensating for the low-frequency errors in space and time, the satellite attitude determined by the star camera is , where each matrix can be expressed as follows.

[0086]

[0087] in, They are the directions of the X-axis, Y-axis, and Z-axis of the satellite body in the inertial system determined by the star sensor measurement data. They are the directions of the X-axis, Y-axis, and Z-axis of the satellite body in the inertial system determined by the star camera after time and space error compensation.

[0088] The multi-source attitude sensor data fusion model is further constructed as shown in the following formula.

[0089]

[0090] in, is the rotation matrix from the inertial coordinate system to the satellite body coordinate system, are the three Euler angles corresponding to the rotation matrix, It is the 3*3 identity matrix.

[0091] By expressing the above relationship in symbols, we can get the following expression.

[0092]

[0093] Since the above equation is a nonlinear equation about the rotation angle, it needs to be linearized to obtain the following equation.

[0094]

[0095] in, Calculates the initial value for the coordinates calculated using the current initial value of Euler angles, is the correction of Euler angle.

[0096] According to the above linearized equations, using the measurement data, an iterative process is performed to obtain the optimal solution of the Euler angles from the inertial coordinate system that integrates the star sensor and star camera data to the satellite body coordinate system.

[0097] Step 5: perform a cross-check on the relative attitude measurement data based on the high-frequency attitude measurement data obtained by the fiber optic gyroscope and the laser gyroscope to extract high-precision, high-frequency relative attitude data.

[0098] Furthermore, the high-frequency attitude measurement data of the satellite obtained by high-precision sensors such as fiber optic gyroscopes and laser gyroscopes are down-sampled, cross-checked, and analyzed for time-frequency characteristics, and appropriate data are selected to obtain processed gyroscope data.

[0099] As a specific implementation of this embodiment, since too high a sampling frequency will lead to significant data noise and susceptibility to noise interference, the original high-frequency attitude data needs to be down-sampled to obtain measurement data with a reasonable sampling frequency and high accuracy. High-frequency attitude measurement data is the change in satellite attitude within a certain time interval. The absolute attitude of the satellite (filtering out the inherently stable attitude of the satellite) and at sampling intervals Relative posture change obtained , sampling interval Relative posture change obtained The relationship between can be expressed as follows.

[0100]

[0101] in, are the absolute attitude amplitude, frequency and initial phase.

[0102] There is a mapping relationship between the frequency of the high-frequency original data and the frequency-reduced data, and the amplitude is unchanged. , ,……, The amplitude attributes of the relative posture data are mapped, and the consistency of the time-frequency characteristics of each down-converted data with the time-frequency characteristics of the original data is detected. The data with a difference less than the threshold and the longest sampling interval are selected as the most reasonable high-frequency posture data for subsequent data processing.

[0103] Step 6: Based on the bidirectional dynamic unscented Kalman filter, the high-frequency attitude data of the star sensor, star camera and gyroscope are integrated to improve the attitude determination accuracy of the agile satellite.

[0104] As a specific implementation method of this embodiment, a bidirectional dynamic unscented Kalman filter is designed. The initial filtering adopts a forward filtering process with increasing time, and the weights of the multi-star camera and the star sensor are set to a larger value to improve the error estimation accuracy of the gyro data. The backward filtering uses the result of the initial filtering as the input value, and performs filtering from back to front in time. The weight of the gyro observation data is increased in this filtering to improve the attitude solution accuracy. Then, a fixed interval smoothing filter is performed, and the weight of the multi-source attitude data is dynamically adjusted to increase the contribution of the relative attitude to the fused attitude, thereby realizing high-frequency and high-precision multi-source attitude data fusion, which is suitable for a variety of imaging modes of agile satellites.

[0105] The present invention aims at the fusion problem between different sensor measurement data, and studies a method for multi-source attitude sensor data fusion. The present invention aims at the characterization problem of thermal deformation errors generated by agile optical satellites under different conditions, and studies a new modeling and compensation method. The present invention aims at the problem of establishing a relationship model between spatiotemporal low-frequency errors and thermal environment factors, and studies an innovative method for applying machine learning technology to the field of satellite attitude research. The present invention aims at the problem of fusing high-frequency attitude data of star sensors, star cameras, and gyroscopes, and designs a bidirectional dynamic unscented Kalman filter. The present invention improves the attitude determination accuracy of agile satellites by comprehensively considering thermal deformation errors and multi-source benchmark differences, so that satellite data can meet high-precision processing requirements.

[0106] Embodiment 2.

[0107] The present invention also provides an agile satellite attitude fusion system that takes into account multi-source benchmark differences, the system comprising: a thermal deformation factor calculation module, for calculating the satellite attitude, satellite geodetic coordinates, solar altitude angle and azimuth angle and solar pointing parameters in the orbital coordinate system corresponding to the attitude data measurement time according to the orbit measurement data, attitude measurement data, installation parameters and ephemeris file of the agile satellite, and outputting them as thermal deformation factors; a spatiotemporal error extraction module, for calculating the multi-star sensor attitude parameters and the multi-star camera attitude parameters according to the multi-star sensor attitude determination model and the multi-star camera attitude determination model using the original observation data of a single star sensor and a star camera, and calculating the difference between the two types of attitude parameters, and then performing The multi-source data fusion processing module is used to fuse the attitude parameters determined by the multi-star sensors and multi-star cameras after the space-time error compensation to obtain high-precision absolute attitude parameters, and adopts a bidirectional dynamic unscented Kalman filter to perform absolute and relative attitude data fusion filtering processing to obtain high-precision and high-frequency attitude data results.

[0108] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An agile satellite attitude fusion method taking into account multi-source reference differences, characterized in that: The following steps are involved: The thermal environment factor is calculated based on the satellite measurement data and the satellite sensor installation parameters; wherein the satellite measurement data includes: agile satellite attitude measurement data and orbit measurement data; the thermal environment factor includes: satellite attitude, satellite geodetic coordinates, solar altitude angle and azimuth angle and solar pointing parameters in the orbital coordinate system; Extracting spatiotemporal low-frequency errors based on observational data from star cameras and star sensors; The thermal environment factors are normalized, and a relationship model between the spatiotemporal low-frequency error and the thermal environment factors is constructed using a multi-layer perceptron network to obtain the compensated attitude data of the star sensor and the star camera; wherein the multi-layer perceptron network comprises an input layer, a hidden layer and an output layer, and the normalized thermal environment factors are used as the input of the multi-layer perceptron network, and the spatiotemporal low-frequency error is used as the output of the multi-layer perceptron network, and model training is performed to establish a relationship model between the spatiotemporal low-frequency error and the thermal environment factors; the thermal environment transformation factor corresponding to any measurement data of the star camera is calculated, and the thermal environment transformation factor corresponding to any measurement data of the star camera is input into the relationship model to obtain the spatiotemporal low-frequency error compensation value; Based on the compensated attitude data of the star sensor and the star camera, a multi-source absolute attitude sensor data fusion model is established; A bidirectional dynamic unscented Kalman filter is used to perform absolute and relative attitude data fusion filtering on the processed gyro data and the fusion results of the star sensor and star camera to obtain high-precision attitude data results.

2. The agile satellite attitude fusion method taking into account multi-source reference differences according to claim 1, characterized in that: The process of obtaining the satellite attitude in the orbital coordinate system includes: Based on the orbit measurement data, the position and velocity of the satellite at the current moment in the earth-fixed coordinates are obtained; Based on the earth's rotation parameters, the position and velocity of the satellite in the earth-fixed coordinate system are converted to the inertial coordinate system to obtain the attitude rotation matrix from the satellite orbital coordinate system to the inertial coordinate system; Based on the attitude rotation matrix from the satellite orbital coordinate system to the inertial coordinate system, a rotation matrix from the satellite body to the orbital coordinate system is obtained; Calculate the yaw angle, pitch angle and roll angle of the satellite at the current moment based on the rotation matrix of the satellite body to the orbital coordinate system; The satellite attitude in the orbital coordinate system is characterized based on the yaw angle, pitch angle and roll angle of the satellite at the current moment.

3. The agile satellite attitude fusion method taking into account multi-source reference differences according to claim 1, characterized in that: The process of extracting spatiotemporal low-frequency errors based on the observation data of star cameras and star sensors includes: Constructing a multi-star sensor attitude fusion equation based on the observation data of the star sensor, and solving the multi-star sensor attitude fusion equation to obtain the star sensor attitude parameters; Constructing a multi-star camera attitude fusion equation based on the observation data of the star camera, and solving the multi-star camera attitude fusion equation to obtain the star camera attitude parameters; Obtaining an attitude parameter difference based on the star sensor attitude parameter and the star camera attitude parameter; The random error filtering process is performed on the attitude parameter differences to extract the spatiotemporal low-frequency errors.

4. The agile satellite attitude fusion method taking into account multi-source reference differences according to claim 3 is characterized in that: The spatiotemporal low-frequency errors include: yaw angle spatiotemporal low-frequency error, pitch angle spatiotemporal low-frequency error and roll angle spatiotemporal low-frequency error Among them, the expression of the spatiotemporal low-frequency error of the yaw angle is: In the formula, is the extracted spatiotemporal low-frequency error of the yaw angle, are the distance weight factor and the error similarity factor, respectively. Represents the index of the current spatiotemporal low-frequency error data, is the size of the window for data filtering, ss represents the star sensor, sc represents the star camera, represents the yaw angle parameter difference, The index representing the current yaw angle parameter difference.

5. The agile satellite attitude fusion method taking into account multi-source reference differences according to claim 1, characterized in that: The expression of the multi-source absolute attitude sensor data fusion model is: In the formula, is the rotation matrix from the inertial coordinate system to the satellite body coordinate system, are the three Euler angles corresponding to the rotation matrix from the inertial coordinate system to the satellite body coordinate system, is the 3*3 identity matrix, They are the directions of the X-axis, Y-axis, and Z-axis of the satellite body in the inertial system determined by the star sensor measurement data. They are the directions of the X-axis, Y-axis, and Z-axis of the satellite body in the inertial system determined by the star camera after compensating for the low-frequency errors in space and time.

6. The agile satellite attitude fusion method considering multi-source reference differences according to claim 1, characterized in that: The process of obtaining processed gyro data includes: Get raw high-frequency gesture data; The original high-frequency attitude data is sequentially subjected to frequency reduction processing and cross-validation to obtain processed gyroscope data.

7. An agile satellite attitude fusion system taking into account multi-source reference differences, characterized in that: For implementing the method of claim 1, the system comprises: The thermal deformation factor calculation module is used to calculate the satellite attitude, satellite geodetic coordinates, solar altitude angle and azimuth angle and solar pointing parameters in the orbital coordinate system corresponding to the attitude data measurement time according to the orbit measurement data, attitude measurement data, installation parameters and ephemeris file of the agile satellite, and output them as thermal deformation factors; The spatiotemporal low-frequency error extraction module is used to use the original observation data of a single star sensor and a star camera to calculate the attitude parameters of the multi-star sensor and the multi-star camera according to the attitude determination model of the multi-star sensor and the multi-star camera, and calculate the difference between the two types of attitude parameters, perform filtering processing, and obtain the spatiotemporal low-frequency error between the two types of attitude sensors; The spatiotemporal low-frequency error modeling and compensation module is used to train the model according to the set multi-layer perceptron network, use the normalized thermal deformation factor and the spatiotemporal low-frequency error extracted by denoising to obtain the spatiotemporal low-frequency error model, and perform spatiotemporal low-frequency error compensation on the measured data to obtain the attitude parameters after the benchmark is unified; The multi-source data fusion processing module is used to fuse the attitude parameters determined by the multi-star sensors and multi-star cameras after the temporal and spatial low-frequency error compensation to obtain high-precision absolute attitude parameters, and adopts a bidirectional dynamic unscented Kalman filter to perform absolute and relative attitude data fusion filtering processing to obtain high-precision and high-frequency attitude data results.

Images