Onboard IMU Data Compensation Method for Inertial Navigation in Intelligent Driving

Abstract

This application relates to the field of positioning and navigation technology, specifically to a method for compensating onboard IMU data for inertial navigation in intelligent driving. The method includes: acquiring various attitude data of the onboard IMU through a rotating platform, obtaining reference quantities, establishing an objective function, and using an optimization algorithm to obtain the offline calibrated rotation error matrix and scaling factor matrix; acquiring raw measurements during vehicle operation, defining the system state vector, analyzing the Taylor expansion approximation relationship between the cross product matrix corresponding to the small error angle and the rotation error matrix, constructing discrete-time state equations and observation equations based on the IMU kinematic model, and using extended Kalman filtering to obtain the small error angle after each update; and combining the raw measurements at each moment to perform real-time compensation of the onboard IMU data. This application aims to improve the accuracy and robustness of inertial navigation, meeting the requirements of intelligent driving for continuous and stable positioning.

Description

Technical Field

[0001] This application relates to the field of positioning and navigation technology, specifically to a method for compensating onboard IMU data for intelligent driving inertial navigation. Background Technology

[0002] The inertial measurement unit (IMU) is a core sensor in a vehicle, composed of accelerometers and gyroscopes. It measures the vehicle's three-dimensional acceleration and angular velocity in real time, thereby sensing the vehicle's motion state. Intelligent driving inertial navigation technology, based on data provided by the IMU, calculates the vehicle's real-time position, velocity, and attitude using inertial navigation algorithms, continuing to provide accurate motion status even in the event of GNSS signal loss. However, due to the influence of hardware manufacturing precision and installation errors, the IMU's raw measurement data may deviate from the actual motion state, and these errors accumulate rapidly during integration, causing navigation results to drift. In severe cases, this can affect intelligent driving decisions. Therefore, error compensation techniques are needed to eliminate these deviations and ensure the high accuracy and reliability of inertial navigation.

[0003] The physical mounting of the three-axis accelerometers and gyroscopes inside the IMU is often not perfectly orthogonal. Furthermore, the entire device may be subject to slight rotation or tilting on the vehicle body. This inter-axis non-orthogonality and mounting errors cause the motion component of one axis to interfere with the measurements of other axes, thus affecting the accuracy of attitude estimation and position integration. Traditional methods typically obtain fixed error parameters through offline calibration (such as the six-sided method); however, this fails to adequately consider vehicle dynamics, causing error parameters to drift over time or with changes in complex road conditions. The discrepancy between offline calibration and actual operating conditions ultimately makes it difficult for the compensation model to continuously track error changes. Especially during long-term operation or under complex road conditions, the accuracy of inertial navigation and positioning significantly decreases, failing to meet the continuous high-precision positioning requirements of intelligent driving systems. Summary of the Invention

[0004] In view of the above, it is necessary to provide a method for compensating onboard IMU data for inertial navigation in intelligent driving to solve the above problems.

[0005] One embodiment of this application provides a method for onboard IMU data compensation for intelligent driving inertial navigation, the method comprising:

[0006] The original calibration measurements for each group of vehicle-mounted IMUs in various orientations are obtained by rotating the platform, and the corresponding calibration reference values ​​are obtained. Based on the difference between the original calibration measurements and the corresponding reference values, and combined with the inter-axis non-orthogonality of the vehicle-mounted IMU, an objective function is established, and an optimization algorithm is used to obtain the rotation error matrix and scaling factor matrix of the offline calibration.

[0007] During vehicle operation, the raw measurements of the on-board IMU at each moment are acquired, a system state vector containing the small error angle is defined, the Taylor expansion approximation relationship between the cross product matrix and the rotation error matrix corresponding to the small error angle is analyzed, and based on the IMU kinematic model, discrete-time state equations and observation equations are constructed to predict the system state vector at the next moment; based on the difference between the actual measurement value and the predicted value, the extended Kalman filter is used to obtain the small error angle after each update.

[0008] Based on the scale factor matrix obtained from offline calibration and the cross product matrix corresponding to the small error angle parameter after each update, combined with the original measurements at each time point, the on-board IMU data is compensated in real time.

[0009] The specific formula for the objective function is as follows: In the formula, min represents the minimization function; S represents the rotation error matrix; S represents the scaling factor matrix; N represents the number of data sets acquired for each posture. This represents the i-th set of original calibration measurements for each pose; This represents the calibration reference value corresponding to the original measurement of the i-th group of calibration for each posture.

[0010] The objective function is constrained by the fact that the rotation error matrix is ​​an orthogonal matrix.

[0011] The system state vector also includes the following elements: the vehicle's position vector, instantaneous velocity vector, and attitude angle vector.

[0012] Specifically, the construction of the discrete-time state equation and the observation equation includes:

[0013] Discrete-time state equations: ,in, , These represent the system state vector at time k and the original measurements taken during IMU operation, respectively. Represents the state transition function; Indicates the overall system noise; This represents the predicted system state vector at time k+1;

[0014] Observation equation: ,in, Denotes the observation vector at time k. Represents the observation matrix. Indicates observation noise. This represents the system state vector at time k.

[0015] Among them, the update formula for the instantaneous velocity vector in the system state vector is: In the formula: Indicates vehicle attitude The corresponding rotation matrix, Represents gravitational acceleration. Indicates the sampling interval of the IMU. This represents the acceleration after error compensation at time k, as measured by the IMU. This represents the instantaneous velocity vector of the vehicle at time k. This represents the predicted instantaneous velocity vector of the vehicle at time k+1.

[0016] In the system state vector, the update process of the position vector is as follows: obtain the product between the instantaneous velocity vector of the vehicle at each moment and the IMU sampling interval, add it to the vehicle position vector at each moment, and obtain the predicted vehicle position vector for the next moment.

[0017] Among them, the update formula for the attitude vector in the system state vector is: ;in: This represents the angular velocity measured by the IMU at time k after error compensation. This represents the vehicle attitude vector predicted at time k+1; This represents the vehicle's attitude vector at time k; Indicates vehicle attitude The corresponding rotation matrix; This indicates the sampling interval of the IMU.

[0018] In the system state vector, the update process of the small error angle vector is as follows: the sum of Gaussian white noise and the small error angle vector at each time moment is used as the predicted small error angle vector for the next time moment.

[0019] Specifically, the real-time compensation of the vehicle-mounted IMU data includes:

[0020] Calculate the sum of the cross product matrix and the identity matrix corresponding to the small error angle obtained in each update, and multiply it with the scaling factor matrix obtained from offline calibration.

[0021] Multiply the result of the multiplication with the angular velocity measured in the original IMU operation to obtain the compensated angular velocity;

[0022] The multiplication result is multiplied by the acceleration measured in the original IMU operation to obtain the compensated acceleration.

[0023] This application has at least the following beneficial effects:

[0024] This application acquires raw calibration measurements for each group of vehicle-mounted IMUs in various attitudes using a rotating platform, and obtains corresponding calibration reference values. This helps provide calibration data and determine the relationship between IMU measurements and reference values. The raw calibration measurements in different attitudes provide data on various angles and directions, which are the basis for model building and error calculation. Based on the differences between the raw calibration measurements and the corresponding reference values, and considering the inter-axis non-orthogonality of the vehicle-mounted IMU, an objective function is established. An optimization algorithm is then used to obtain the rotation error matrix and scale factor matrix for offline calibration. This helps to correct the systematic errors of the IMU through the optimization algorithm, resulting in the rotation error matrix and scale factor matrix. These matrices represent the non-ideal errors in IMU measurements. The transformed matrix provides an accurate reference for subsequent error compensation. During vehicle operation, the raw measurements of the IMU at each moment are acquired, and a system state vector containing the small error angle is defined. The Taylor expansion approximation relationship between the cross product matrix corresponding to the small error angle and the rotation error matrix is ​​analyzed. The beneficial effect is that by analyzing the small error angle and combining the Taylor expansion approximation relationship, the complex error calculation process is simplified, providing an effective theoretical framework for subsequent Kalman filtering and system state updates. Based on the IMU kinematic model, discrete-time state equations and observation equations are constructed. The IMU kinematic model is used to predict the system state, providing a preliminary estimate of the state at the next moment in the absence of real-time observation data. The prediction step provides the foundation for subsequent update steps. Based on the difference between the actual measured value and the predicted value, an extended Kalman filter is used to obtain the small error angle after each update. By processing the difference between the actual measured value and the predicted value, the system state estimate can be corrected, and the small error angle after each update can be obtained. EKF, combined with the characteristics of nonlinear systems, can optimize state estimation in real time and improve accuracy. Finally, based on the scale factor matrix obtained from offline calibration and the cross product matrix corresponding to the small error angle parameters after each update, combined with the original measurements at each time point, the onboard IMU data is compensated in real time. The beneficial effect is that through real-time compensation, the IMU error is corrected to a minimum, providing high-precision attitude estimation. This step combines the results of offline calibration and the real-time updated data to dynamically calibrate the IMU measurements, ensuring the accuracy of the compensated IMU data. This application solves the systematic static error through offline calibration and tracks the dynamic error drift in real time with online EKF self-calibration, balancing calibration accuracy and environmental adaptability; the small angle parameterization design reduces computational complexity, making real-time estimation feasible; the entire life cycle covers error sources from factory to operation, effectively eliminating the influence of inter-axle crosstalk, significantly improving the accuracy and robustness of inertial navigation, and meeting the requirements of intelligent driving for continuous and stable positioning. Attached Figure Description

[0025] Figure 1A flowchart of the onboard IMU data compensation method for intelligent driving inertial navigation provided in this application;

[0026] Figure 2 A schematic diagram illustrating the acquisition of compensated vehicle-mounted IMU data provided in this application. Detailed Implementation

[0027] In the description of the embodiments in this application, the words "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary," "or," and "for example" is intended to present the relevant concepts in a specific manner.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this application's specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0029] It should also be noted that the terms "first" and "second" in this application and its accompanying drawings are used to distinguish similar objects, rather than to describe a specific order or sequence. The methods disclosed in the embodiments of this application or the methods shown in the flowcharts include one or more steps for implementing the method. Without departing from the scope of protection of this application, the execution order of multiple steps can be interchanged, and some steps can also be deleted.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0031] This application proposes an onboard IMU data compensation method for inertial navigation in intelligent driving, which is applied in the field of positioning and navigation technology. (See attached document.) Figure 1 The method includes the following steps:

[0032] Step 1: Obtain the original calibration measurements for each group of vehicle-mounted IMUs in various orientations through a rotation platform, and obtain the corresponding calibration reference values; based on the difference between the original calibration measurements and the corresponding reference values, and combined with the inter-axis non-orthogonality of the vehicle-mounted IMU, establish an objective function, and use an optimization algorithm to obtain the rotation error matrix and scaling factor matrix of the offline calibration.

[0033] The physical axes of the IMU's internal three-axis accelerometers and gyroscopes cannot be perfectly orthogonal during manufacturing and assembly due to mechanical precision limitations. Furthermore, there are inherent initial deviations when the entire unit is installed onto the vehicle body. These systematic, static errors cause crosstalk between motion components of different axes in the measurements. Without prior calibration, these errors directly amplify deviations in subsequent attitude estimation and position integration. Since these errors are relatively stable and fixed in value shortly after the equipment leaves the factory, they are difficult to eliminate in one go through online dynamic calibration. Therefore, offline precision calibration is necessary to establish a foundation for error compensation before commissioning.

[0034] The IMU to be calibrated is fixed on the 3D rotating platform using a rigid fixture to ensure that the relative position of the IMU and the platform coordinate system is fixed; the IMU data acquisition module is connected to the rotating platform control system to realize the synchronous recording of the original measurement and the platform attitude.

[0035] Design at least 30 reference poses covering the entire pose space, including:

[0036] Static posture: For example, horizontal placement, pitch angle range from arrive Roll angle range from arrive yaw angle range from arrive These attitudes should ensure that the accelerometer can detect the projection of gravity onto different axes.

[0037] Dynamic attitude: Rotating platform with arrive The gyroscope rotates at a constant speed around the x / y / z axes to ensure that it can detect the known angular velocity.

[0038] The rotating platform is controlled to switch to preset postures sequentially. Data is collected after each posture has been stable for 5 seconds, and the original calibration measurements of the IMU in the i-th set of each posture are recorded. This includes: the accelerometer output of the i-th set of acceleration measurement vectors for each attitude. The gyroscope outputs the i-th set of angular velocity measurement vectors for each attitude. , recorded as ,in This indicates transpose. The calibration reference value corresponding to the i-th group of original calibration measurements for each orientation is synchronously recorded from the output of the rotating platform. This includes: the i-th set of acceleration reference vectors for each attitude under static orientation. The acceleration reference vector is determined by the projection of the gravity vector onto the platform coordinate system in each direction; during dynamic rotation, the i-th set of acceleration reference vectors for each attitude is used. The rotational angular velocity is determined by setting the rotational angular velocity in each direction of the platform coordinate system, and is denoted as . It should be noted that the calibration of the original measurement Calibration reference quantity All are 3×2 matrices.

[0039] For each group of attitudes, all raw calibration measurements are subjected to time window mean filtering. In this application, the window length is 100ms to eliminate high-frequency noise, obtain smoothed raw calibration measurements, and discard those that deviate from the reference value. The above outliers ensure the validity of the data used for calibration.

[0040] Based on the principle that the original calibration measurement should approximate the reference value after error matrix transformation, this application solves the error matrix through constrained optimization. Since the original IMU calibration measurement is affected by inter-axis non-orthogonality, i.e., rotation error and scale factor error, the correction relationship is as follows: ,in, Let S represent the calibration reference value for the i-th group of each attitude, and let S represent the scaling factor matrix, which is used to describe the calibration factor error of the IMU sensors (accelerometer and gyroscope), that is, the scaling deviation between the original output value of the sensor and the actual physical quantity. The scaling factor matrix is ​​a 3×3 matrix. This represents the rotational error matrix, used to describe the non-orthogonality between shafts and installation deviations, satisfying constraints. (Orthogonality) Because the rotation error matrix needs to ensure that the length and angle of the vectors remain unchanged during the transformation process, the rotation error matrix is ​​also a 3×3 matrix.

[0041] The optimization objective is to achieve optimal calibration by minimizing the sum of squares of the deviations between the calibrated measurements and the reference values. That is, the objective function is: ; This application employs a constrained Levenberg-Marquardt algorithm for iterative solution, initializing... (Ideal orthogonal) (Ideal scale); In each iteration, the rotation error matrix is ​​parameterized in axis-angle form (3 degrees of freedom), combined with the 3 diagonal elements of the scale factor matrix, resulting in a total of 6 parameters to be estimated; the objective function is minimized through gradient descent, while simultaneously forcing the rotation error matrix to satisfy orthogonal constraints; after convergence, the rotation error matrix is ​​obtained. and scaling factor matrix .

[0042] Will and All data written to the IMU preprocessing module serves as the initial compensation matrix. This initial compensation matrix is ​​used for hardware mapping correction before the equipment's first commissioning, eliminating most systematic static errors at the source. It initially maps the original IMU calibration measurements to the ideal axis system, providing a low-error basis for subsequent online dynamic calibration and preventing excessive initial errors from causing difficulties in online algorithm convergence or decreased accuracy.

[0043] Step 2: During vehicle operation, acquire the raw measurements of the onboard IMU at each moment, define the system state vector including the small error angle, analyze the Taylor expansion approximation relationship between the cross product matrix and the rotation error matrix corresponding to the small error angle, construct the discrete-time state equation and observation equation based on the IMU kinematic model, and predict the system state vector for the next moment; based on the difference between the actual measurement value and the predicted value, use extended Kalman filtering to obtain the small error angle after each update.

[0044] While offline calibration can determine the initial compensation matrix, environmental factors such as temperature changes, continuous vibration, and minor collisions during actual vehicle operation can cause IMU axis non-orthogonality and installation errors to drift slowly over time. These dynamic changes cannot be fully covered by static calibration in the laboratory. If only the initial compensation is relied upon, the accumulation of errors after long-term operation will significantly reduce navigation accuracy. Therefore, it is necessary to use online methods to track and correct these dynamic errors in real time to adapt to the error variation characteristics under complex operating conditions.

[0045] The coupling relationship between IMU measurements and attitude / installation errors is nonlinear. The Extended Kalman Filter (EKF) can recursively estimate the state through a linearized approximation of the nonlinear system. Simultaneously, it can integrate multi-source observation information such as GNSS and odometer readings as constraints, effectively suppressing error accumulation. Compared to the traditional EKF, this application uses small-angle parameters for error... Incorporating these extended states into the system state vector enables joint dynamic estimation of attitude, position, velocity, and installation error parameters, rather than just estimating conventional motion states. This allows for real-time tracking of error changes caused by factors such as temperature and vibration, and provides online self-calibration capabilities. Traditional EKF systems typically do not use these error parameters as state variables, making it difficult to adapt to dynamic error drift.

[0046] Specifically, the system state vector is defined as follows: , where x represents the system state vector; Represents the position vector of the vehicle body (three-dimensional). Represents the instantaneous velocity vector of the vehicle body (three-dimensional). Represents the vehicle's attitude angle vector (3D) This represents a small-angle error vector (three-dimensional), used to address small-angle disturbances caused by inter-axis non-orthogonality and installation deviations.

[0047] When performing preliminary compensation on the raw measurements from the IMU operation, a small angle error parameter is introduced. The effect of small-angle rotation. Utilizing the principle of approximation, that is, when... When the rotation error matrix is ​​in radians, it can be approximated by a Taylor expansion as follows: In the formula, Represents a unit vector. express The cross product matrix represents the transformation of vector cross product operations into matrix multiplication. ;in, , , These represent the components of the small error angle on the x-axis, y-axis, and z-axis of the IMU coordinate system, respectively.

[0048] Based on the IMU kinematic model and the dynamic characteristics of error parameters, a discrete-time state equation is established: ,in, This represents the predicted system state vector at time k+1; , Let K represent the system state vector at time k and the initial measurements. In the initial system state vector, the position is set to the coordinates output by GNSS, the velocity is set to a zero vector, and the attitude angles are calculated using the gravity vector measured by the accelerometer in a stationary state, namely pitch and roll angles, and the yaw angle is set to 0. The initial values ​​of small angles are obtained by extracting the installation rotation error matrix from the offline calibration. The antisymmetric part is obtained, and the subsequent state vector is updated by Kalman filtering; The overall system noise is represented by a zero-mean Gaussian distribution, and its covariance matrix is ​​preset to be the variance of the inherent noise in the IMU. It represents the state transition function, which describes the evolution of the state over time. The core of the derivation is based on the calibrated IMU measurements.

[0049] For the update of each element of the system state vector, the speed The update satisfies: ,in: Indicates vehicle attitude The corresponding rotation matrix, which can be used for Euler angle or quaternion transformation, is used to transform IMU measurements from the IMU coordinate system to the navigation coordinate system. Represents gravitational acceleration. Indicates the sampling interval of the IMU. This represents the acceleration after error compensation at time k, as measured by the IMU. This represents the instantaneous velocity vector of the vehicle at time k. This represents the predicted instantaneous velocity vector of the vehicle at time k+1. The velocity increment is obtained by integrating the effective acceleration within this sampling interval, ultimately updating the velocity from the current time to the next time step.

[0050] Location and posture The update is also based on the derivation of velocity integral and angular velocity integral, with the formulas as follows: , ,in: Represents the vehicle's position vector at time k; This represents the predicted vehicle position vector at time k+1; This represents the angular velocity measured by the IMU at time k after error compensation. This represents the vehicle attitude vector predicted at time k+1; Let k represent the vehicle attitude vector at time k.

[0051] Error parameters The update needs to account for the slow error drift, which is approximately a random walk process. ,in This represents Gaussian white noise with a preset variance of 1. This reflects the slowness of error drift, making The estimate can both track real, slow changes and be free from high-frequency noise interference; Let represent the small angle vector of error at time k; This represents the small angle vector of the predicted error at time k+1.

[0052] In this embodiment, state estimation is constrained by observations provided by external sensors. The velocity measured by GNSS is selected. As observed values, the observation equation is: ,in Denotes the observation vector at time k, i.e. The implementer may also choose any other observation for analysis; Representing the observation matrix, the velocity components in the system state vector are extracted and correlated with the observed values, i.e. , which correspond to the three-dimensional position, velocity, attitude angle and error, respectively, and 0 indicates no correlation. The observed noise is represented by a zero-mean Gaussian distribution, and its covariance matrix is ​​preset to be the variance of the inherent noise in the GNSS sensor.

[0053] Then, EKF iteration is used to predict and update the error, with the prediction phase using state equations. The prior state at time t is used to calculate the prior covariance matrix using the Jacobian matrix; in the update phase, the Kalman filter is first calculated, and then the observed values ​​are used. The state is updated to obtain the posterior estimate, and finally the covariance matrix is ​​updated. The EKF iteration formula is a well-known technique, and its details will not be elaborated further.

[0054] Step 3: Based on the scale factor matrix obtained from offline calibration and the cross product matrix corresponding to the small error angle parameter after each update, combined with the original measurements at each time point, perform real-time compensation on the vehicle-mounted IMU data.

[0055] After each EKF iteration, the updated small-angle error parameters are extracted from the posterior state. As the input for the next round of IMU measurement calibration, through continuous iteration, the error parameters will gradually converge to the true error value, realizing dynamic tracking and calibration of installation error. The installation error parameters are incorporated into the real-time estimation closed loop, ultimately achieving adaptive correction of error according to vehicle operating status.

[0056] By estimating dynamically changing small-angle error parameters in real time using EKF extended states, dynamic correction of offline initial compensation is achieved, ensuring that IMU measurements remain consistent with real physical motion throughout the vehicle's entire lifecycle (including environmental disturbances, hardware aging, and other scenarios). Finally, through online convergence of error parameters, high-precision IMU correction data is provided for subsequent attitude calculation and position integration, reducing navigation drift caused by dynamic errors and improving the long-term stability and accuracy of the vehicle navigation system under complex conditions.

[0057] After offline calibration and online EKF dynamic estimation, the parameters required for error compensation—the scale factor matrix and the real-time small-angle error vector—have been obtained. However, if these parameters are not applied to the IMU's raw data in real time, inter-axis non-orthogonality and installation errors will continue to cause cross-axis measurement, leading to deviations in attitude and position calculations. Furthermore, when the IMU operates alone, errors accumulate over time, requiring the integration of observation information from external sensors to constrain drift. Therefore, real-time compensation and correction of the raw data, along with the fusion of external information, are essential to ensure that error suppression is effectively implemented in the actual navigation output.

[0058] Furthermore, the scaling factor matrix obtained through offline calibration matches the original measurements with the actual physical quantities. When the magnitude of the small error angle is much smaller than 1 radian, the rotation error matrix is ​​approximately... The original measurements are rotated and mapped from the actual axis system with deviations to an ideal orthogonal axis system, eliminating crosstalk between axes and ensuring that the measurements of each axis only reflect the true motion components in the corresponding direction.

[0059] The acceleration measured by the calibrated IMU is: Angular velocity is: ,in, This represents the scale factor matrix obtained from offline calibration. The cross product matrix representing small angular errors. Represents the acceleration measurement vector at time i; Let i represent the angular velocity measurement vector at time i.

[0060] The corrected IMU data is input into the main navigation filter, and external observation information is introduced as a constraint. In this embodiment, the velocity output from GNSS is used. During the fusion process, the main navigation filter suppresses the error accumulation that occurs when the IMU operates alone through state prediction and observation updates.

[0061] After fusion, the filter outputs real-time high-precision navigation parameters: vehicle position. ,speed and posture This serves as the positioning and attitude reference for the intelligent driving system. A schematic diagram illustrating the acquisition of compensated onboard IMU data is shown below. Figure 2 As shown.

[0062] The core purpose of real-time compensation is to apply the error parameters of offline calibration and online dynamic estimation to the raw IMU data in real time, eliminate measurement deviations caused by inter-axis non-orthogonality and installation errors, and ensure that the IMU data input to the navigation filter accurately reflects the vehicle's motion state. The fusion output solves the error accumulation problem when the IMU works alone by using the observation constraints of external sensors, and ultimately provides continuous, stable and high-precision attitude and position information for the inertial navigation of the intelligent driving system, meeting the stringent requirements of the vehicle for positioning accuracy and robustness under complex road conditions.

[0063] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description; sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0064] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for compensating onboard IMU data for inertial navigation in intelligent driving, characterized in that, The method includes the following steps: The original calibration measurements for each group of vehicle-mounted IMUs in various orientations are obtained by rotating the platform, and the corresponding calibration reference values ​​are obtained. Based on the difference between the original calibration measurements and the corresponding reference values, and combined with the inter-axis non-orthogonality of the vehicle-mounted IMU, an objective function is established, and an optimization algorithm is used to obtain the rotation error matrix and scaling factor matrix of the offline calibration. During vehicle operation, the raw measurements of the on-board IMU at each moment are acquired, a system state vector containing the small error angle is defined, the Taylor expansion approximation relationship between the cross product matrix and the rotation error matrix corresponding to the small error angle is analyzed, and based on the IMU kinematic model, discrete-time state equations and observation equations are constructed to predict the system state vector at the next moment; based on the difference between the actual measurement value and the predicted value, the extended Kalman filter is used to obtain the small error angle after each update. Based on the scale factor matrix obtained from offline calibration and the cross product matrix corresponding to the small error angle parameter after each update, combined with the original measurements at each moment, the on-board IMU data is compensated in real time. The specific formula for the objective function is as follows: In the formula, min represents the minimization function; S represents the rotation error matrix; S represents the scaling factor matrix; N represents the number of data sets acquired for each posture. This represents the i-th set of original calibration measurements for each pose; The calibration reference value represents the original measurement corresponding to the i-th group of calibration for each posture; the constraint condition of the objective function is that the rotation error matrix is ​​an orthogonal matrix.

2. The onboard IMU data compensation method for intelligent driving inertial navigation as described in claim 1, characterized in that, The system state vector also includes the following elements: the vehicle's position vector, instantaneous velocity vector, and attitude angle vector.

3. The onboard IMU data compensation method for intelligent driving inertial navigation as described in claim 1, characterized in that, The construction of the discrete-time state equation and observation equation is specifically as follows: Discrete-time state equations: ,in, , These represent the system state vector at time k and the original measurements taken during IMU operation, respectively. Represents the state transition function; Indicates the overall system noise; This represents the predicted system state vector at time k+1; Observation equation: ,in, Denotes the observation vector at time k. Represents the observation matrix. Indicates observation noise. This represents the system state vector at time k.

4. The onboard IMU data compensation method for intelligent driving inertial navigation as described in claim 2, characterized in that, The update formula for the instantaneous velocity vector in the system state vector is: In the formula: Indicates vehicle attitude The corresponding rotation matrix, Represents gravitational acceleration. Indicates the sampling interval of the IMU. This represents the acceleration after error compensation at time k, as measured by the IMU. This represents the instantaneous velocity vector of the vehicle at time k. This represents the predicted instantaneous velocity vector of the vehicle at time k+1.

5. The onboard IMU data compensation method for intelligent driving inertial navigation as described in claim 2, characterized in that, In the system state vector, the update process of the position vector is as follows: obtain the product between the instantaneous velocity vector of the vehicle at each moment and the IMU sampling interval, add it to the vehicle position vector at each moment, and obtain the predicted vehicle position vector for the next moment.

6. The onboard IMU data compensation method for intelligent driving inertial navigation as described in claim 2, characterized in that, In the system state vector, the update formula for the attitude vector is: ;in: This represents the angular velocity measured by the IMU at time k after error compensation. This represents the vehicle attitude vector predicted at time k+1; This represents the vehicle's attitude vector at time k; Indicates vehicle attitude The corresponding rotation matrix; This indicates the sampling interval of the IMU.

7. The onboard IMU data compensation method for intelligent driving inertial navigation as described in claim 1, characterized in that, In the system state vector, the update process of the small error angle vector is as follows: the sum of Gaussian white noise and the small error angle vector at each time step is used as the predicted small error angle vector for the next time step.

8. The onboard IMU data compensation method for intelligent driving inertial navigation as described in claim 1, characterized in that, The real-time compensation of the vehicle-mounted IMU data specifically includes: Calculate the sum of the cross product matrix and the identity matrix corresponding to the small error angle obtained in each update, and multiply it with the scaling factor matrix obtained from offline calibration. Multiply the result of the multiplication with the angular velocity measured in the original IMU operation to obtain the compensated angular velocity; The multiplication result is multiplied by the acceleration measured in the original IMU operation to obtain the compensated acceleration.

Images