Intelligent multi-source navigation method based on inertia and Beidou / visual information fusion

Abstract

The invention relates to an intelligent multi-source navigation method based on inertia and Beidou / visual information fusion, which comprises the following steps: firstly, establishing a multi-source information system of an inertial navigation system, and embedding an IMU (Inertial Measurement Unit) multi-parameter online calibration mechanism based on carrier dynamics in inertial navigation solution; establishing a tight coupling fusion framework taking an INS resolving result as a reference, and constructing tight combination measurement with the original observed quantity of the BDS and constructing re-projection error measurement with the visual feature points by utilizing INS high-frequency pose information; an intelligent adaptive filter based on inertial dynamics error model constraint is adopted, an INS error equation is used as a state prediction model, a Sage-Husa algorithm is introduced to estimate system noise statistical characteristics in real time, estimated error parameters are fed back to INS calculation through closed-loop correction, real-time correction of navigation results is achieved, and the system noise statistical characteristics are estimated in real time. Through deep tight coupling fusion and intelligent adaptive processing, navigation precision and system reliability in a complex environment are effectively improved.

Description

Technical Field

[0001] This invention relates to an intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information, belonging to the field of multi-source navigation technology. Background Technology

[0002] Navigation and positioning technology is a key support for fields such as aviation, aerospace, navigation, and unmanned systems. Due to the limitations of its own principles, a single navigation system is difficult to maintain continuous high accuracy and high reliability in complex environments, making multi-source information fusion technology an inevitable choice.

[0003] Inertial navigation systems (INS) have become the core of integrated navigation due to their autonomy, high-frequency output, and strong anti-interference capabilities. However, their integral-based solution mechanism leads to the accumulation of sensor errors over time, making them unsuitable for long-term independent operation. Combining INS with the BeiDou Navigation Satellite System (BDS) can effectively suppress error divergence. Traditional loosely combined algorithms are simple but lack robustness; while tightly combined algorithms improve performance by fusing raw observations, they fail in environments with long-term satellite signal rejection. Visual navigation calculates relative pose through image features and, when combined with INS, can compensate for inertial drift. However, it suffers from scale uncertainty and is susceptible to changes in environmental texture and lighting, making feature tracking stability and real-time performance challenges.

[0004] The existing problems with multi-source fusion navigation are: (1) Most of them adopt loose coupling or federated filtering structures and fail to establish a deep fusion architecture with INS mechanical arrangement as the core and BDS original observations and visual features as tightly coupled measurement sources; (2) The fusion algorithm adopts a fixed noise statistical model and lacks an adaptation mechanism for time-varying factors such as the degradation of BDS observation quality and changes in the stability of visual feature tracking, which leads to a decrease in estimation accuracy or even filter divergence in dynamic environments; (3) There is a lack of intelligent fault detection and isolation mechanism based on INS short-term forecast, which cannot achieve autonomous optimization of fusion weights and seamless switching of navigation modes when the performance of BDS / visual sensors degrades or fails; (4) There is a lack of effective online calibration and compensation methods for deep parameters such as IMU scaling factor error that affect the core solution accuracy of INS. Summary of the Invention

[0005] The technical problem solved by this invention is: addressing the lack of effective calibration and compensation methods in the existing technology, this invention proposes an intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information.

[0006] The present invention solves the above-mentioned technical problem through the following technical solution: An intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information includes: Real-time acquisition of angular and velocity increments of the inertial measurement unit (IMU) in the inertial navigation system, and acquisition of raw observation images and visual time-series images of the BeiDou Navigation Satellite System (BDS); An inertial navigation differential equation is established. After embedding an IMU multi-parameter calibration and compensation mechanism in the inertial recursive solution process, the inertial recursive solution is performed. The zero bias and scale factor errors of the gyroscope and accelerometer are estimated and compensated in real time to obtain the current attitude, velocity and position information of the carrier. Based on the carrier's attitude, velocity, and position information, the high-frequency pose information obtained through inertial recursion calculation is used to construct a tight combination measurement with the original BDS observations, and a reprojection error measurement is constructed with visual feature points. An adaptive filter is constructed using the inertial dynamics error model as a constraint, and the inertial dynamics error model is used as the state prediction model. Tightly combined measurements are used as observation update quantities to estimate and adjust the system noise of the inertial navigation system in real time. An adaptive filter is used to estimate the IMU error and navigation error parameters. The results are then fed back to the inertial dynamics navigation solution through closed-loop correction. The navigation solution results are corrected and error suppressed in real time to obtain the corrected navigation results.

[0007] When using Kalman filtering to achieve real-time estimation and feedback compensation of the zero bias and scale factor errors of the gyroscope and accelerometer, the IMU error parameters are augmented to the inertial navigation error state vector to achieve real-time observability estimation. The augmented and expanded state vector is as follows:

[0008] In the formula, f For misalignment angle, dv For speed error, d P is the position error. e b For zero bias of the gyroscope, b To achieve zero bias in the accelerometer, δs g and δs a The scaling factor errors of the gyroscope and accelerometer are respectively used to form a 21-dimensional error state vector.

[0009] During the tight-fitting measurement process, when the BeiDou satellite navigation system signal is lost and enters a pure inertial or visual-assisted mode, the physical scale information output by the inertial navigation system velocity is used to perform scale recovery on the scale-free relative displacement calculated by monocular vision. The scale factor required for scale recovery... s for:

[0010] In the formula,v ins For the speed of INS solution, Δ t For time intervals, Δp The scale-free displacement vector computed for the visual camera; The scale factor is used to normalize the visually estimated displacement and map to maintain the scale accuracy of the inertial navigation system during periods without satellite signals.

[0011] In the process of obtaining high-frequency pose information through inertial recursive calculation, the high-frequency pose information is the predicted pixel coordinates, which is obtained through the pose matrix of the previous time step obtained by inertial recursive calculation. With position The current moment and The data collected by the visual front end of the inertial navigation system k The 3D coordinates of feature points in the time-lapse camera coordinate system are transformed to... k Obtain the predicted pixel coordinates in the camera coordinate system at time +1. for:

[0012] In the formula, K For the camera intrinsic parameter matrix, For the camera-to-IMU mounting matrix, Its inverse matrix, , This is the depth estimate.

[0013] The method for obtaining the visual reprojection error is as follows: Based on predicted pixel coordinates , k The actual pixel coordinates observed at time +1 The calculation method is as follows:

[0014] In the formula, This refers to visual reprojection error.

[0015] After using compact combination measurements as observation update quantities, the Sage-Husa adaptive algorithm is introduced to estimate system noise in real time. The method is as follows: The system process noise covariance matrix is ​​estimated and adjusted in real time using the Sage-Husa adaptive algorithm. Q Covariance matrix of measurement noise R Based on the dual-matrix output as system noise, the process noise covariance matrix... Q Covariance matrix of measurement noise R They are respectively:

[0016] In the formula, for k Filtered information at any given time For the measurement matrix, P The state estimation error covariance matrix is... K k Here is the filter gain matrix. F k|k-1 Here is the state transition matrix. d k It is a forgetting factor.

[0017] The adaptive filter is equipped with a fault detection and isolation FDI mechanism. It takes the short-term prediction results of the inertial navigation system as input, calculates the chi-square test statistic of the Beidou satellite navigation system and visual measurement information, and determines that the chi-square test statistic of any information source is continuously exceeded by a preset threshold. The measurement noise covariance matrix R in the adaptive filter is then adaptively adjusted.

[0018] The navigation error parameters are used for all-state feedback correction. The navigation error parameters estimated by the adaptive filter include the misalignment angle. f Speed ​​error d v Position error d P These are all used to correct the corresponding state variables in the inertial navigation solution, and the estimated IMU error is fed back to the preprocessing stage of the IMU raw data for real-time compensation.

[0019] The adaptive filter employs a closed-loop correction architecture, and the method for error suppression using this architecture is as follows: The estimated IMU error parameter ε b , b ,δκ g ,δκ a The data is fed back to the INS mechanical orchestration front end in real time for compensation, and the navigation error parameters φ and δ are also fed back. v δ P The feedback correction method is input into the INS solution to form a fully closed-loop error suppression mechanism.

[0020] The advantages of this invention compared to the prior art are: (1) The present invention provides an intelligent multi-source navigation method based on the fusion of inertial and Beidou / visual information. With inertial navigation as the core, it can deeply integrate Beidou and visual information and has the capabilities of online calibration, adaptive filtering and intelligent decision-making. It can achieve continuous, robust and high-precision navigation and positioning in complex environments. (2) This invention suppresses the divergence of inertial sensor errors by using an online calibration and compensation mechanism for multiple parameters of the IMU based on carrier dynamics; and establishes a tightly coupled fusion architecture with inertial navigation as a unified spatiotemporal reference to achieve deep fusion of BeiDou original observations and visual features in the observation domain. (3) The present invention is based on the intelligent multi-source navigation method of inertial and Beidou / visual information fusion. It adopts an intelligent filter based on the ψ angle error model and Sage-Husa adaptive algorithm, which tracks the time-varying noise characteristics in real time and has the ability to autonomously diagnose and isolate faults. Through full closed-loop correction, the error estimate is fed back to each link of INS solution, which optimizes the sensor front end to navigation output and significantly improves high-precision navigation, system robustness and autonomous reliability in complex environments. Attached Figure Description

[0021] Figure 1 A flowchart of the intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information provided by this invention; Figure 2 A schematic diagram of the tightly coupled deep fusion framework provided by the present invention; Figure 3 The schematic diagram of the intelligent adaptive filtering and closed-loop correction provided by this invention. Detailed Implementation

[0022] An intelligent multi-source navigation method based on the fusion of inertial navigation system (INS) and BeiDou / visual information is proposed. First, a multi-source information system of the INS is established, embedding an online multi-parameter calibration mechanism of the IMU based on carrier dynamics into the INS solution. A tightly coupled fusion architecture based on the INS solution results is established, utilizing INS high-frequency pose information and original BDS observations to construct tightly combined measurements, and constructing reprojection error measurements with visual feature points. An intelligent adaptive filter based on inertial dynamics error model constraints is employed, using the INS error equation as the state prediction model. The Sage-Husa algorithm is introduced to estimate the system noise statistics in real time. This invention feeds back the estimated error parameters to the INS solution through closed-loop correction, achieving real-time correction of the navigation results. Through deep tightly coupled fusion and intelligent adaptive processing, the navigation accuracy and system reliability in complex environments are significantly improved.

[0023] The intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information involves the following steps: Real-time acquisition of angular and velocity increments of the inertial measurement unit (IMU) in the inertial navigation system, and acquisition of raw observation images and visual time-series images of the BeiDou Navigation Satellite System (BDS); An inertial navigation differential equation is established. After embedding an IMU multi-parameter calibration and compensation mechanism in the inertial recursive solution process, the inertial recursive solution is performed. The zero bias and scale factor errors of the gyroscope and accelerometer are estimated and compensated in real time to obtain the current attitude, velocity and position information of the carrier. Based on the carrier's attitude, velocity, and position information, the high-frequency pose information obtained through inertial recursion calculation is used to construct a tight combination measurement with the original BDS observations, and a reprojection error measurement is constructed with visual feature points. An adaptive filter is constructed using the inertial dynamics error model as a constraint, and the inertial dynamics error model is used as the state prediction model. Tightly combined measurements are used as observation update quantities to estimate and adjust the system noise of the inertial navigation system in real time. An adaptive filter is used to estimate IMU error and navigation error parameters, which are then fed back to the inertial dynamics navigation solution through closed-loop correction. The navigation solution results are then corrected and error suppressed in real time to obtain the corrected navigation results.

[0024] When using Kalman filtering to achieve real-time estimation and feedback compensation of the zero bias and scale factor errors of the gyroscope and accelerometer, the IMU error parameters are augmented to the inertial navigation error state vector to achieve real-time observability estimation. The augmented and expanded state vector is as follows:

[0025] In the formula, f For misalignment angle, dv For speed error, d P is the position error. e b For zero bias of the gyroscope, b To achieve zero bias in the accelerometer, δs g and δs a The scaling factor errors of the gyroscope and accelerometer are respectively used to form a 21-dimensional error state vector.

[0026] During tight-fitting measurement, when the BeiDou Navigation Satellite System signal is lost and enters pure inertial or vision-assisted mode, the physical scale information from the velocity output of the inertial navigation system is used to perform scale recovery on the scale-free relative displacement calculated by monocular vision. The scale factor required for scale recovery... s for:

[0027] In the formula, v ins For the speed of INS solution, Δ t For time intervals, ΔpThe scale-free displacement vector computed for the visual camera; The scale factor is used to normalize the visually estimated displacement and map to maintain the scale accuracy of the inertial navigation system during periods without satellite signals.

[0028] In the process of obtaining high-frequency pose information through inertial recursive calculation, the high-frequency pose information is the predicted pixel coordinates, which is obtained through the pose matrix of the previous time step obtained by inertial recursive calculation. With position The current moment and The data collected by the visual front end of the inertial navigation system k The 3D coordinates of feature points in the time-lapse camera coordinate system are transformed to... k Obtain the predicted pixel coordinates in the camera coordinate system at time +1. for:

[0029] In the formula, K For the camera intrinsic parameter matrix, For the camera-to-IMU mounting matrix, Its inverse matrix, , This is the depth estimate.

[0030] The method for obtaining visual reprojection error is as follows: Based on predicted pixel coordinates , k The actual pixel coordinates observed at time +1 The calculation method is as follows:

[0031] In the formula, This refers to visual reprojection error.

[0032] After using compact combination measurements as observation update quantities, the Sage-Husa adaptive algorithm is introduced to estimate system noise in real time. The method is as follows: The system process noise covariance matrix is ​​estimated and adjusted in real time using the Sage-Husa adaptive algorithm. Q Covariance matrix of measurement noise R Based on the dual-matrix output as system noise, the process noise covariance matrix... Q Covariance matrix of measurement noise R They are respectively:

[0033] In the formula, for k Filtered information at any given time For the measurement matrix,P The state estimation error covariance matrix is... K k Here is the filter gain matrix. F k|k-1 Here is the state transition matrix. d k It is a forgetting factor.

[0034] The adaptive filter is equipped with a fault detection and isolation FDI mechanism. It takes the short-term prediction results of the inertial navigation system as input and calculates the chi-square test statistic of the Beidou satellite navigation system and visual measurement information. When the chi-square test statistic of any information source continuously exceeds the preset threshold, it is determined to be faulty, and the measurement noise covariance matrix R in the adaptive filter is adaptively adjusted.

[0035] Navigation error parameters are used for full-state feedback correction. The navigation error parameters estimated by the adaptive filter include the misalignment angle. f Speed ​​error dv Position error δP These are all used to correct the corresponding state variables in the inertial navigation solution, and the estimated IMU error is fed back to the preprocessing stage of the IMU raw data for real-time compensation.

[0036] The adaptive filter employs a closed-loop correction architecture. The method for error suppression using this closed-loop correction architecture is as follows: The estimated IMU error parameter ε b , b ,δκ g ,δκ a The error is fed back to the INS mechanical orchestration front end in real time for compensation. At the same time, the navigation error parameters φ, δv, and δP are input into the INS calculation through feedback correction, forming a closed-loop error suppression mechanism.

[0037] The following description, in conjunction with the accompanying drawings and preferred embodiments, provides further details: In the current embodiment, such as Figure 1 As shown, the intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information includes the following specific steps: S1. Establish a multi-source information acquisition and processing system with the inertial navigation system (INS) as the solution benchmark to acquire in real time the angular / velocity increments of the inertial measurement unit (IMU), the raw observations of the BeiDou satellite navigation system (BDS), and visual time-series images. S2. Based on the inertial navigation differential equation, an IMU multi-parameter calibration and compensation mechanism based on the dynamic characteristics of the carrier is embedded in the inertial recursive solution. The zero bias and scale factor errors of the gyroscope and accelerometer are estimated and compensated in real time to obtain the carrier attitude, velocity and position. IMU multi-parameter calibration and compensation employs online estimation using Kalman filtering, augmenting the IMU error parameters into the inertial navigation error state vector. Real-time estimation is then performed based on system observability, resulting in the expanded state vector:

[0038] In the formula, f For misalignment angle, dv For speed error, d P is the position error. e b For zero bias of the gyroscope, b To achieve zero bias in the accelerometer, δs g and δs a The scaling factor errors of the gyroscope and accelerometer are respectively used to form a 21-dimensional error state vector.

[0039] S3. Establish a tightly coupled fusion framework with the inertial navigation solution results as a unified spatiotemporal reference. Through the high-frequency pose information solved by INS, construct a tightly coupled measurement with the original BDS observations and construct a reprojection error measurement with visual feature points. During visual / INS tightly coupled measurements, when the BDS signal is lost and the system enters a pure inertial / visual-assisted mode, the physical scale information output by the INS velocity is used to perform scale recovery on the scale-free relative displacement calculated by monocular vision. Its scale factor... s for:

[0040] In the formula, v ins For the speed of INS solution, Δ t For time intervals, Δp The scale-free displacement vector calculated by the visual camera is used to normalize the visually estimated displacement and map, maintaining the scale accuracy of navigation during periods without satellite signals.

[0041] The pose matrix of the previous time step calculated by INS With position and the current moment and The data obtained through the visual front end will be in k The 3D coordinates of feature points in the time-lapse camera coordinate system are transformed to... k In the camera coordinate system at time +1, the predicted pixel coordinates are obtained. ,

[0042] andk The actual pixel coordinates observed at time +1 Constituting visual reprojection error:

[0043] In the formula, K For the camera intrinsic parameter matrix, For the camera-to-IMU mounting matrix, Its inverse matrix, , This is the depth estimate.

[0044] S4. An intelligent adaptive filter based on the inertial dynamics error model constraint is adopted, with the INS error equation as the state prediction model and compact combination measurement as the observation update. The Sage-Husa adaptive algorithm is introduced to estimate and adjust the system noise in real time. In S4, the Sage-Husa adaptive algorithm is used to estimate and adjust the system process noise covariance matrix in real time. Q Covariance matrix of measurement noise R The recursive formula is:

[0045] In the formula, for k Filtered information at any given time For the measurement matrix, P The state estimation error covariance matrix is... K k Here is the filter gain matrix. F k|k-1 Here is the state transition matrix. d k It is a forgetting factor.

[0046] The S4 intelligent adaptive filter includes a fault detection and isolation FDI mechanism based on inertial information: using the short-term forecast results of INS as a reference, the chi-square test statistics of BDS and visual measurement information are calculated respectively. When the statistics of a certain information source continuously exceed the preset threshold, it is determined to be faulty, and its measurement noise covariance R in the filter is adaptively adjusted.

[0047] S5. The IMU error and navigation error parameters estimated by the adaptive filter are fed back to the INS navigation solution through closed-loop correction to correct and suppress the error of the INS solution in real time, so as to obtain high-precision navigation results.

[0048] The navigation error parameters estimated by the filter, the misalignment angle f Speed ​​error dv Position error δPThe corresponding state variables in the INS recursive solution are corrected, and the estimated IMU error is fed back to the preprocessing stage of the raw IMU data for real-time compensation. The intelligent adaptive filter adopts a closed-loop correction architecture, specifically: the estimated IMU error parameter ε is used to correct the state variables in the INS recursive solution. b , b ,δκ g ,δκ a The error is fed back to the INS mechanical orchestration front end in real time for compensation. At the same time, the navigation error parameters φ, δv, and δP are input into the INS calculation through feedback correction, forming a closed-loop error suppression mechanism.

[0049] This embodiment is based on an intelligent multi-source navigation method that integrates inertial and BeiDou / visual information. It uses an intelligent filter based on the ψ-angle error model and the Sage-Husa adaptive algorithm to track time-varying noise characteristics in real time and has the ability to autonomously diagnose and isolate faults. Through full closed-loop correction, the error estimate is fed back to each link of the INS solution, optimizing the sensor front-end to the navigation output. This significantly improves high-precision navigation, system robustness, and autonomous reliability in complex environments.

[0050] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solutions of the present invention by utilizing the methods and techniques disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention shall fall within the protection scope of the technical solutions of the present invention.

[0051] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. An intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information, characterized in that... include: Real-time acquisition of angular and velocity increments of the inertial measurement unit (IMU) in the inertial navigation system, and acquisition of raw observation images and visual time-series images of the BeiDou Navigation Satellite System (BDS); An inertial navigation differential equation is established. After embedding an IMU multi-parameter calibration and compensation mechanism in the inertial recursive solution process, the inertial recursive solution is performed. The zero bias and scale factor errors of the gyroscope and accelerometer are estimated and compensated in real time to obtain the current attitude, velocity and position information of the carrier. Based on the carrier's attitude, velocity, and position information, the high-frequency pose information obtained through inertial recursion calculation is used to construct a tight combination measurement with the original BDS observations, and a reprojection error measurement is constructed with visual feature points. An adaptive filter is constructed using the inertial dynamics error model as a constraint, and the inertial dynamics error model is used as the state prediction model. Tightly combined measurements are used as observation update quantities to estimate and adjust the system noise of the inertial navigation system in real time. An adaptive filter is used to estimate the IMU error and navigation error parameters. The results are then fed back to the inertial dynamics navigation solution through closed-loop correction. The navigation solution results are corrected and error suppressed in real time to obtain the corrected navigation results.

2. The intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information according to claim 1, characterized in that: When using Kalman filtering to achieve real-time estimation and feedback compensation of the zero bias and scale factor errors of the gyroscope and accelerometer, the IMU error parameters are augmented to the inertial navigation error state vector to achieve real-time observability estimation. The augmented and expanded state vector is as follows: In the formula, φ For misalignment angle, δv For speed error, δ P is the position error. ε b For zero bias of the gyroscope, b To achieve zero bias in the accelerometer, δ s g and δs a The scaling factor errors of the gyroscope and accelerometer are respectively used to form a 21-dimensional error state vector.

3. The intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information according to claim 1, characterized in that: During the tight-fitting measurement process, when the BeiDou satellite navigation system signal is lost and enters a pure inertial or visual-assisted mode, the physical scale information output by the inertial navigation system velocity is used to perform scale recovery on the scale-free relative displacement calculated by monocular vision. The scale factor required for scale recovery... s for: In the formula, v ins For the speed of INS solution, Δ t For time intervals, Δp The scale-free displacement vector computed for the visual camera; The scale factor is used to normalize the visually estimated displacement and map to maintain the scale accuracy of the inertial navigation system during periods without satellite signals.

4. The intelligent multi-source navigation method based on inertial and BeiDou / visual information fusion according to claim 1, characterized in that: In the process of obtaining high-frequency pose information through inertial recursive calculation, the high-frequency pose information is the predicted pixel coordinates, which is obtained through the pose matrix of the previous time step obtained by inertial recursive calculation. With position The current moment and The data collected by the visual front end of the inertial navigation system k The 3D coordinates of feature points in the time-lapse camera coordinate system are transformed to... k Obtain the predicted pixel coordinates in the camera coordinate system at time +1. for: In the formula, K For the camera intrinsic parameter matrix, For the camera-to-IMU mounting matrix, Its inverse matrix, , This is the depth estimate.

5. The intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information according to claim 4, characterized in that: The method for obtaining the visual reprojection error is as follows: Based on predicted pixel coordinates , k The actual pixel coordinates observed at time +1 The calculation method is as follows: In the formula, This refers to visual reprojection error.

6. The intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information according to claim 4, characterized in that: After using compact combination measurements as observation update quantities, the Sage-Husa adaptive algorithm is introduced to estimate system noise in real time. The method is as follows: The system process noise covariance matrix is ​​estimated and adjusted in real time using the Sage-Husa adaptive algorithm. Q Covariance matrix of measurement noise R Based on the dual-matrix output as system noise, the process noise covariance matrix... Q Covariance matrix of measurement noise R They are respectively: In the formula, for k Filtered information at any given time For the measurement matrix, P The state estimation error covariance matrix is... K k Here is the filter gain matrix. Φ k|k-1 Here is the state transition matrix. d k It is a forgetting factor.

7. The intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information according to claim 6, characterized in that: The adaptive filter is equipped with a fault detection and isolation FDI mechanism. It takes the short-term prediction results of the inertial navigation system as input, calculates the chi-square test statistic of the Beidou satellite navigation system and visual measurement information, and determines that the chi-square test statistic of any information source is continuously exceeded by a preset threshold. The measurement noise covariance matrix R in the adaptive filter is then adaptively adjusted.

8. The intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information according to claim 6, characterized in that: The navigation error parameters are used for all-state feedback correction. The navigation error parameters estimated by the adaptive filter include the misalignment angle. φ Speed ​​error δ v Position error δ P These are all used to correct the corresponding state variables in the inertial navigation solution, and the estimated IMU error is fed back to the preprocessing stage of the IMU raw data for real-time compensation.

9. The intelligent multi-source navigation method based on the fusion of inertial and BeiDou / visual information according to claim 7, characterized in that: The adaptive filter employs a closed-loop correction architecture, and the method for error suppression using this architecture is as follows: The estimated IMU error parameter ε b , b ,δκ g ,δκ a The data is fed back to the INS mechanical orchestration front end in real time for compensation, and the navigation error parameters φ and δ are also fed back. v δ P The feedback correction method is input into the INS solution to form a fully closed-loop error suppression mechanism.

Images