Multi-source error compensation and high-precision fusion method for satellite-inertial integrated navigation

Abstract

A multi-source error compensation and high-precision fusion method for satellite inertial integrated navigation belongs to the technical field of integrated navigation, and mainly comprises the following steps: constructing an inertial error model containing gyroscopic drift and accelerometer zero offset, and realizing online calibration compensation of installation errors through observability analysis; a multi-rate Kalman filtering framework is designed, and synchronous fusion of inertial navigation high-frequency data and star sensor low-frequency data is completed; a historical fixed star vector equivalent compensation strategy is introduced, and the filtering weight is dynamically adjusted in combination with an astronomical attitude determination error model, so that stable attitude determination under a complex observation condition is realized. According to the method, the fusion precision, autonomy and anti-interference performance of integrated navigation can be effectively improved, stable and reliable attitude determination support is provided for aircrafts and the like with strict requirements on navigation performance, and the method has wide engineering application value.

Description

Technical Field

[0001] This invention relates to the field of integrated navigation technology, and in particular to a multi-source error compensation and high-precision fusion method for satellite-inertial integrated navigation. It is applicable to aircraft and other carriers with stringent requirements for navigation autonomy, accuracy and anti-interference, and can provide stable and reliable attitude support in complex environments where aircraft navigation fails or is interfered with. Background Technology

[0002] The flight characteristics of aircraft and other carriers place extremely high demands on the autonomy, anti-interference, and accuracy of navigation systems. Among the current mainstream navigation solutions, inertial navigation systems (INS) can output attitude and position information in real time through built-in multi-axis gyroscopes and accelerometers, but errors accumulate over time; satellite navigation systems can provide high-precision position information, but are susceptible to electromagnetic interference and may fail; celestial navigation systems (CNS) are based on stellar observations, their errors do not drift over time and they have strong anti-interference capabilities, but they are limited by imaging periods and star map recognition and processing procedures, resulting in insufficient real-time performance, and they cannot meet the continuous output requirements of dynamic navigation when operating alone.

[0003] Existing inertial / astronomical integrated navigation technologies largely rely on satellites to provide initial position information, resulting in insufficient initial alignment accuracy in scenarios without satellite support. Furthermore, the highly dynamic flight of the carrier can easily lead to installation angle deviations between the inertial sensor and the star sensor, and the number of observed stars may dynamically change, affecting the stability of the integrated navigation. In addition, the output data from the inertial sensor and the star sensor exhibit time asynchrony, making it difficult for traditional fusion methods to balance real-time performance and accuracy, thus failing to fully leverage the complementary advantages of both.

[0004] In summary, there is an urgent need for a heterogeneous data fusion method that can solve the problems of time synchronization, error compensation, and dynamic adaptation, so as to provide high-precision, high-reliability, and highly adaptable technical support for the combined navigation application of star sensors and inertial sensors in aircraft and other carriers, and ensure the autonomy and stable output of the aircraft navigation system in complex environments. Summary of the Invention

[0005] This disclosure provides a multi-source error compensation and high-precision fusion method for satellite-inertial integrated navigation to solve technical problems such as insufficient initial alignment accuracy of inertial / astronomical integrated navigation in scenarios without satellite support, the impact of installation angle deviation caused by high-motion flight and dynamic changes in the number of star observations on navigation stability, and the difficulty in balancing real-time performance and accuracy due to asynchronous data time. This enables high precision, high reliability and strong adaptability of the integrated navigation system during long-endurance flights of aircraft and other carriers, ensuring the autonomous and stable output of the navigation system.

[0006] The main steps of this method include: S1: Based on the known spatial coordinate system relationship, the angular velocity and specific force data of the inertial sensor are collected to calculate the initial attitude; the star sensor captures a star map, and star vectors are obtained through star point extraction and matching to calculate attitude data; S2, establish an inertial error model including gyroscope drift and accelerometer zero bias; through observability analysis, online calibration is used to compensate for installation errors, and astronomical attitude determination errors are modeled in combination with the dynamic characteristics of star number, and measurement noise is quantified; S3, design a multi-rate filtering framework to synchronize high-frequency inertial navigation and low-frequency star-sensor data, use navigation calculation, device and installation errors as state variables, attitude difference as measurement, and iteratively correct the output navigation parameters; S4. When the number of stars is insufficient, the attitude determination is completed by equivalent compensation using historical star vectors. The weight of filtered measurement noise is dynamically adjusted based on the astronomical attitude determination error model to improve the fusion stability under complex conditions.

[0007] Furthermore, step S1 specifically includes: Based on the known spatial coordinate system relationship between the inertial sensor and the star sensor, raw data from both types of sensors were collected respectively. The inertial sensor outputs the carrier's angular velocity and specific force data in real time to calculate the initial attitude and position; The star sensor captures images of the night sky. After star point extraction, star map recognition and matching, the vector information of stars in the inertial frame is obtained, and then the attitude and position data are calculated.

[0008] Furthermore, step S2 specifically includes: An error model for the inertial sensor is established, including gyroscope constant drift, accelerometer bias, and random noise. To address potential installation angle deviations between star sensors and inertial sensors during carrier flight, an online calibration algorithm is employed through observability analysis to estimate and compensate for installation error angles in real time. By combining the dynamic characteristics of the number of observed stars, an astronomical attitude determination error model is established to quantify the measurement noise under different observation conditions.

[0009] Furthermore, step S3 specifically includes: To address the issue of inconsistent data update rates between the two types of sensors, a multi-rate Kalman filter framework is designed to perform time synchronization processing between the high-frequency data from the inertial sensor and the low-frequency data from the star sensor. Using inertial navigation calculation error, device error, and installation error as state variables, and the attitude and position difference between the two types of sensors as measurement variables, error estimation and real-time correction are achieved through filtering iteration, and high-precision fused navigation parameters are output.

[0010] Furthermore, step S4 specifically includes: When the number of stars observed by the star sensor is insufficient, historical star vector information is introduced for equivalent compensation, and astronomical attitude determination is completed by combining the current observation data, thus breaking the limit of the minimum number of observed stars. By dynamically adjusting the filter measurement noise weights based on the astronomical attitude determination error model, the filter's adaptability to observation information of different quality is optimized, thereby improving the fusion stability under complex observation conditions.

[0011] Compared with the prior art, the beneficial effects of this disclosure are: ① solving the time asynchrony problem between inertial sensors and star sensors by multi-rate Kalman filtering, giving full play to the real-time advantages of inertial sensors and the high-precision characteristics of star sensors, realizing the synchronous fusion of heterogeneous data, and overcoming the defects of single sensors; ② achieving high precision of the integrated navigation system during long-endurance flight of aircraft and other carriers; ③ improving the stability of fusion under complex conditions. Attached Figure Description

[0012] The above and other objects, features and advantages of this disclosure will become more apparent from the more detailed description of exemplary embodiments of this disclosure taken in conjunction with the accompanying drawings, in which the same reference numerals generally represent the same components.

[0013] Figure 1 This is a system flowchart of a multi-source error compensation and high-precision fusion method for satellite-inertial navigation according to the present disclosure. Detailed Implementation

[0014] Preferred embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0015] This disclosure provides a method for multi-source error compensation and high-precision fusion of satellite-inertial navigation systems. In one exemplary implementation, the process is shown in the attached figure. Figure 1 As shown, it includes the following steps: 1. Data Acquisition and Preprocessing. Based on the known spatial coordinate system relationship, the angular velocity and specific force data of the inertial sensor are acquired to calculate the initial attitude; the star sensor captures a star map, and star vectors are obtained through star point extraction and matching to calculate high-precision attitude data; 2. Error Modeling and Compensation. An inertial error model incorporating gyroscope drift and accelerometer bias is established; online calibration and compensation for installation errors are achieved through observability analysis; astronomical attitude determination errors are modeled in conjunction with the dynamic characteristics of star number; and measurement noise is quantified. 3. Multi-rate Kalman filter fusion. A multi-rate filter framework is designed to synchronize high-frequency inertial navigation data with low-frequency star-sensor data. Navigation calculation, device and installation errors are used as state variables, attitude difference is used as a measurement, and high-precision navigation parameters are output through iterative correction. 4. Adaptive optimization and adaptation. When the number of stars is insufficient, the attitude determination is completed by equivalent compensation using historical star vectors; the weight of filtered measurement noise is dynamically adjusted based on the astronomical attitude determination error model to improve the fusion stability under complex conditions.

[0016] Specifically, such as Figure 1 As shown, in this embodiment, step 1 is implemented as follows: (1) Based on the preset spatial coordinate system transformation relationship between the inertial sensor and the star sensor, a multi-source sensor synchronous acquisition system was built. First, the physical installation and calibration of the inertial sensor and the star sensor were completed, and the rotation matrix and translation vector between their coordinate systems were determined. Second, the synchronous trigger was started, and the inertial sensor was controlled to output the raw data of the carrier angular velocity and specific force at the set sampling frequency. At the same time, the star sensor was driven to capture star images according to the preset exposure time and frame rate to ensure that the timestamps of the two types of sensor data were strictly aligned, providing a consistent data source basis for subsequent fusion calculation.

[0017] (2) The angular velocity and specific force data collected by the inertial sensor are preprocessed, and zero bias compensation, scaling factor calibration and temperature error correction are completed in sequence. Based on the effective data after preprocessing, the strapdown inertial navigation solution algorithm is adopted to solve the real-time attitude angle of the carrier through quaternion differential equations iteratively. Combined with the initial alignment strategy, the coarse alignment and fine alignment of the inertial navigation system are completed by using the constraints of the gravity vector and the Earth's rotation angular velocity vector. Finally, the initial attitude and position information of the carrier are calculated, providing an initial reference benchmark for the high-precision calculation of the star sensor.

[0018] (3) Star points are extracted from the star images captured by the star sensor. Gray-scale thresholding and centroid fitting algorithms are used to obtain the precise pixel coordinates of the star points. Star map recognition and matching are carried out based on star catalog data. The extracted star point features are associated with the inertial frame vectors of known stars through triangle matching. After matching, the least squares attitude determination algorithm is used to construct the transformation relationship between the observation vector and the reference vector, and the high-precision attitude matrix of the carrier in the inertial frame is calculated. Combined with the initial position information of the inertial navigation system, the position parameters of the carrier are further corrected, and the attitude and position data that meet the requirements of high-precision navigation are output.

[0019] Specifically, such as Figure 1 As shown, in this embodiment, step 2 is implemented as follows: (1) Based on the output error characteristics of the inertial sensor, an inertial error model including systematic error and random error is constructed. Among them, the systematic error term covers the gyroscope constant drift, accelerometer zero bias and scaling factor error, and the random error term uses a random walk model to characterize the white noise interference of the gyroscope and accelerometer. Based on laboratory calibration data and field test data, the fitting and verification of the error model parameters are completed, providing a mathematical basis for error compensation in subsequent inertial navigation calculations.

[0020] (2) To address the deviation in the installation angle of the inertial sensor and the star sensor caused by vibration and deformation during the flight of the carrier, an error observability analysis was conducted. The angular motion and linear motion of the carrier were selected as the observed state variables, and a state equation including the installation error angle was constructed. An extended Kalman filter online calibration algorithm was adopted, using the deviation between the inertial navigation solution value and the star sensor attitude determination result as the measurement input. The magnitude of the installation error angle was estimated in real time and fed back to the coordinate system transformation stage to achieve dynamic compensation of the installation error.

[0021] (3) Based on the dynamic variation characteristics of the number of observed stars, an astronomical attitude determination error model was established to clarify the mapping relationship between the number of observed stars and the attitude determination accuracy. For factors such as star point extraction error and star map matching mismatch rate under different observation scenarios, statistical analysis methods were used to quantify the measurement noise variance of the star sensor. Through simulation and measured data verification, the range of measurement noise values ​​under different numbers of observed stars was determined, providing a basis for setting the noise parameters of the integrated navigation filtering algorithm.

[0022] Specifically, such as Figure 1 As shown, in this embodiment, step 3 is implemented as follows: (1) To address the rate difference between high-frequency sampling of the inertial sensor and low-frequency output of the star sensor, a multi-rate Kalman filter framework is constructed. An interpolation extrapolation algorithm is used to extend the time scale of the low-frequency attitude data of the star sensor so that it corresponds one-to-one with the timestamp of the high-frequency data of the inertial sensor. At the same time, based on the data validity judgment criterion, abnormal sampling points of the two types of sensors are eliminated, and the spatiotemporal synchronization of the high-frequency angular velocity and force data of the inertial navigation system and the low-frequency attitude data of the star sensor is completed, providing a consistent input for subsequent fusion calculation.

[0023] (2) The inertial navigation calculation error, sensor device error, and installation error are integrated into a filtered state vector. The difference between the inertial navigation calculated attitude and position and the star sensor high-precision measurement results is used as the measurement to establish the filtered state equation and measurement equation. Through Kalman filtering iterative calculation, various errors are estimated and compensated in real time. The error correction is fed back to the inertial navigation calculation link to iteratively optimize the attitude, position, and velocity information of the carrier, and finally output fused navigation parameters that meet the requirements of high-precision navigation.

[0024] Specifically, such as Figure 1As shown, in this embodiment, step 4 is implemented as follows: (1) When the star sensor is affected by factors such as obstruction and light interference, and the number of observed stars is lower than the attitude determination threshold, the historical star vector compensation mechanism is activated. High-precision star vector data that has been observed in the previous period and verified for effectiveness are retrieved and combined with the vector information of the current small number of observed star points to construct an expanded star vector set; based on the weighted fusion algorithm, different confidence weights are assigned to the historical vectors and the current observed vectors, and they are substituted into the attitude calculation model to complete the matrix solution, breaking through the limitation of the traditional attitude determination on the minimum number of observed stars, and realizing continuous attitude determination output under the condition of low number of stars.

[0025] (2) Based on the established astronomical attitude determination error model, key influencing factors such as the number of observed stars and the accuracy of star point extraction are extracted, and a mapping function between measurement noise variance and observation conditions is constructed. During the filtering iteration process, the number of stars and star point quality in the current observation scene are identified in real time, and the measurement noise weight is dynamically adjusted according to the mapping function: when the number of observed stars is sufficient and the star point quality is high, the measurement noise weight is reduced to enhance the correction effect of star-sensitive data on filtering; when the observation conditions are poor, the measurement noise weight is increased to suppress the interference of low-quality observation data and improve the stability and robustness of integrated navigation filtering under complex conditions.

[0026] The above technical solutions are merely exemplary embodiments of the present invention. For those skilled in the art, based on the application methods and principles disclosed in the present invention, it is easy to make various types of improvements or modifications, and not limited to the methods described in the specific embodiments of the present invention. Therefore, the methods described above are merely preferred and not restrictive.

Claims

1. A multi-source error compensation and high-precision fusion method for satellite-inertial navigation systems, characterized in that, Includes the following steps: S1: Based on the known spatial coordinate system relationship, the angular velocity and specific force data of the inertial sensor are collected to calculate the initial attitude; the star sensor captures a star map, and star vectors are obtained through star point extraction and matching to calculate attitude data; S2, establish an inertial error model including gyroscope drift and accelerometer zero bias; through observability analysis, online calibration is used to compensate for installation errors, and astronomical attitude determination errors are modeled in combination with the dynamic characteristics of star number, and measurement noise is quantified; S3, design a multi-rate filtering framework to synchronize high-frequency inertial navigation and low-frequency star-sensor data, use navigation calculation, device and installation errors as state variables, attitude difference as measurement, and iteratively correct the output navigation parameters; S4. When the number of stars is insufficient, the attitude determination is completed by equivalent compensation using historical star vectors. The weight of filtered measurement noise is dynamically adjusted based on the astronomical attitude determination error model to improve the fusion stability under complex conditions.

2. The method according to claim 1, characterized in that, Step S1 specifically includes: S11. Based on the preset spatial coordinate system transformation relationship between the inertial sensor and the star sensor, a multi-source sensor synchronous acquisition system is built, including: completing the physical installation and calibration of the inertial sensor and the star sensor, and determining the rotation matrix and translation vector between their coordinate systems; secondly, starting the synchronous trigger, controlling the inertial sensor to output the carrier angular velocity and specific force raw data at a set sampling frequency, and simultaneously driving the star sensor to capture starry sky images according to the preset exposure time and frame rate, ensuring that the timestamps of the two types of sensor data are strictly aligned; S12 preprocesses the angular velocity and specific force data collected by the inertial sensor, sequentially performing zero bias compensation, scaling factor calibration, and temperature error correction. This includes: based on the preprocessed effective data, using a strapdown inertial navigation solution algorithm, iteratively solving the carrier's real-time attitude angles through quaternion differential equations; combining the initial alignment strategy, using the constraints of the gravity vector and the Earth's rotation angular velocity vector, completing the coarse and fine alignment of the inertial navigation system, and calculating the carrier's initial attitude and position information, providing an initial reference benchmark for the star sensor's calculation. S13. Star points are extracted from the star image captured by the star sensor. Gray-scale thresholding and centroid fitting algorithms are used to obtain the precise pixel coordinates of the star points. Star image recognition and matching are performed based on star catalog data. The extracted star point features are associated with the inertial frame vectors of known stars using the triangle matching method. After matching, the least squares attitude determination algorithm is used to construct the transformation relationship between the observation vector and the reference vector, and the attitude matrix of the carrier in the inertial frame is calculated. Combined with the initial position information of the inertial navigation system, the position parameters of the carrier are further corrected, and the attitude and position data that meet the navigation requirements are output.

3. The method according to claim 1, characterized in that, Step S2 specifically includes: To address the output error characteristics of inertial sensors, an inertial error model was constructed that includes systematic and random errors. The systematic error term includes gyroscope constant drift, accelerometer zero bias, and scaling factor error. The random error term uses a random walk model to characterize the white noise interference of the gyroscope and accelerometer. Based on laboratory calibration data and field test data, the parameters of the error model were fitted and validated. S22 addresses the installation angle deviation between inertial sensors and star sensors caused by vibration and deformation during carrier flight. It conducts error observability analysis, selects carrier angular motion and linear motion as observed state variables, and constructs a state equation that includes the installation error angle. It adopts an extended Kalman filter online calibration algorithm, using the deviation between the inertial navigation solution value and the star sensor attitude determination result as the measurement input, to estimate the magnitude of the installation error angle in real time and feed it back to the coordinate system transformation stage to achieve dynamic compensation of installation error. S23. Combining the dynamic changes in the number of observed stars, an astronomical attitude determination error model is established to clarify the mapping relationship between the number of observed stars and the attitude determination accuracy. For the influencing factors under different observation scenarios, including star point extraction error and star map matching mismatch rate, statistical analysis methods are used to quantify the measurement noise variance of the star sensor. Through simulation and measured data verification, the range of measurement noise values ​​under different numbers of observed stars is determined as the basis for setting the noise parameters of the integrated navigation filtering algorithm.

4. The method according to claim 1, characterized in that, Step S3 specifically includes: S31. To address the rate difference between high-frequency sampling of the inertial sensor and low-frequency output of the star sensor, a multi-rate Kalman filter framework is constructed. An interpolation extrapolation algorithm is used to extend the time scale of the low-frequency attitude data of the star sensor so that it corresponds one-to-one with the timestamp of the high-frequency data of the inertial sensor. At the same time, based on the data validity judgment criterion, abnormal sampling points of the two types of sensors are eliminated to achieve spatiotemporal synchronization of the high-frequency angular velocity and force data of the inertial navigation system and the low-frequency attitude data of the star sensor. S32 integrates inertial navigation calculation errors, sensor device errors, and installation errors into a filtered state vector. It uses the difference between the inertial navigation calculated attitude and position and the high-precision measurement results of the star sensor as the measurement, establishes the filtered state equation and measurement equation, and estimates and compensates for various errors in real time through Kalman filtering iterative calculation. The error correction is fed back to the inertial navigation calculation link to iteratively optimize the attitude, position, and velocity information of the carrier, and finally outputs fused navigation parameters that meet the requirements of high-precision navigation.

5. The method according to any one of claims 1-4, characterized in that, Step S4 specifically includes: S41. When the star sensor is affected by factors such as obstruction and light interference, and the number of observed stars is lower than the attitude determination threshold, the historical star vector compensation mechanism is activated. Specifically, this includes: retrieving high-precision star vector data from previous observations that have been verified for effectiveness, combining it with the vector information of the current small number of observed star points, and constructing an expanded set of star vectors; assigning different confidence weights to historical vectors and current observed vectors based on a weighted fusion algorithm, substituting them into the attitude calculation model to complete the matrix solution, and realizing continuous attitude determination output under low star count conditions. S42, based on the established astronomical attitude determination error model, extracts key influencing factors including the number of observed stars and the accuracy of star point extraction, and constructs a mapping function between measurement noise variance and observation conditions. During the filtering iteration process, the number of stars and star point quality in the current observation scene are identified in real time, and the measurement noise weight is dynamically adjusted according to the mapping function: when the number of observed stars is sufficient and the star point quality is high, the measurement noise weight is reduced to enhance the correction effect of star-sensitive data on filtering; when the observation conditions are poor, the measurement noise weight is increased to suppress the interference of low-quality observation data and improve the stability and robustness of integrated navigation filtering under complex conditions.

Images