CN-121994277-A - Calibration method and calibration system for inertial measurement unit and computer program product

CN121994277ACN 121994277 ACN121994277 ACN 121994277ACN-121994277-A

Abstract

The application provides a calibration method, a calibration system and a computer program product of an inertial measurement unit, and relates to the technical field of filter estimation algorithms; the calibration method of the inertial measurement unit comprises the steps of constructing a first error mathematical model aiming at acceleration and a second error mathematical model aiming at angular velocity, integrating a first error parameter and a second error parameter into a target state vector, constructing a state equation of the first state vector changing along with time, a state equation of the second state vector changing along with time, an observation equation of gravity constraint under a static scene and an observation equation of external sensor true value constraint under a dynamic scene, executing an unscented Kalman filtering process on the original data of the inertial measurement unit at each moment, and dynamically adjusting process noise covariance and observation noise covariance in the unscented Kalman filtering frame. By the embodiment of the application, the real-time calibration under static and dynamic scenes is realized, and special static calibration equipment or forced static stages are not needed to be relied on.

Inventors

TAN FEIYANG
ZHANG GUANGYA
ZHANG YAN
NING ZHONGLIN
XIE GUOWEN

Assignees

上汽通用五菱汽车股份有限公司

Dates

Publication Date: 20260508
Application Date: 20251231

Claims (10)

1.A method of calibrating an inertial measurement unit, comprising: constructing a first error mathematical model for acceleration, wherein the first error mathematical model comprises a first zero offset error, a first scale factor and first inter-axis coupling which are used as first error parameters; Constructing a second error mathematical model for the angular velocity, wherein the second error mathematical model comprises a second zero offset error, a second scale factor and second inter-axis coupling which are used as second error parameters; integrating the first error parameter and the second error parameter into a target state vector; Based on a preset unscented Kalman filtering framework, constructing a state equation of the first state vector changing along with time, constructing a state equation of the second state vector changing along with time, constructing a gravity constraint observation equation under a static scene, and constructing an external sensor true value constraint observation equation under a dynamic scene; For the original data of the inertial measurement unit at each moment, performing an unscented Kalman filtering process; And dynamically adjusting the process noise covariance and the observed noise covariance in the unscented Kalman filtering frame according to the motion state of the inertial measurement unit or the quality of external reference information.
2. The method of calibrating an inertial measurement unit according to claim 1, wherein the first mathematical model of error is: Wherein, the A first inter-axis coupling matrix for a first scale factor; Outputting an acceleration value for the accelerometer; Is the true acceleration; Is the first zero offset error; Is white gaussian noise.
3. The method of calibrating an inertial measurement unit according to claim 2, wherein the second mathematical model of error is: Wherein, the Is a second scale factor and an inter-axis coupling matrix; outputting an angular velocity value for the gyroscope; Is the true angular velocity; Is the second zero offset error; Is white gaussian noise.
4. A method of calibrating an inertial measurement unit according to claim 3, wherein integrating the first error parameter and the second error parameter into a target state vector comprises: 。
5. A method of calibrating an inertial measurement unit according to claim 3, wherein integrating the first error parameter and the second error parameter into a target state vector comprises: 。
6. The method according to claim 1, wherein constructing a state equation of a first state vector over time, constructing a state equation of a second state vector over time, constructing an observation equation of a gravity constraint in a static scene, and constructing an observation equation of an external sensor true value constraint in a dynamic scene based on a preset unscented kalman filter frame, comprises: Assuming that the first and second error parameters change smoothly over time, the fitted curve may be expressed as: Setting up To vary the covariance of the process noise, diagonal elements correspond to the time-varying rates of the parameters Wherein is An offset; Accelerometer gravity mode length constraint in static scene Scalar observation equations can be constructed: , N is 9; providing motion truth in dynamic scenarios in combination with external sensors Constructing an angular velocity observation equation: , 。
7. The method according to claim 1, wherein the performing an unscented kalman filter process for the raw data of the inertial measurement unit at each time instant comprises: According to the current state Sum covariance 2N+1 Sigma points were generated, n being 9: Wherein, the K is a positive definite parameter range [3, n ], n is 9,W and is a weighting value; the positive definite matrix is decomposed by square root method, Along the generated lower triangular matrix Symmetrically sampling each column vector direction to generate Sigma points, wherein L is a positive definite matrix; according to the propagation formula Transfer with Sigma points, and curve fitting Weighted calculation of all Sigma Point means Calculating Sigma Point covariance Q is process noise; Setting an observation model Mapping Sigma points to an observation space; The observation point is expressed as Based on the calculation result of the prediction step, the observation point statistics can be expressed as Mean value of Covariance (covariance) The cross-covariance of the predicted and observed values may be expressed as R is observation noise; Computing gain of unscented Kalman filter Updating the predicted value: 。
8. The method according to claim 1, wherein dynamically adjusting the process noise covariance and the observed noise covariance in the unscented kalman filter frame according to the motion state of the inertial measurement unit or the quality of the external reference information, comprises: Aiming at process noise, diagonal elements are set according to parameter time-varying characteristics, and the random walk coefficient of zero offset is 0.01; Aiming at observation noise, taking the square of a gravity module length measurement error by a static scene and taking the square of the precision of an external sensor by a dynamic scene; the accelerometer and gyroscope with the inertial measurement unit are separately calibrated, and the state dimension of each filtering calculation is reduced from 18 dimensions to 9 dimensions.
9. Calibration system of an inertial measurement unit, for implementing a method according to any one of claims 1 to 8, the system comprising: a first construction module, configured to construct a first error mathematical model for acceleration, where the first error mathematical model includes a first zero offset error, a first scale factor, and a first inter-axis coupling as first error parameters; the first integration module is used for integrating the first error parameters into a first state vector; The second construction module is used for constructing a second error mathematical model aiming at the angular velocity, and the second error mathematical model comprises a second zero offset error, a second scale factor and second inter-axis coupling which are second error parameters; the second integration module is used for integrating the second error parameter into a second state vector; the third construction module is used for constructing a state equation of the first state vector changing along with time based on a preset unscented Kalman filtering frame, constructing a state equation of the second state vector changing along with time, constructing a gravity constraint observation equation under a static scene and constructing an external sensor true value constraint observation equation under a dynamic scene; the execution module is used for executing an unscented Kalman filtering process on the original data of the inertial measurement unit at each moment; And the dynamic adjustment module is used for dynamically adjusting the process noise covariance and the observed noise covariance in the unscented Kalman filter frame according to the motion state of the inertial measurement unit or the quality of external reference information.
10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the steps of the method as claimed in any one of claims 1-8.

Description

Calibration method and calibration system for inertial measurement unit and computer program product Technical Field The invention relates to the technical field of filter estimation algorithms, in particular to a calibration method, a calibration system and a computer program product of an inertial measurement unit. Background The inertial measurement unit is a core sensor device for measuring the motion state of an object in a three-dimensional space, and three-axis acceleration and three-axis angular velocity data of the object are obtained in real time mainly through elements such as an integrated accelerometer and a gyroscope. IMU is a key sensor in the fields of unmanned aerial vehicle, robot, autopilot, etc., but its measurement accuracy is affected by various errors. The traditional calibration method generally depends on a high-precision turntable and a strict laboratory environment, is complex in process, cannot perform online estimation and compensation on time-varying error parameters in the running process of equipment, and is difficult to meet the requirements of dynamic and long-endurance application. Disclosure of Invention Aiming at the defects of the prior art, the application provides a calibration method, a calibration system and a computer program product of an inertial measurement unit, which are used for solving the technical problems of static dependence, expensive equipment, error accumulation and poor nonlinear processing of an IMU calibration method in the prior art. In a first aspect, an embodiment of the application provides a calibration method of an inertial measurement unit, which comprises the steps of constructing a first error mathematical model for acceleration, wherein the first error mathematical model comprises a first zero bias error serving as a first error parameter, a first scale factor and first inter-axis coupling, constructing a second error mathematical model for angular velocity, wherein the second error mathematical model comprises a second zero bias error serving as a second error parameter, a second scale factor and second inter-axis coupling, integrating the first error parameter and the second error parameter into a target state vector, constructing a state equation of a first state vector changing with time based on a preset unscented Kalman filter frame, constructing a state equation of a second state vector changing with time, constructing an observation equation of a gravity constraint in a static scene and an observation equation of an external sensor true value constraint in a dynamic scene, performing a unscented Kalman filter process on raw data of the inertial measurement unit at each moment, and dynamically adjusting a noise-free covariance process in the unscented Kalman filter frame according to the motion state of the inertial measurement unit or the quality of external reference information. As an alternative embodiment, the first error mathematical model is: Wherein, the A first inter-axis coupling matrix for a first scale factor; Outputting an acceleration value for the accelerometer; Is the true acceleration; Is the first zero offset error; Is white gaussian noise. As an alternative embodiment, the second mathematical error model is: Wherein, the Is a second scale factor and an inter-axis coupling matrix; outputting an angular velocity value for the gyroscope; Is the true angular velocity; Is the second zero offset error; Is white gaussian noise. As an alternative embodiment, integrating the first error parameter and the second error parameter into a target state vector comprises: As an alternative embodiment, integrating the first error parameter and the second error parameter into a target state vector comprises: As an alternative embodiment, based on a preset unscented kalman filter framework, constructing a state equation of a first state vector changing with time, constructing a state equation of a second state vector changing with time, constructing an observation equation of gravity constraint in a static scene, and constructing an observation equation of true value constraint of an external sensor in a dynamic scene, including: Assuming that the first and second error parameters change smoothly over time, the fitted curve may be expressed as: Setting up To vary the covariance of the process noise, diagonal elements correspond to the time-varying rates of the parametersWherein isAn offset; Accelerometer gravity mode length constraint in static scene Scalar observation equations can be constructed: , N is 9; providing motion truth in dynamic scenarios in combination with external sensors Constructing an angular velocity observation equation: , as an alternative embodiment, the performing the unscented kalman filter process for the raw data of the inertial measurement unit at each moment includes: According to the current state Sum covariance2N+1 Sigma points were generated, n being 9: Wherein, the K is a positive definite p