An Adaptive Lightweight Online Calibration Method for Extrinsic Parameters of an In-Vehicle Integrated Navigation System

Abstract

This invention relates to an adaptive, lightweight, online calibration method for extrinsic parameters of a vehicle-mounted integrated navigation system. The method includes: Step 1, augmenting the navigation system extrinsic parameters to a Kalman filter to construct a system error model; Step 2, obtaining the observability analysis results of the extrinsic parameters through observability matrix analysis; Step 3, grouping the extrinsic parameters according to the required maneuvering mode to achieve piecewise decoupled calibration and Kalman filter dimensionality reduction; Step 4, establishing corresponding state equations and measurement equations through the system error model, and then constructing an adaptive feedback factor; Step 5, using the adaptive feedback factor to reflect the accuracy of error estimation, determining whether to provide feedback to the extrinsic parameters and the magnitude of the feedback. This invention solves the technical problems in the prior art, such as the signal being easily affected by external factors, leading to reduced navigation accuracy, and the system extrinsic parameter calibration relying on the vehicle performing specific continuous maneuvers.

Description

Technical Field

[0001] This invention relates to the field of parameter calibration technology, and in particular to an adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system. Background Technology

[0002] In recent years, integrated navigation using Global Navigation Satellite System (GNSS) and Inertial Navigation System (INS) has become a common navigation solution for passenger vehicles. However, GNSS signals are susceptible to interference from the external environment, leading to positioning drift or even signal loss, thus reducing navigation accuracy. To address this, odometers (ODO) and non-holonomic constraints (NHC) are typically used to suppress positioning drift during GNSS signal interference. However, in practical applications, spatial errors often exist between the Inertial Measurement Unit (IMU), the GNSS receiver, and the vehicle coordinate system. Ignoring or improperly handling these errors can reduce the effectiveness of GNSS, ODO, and NHC in assisting the navigation system, and even adversely affect navigation accuracy. Existing methods rely on specific maneuvers performed by the vehicle to ensure parameter estimation accuracy. When the vehicle's trajectory dynamic excitation is insufficient, parameter estimation is easily interfered with, leading to decreased accuracy. In addition, current research on the observability analysis of extrinsic parameters of INS / GNSS / ODO integrated navigation systems under different maneuvers is still relatively limited, and there is also little research on the mechanism of the influence of instantaneous vehicle motion on the observability of system state. Summary of the Invention

[0003] In view of the above problems, the present invention provides an adaptive lightweight online calibration method for extrinsic parameters of a vehicle-mounted integrated navigation system, in order to solve the technical problems in the prior art, such as the signal being easily affected by external factors, resulting in reduced navigation accuracy, and the reliance on the carrier to perform specific maneuvers.

[0004] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of a vehicle-mounted integrated navigation system. The method includes: Step 1, augmenting the navigation system extrinsic parameters to a Kalman filter, including the lever arm error between the GNSS antenna and the IMU, and the lever arm error between the IMU and the vehicle body, installation deviation angle error, and odometer scaling factor error, to construct a system error model; Step 2, obtaining the system's observability matrix based on the piecewise linear time-invariant system observability analysis method and the error system model, and obtaining the extrinsic parameter observability analysis results through observability matrix analysis; Step 3, based on the extrinsic parameter observability... Based on the observational analysis results, the external parameters are grouped according to the required maneuvering mode to achieve segmented decoupling calibration and Kalman filter dimensionality reduction, thereby improving the real-time performance and accuracy of the external parameter calibration. Step 4: Based on the Kalman filter state variables selected during the segmented dimensionality reduction process, the corresponding state equations and measurement equations are established through the system error model, and then an adaptive feedback factor is constructed. Step 5: Through the adaptive feedback factor, the accuracy of the error estimation is reflected, and it is determined whether to provide feedback to the external parameters and the magnitude of the feedback, thereby improving the accuracy and reliability of the external parameter estimation and realizing the adaptive online calibration of the external parameters.

[0005] Furthermore, the method also includes: Step 0, acquiring sensor information through ODO, IMU, and GNSS, and obtaining navigation information through INS, for subsequent external parameter calibration processing.

[0006] Furthermore, the method also includes: step 6, after completing the adaptive calibration of the extrinsic parameters of the vehicle-mounted integrated navigation system, outputting navigation results for vehicle use.

[0007] Furthermore, step 1 includes: Step 11, calculating the position measurement equation between INS and GNSS based on the boom arm between the GNSS antenna and the IMU: ,in, Let be the coordinate transformation matrix from system b to system n. The pole arm between the GNSS antenna and the IMU. For attitude error, This refers to the boom arm error between the GNSS antenna and the IMU. This is the coordinate transformation matrix. To measure position noise; Step 12, based on the speed measurement provided by the odometer and the vehicle non-integrity constraints, obtain the actual speed equation of the vehicle in the vehicle coordinate system:

[0008] ,

[0009] in, The scale factor error of the odometer. This represents the actual speed of the vehicle in the vehicle coordinate system. For speed measurement noise; Step 13, set the rear axle center of the non-drive wheel of the vehicle or the contact point between the non-drive wheel and the ground as reference point P, and calculate the vehicle's velocity equation at point P in the vehicle coordinate system using the IMU: ,

[0010] in, For installation deviation angle error, The coordinate transformation matrix is ​​composed of the installation angles. For attitude error, For speed error, For angular velocity error, For the odometer lever arm, For the odometer lever arm error; Step 14, combining the actual vehicle velocity equation in the vehicle coordinate system and the vehicle velocity equation at point P in the vehicle coordinate system, the velocity measurement equation between INS and ODO is obtained as follows: ,in, Coordinate transformation matrix, V n The velocity in the n-system is Let be the rotational angular velocity of the carrier; Step 15, establish the navigation system state variables as: ,in, For attitude error, For speed error, For positional error, and These are the zero bias values ​​for the gyroscope and the accelerometer, respectively. and These represent the residual pitch and yaw installation angles to be estimated between the B and M series, respectively. The scaling factor error of the odometer; Step 16, model all external parameters to be estimated as random constants, then the state equation of the filter is: ,in, and This refers to the non-zero block matrices in the state transition matrix and system noise matrix of a traditional 15-dimensional loosely combined EKF filter. For the system noise of INS; Step 17, construct the system error model from the state variables of the navigation system, the transition equation of the filter, the position measurement equation and the velocity measurement equation.

[0011] Furthermore, step 2 includes: step 21, obtaining the system's state matrix by combining the state equation of the filter with the observability analysis method for piecewise linear time-invariant systems. Step 22: Assuming the position measurement information and velocity measurement information are synchronized at the same frequency, merge the position measurement equation and velocity measurement equation to form the system's total observation matrix H; Step 23: From the total observation matrix H, the total observability matrix of the system from the beginning to the r-th time period is obtained as follows: , ,in, For the first The length of each time period Let be the observability matrix at time j. Let J be the constant state matrix of the system in the j-th time period. This is the constant measurement matrix for the j-th time period of the system; Step 24, based on the PWCS observability analysis method, through the observability matrix... The observability of the state variables is analyzed to obtain the observability analysis results of the external parameters.

[0012] Furthermore, the results of the external parameter observability analysis include: when the vehicle is stationary, all system external parameters are unobservable; when the vehicle is moving at a constant speed in a straight line, both the pitch installation angle error and the odometer scaling factor error are observable; when the vehicle is moving at varying speeds in a straight line, the heading installation angle error is observable; when the vehicle is moving at an angle around a certain direction, it will generate dynamic excitation on the GNSS antenna mast arm error and the P-point mast arm error in the other two orthogonal directions, thereby being separated and observed by the system.

[0013] Furthermore, step 3 includes: step 31, grouping the external parameters according to the required maneuvering mode based on the external parameter observability analysis results; step 32, decoupling and calibrating the first group of external parameters, the first group of external parameters including the GNSS arm error in the x-direction. GNSS lever arm error in the y-direction GNSS lever arm error in the z-direction Step 33: After the first set of extrinsic parameters is calibrated, accurate pose information is obtained through INS / GNSS combination to assist in the decoupling calibration of the second set of extrinsic parameters, which includes pitch installation angle error. Heading installation angle error Scale factor error of odometer Step 34: After the second set of extrinsic parameters is calibrated, the third set of extrinsic parameters is decoupled and calibrated. The third set of extrinsic parameters includes the ODO lever arm error in the x-direction. ODO lever arm error in the y direction ODO lever arm error in the z-direction Step 35: When the mean and standard deviation of a certain external parameter change less than the preset value, the calibration of the external parameter ends; otherwise, proceed to step 32 to recalibrate the external parameter.

[0014] Furthermore, the maneuvering modes include stationary, uniform, variable speed linear motion, and angular motion.

[0015] Furthermore, step 4 includes: step 41, establishing the corresponding state equation and measurement equation through the system error model to obtain the observation equation. ,in, To extract the observability matrix of the system, Let Z be the initial state value and Z be the observation vector; Step 42, for Perform SVD decomposition to obtain the observability of different state variables. Then, by restricting the definition interval of the local observability matrix to the piecewise interval, the observability of different state quantities at the current time can be obtained; Step 43, according to the function... Construct an instantaneous adaptive feedback factor, where, Let be the coefficient value of the i-th state variable at time j in the observability matrix. for The reference threshold.

[0016] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of a vehicle-mounted integrated navigation system. It is mainly used to solve the problems of existing technologies, such as the need for the vehicle to perform specific maneuvers to ensure the accuracy of parameter estimation, the susceptibility of parameter estimation to interference and the decrease in accuracy when the dynamic excitation of the vehicle trajectory is insufficient, and the relatively limited observability analysis of extrinsic parameters of INS / GNSS / ODO integrated navigation systems under different maneuvers, as well as the lack of research on the influence mechanism of the instantaneous motion of the vehicle on the observability of the system state. Attached Figure Description

[0017] Figure 1 A flowchart of an adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system provided by the present invention;

[0018] Figure 2 Flowchart of another adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system provided by the present invention;

[0019] Figure 3 This is a flowchart of the method for constructing a systematic error model provided by the present invention;

[0020] Figure 4 This is a flowchart of the method for obtaining extrinsic parameter observability analysis results through observability matrix analysis provided by the present invention;

[0021] Figure 5 This is a flowchart of the segmented decoupling calibration and Kalman filter dimensionality reduction method provided by the present invention;

[0022] Figure 6 This is a flowchart of the method for constructing adaptive feedback factors provided by the present invention. Detailed Implementation

[0023] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0024] Example 1:

[0025] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system, which is as follows: Figure 1 As shown, the method includes the following steps.

[0026] Step 1: Augment the extrinsic parameters of the navigation system to the Kalman filter, including the lever error between the GNSS antenna and the IMU, the lever error between the IMU and the vehicle body, the installation deviation angle error, and the scaling factor error of the odometer, and construct a system error model;

[0027] Step 2: Based on the observability analysis method for piecewise linear time-invariant systems and combined with the error system model, obtain the observability matrix of the system, and obtain the observability analysis results of the extrinsic parameters through observability matrix analysis;

[0028] Step 3: Based on the results of the external parameter observability analysis, the external parameters are grouped according to the required maneuvering mode to achieve segmented decoupling calibration and Kalman filter dimensionality reduction, thereby improving the real-time performance and accuracy of the external parameter calibration.

[0029] Step 4: Based on the state variables of the Kalman filter selected during the piecewise dimensionality reduction process, establish the corresponding state equations and measurement equations through the system error model, and then construct the adaptive feedback factor.

[0030] Step 5: By using an adaptive feedback factor to reflect the accuracy of error estimation, it determines whether to provide feedback to the external parameters and the extent of such feedback, thereby improving the accuracy and reliability of external parameter estimation and achieving adaptive online calibration of the external parameters.

[0031] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of a vehicle-mounted integrated navigation system. It is mainly used to solve the problems of existing technologies, such as the need for the vehicle to perform specific maneuvers to ensure the accuracy of parameter estimation, the susceptibility of parameter estimation to interference and the decrease in accuracy when the dynamic excitation of the vehicle trajectory is insufficient, and the relatively limited observability analysis of extrinsic parameters of INS / GNSS / ODO integrated navigation systems under different maneuvers, as well as the lack of research on the influence mechanism of the instantaneous motion of the vehicle on the observability of the system state.

[0032] Example 2:

[0033] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system, such as... Figure 2 As shown, the method includes the following steps.

[0034] Step 0: Obtain sensor information through ODO, IMU, and GNSS, and obtain navigation information through INS for later external parameter calibration processing.

[0035] The IMU is an inertial measurement unit, a sensing device used to measure and report the acceleration and angular velocity of an object. The ODO provides vehicle speed, the GNSS provides positioning, navigation, and time information, and the INS provides position, velocity, and attitude information.

[0036] Step 1: Augment the system extrinsic parameters to the Kalman filter, including the lever error between the GNSS antenna and the IMU, the lever error between the IMU and the vehicle body, the installation deviation angle error, and the odometer scaling factor error, and construct a system error model;

[0037] Step 2: Based on the observability analysis method for piecewise linear time-invariant systems and combined with the error system model, obtain the observability matrix of the system, and obtain the observability analysis results of the extrinsic parameters through observability matrix analysis;

[0038] Step 3: Based on the results of the external parameter observability analysis, the external parameters are grouped according to the required maneuvering mode to achieve segmented decoupling calibration and Kalman filter dimensionality reduction, thereby improving the real-time performance and accuracy of the external parameter calibration.

[0039] Step 4: Based on the state variables of the Kalman filter selected during the piecewise dimensionality reduction process, establish the corresponding state equations and measurement equations through the system error model, and then construct the adaptive feedback factor.

[0040] Step 5: By using an adaptive feedback factor to reflect the accuracy of error estimation, it determines whether to provide feedback to the external parameters and the extent of such feedback, thereby improving the accuracy and reliability of external parameter estimation and achieving adaptive online calibration of the external parameters.

[0041] Step 6: After completing the adaptive calibration of the extrinsic parameters of the vehicle navigation system, output the navigation results for vehicle use.

[0042] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of a vehicle-mounted integrated navigation system. It is mainly used to solve the problems of existing technologies, such as the need for the vehicle to perform specific maneuvers to ensure the accuracy of parameter estimation, the susceptibility of parameter estimation to interference and the decrease in accuracy when the dynamic excitation of the vehicle trajectory is insufficient, and the relatively limited observability analysis of extrinsic parameters of INS / GNSS / ODO integrated navigation systems under different maneuvers, as well as the lack of research on the influence mechanism of the instantaneous motion of the vehicle on the observability of the system state.

[0043] Example 3:

[0044] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system, such as... Figure 1 As shown, the method includes the following steps.

[0045] Step 1: Add system extrinsic parameters to the Kalman filter, including the lever error between the GNSS antenna and the IMU, the lever error between the IMU and the vehicle body, the installation deviation angle error, and the odometer scaling factor error, to construct a system error model;

[0046] Because when vehicle sensors are actually installed, there is usually a lever between the GNSS antenna and the IMU. This results in a fixed spatial deviation between the INS calculated position and the GNSS receiver output position; the linkage between the IMU and the vehicle body, the installation deviation angle, and the scaling factor error of the odometer also affect the system's overall accuracy. To further improve the fusion effect of multi-source sensor information, an error system model needs to be established to estimate and calibrate the above-mentioned external parameters. Figure 3 As shown, step 1 includes the following steps.

[0047] Step 11: Based on the boom arm between the GNSS antenna and the IMU, calculate the position measurement equation between the INS and GNSS antennas:

[0048] ,

[0049] in, Let be the coordinate transformation matrix from system b to system n. The pole arm between the GNSS antenna and the IMU. For attitude error, This refers to the boom arm error between the GNSS antenna and the IMU. This is the coordinate transformation matrix. Noise for position measurement;

[0050] Step 12: Based on the speed measurement provided by the odometer and the vehicle non-integrity constraints, obtain the actual speed equation of the vehicle in the vehicle coordinate system:

[0051] ,

[0052] in, The scale factor error of the odometer. This represents the actual speed of the vehicle in the vehicle coordinate system. For speed measurement noise;

[0053] The vehicle coordinate system refers to a coordinate system established with the vehicle's center of mass as the center, also known as the m-system. In addition, autonomous driving typically involves the following coordinate systems: geocentric coordinate system (e-system), inertial coordinate system (i-system), world coordinate system (w-system), navigation coordinate system (n-system), IMU coordinate system (b-system), camera coordinate system, and LiDAR coordinate system. These coordinate systems are applicable to the output information of different sensors and can be transformed between them using matrices, allowing for effective fusion of the output information from various sensors.

[0054] Step 13: Set the rear axle center of the non-drive wheels or the contact point between the non-drive wheels and the ground as reference point P. The IMU then calculates the vehicle's velocity equation at point P in the vehicle coordinate system as follows:

[0055] ,

[0056] in, For installation deviation angle error, The coordinate transformation matrix is ​​composed of the installation angles. For attitude error, For speed error, For angular velocity error, For the odometer lever arm, This refers to the odometer lever arm error;

[0057] Step 14: Combining the actual velocity equation of the vehicle in the vehicle coordinate system and the velocity equation of the vehicle at point P in the vehicle coordinate system, the velocity measurement equation between INS and ODO is obtained as follows:

[0058] ,in, Coordinate transformation matrix, V n The velocity in the n-system is The rotational angular velocity of the carrier;

[0059] Step 15: Based on the position measurement equation and velocity measurement equation, obtain the navigation system state variables: ,in, For attitude error, For speed error, For positional error, and These are the zero bias values ​​for the gyroscope and the accelerometer, respectively. and These represent the residual pitch and yaw installation angles to be estimated between the B and M series, respectively. This refers to the scale factor error of the odometer.

[0060] Step 16: Model all external parameters to be estimated as random constants. Then the state equation of the filter is: ,in, and This refers to the non-zero block matrices in the state transition matrix and system noise matrix of a traditional 15-dimensional loosely combined EKF filter. This refers to the system noise of the INS.

[0061] The system noise of the INS is the white noise of the gyroscope and accelerometer.

[0062] Step 17: Construct a system error model from the state variables of the navigation system, the transition equations of the filter, the position measurement equations, and the velocity measurement equations.

[0063] Step 2: Based on the observability analysis method for piecewise linear time-invariant systems and combined with the error system model, obtain the observability matrix of the system, and obtain the observability analysis results of the extrinsic parameters through observability matrix analysis;

[0064] During the filtering process, by performing observability analysis on the state variables, the convergence characteristics of each state variable can be reasonably predicted and judged. Therefore, it is necessary to analyze the observability of the navigation system's extrinsic parameters, such as... Figure 4 As shown, step 2 includes the following steps.

[0065] Step 21: Based on the observability analysis method for piecewise linear time-invariant systems and combined with the state equation of the filter, obtain the system's state matrix. ;

[0066] The integrated navigation system is a continuously time-varying system and cannot be directly analyzed using the observability matrix. However, according to the observability analysis method for piecewise linear time-invariant systems (PWCS), the system can be considered time-invariant in a short time interval. Combining the state equations of the filters, the system's state matrix is: .

[0067] Step 22: Assuming that the position measurement information and the velocity measurement information are synchronized at the same frequency, the position measurement equation and the velocity measurement equation are merged to form the total observation matrix H of the system;

[0068] For ease of analysis, it is assumed that the position measurement information and the velocity measurement information are synchronized at the same frequency. The position measurement equation is then compared with... Velocity measurement matrix under system The combined observation matrix constitutes the system. .

[0069] Step 23, from the total observation matrix H, the total observability matrix of the system from the beginning to the r-th time period is obtained as follows:

[0070] , ,

[0071] in, For the first The length of each time period Let be the observability matrix at time j. Let J be the constant state matrix of the system in the j-th time period. Let be the constant measurement matrix for the j-th time period of the system;

[0072] Because the function in this step contains an exponential term, making the calculation process complex, we can introduce, provided the theorem is satisfied, an exponential term is present. Perform system observability analysis to extract the system's stripped Observability Matrix (SOM).

[0073] Step 24: Based on the PWCS observability analysis method, use the observability matrix... The observability of the state variables is analyzed to obtain the observability analysis results of the external parameters.

[0074] Based on the properties of the PWCS observability analysis method, the observability of a system is determined solely by the effective time period that increases the rank of the observability matrix. Therefore, it is sufficient to analyze the observability of state variables using the observability matrix for certain time periods. Since the vehicle's speed is much smaller than the Earth's principal radius of curvature, the angular rate of rotation of the geographic system relative to the Earth system due to the vehicle's speed is approximately zero when calculating the observability matrix. Furthermore, compared to the vehicle's own motion, the Earth's rotational angular velocity is... The value is on the order of rad / s, contributing very little to the observability, and therefore is ignored in the subsequent analysis. The effects of external parameter observability analysis include: when the vehicle is stationary, all system external parameters are unobservable; when the vehicle is moving at a constant speed in a straight line, the pitch installation angle error and the odometer scaling factor error are observable; when the vehicle is moving at varying speeds in a straight line, the heading installation angle error is observable; when the vehicle moves in an angular direction, it will dynamically excite the GNSS antenna mast arm error and the P-point mast arm error in the other two orthogonal directions, thus allowing them to be separated and observed by the system.

[0075] Step 3: Based on the results of the external parameter observability analysis, the external parameters are grouped according to the required maneuvering mode to achieve segmented decoupling calibration and Kalman filter dimensionality reduction, thereby improving the real-time performance and accuracy of the external parameter calibration.

[0076] Step 4: Based on the state variables of the Kalman filter selected during the piecewise dimensionality reduction process, establish the corresponding state equations and measurement equations through the system error model, and then construct the adaptive feedback factor.

[0077] Step 5: By using an adaptive feedback factor to reflect the accuracy of error estimation, it determines whether to provide feedback to the external parameters and the extent of such feedback, thereby improving the accuracy and reliability of external parameter estimation and achieving adaptive online calibration of the external parameters.

[0078] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of a vehicle-mounted integrated navigation system. It is mainly used to solve the problems of existing technologies, such as the need for the vehicle to perform specific maneuvers to ensure the accuracy of parameter estimation, the susceptibility of parameter estimation to interference and the decrease in accuracy when the dynamic excitation of the vehicle trajectory is insufficient, and the relatively limited observability analysis of extrinsic parameters of INS / GNSS / ODO integrated navigation systems under different maneuvers, as well as the lack of research on the influence mechanism of the instantaneous motion of the vehicle on the observability of the system state.

[0079] Example 4:

[0080] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system, such as... Figure 1 As shown, the method includes the following steps.

[0081] Step 1: Augment the system extrinsic parameters to the Kalman filter, including the lever error between the GNSS antenna and the IMU, the lever error between the IMU and the vehicle body, the installation deviation angle error, and the odometer scaling factor error, and construct a system error model;

[0082] Step 2: Based on the observability analysis method for piecewise linear time-invariant systems and combined with the error system model, obtain the observability matrix of the system, and obtain the observability analysis results of the extrinsic parameters through observability matrix analysis;

[0083] Step 3: Based on the results of the external parameter observability analysis, the external parameters are grouped according to the required maneuvering mode to achieve segmented decoupling calibration and Kalman filter dimensionality reduction, thereby improving the real-time performance and accuracy of the external parameter calibration.

[0084] Based on the observability analysis results, the external parameters are grouped according to the required maneuvering modes to achieve segmented decoupled calibration, thereby reducing the dimensionality of the 24-dimensional filter to 18 dimensions and improving the real-time performance of the external parameter calibration. For example... Figure 5 As shown, step 3 includes the following.

[0085] Step 31: Based on the results of the external parameter observability analysis, group each external parameter according to the required maneuver mode;

[0086] The maneuvering modes include stationary, uniform, variable speed linear motion, and angular motion.

[0087] Step 32: Decouple and calibrate the first set of extrinsic parameters, which includes the GNSS arm error in the x-direction. GNSS lever arm error in the y-direction GNSS lever arm error in the z-direction ;

[0088] Since the calibration of the mast arms between IMU / GNSS antennas is less dependent on odometer information and requires angular motion excitation, the first stage only involves... , , Perform calibration.

[0089] Step 33: After the first set of extrinsic parameters is calibrated, accurate pose information is obtained through INS / GNSS combination to assist in the decoupling calibration of the second set of extrinsic parameters, which includes pitch installation angle error. Heading installation angle error Scale factor error of odometer ;

[0090] After the mast arm calibration between the IMU / GNSS antennas is completed, relatively accurate pose information can be obtained through INS / GNSS combination, thereby assisting in the parameter calibration between the IMU / ODO. Combined with the observability analysis results, , , The dependence on maneuverability is relatively weak, so the second stage involves calibrating all three.

[0091] Step 34: After the second set of extrinsic parameters is calibrated, the third set of extrinsic parameters is decoupled and calibrated. The third set of extrinsic parameters includes the ODO lever arm error in the x-direction. ODO lever arm error in the y direction ODO lever arm error in the z-direction ;

[0092] The third stage , , Calibration is then performed. This completes the piecewise decoupled calibration and achieves filter dimensionality reduction through decoupling. At this point, the system navigation obtains accurate convergence results for the extrinsic parameters, but to achieve a higher precision convergence state, iterative optimization can be used.

[0093] Step 35: When the mean and standard deviation of a certain external parameter change less than the preset value, the calibration of the current external parameter ends and the next cycle begins to calibrate the external parameter.

[0094] For any stage, when the mean and standard deviation of a certain external parameter show low variation, the parameter is considered converged, and the calibration of the corresponding external parameter ends. Once all three external parameters for the current stage have been calibrated, the calibration process for that stage ends, and the calibration for the next stage begins. Steps 32 to 34 constitute one cyclic calibration process. After the entire calibration process is completed, the process returns to step 32 to start a new calibration process. If the value is not lower than a preset value, the next cyclic calibration process will not begin.

[0095] Step 4: Based on the state variables of the Kalman filter selected during the piecewise dimensionality reduction process, establish the corresponding state equations and measurement equations through the system error model, and then construct the adaptive feedback factor.

[0096] Because vehicles exhibit dynamic and time-varying characteristics in actual motion, rank analysis of the observability matrix is ​​insufficient for timely assessment of the vehicle's observability at the current moment. Therefore, this application constructs an adaptive feedback factor in conjunction with the observability calculation process. For example... Figure 6 As shown, step 4 includes the following steps.

[0097] Step 41: Establish the corresponding state equation and measurement equation through the system error model to obtain the observation equation. ,in, To extract the observability matrix of the system, Let Z be the initial state value and Z be the observation vector.

[0098] The construction of the adaptive feedback factor here needs to be combined with the observability and observability results. As mentioned above, observability and observability analysis need to be based on the state equation and the measurement equation, so the state equation and the measurement equation are required here.

[0099] Step 42, for Perform SVD decomposition to obtain the observability of different state variables. By further restricting the definition interval of the local observability matrix to the segmented interval, the observability of different state quantities at the current time can be obtained;

[0100] The SVD is a type of singular value decomposition, an important matrix decomposition in linear algebra. It can represent a complex matrix as the product of three smaller, simpler submatrices, which describe important properties of the larger matrix. The different state variables are the Kalman filter state variables selected during the piecewise dimensionality reduction process in step 3.

[0101] Step 43, according to the function By combining vehicle maneuverability and observability, an instantaneous adaptive feedback factor is constructed, in which... Let be the coefficient value of the i-th state variable at time j in the observability matrix. for The reference threshold.

[0102] Step 5: By using an adaptive feedback factor to reflect the accuracy of error estimation, it determines whether to provide feedback to the external parameters and the extent of such feedback, thereby improving the accuracy and reliability of external parameter estimation and achieving adaptive online calibration of the external parameters.

[0103] Through a function containing an adaptive feedback factor While describing the observability of state variables, it also reflects the accuracy of error estimation by the filter in indirect filtering, thus determining whether to provide error feedback and the magnitude of such feedback. Let be the optimal estimate of the state variables at time j. These are estimates of the state variables from the previous time step. Let be the state quantity estimated by the filter at time j. The adaptive feedback factor is obtained at time j.

[0104] This invention provides an adaptive lightweight online calibration method for extrinsic parameters of a vehicle-mounted integrated navigation system. It is mainly used to solve the problems of existing technologies, such as the need for the vehicle to perform specific maneuvers to ensure the accuracy of parameter estimation, the susceptibility of parameter estimation to interference and the decrease in accuracy when the dynamic excitation of the vehicle trajectory is insufficient, and the relatively limited observability analysis of extrinsic parameters of INS / GNSS / ODO integrated navigation systems under different maneuvers, as well as the lack of research on the influence mechanism of the instantaneous motion of the vehicle on the observability of the system state.

[0105] In summary, this invention presents an adaptive lightweight online calibration method for extrinsic parameters of a vehicle-mounted integrated navigation system. This method constructs a system error model and an observability model for the vehicle-mounted INS / GNSS / ODO integrated navigation system. It analyzes the observability of the system's extrinsic parameters for maneuvers such as stationary, uniform, and variable-speed linear and angular motions of the vehicle, providing a theoretical basis for subsequent work. Simultaneously, a lightweight calibration architecture based on observability analysis is designed. Combining the observability analysis results, the extrinsic parameters are calibrated in a segmented manner, achieving dimensionality reduction of the filter and decoupling of the extrinsic parameters, thus improving the real-time performance and accuracy of the extrinsic parameter calibration. Furthermore, an adaptive extrinsic parameter estimation method based on observability feedback is proposed. Through a dynamic feedback mechanism, adaptive estimation of the extrinsic parameters is achieved, suppressing the accumulation of errors in extrinsic parameter calibration under low-observable maneuvers, and improving calibration accuracy and robustness.

[0106] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. An adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system, characterized in that, The method includes: Step 1: Augment the extrinsic parameters of the navigation system to the Kalman filter, including the lever error between the GNSS antenna and the IMU, the lever error between the IMU and the vehicle body, the installation deviation angle error, and the scaling factor error of the odometer, and construct a system error model; Step 2: Based on the observability analysis method for piecewise linear time-invariant systems and combined with the error system model, obtain the observability matrix of the system, and obtain the observability analysis results of the extrinsic parameters through observability matrix analysis; Step 3: Based on the results of the external parameter observability analysis, the external parameters are grouped according to the required maneuvering mode to achieve segmented decoupling calibration and Kalman filter dimensionality reduction, thereby improving the real-time performance and accuracy of the external parameter calibration. Step 4: Based on the state variables of the Kalman filter selected during the piecewise dimensionality reduction process, establish the corresponding state equations and measurement equations through the system error model, and then construct the adaptive feedback factor. Step 5: By using an adaptive feedback factor to reflect the accuracy of error estimation, it determines whether to provide feedback to the external parameters and the extent of such feedback, thereby improving the accuracy and reliability of external parameter estimation and achieving adaptive online calibration of the external parameters.

2. The adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system according to claim 1, characterized in that, The method further includes: Step 0: Obtain sensor information through ODO, IMU, and GNSS, and obtain navigation information through INS for later external parameter calibration processing.

3. The adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system according to claim 1, characterized in that, The method further includes: Step 6: After completing the adaptive calibration of the extrinsic parameters of the vehicle navigation system, output the navigation results for vehicle use.

4. The adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system according to claim 1, characterized in that, Step 1 includes: Step 11: Based on the boom arm between the GNSS antenna and the IMU, calculate the position measurement equation between the INS and GNSS antennas: , in, Let be the coordinate transformation matrix from system b to system n. The arm connecting the GNSS antenna and the IMU. For attitude error, This refers to the boom arm error between the GNSS antenna and the IMU. This is the coordinate transformation matrix. Noise for position measurement; Step 12: Based on the speed measurement provided by the odometer and the vehicle non-integrity constraints, obtain the actual speed equation of the vehicle in the vehicle coordinate system: ， in, The scale factor error of the odometer. This represents the actual speed of the vehicle in the vehicle coordinate system. For speed measurement noise; Step 13: Set the rear axle center of the non-drive wheels or the contact point between the non-drive wheels and the ground as reference point P. The IMU then calculates the vehicle's velocity equation at point P in the vehicle coordinate system as follows: ， in, For installation deviation angle error, The coordinate transformation matrix is ​​composed of the installation angles. For attitude error, For speed error, For angular velocity error, For the odometer lever arm, This refers to the odometer lever arm error; Step 14: Combining the actual velocity equation of the vehicle in the vehicle coordinate system and the velocity equation of the vehicle at point P in the vehicle coordinate system, the velocity measurement equation between INS and ODO is obtained as follows: ， in, The coordinate transformation matrix is ​​composed of the installation angles. V is the coordinate transformation matrix from the n-system to the b-system. n The velocity in the n-system is The rotational angular velocity of the carrier; Step 15, establish the navigation system state variables as follows: ， in, For attitude error, For speed error, For positional error, and These are the zero bias values ​​for the gyroscope and the accelerometer, respectively. and These represent the residual pitch and yaw installation angles to be estimated between the B and M series, respectively. This refers to the scale factor error of the odometer. Step 16: Model all external parameters to be estimated as random constants. Then the state equation of the filter is: ,in, and This refers to the non-zero block matrices in the state transition matrix and system noise matrix of a traditional 15-dimensional loosely combined EKF filter. This refers to the system noise of the INS. Step 17: Construct a system error model from the state variables of the navigation system, the transition equations of the filter, the position measurement equations, and the velocity measurement equations.

5. The adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system according to claim 4, characterized in that, Step 2 includes: Step 21: Based on the observability analysis method for piecewise linear time-invariant systems and combined with the state equation of the filter, obtain the system's state matrix. ; Step 22: Assuming that the position measurement information and the velocity measurement information are synchronized at the same frequency, the position measurement equation and the velocity measurement equation are merged to form the total observation matrix H of the system; Step 23, from the total observation matrix H, the total observability matrix of the system from the beginning to the r-th time period is obtained as follows: ， ， in, For the first The length of each time period Let be the observability matrix at time j. Let J be the constant state matrix of the system in the j-th time period. Let be the constant measurement matrix for the j-th time period of the system; Step 24: Based on the PWCS observability analysis method, use the observability matrix... The observability of the state variables is analyzed to obtain the observability analysis results of the external parameters.

6. The adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system according to claim 5, characterized in that, The results of the external parameter observability analysis include: When the vehicle is stationary, all external parameters of the system are unobservable; When the vehicle is moving at a constant speed in a straight line, both the pitch installation angle error and the odometer scale factor error can be observed. When the vehicle is moving in a straight line with varying speeds, the heading installation angle error is observable. When the vehicle moves angularly in a certain direction, it will generate dynamic excitation on the GNSS antenna arm error and the arm error at point P in the other two orthogonal directions, thus allowing them to be separated and observed by the system.

7. The adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system according to claim 1, characterized in that, Step 3 includes: Step 31: Based on the results of the external parameter observability analysis, group each external parameter according to the required maneuver mode; Step 32: Decouple and calibrate the first set of extrinsic parameters, which includes the GNSS arm error in the x-direction. GNSS lever arm error in the y-direction GNSS lever arm error in the z-direction ; Step 33: After the first set of extrinsic parameters is calibrated, accurate pose information is obtained through INS / GNSS combination to assist in the decoupling calibration of the second set of extrinsic parameters, which includes pitch installation angle error. Heading installation angle error Scale factor error of odometer ; Step 34: After the second set of extrinsic parameters is calibrated, the third set of extrinsic parameters is decoupled and calibrated. The third set of extrinsic parameters includes the ODO lever arm error in the x-direction. ODO lever arm error in the y direction ODO lever arm error in the z-direction ; Step 35: When the mean and standard deviation of a certain external parameter change less than the preset value, the calibration of the external parameter is terminated; otherwise, proceed to step 32 to recalibrate the external parameter.

8. The adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system according to claim 7, characterized in that, The maneuvering modes include stationary, uniform, variable speed linear motion, and angular motion.

9. The adaptive lightweight online calibration method for extrinsic parameters of an in-vehicle integrated navigation system according to claim 5, characterized in that, Step 4 includes: Step 41: Establish the corresponding state equation and measurement equation through the system error model to obtain the observation equation. ,in, To extract the observability matrix of the system, Let Z be the initial state value and Z be the observation vector. Step 42, for Perform SVD decomposition to obtain the observability of different state variables. By further restricting the definition interval of the local observability matrix to the segmented interval, the observability of different state quantities at the current time can be obtained; Step 43, according to the function Construct an instantaneous adaptive feedback factor, where, Let be the coefficient value of the i-th state variable at time j in the observability matrix. for The reference threshold.

Images