Wheel-mounted triaxial accelerometer axis alignment error calibration method and system

Abstract

The application provides a wheel-mounted three-axis accelerometer axis alignment error calibration method and system, and relates to the technical field of wheel motion measurement. The method comprises the following steps: acquiring original three-axis acceleration data in the wheel motion process; preprocessing the original three-axis acceleration data to construct an effective data set; constructing an accelerometer coordinate system and a wheel coordinate system, and constructing an attitude transformation relationship from the accelerometer coordinate system to the wheel coordinate system; constructing a cost function according to the attitude transformation relationship and the effective data set; outputting an axis alignment error angle estimation value through a preset nonlinear optimization algorithm; acquiring real-time accelerometer data, and projecting and transforming the real-time accelerometer data according to the axis alignment error angle estimation value and the attitude transformation relationship to obtain a plurality of transformed acceleration components; calculating a root mean square error, verifying the axis alignment error angle estimation value, and generating an installation error calibration parameter; and calibrating the real-time accelerometer data according to the installation error calibration parameter.

Description

Technical Field

[0001] This invention relates to the field of wheel motion measurement technology, and in particular to a method and system for calibrating the alignment error of a triaxial accelerometer mounted on a wheel. Background Technology

[0002] As a portable distance measuring tool, the distance measuring wheel is widely used in engineering surveying, road mapping, stadium marking, and traffic accident scene investigation. Its core objective is to achieve high-precision measurement of the rolling distance of wheeled devices. With the development of microelectromechanical systems (MEMS) technology, a solution that directly mounts a triaxial accelerometer onto the wheel and calculates the rotation angle and mileage by sensing the periodic changes in the gravity component as the wheel rotates is gaining popularity. This solution eliminates the need for a mechanical connection to the axle, offering advantages such as flexible installation, small size, low power consumption, and strong vibration and contamination resistance. It is well-suited for rapid on-site deployment and portable applications, possessing significant engineering application value.

[0003] Currently, most mainstream distance measuring wheels use wheel-axle measurement solutions such as mechanical encoders, photoelectric encoders, or Hall sensors to calculate the mileage by detecting the number of wheel rotations and combining the wheel radius. For accelerometer installation error calibration, existing technologies mainly include two paths: one is to achieve precise alignment and installation of the accelerometer through structural design at the factory stage, and the other is to carry out on-site calibration using wheel-mounted inertial measurement units.

[0004] However, existing technologies have many problems: First, factory precision installation can only guarantee the axis alignment accuracy in the factory state, which cannot adapt to the usage requirements of arbitrary installation on site and is difficult to support rapid on-site deployment and tool-free installation; Second, the on-site calibration scheme based on inertial measurement unit requires the introduction of additional sensing units such as gyroscope, which increases the hardware cost compared with the pure triaxial accelerometer scheme and is not conducive to low-cost promotion and application. Summary of the Invention

[0005] To address the limitations of existing technologies that require precise factory installation, making them unsuitable for arbitrary on-site installation and hindering rapid deployment and tool-free installation; and to address the technical challenges of field calibration solutions based on inertial measurement units (IMUs) which require additional sensing units such as gyroscopes, increasing hardware costs compared to pure triaxial accelerometer solutions, thus hindering low-cost promotion and application.

[0006] The technical solution provided by this invention is as follows: The first aspect of this invention provides a method for calibrating the shaft alignment error of a wheel-mounted triaxial accelerometer, comprising: S1: Acquire raw triaxial acceleration data during wheel motion; S2: Preprocess the raw triaxial acceleration data and construct an effective dataset based on the preprocessed raw triaxial acceleration data; S3: Based on the valid dataset, construct the accelerometer coordinate system and the wheel coordinate system, and construct the attitude transformation relationship from the accelerometer coordinate system to the wheel coordinate system; S4: Construct the cost function based on the pose transformation relationship and the effective dataset; S5: By using a preset nonlinear optimization algorithm, the minimum value of the cost function is iteratively solved, and the estimated value of the axis alignment error angle is output. S6: Acquire real-time accelerometer data, and perform projection transformation on the real-time accelerometer data based on the axis alignment error angle estimate and attitude transformation relationship to obtain multiple transformed acceleration components; S7: Calculate the root mean square error based on each transformed acceleration component, and verify the shaft alignment error angle estimate based on the root mean square error to generate installation error calibration parameters. S8: Calibrate the real-time accelerometer data according to the installation error calibration parameters.

[0007] A second aspect of the present invention provides a wheel-mounted triaxial accelerometer shaft alignment error calibration system, comprising: processor.

[0008] The memory stores computer-readable instructions, which, when executed by a processor, implement the wheel-mounted triaxial accelerometer shaft alignment error calibration method as described in the first aspect.

[0009] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the wheel-mounted triaxial accelerometer shaft alignment error calibration method of the first aspect.

[0010] The beneficial effects of the technical solution provided by this invention include: In this embodiment of the invention, firstly, by utilizing triaxial acceleration data collected during the normal rolling of the wheel, a least-squares cost function is constructed with the theoretical acceleration component in the wheel rotation axis direction always being zero as a physical constraint. This is combined with a nonlinear optimization algorithm to automatically solve for the shaft alignment installation error angle. No factory-installed precision tooling or on-site positioning reference is required. This supports automatic calibration after on-site installation in any orientation of the accelerometer, enabling existing factory-installed precision installation solutions to adapt to any on-site installation and allowing for rapid deployment. Secondly, by using only a single triaxial accelerometer as the sole sensing unit to complete all calibration calculations and subsequent odometer calculations, no additional gyroscopes or other devices are needed to form an inertial measurement unit. This effectively reduces hardware costs while ensuring calibration accuracy and ranging performance, lowering the cost of existing inertial measurement unit calibration schemes and facilitating low-cost promotion. Attached Figure Description

[0011] Figure 1This is a flowchart illustrating a method for calibrating the alignment error of a triaxial accelerometer mounted on a wheel, as provided in an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of a wheel-mounted triaxial accelerometer shaft alignment error calibration system provided in an embodiment of the present invention. Detailed Implementation

[0013] Reference manual attached Figure 1 The diagram shows a flowchart of a method for calibrating the shaft alignment error of a wheel-mounted triaxial accelerometer according to an embodiment of the present invention.

[0014] This invention provides a method for calibrating the shaft alignment error of a wheel-mounted triaxial accelerometer, which may include the following steps: S1: Acquire raw triaxial acceleration data during wheel movement.

[0015] S2: Preprocess the raw triaxial acceleration data and construct an effective dataset based on the preprocessed raw triaxial acceleration data.

[0016] In one possible implementation, S2 specifically includes sub-steps S201 to S203: S201: The original triaxial acceleration data is filtered using a low-pass filtering algorithm to obtain denoised triaxial acceleration data.

[0017] It should be noted that by using a low-pass filtering algorithm to filter out the high-frequency noise components introduced by wheel vibration and road impact, the effective low-frequency acceleration signals corresponding to gravity and wheel rotation can be retained, significantly improving the data signal-to-noise ratio and avoiding the deviation caused by high-frequency interference in the subsequent cost function construction and optimization solution, thus providing a high-quality data foundation for the accurate estimation of axle alignment error.

[0018] S202: By using the amplitude threshold discrimination method, outliers are removed from the denoised triaxial acceleration data to obtain multiple valid data segments.

[0019] It should be noted that the amplitude threshold discrimination method can quickly remove outliers that exceed the physical reasonable range, effectively eliminating jump data caused by instantaneous impacts, avoiding outlier samples from pulling the optimization direction of the least squares cost function, ensuring the stability and accuracy of the error angle solution process, and the algorithm has low computational load and is easy to embed into a system. At the same time, by screening data segments under stable rolling conditions, it can exclude invalid data that does not meet the theoretical constraints of the rotation axis acceleration, such as sharp turns and violent bumps, ensuring that all samples participating in the calibration calculation meet the physical premises.

[0020] S203: Perform time-series splicing and indexing of each valid data segment to obtain a valid dataset.

[0021] It should be noted that temporal splicing and unified indexing of valid data fragments can integrate scattered valid samples into a structured valid dataset, which facilitates subsequent batch construction of residual terms and iterative optimization calculations, improves the execution efficiency of the calibration algorithm, and also facilitates data traceability and verification, ensuring the reproducibility of the calibration process.

[0022] S3: Based on the valid dataset, construct the accelerometer coordinate system and the wheel coordinate system, and construct the attitude transformation relationship from the accelerometer coordinate system to the wheel coordinate system.

[0023] Specifically, the transformation relationship from the accelerometer coordinate system to the wheel coordinate system is constructed, and the formula is as follows:

[0024] in, This represents the direction cosine transformation matrix from the accelerometer coordinate system to the wheel coordinate system. c Indicates the first attitude angle. i The second attitude angle is represented by sin, which represents the sine function, and cos, which represents the cosine function. Indicates circling x The fundamental single-axis rotation transformation matrix for the second attitude angle of the axis. Indicates circling y The basic rotational transformation matrix of the first attitude angle of the axis.

[0025] S4: Construct the cost function based on the pose transformation relationship and the effective dataset.

[0026] In one possible implementation, S4 specifically includes sub-steps S401 to S404: S401: Based on the attitude transformation relationship, the effective dataset is projected onto the wheel coordinate system to obtain multiple acceleration components, and multiple target axial acceleration components parallel to the wheel rotation axis are extracted from each acceleration component.

[0027] Specifically, the formula for calculating the axial acceleration component of the target parallel to the wheel's rotation axis in the projected wheel coordinate system is as follows:

[0028] in, Representing the wheel coordinate system z axial acceleration components, Represents the accelerometer coordinate system x The acceleration value of the shaft. Represents the accelerometer coordinate system y The acceleration value of the shaft. Represents the accelerometer coordinate system z The acceleration value of the shaft.

[0029] It should be noted that by mapping the raw accelerometer data to the standard wheel coordinate system through mathematical transformation, a direct geometric relationship is established between the sensor output and the wheel rotation axis. This provides an accurate and physically meaningful coordinate reference for error quantification using physical constraints, thereby ensuring the targeted nature of subsequent processing.

[0030] S402: Calculate the difference between each target axial acceleration component and zero to obtain multiple single-sampling point residual terms.

[0031] Specifically, considering the constraint that the theoretical value of the acceleration in the direction of the wheel's rotation axis is 0, the residual of a single sampling point is defined as follows:

[0032] in, Indicates the first i The axial residual term corresponding to each valid sampling point Indicates the first i The acquisition time corresponding to each valid sampling point.

[0033] It should be noted that calculating the difference between the target axial component and the theoretical zero value can specifically quantify the measurement deviation at each sampling moment, transforming the abstract physical constraint (no acceleration in the direction of the rotation axis) into a clear and calculable numerical error index, providing a direct calculation object for subsequent mathematical optimization.

[0034] S403: Arrange the residual terms of each single sampling point according to the sampling time sequence corresponding to the residual terms of the single sampling point to obtain the residual vector.

[0035] Specifically, the residuals of all sampling points are squared and summed to generate a least-squares cost function with the rotation angle as the independent variable.

[0036] The residual vector is composed of the residuals of all sampling points:

[0037] in, This represents the axial residual vector composed of the axial residual terms of all valid sampling points arranged in time sequence, with the superscript indicating that... T This represents the vector transpose operator. N This represents the total number of valid sampling points used in constructing the objective function.

[0038] It should be noted that organizing the discrete sampling point residual terms into a unified residual vector according to the time sequence integrates the scattered time series data into a structured mathematical object. This not only facilitates data management, but also lays the mathematical foundation for batch processing and overall error assessment using matrix operations.

[0039] S404: Determine the cost function based on the residual vector.

[0040] Specifically, a least-squares cost function is constructed based on the sum of squared residuals:

[0041] in, Let represent the robust axial residual objective function, which is an optimization objective in the form of least squares. Indicates the first i The axial residual vector of each sampling point.

[0042] Furthermore, the preset attitude parameter is the roll angle. c and pitch angle i And according to the first wheel coordinate system y Axis rotation i Corner, then around x Axis rotation c The order of the angles constructs the transformation matrix from the accelerometer coordinate system to the wheel coordinate system, that is:

[0043] Based on this, the corresponding projection component expressions and cost functions are derived.

[0044] It should be noted that by constructing a cost function, this step integrates all residual information into a global optimization objective, enabling the algorithm to systematically search for and determine the optimal axis alignment parameters that minimize the total error through mathematical optimization methods (such as least squares), thereby ensuring the global optimality of the calibration results and the convergence of the calculation process.

[0045] S5: By using a preset nonlinear optimization algorithm, the minimum value of the cost function is iteratively solved, and the estimated value of the axis alignment error angle is output.

[0046] In one possible implementation, S5 specifically includes sub-steps S501 to S504: S501: Set the initial value of the preset attitude parameter iteration, the preset value boundary, the preset iteration convergence threshold, and the preset maximum number of iterations to obtain the current preset attitude parameters and iteration control conditions.

[0047] It should be noted that by pre-setting the initial values, boundary values, and convergence conditions for iteration, a complete set of operational constraints and a starting point are provided for the nonlinear optimization algorithm. This effectively prevents the algorithm from falling into an unsolvable state or diverging, while balancing computational accuracy and speed, ensuring that the entire optimization process starts stably within a controllable range.

[0048] S502: Input the current preset attitude parameters into the attitude transformation relationship, and calculate the current cost function value and gradient value corresponding to the current preset attitude parameters.

[0049] It should be noted that by calculating the cost function value and gradient corresponding to the current attitude parameters in real time, the algorithm is provided with an indicator to quantitatively evaluate the magnitude of the current position error and the direction for the next optimization step. This feedback mechanism provides a key mathematical basis for accurately calculating the parameter update amount in subsequent steps, ensuring the correctness of the iterative correction.

[0050] S503: Calculate the update amount of the preset attitude parameters through a preset nonlinear optimization algorithm, and update the current preset attitude parameters according to the update amount to obtain the updated preset attitude parameters.

[0051] It should be noted that by using a nonlinear optimization algorithm to automatically calculate and update parameters based on gradient information, the alignment error of the axis is automatically corrected. By gradually approximating the optimal solution through mathematical iteration, the tediousness of manual trial and error is effectively avoided, and the automation level of the calibration process and the accuracy of the final parameter estimation are significantly improved.

[0052] S504: Input the updated preset attitude parameters into the cost function to obtain the updated cost function value. Based on the updated preset attitude parameters, the updated cost function value, the preset iteration convergence threshold, and the current iteration number, determine whether the current iteration meets the convergence condition. If yes, determine the rotation angle corresponding to the updated preset attitude parameters as the estimated axis alignment error angle. Otherwise, return to step S502.

[0053] It should be noted that by strictly controlling the termination time of the algorithm through logical judgment, it is ensured that the final output axis alignment error angle estimate meets the preset accuracy requirements, and the calculation can be stopped in time after the target is reached to avoid invalid redundant iterations. This effectively improves the execution efficiency of the algorithm while ensuring the accuracy of the results.

[0054] S6: Acquire real-time accelerometer data, and perform projection transformation on the real-time accelerometer data based on the axis alignment error angle estimate and attitude transformation relationship to obtain multiple transformed acceleration components.

[0055] In one possible implementation, S6 specifically includes sub-steps S601 to S603: S601: Determine the attitude transformation matrix based on the estimated axis alignment error angle and the attitude transformation relationship.

[0056] It should be noted that by converting the estimated axis alignment error angle into a precise attitude transformation matrix, a definite mapping relationship from the sensor coordinate system to the wheel coordinate system is established, and the abstract calibration parameters are solidified into a specific mathematical model, providing a standardized calculation tool for subsequent data correction.

[0057] S602: Project the effective dataset onto the wheel coordinate system to obtain the acceleration components in the wheel coordinate system.

[0058] It should be noted that by using the attitude transformation matrix to project the original acceleration data onto the wheel coordinate system, the coordinate system inconsistency problem caused by installation attitude deviation is completely eliminated from the data source, and the original sensor output is restored to ideal data that conforms to the geometric motion characteristics of the wheel.

[0059] S603: Extract the target axial acceleration component parallel to the wheel rotation axis from the acceleration components to obtain the transformed acceleration components.

[0060] It should be noted that the precise extraction of the target axial acceleration component after projection correction effectively eliminates the interference of gravity component coupling caused by installation errors, ensuring that the input signal used for mileage calculation purely reflects the real rotational motion of the wheel, thereby significantly improving the accuracy and reliability of the measurement.

[0061] S7: Calculate the root mean square error based on each transformed acceleration component, and verify the shaft alignment error angle estimate based on the root mean square error to generate installation error calibration parameters.

[0062] In one possible implementation, S7 specifically includes sub-steps S701 to S703: S701: Perform a square operation on each transformed acceleration component, and obtain the root mean square error based on the square operation result.

[0063] It should be noted that by calculating the root mean square error, the overall deviation level across all sampling points can be comprehensively quantified, effectively avoiding the bias caused by fluctuations in a single data point, and providing an objective, accurate, and representative global indicator for evaluating the calibration effect.

[0064] S702: Compare the root mean square error with the preset verification conditions to obtain the verification result of the axis alignment error angle estimate.

[0065] It should be noted that by automatically comparing the error index with the preset threshold, the calibration results are automatically inspected and judged, ensuring that only the estimated value that meets the specific accuracy standard can be accepted, thereby effectively avoiding measurement errors caused by unqualified parameters in practical applications.

[0066] S703: When the verification result is passed, the shaft alignment error angle estimate is determined as the installation error calibration parameter.

[0067] Specifically, the quasi-Newton method is used to evaluate the cost function. Perform a nonlinear minimization solution. Specifically, this includes: Parameter initialization: Set the number of iterations Initial value of rotation angle iteration Set angle boundaries: , Convergence threshold Maximum number of iterations .

[0068] Iterative update formula:

[0069] in, Indicates the current iteration step. Indicates the first attitude angle after the first... k The result obtained after the nth iteration update k +1 axis alignment error attitude parameter value Indicates the second attitude angle after the first k The result obtained after the nth iteration update k +1 axis alignment error attitude parameter value Indicates the first attitude angle after the first... k The axis alignment error attitude parameter values ​​for the next iteration. Indicates the second attitude angle after the first k The axis alignment error attitude parameter values ​​for the next iteration. For the first Approximate value of the Hessian matrix in the next iteration Cost function The gradient vector at the current angle.

[0070] Convergence criterion: When or When the iteration ends, retrieve the current value. This serves as the final estimate of the axis alignment error angle. The estimation result is: , , This represents the estimated value of the first attitude angle. This represents the estimated value of the second attitude angle. e This represents the preset iteration convergence threshold.

[0071] Using the estimated , Project all the original data onto the wheel coordinate system and calculate the z-axis acceleration of the projected b-frame. Before calibration, i.e. The root mean square error (true value taken as 0) is 5.7 m / s². After calibration, that is... The root mean square error (true value taken as 0) was reduced to 0.82 m / s².

[0072] Furthermore, utilizing the projected and The wheel rotation angle is calculated by sequentially passing the low-pass filter, removing the low-frequency centripetal force component, and calculating the arctangent.

[0073] For example, define a unit quaternion. Characterizes the attitude deflection of the accelerometer coordinate system relative to the wheel coordinate system, and satisfies the normalization constraints. .

[0074] Deriving the coordinate transformation matrix from unit quaternions :

[0075] in, Representing unit quaternions q Let be the coordinate transformation matrix from the accelerometer coordinate system to the wheel coordinate system for the attitude parameters. , , , These represent the four components of the unit quaternion.

[0076] The original output vector of the accelerometer Left multiplication of the transformation matrix By projecting the image onto the wheel coordinate system, the acceleration components along the wheel's rotation axis are obtained. :

[0077] in, This represents the acceleration component in the direction of the wheel's rotation axis.

[0078] based on Given a constraint with a theoretical value of 0, construct a cost function with quaternions as independent variables.

[0079] Define the residual of a single sampling point:

[0080] in, Indicates the first The time corresponding to each sampling point.

[0081] The residual vector is composed of the residuals of all sampling points:

[0082] in, This represents the total number of valid sampling points.

[0083] Constructing a least-squares cost function based on the sum of squared residuals:

[0084] The Levenberg-Marquardt (LM) algorithm is used to evaluate the cost function. Perform a nonlinear minimization solution. This includes the following steps: Parameter initialization: Set the number of iterations Quaternion iteration initial value Set a convergence threshold Maximum number of iterations The initial value of the damping coefficient is 0.01.

[0085] Iterative update formula:

[0086] in, Indicates the current iteration step. Indicates the first Quaternion in the next iteration This represents the residual vector formed by the residuals of all sampling points. Represents the residual vector with respect to Jacobian matrix, This represents the damping coefficient, used to balance gradient descent and Gaussian-Newton characteristics. Represents the identity matrix.

[0087] Convergence criterion: When or When the iteration ends, retrieve the current value. As the final quaternion estimate, the estimation result is: .

[0088] Using the estimated Project all the original data onto the wheel coordinate system and calculate the z-axis acceleration of the projected b-frame. After calibration, that is... The root mean square error (true value taken as 0) was reduced to 0.83 m / s². Furthermore, using the projected... and By sequentially applying low-pass filtering, removing low-frequency centripetal force components, calculating the arctangent of the wheel rotation angle, and multiplying it by the radius, the mileage was obtained as 492.219m, with an error of only 0.08m, indicating that the method is effective.

[0089] For example, define a rotation vector The attitude deflection of the accelerometer coordinate system relative to the wheel coordinate system is physically represented by the direction of the rotation axis relative to the wheel coordinate system. The directions are the same, and the rotation angle is equal to... Length of the module , .

[0090] The attitude transformation matrix is ​​calculated from the rotation vector based on the Rodrigues rotation formula. :

[0091] in, It is a 3-order identity matrix. express The corresponding antisymmetric matrix:

[0092] The original output vector of the accelerometer Left multiplication of the transformation matrix By projecting the image onto the wheel coordinate system, the acceleration components along the wheel's rotation axis are obtained. .

[0093] based on Given a constraint with a theoretical value of 0, construct a cost function with the rotation vector as the independent variable:

[0094] The Levenberg-Marquardt (LM) algorithm is used to evaluate the cost function. Perform a nonlinear minimization solution. This includes the following steps: Parameter initialization: Set the number of iterations Initial value of rotation vector iteration Set a convergence threshold Maximum number of iterations The initial value of the damping coefficient is 0.01.

[0095] Iterative update formula:

[0096] in, Indicates the current iteration step. Indicates the first The rotation vector of the next iteration. This represents the residual vector formed by the residuals of all sampling points. Represents the residual vector with respect to Jacobian matrix, This represents the damping coefficient, used to balance gradient descent and Gaussian-Newton characteristics. Represents the identity matrix.

[0097] Convergence criterion: When or When the iteration ends, retrieve the current value. As the final quaternion estimate, the estimation result is: .

[0098] Using the estimated Project all the original data onto the wheel coordinate system and calculate the z-axis acceleration of the projected b-frame. After calibration, that is... The root mean square error (true value taken as 0) was reduced to 0.83 m / s².

[0099] Furthermore, utilizing the projected... and By sequentially applying low-pass filtering, removing low-frequency centripetal force components, calculating the arctangent of the wheel rotation angle, and multiplying it by the radius, the mileage was obtained as 492.219m, with an error of only 0.08m, indicating that the method is effective.

[0100] The estimated shaft alignment installation error is used to project and transform the real-time acquired accelerometer data, and the transformed acceleration components are used to calculate the wheel angle and mileage.

[0101] After calibration, the estimated , The data is stored in the microcontroller. During subsequent normal mileage measurement, for each frame of accelerometer data acquired, a corresponding projection transformation is performed according to the selected rotation sequence, and the transformed data is then used... , The instantaneous wheel rotation angle is obtained by arctangent calculation, and then the mileage is obtained by integration.

[0102] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0103] It should be noted that after the transformation relationship is verified, the core coordinate transformation matrix and rotation axis direction vector are encapsulated into standardized installation error calibration parameters. This facilitates parameter storage, retrieval, and portability, and also provides a unified input interface for subsequent real-time acceleration data calibration, improving the engineering feasibility and reusability of the solution.

[0104] S8: Calibrate the real-time accelerometer data according to the installation error calibration parameters.

[0105] In one possible implementation, S8 specifically includes sub-steps S801 and S802: S801: Preprocesses real-time accelerometer data.

[0106] It should be noted that by using preprocessing methods such as filtering and denoising, high-frequency vibration noise and outlier interference in the real-time acquired data are effectively removed, significantly improving the signal-to-noise ratio and quality of the signal, providing clean and reliable data input for subsequent real-time calibration calculations, and enhancing the stability of the system under complex operating conditions.

[0107] S802: Calibrate the pre-processed real-time accelerometer data according to the installation error calibration parameters.

[0108] It should be noted that by using pre-determined calibration parameters to dynamically correct real-time data, measurement deviations caused by on-site installation errors can be continuously eliminated, ensuring that the system can output high-precision data conforming to the standard coordinate system under any installation posture, thereby guaranteeing the accuracy and reliability of mileage measurement in practical applications.

[0109] Reference manual attached Figure 2 The diagram shows a structural schematic of a wheel-mounted triaxial accelerometer shaft alignment error calibration system provided by an embodiment of the present invention.

[0110] The present invention also provides a wheel-mounted triaxial accelerometer shaft alignment error calibration system 20, applied to the above-mentioned wheel-mounted triaxial accelerometer shaft alignment error calibration method, comprising: Processor 201.

[0111] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, they implement the wheel-mounted triaxial accelerometer shaft alignment error calibration method as described in the method embodiment.

[0112] The wheel-mounted triaxial accelerometer shaft alignment error calibration system 20 provided by the present invention can perform the above-mentioned wheel-mounted triaxial accelerometer shaft alignment error calibration method and achieve the same or similar technical effects. To avoid repetition, the present invention will not elaborate further.

[0113] This invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the wheel-mounted triaxial accelerometer shaft alignment error calibration method as described in the method embodiment.

[0114] The present invention provides a computer-readable storage medium that can implement the steps and effects of the wheel-mounted triaxial accelerometer shaft alignment error calibration method of the above method embodiments. To avoid repetition, the present invention will not repeat them.

[0115] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for calibrating the shaft alignment error of a triaxial accelerometer mounted on a wheel, characterized in that, include: S1: Acquire raw triaxial acceleration data during wheel motion; S2: Preprocess the raw triaxial acceleration data, and construct an effective dataset based on the preprocessed raw triaxial acceleration data; S3: Based on the valid dataset, construct the accelerometer coordinate system and the wheel coordinate system, and construct the attitude transformation relationship from the accelerometer coordinate system to the wheel coordinate system; S4: Construct a cost function based on the attitude transformation relationship and the effective dataset; S5: By using a preset nonlinear optimization algorithm, the minimum value of the cost function is iteratively solved, and the estimated value of the axis alignment error angle is output. S6: Acquire real-time accelerometer data, and perform projection transformation on the real-time accelerometer data according to the estimated value of the axis alignment error angle and the attitude transformation relationship to obtain multiple transformed acceleration components; S7: Calculate the root mean square error based on each of the transformed acceleration components, and verify the estimated value of the shaft alignment error angle based on the root mean square error to generate installation error calibration parameters; S8: Calibrate the real-time accelerometer data according to the installation error calibration parameters.

2. The method for calibrating the shaft alignment error of a triaxial accelerometer mounted on a wheel according to claim 1, characterized in that, S2 specifically includes: S201: The original triaxial acceleration data is filtered using a low-pass filtering algorithm to obtain denoised triaxial acceleration data; S202: By using the amplitude threshold discrimination method, outliers are removed from the denoised triaxial acceleration data to obtain multiple valid data segments; S203: Perform time-series splicing and indexing on each of the valid data segments to obtain the valid dataset.

3. The method for calibrating the shaft alignment error of a triaxial accelerometer mounted on a wheel according to claim 1, characterized in that, S4 specifically includes: S401: Based on the attitude transformation relationship, the effective dataset is projected onto the wheel coordinate system to obtain multiple acceleration components, and multiple target axial acceleration components parallel to the wheel rotation axis are extracted from each of the acceleration components. S402: Calculate the difference between each of the target axial acceleration components and zero to obtain multiple single-sampling point residual terms; S403: Arrange the residual terms of each single sampling point according to the sampling time sequence corresponding to the single sampling point residual term to obtain the residual vector; S404: Determine the cost function by taking the sum of squared residuals based on the residual vector.

4. The method for calibrating the shaft alignment error of a triaxial accelerometer mounted on a wheel according to claim 1, characterized in that, S5 specifically includes: S501: Set the initial value of the preset attitude parameter iteration, the preset value boundary, the preset iteration convergence threshold, and the preset maximum number of iterations to obtain the current preset attitude parameters and iteration control conditions; S502: Input the current preset attitude parameters into the attitude transformation relationship, and calculate the current cost function value and gradient value corresponding to the current preset attitude parameters; S503: Calculate the update amount of the preset attitude parameters using the preset nonlinear optimization algorithm, and update the current preset attitude parameters according to the update amount to obtain the updated preset attitude parameters; S504: Input the updated preset attitude parameters into the cost function to obtain the updated cost function value, and determine whether the current iteration meets the convergence condition based on the updated preset attitude parameters, the updated cost function value, the preset iteration convergence threshold, and the current iteration number; if yes, determine the rotation angle corresponding to the updated preset attitude parameters as the estimated value of the axis alignment error angle; otherwise, return to step S502.

5. The method for calibrating the shaft alignment error of a triaxial accelerometer mounted on a wheel according to claim 1, characterized in that, S6 specifically includes: S601: Determine the attitude transformation matrix based on the estimated axis alignment error angle and the attitude transformation relationship; S602: Project the effective dataset onto the wheel coordinate system to obtain the acceleration components in the wheel coordinate system; S603: Extract the target axial acceleration component parallel to the wheel rotation axis from the acceleration component to obtain the transformed acceleration component.

6. The method for calibrating the shaft alignment error of a triaxial accelerometer mounted on a wheel according to claim 1, characterized in that, Specifically, S7 includes: S701: Perform a squaring operation on each of the transformed acceleration components, and obtain the root mean square error based on the squaring operation results; S702: Compare the root mean square error with the preset verification conditions to obtain the verification result of the axis alignment error angle estimate; S703: When the verification result is passed, the estimated value of the shaft alignment error angle is determined as the installation error calibration parameter.

7. The method for calibrating the shaft alignment error of a triaxial accelerometer mounted on a wheel according to claim 1, characterized in that, S8 specifically includes: S801: Preprocess the real-time accelerometer data; S802: Calibrate the preprocessed real-time accelerometer data according to the installation error calibration parameters.

8. A wheel-mounted triaxial accelerometer shaft alignment error calibration system, characterized in that, include: processor; A memory storing computer-readable instructions, which, when executed by the processor, implement the wheel-mounted triaxial accelerometer shaft alignment error calibration method as described in any one of claims 1 to 8.

9. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the wheel-mounted triaxial accelerometer shaft alignment error calibration method as described in any one of claims 1 to 7.

Images