A MEMS array gyroscope data fusion method and system based on singular value decomposition

Abstract

The application discloses a MEMS array gyroscope data fusion method and system based on singular value decomposition, and the data fusion method comprises the following steps: S1, MEMS gyroscope array configuration; S2, data acquisition; S3, fusion matrix construction, arranging data according to the sensor serial number priority and row priority filling principle; S4, standard singular value decomposition of the fusion matrix; S5, first principal component screening; S6, fusion matrix reconstruction, obtaining a reconstructed fusion matrix; and S7, average fusion matrix of the reconstructed fusion matrix. The adaptive robust fusion is realized, and the measurement accuracy of the array system is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of sensor data fusion and noise suppression technology, and in particular to a MEMS array gyroscope data fusion method and system based on singular value decomposition. Background Technology

[0002] Technical Challenges: Microelectromechanical systems (MEMS) gyroscopes have been widely used in consumer electronics, drones, autonomous driving, and industrial control due to their advantages such as small size, low cost, and low power consumption. However, due to limitations in manufacturing processes and physical noise mechanisms, the key performance indicators of a single MEMS gyroscope, such as zero-bias stability and noise, are difficult to meet the requirements of high-precision navigation.

[0003] To improve performance, multiple MEMS gyroscopes of the same model are often used in engineering practice to form a redundant array, and data fusion is used to improve the overall output accuracy. The most commonly used method is the arithmetic mean method, which simply averages the output of each channel and is easy to calculate. However, it is susceptible to contamination by abnormal samples and has poor robustness. When individual gyroscopes are affected by electromagnetic interference, resulting in low-frequency sinusoidal disturbances or sudden noise anomalies, the fusion accuracy will drop significantly.

[0004] Current technological status: Singular value decomposition (SVD), as a standard tool in matrix analysis, possesses strict mathematical orthogonality and energy concentration properties. For matrices... Its singular value spectrum satisfy Large singular values ​​correspond to the principal components of the signal, while small singular values ​​correspond to noise and subspace interference. This property provides a theoretical basis for array signal denoising and feature extraction. Singular value decomposition (SVD) has been proven to have excellent low-rank approximation and noise suppression capabilities in fields such as image denoising and speech enhancement, but it has not yet formed a standardized application scheme in the field of MEMS gyroscope fusion.

[0005] Existing MEMS array gyroscope fusion methods suffer from insufficient robustness to abnormal sensor interference and limited improvement in accuracy. Summary of the Invention

[0006] In view of the above problems, the present invention is proposed to provide a MEMS array gyroscope data fusion method and system based on singular value decomposition to overcome or at least partially solve the above problems.

[0007] According to one aspect of the present invention, a MEMS array gyroscope data fusion method based on singular value decomposition is provided, the data fusion method comprising: Step S1: MEMS gyroscope array configuration; Step S2: Collect data; Step S3: Construct a fusion matrix and arrange the data according to the principle of sensor number priority and row priority filling; Step S4: Perform standard singular value decomposition on the fusion matrix; Step S5: Perform the first principal component screening; Step S6: Reconstruct the fusion matrix to obtain the reconstructed fusion matrix; Step S7: Average the reconstructed fusion matrix.

[0008] Optionally, step S1: MEMS gyroscope array configuration specifically includes: use Only MEMS gyroscope chips of the same model are used to build a redundant array, and the sensitive axes of all gyroscopes are arranged in parallel to achieve multi-source synchronous sampling with consistent frequency.

[0009] Optionally, step S2: data acquisition specifically includes: Each collection Angular velocity data from each gyroscope sensor; each gyroscope sensor collects... Each point is fused once, where , are positive integers, and satisfy... , It is a positive integer.

[0010] Optionally, step S3: constructing a fusion matrix and arranging the data according to the sensor number priority and row priority filling principle specifically includes: Data is arranged according to sensor number priority and row priority filling principle: Extract the first sensor sequentially Second sampling data, from the second sensor Sub-sampled data...until all All sampling data from each sensor is retrieved, and the collected data is sequentially filled into the matrix according to the order of "filling the first row of the matrix first, then the second row, and so on..." OK The data matrix in the column forms the fusion matrix. .

[0011] Optionally, step S4: performing standard singular value decomposition on the fusion matrix specifically includes: The numerical linear algebra library is used to perform standard singular value decomposition on the fusion matrix A. The decomposition is as follows: in, Let be a left singular vector matrix, satisfying Orthogonality of columns; For a diagonal singular value matrix, the diagonal elements ,and ; Let be a right singular vector matrix, satisfying Orthogonality of columns This is a transpose.

[0012] Optionally, step S5: performing the first principal component screening specifically includes: Preserve the maximum singular value Construct the truncated singular value matrix in the corresponding principal component subspace. .

[0013] Optionally, step S6: reconstructing the fusion matrix to obtain the reconstructed fusion matrix specifically includes: The formula for calculating the reconstructed matrix based on the truncated singular value matrix is ​​as follows: Reconstructed Noise and abnormal data interference have been filtered out, while retaining the core common characteristics and timing correlation information of each gyroscope channel.

[0014] Optionally, step S7: averaging the reconstructed fusion matrix specifically includes: The reconstructed matrix Calculate the arithmetic mean of all elements; Output fused angular velocity value , accomplish Mapping 3D sensor data to 1D high-precision angular velocity.

[0015] This invention also provides a MEMS array gyroscope data fusion system based on singular value decomposition, applying the aforementioned MEMS array gyroscope data fusion method based on singular value decomposition. The data fusion system includes: Redundant gyroscope array module for MEMS gyroscope array configuration; The synchronous acquisition module is used for data acquisition; The matrix construction module is used to construct a fusion matrix, arranging the data according to the principle of sensor number priority and row priority. The SVD decomposition module is used to perform standard singular value decomposition on the fusion matrix; The principal component screening module is used to perform the first principal component screening; The reconstruction module is used to reconstruct the fusion matrix and obtain the reconstructed fusion matrix. The fusion output module is used to average the reconstructed fusion matrix.

[0016] This invention provides a MEMS gyroscope array data fusion method and system based on singular value decomposition. The data fusion method includes: Step S1: MEMS gyroscope array configuration; Step S2: Data acquisition; Step S3: Constructing a fusion matrix, arranging data according to sensor number priority and row priority filling principle; Step S4: Performing standard singular value decomposition on the fusion matrix; Step S5: Performing first principal component screening; Step S6: Reconstructing the fusion matrix to obtain a reconstructed fusion matrix; Step S7: Averaging the reconstructed fusion matrix. This achieves adaptive robust fusion, significantly improving the measurement accuracy of the array system.

[0017] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a MEMS array gyroscope data fusion method based on singular value decomposition, provided as an embodiment of the present invention. Detailed Implementation

[0020] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0021] The terms "comprising" and "having," and any variations thereof, in the specification, embodiments, claims, and drawings of this invention are intended to cover non-exclusive inclusion, such as including a series of steps or units.

[0022] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0023] Example 1 like Figure 1As shown, a MEMS gyroscope array data fusion method and system based on singular value decomposition is disclosed. The data fusion method includes: step S1: MEMS gyroscope array configuration; step S2: data acquisition; step S3: constructing a fusion matrix, arranging the data according to the sensor number priority and row priority filling principle; step S4: performing standard singular value decomposition on the fusion matrix; step S5: performing first principal component screening; step S6: reconstructing the fusion matrix to obtain a reconstructed fusion matrix; step S7: averaging the reconstructed fusion matrix.

[0024] A method for data fusion calculation of MEMS array gyroscopes based on the singular value decomposition (SVD) algorithm includes the following steps: S1: MEMS gyroscope array configuration: using Only MEMS gyroscope chips of the same model are used to build a redundant array, and the sensitive axes of all gyroscopes are arranged in parallel to achieve multi-source synchronous sampling with consistent frequency; S2: Data Acquisition: Each acquisition Angular velocity data from each gyroscope sensor; each gyroscope sensor collects... Each point is fused once, where , are positive integers, and satisfy... , ( (is a positive integer).

[0025] S3: Fusion Matrix Construction: Arrange data according to sensor number priority and row priority filling principle: extract data from the first sensor in sequence. Second sampling data, from the second sensor Sub-sampled data...until all All sampled data from each sensor are retrieved, and this data is then sequentially filled into the matrix in the order of "filling the first row first, then the second row, and so on..." OK The data matrix in the column forms the fusion matrix. ; S4: Standard Singular Value Decomposition: The standard singular value decomposition of the fused matrix A is performed using a numerical linear algebra library (such as LAPACK). The decomposition formula is: in Let be a left singular vector matrix, satisfying (Orthogonality of columns); A diagonal singular value matrix (diagonal elements) ,and ); Let be a right singular vector matrix, satisfying (Orthogonality) Transpose it; S5: First Principal Component Screening: Retain the largest singular value Construct the truncated singular value matrix in the corresponding principal component subspace. .

[0026] S6: Fusion Matrix Reconstruction: Calculate the reconstructed matrix based on the truncated singular value matrix, using the following formula: Reconstructed Noise and abnormal data interference have been filtered out, while retaining the core common characteristics and timing correlation information of each gyroscope channel.

[0027] S7: Averaging the fusion matrix: Calculate the average of the reconstructed matrix. Calculate the arithmetic mean of all elements and output the fused angular velocity value. ,accomplish Mapping 3D sensor data to 1D high-precision angular velocity.

[0028] In step S1, the sensitive axes of the MEMS gyroscope array are arranged in parallel to ensure the spatial consistency of multi-source measurement data and to ensure the synchronous acquisition of multi-source data to ensure temporal consistency.

[0029] In step S3, the construction of the fusion matrix must strictly follow the principle of "sensor number priority, row priority" to ensure that the spatial correlation and temporal continuity of the data are not disrupted.

[0030] In steps (S5, S6), the maximum singular value is retained. Construct the truncated singular value matrix in the corresponding principal component subspace. Reconstructing the matrix .

[0031] A MEMS array gyroscope data fusion system based on the singular value decomposition (SVD) algorithm includes a redundant gyroscope array module, a synchronous acquisition module, a matrix construction module, an SVD decomposition module, a principal component screening module, a reconstruction module, and a fusion output module.

[0032] Example 2 This embodiment provides a MEMS array gyroscope data fusion method based on the SVD (Singular Value Decomposition) algorithm. It uses an array of 32 gyroscopes, samples at 2kHz, and performs a fusion process every two samples, including: Step 1: Construct a redundant array using 32 MEMS gyroscopes of model ISM330, with the sensitive axes of all gyroscopes arranged along the X-axis of the inertial coordinate system.

[0033] Step 2: Simultaneously acquire X-axis data from 32 gyroscopes at a frequency of 2kHz. (The processing methods for other Y and Z axes are the same as for the X axis), and a fusion process is performed every two data points collected. The collected data will be used to construct an 8×8 dimensional data matrix A. (Example data is as follows, unit: ° / s) Step 3: SVD fusion processing SVD decomposition: The numerical computation library is used to perform standard SVD decomposition on matrix A, yielding... , The left singular matrix , (Orthogonal columns) ); Singular value vectors , It is a diagonal matrix; Right singular matrix , (Orthogonal columns) ).

[0034] Step 4: Preserve the largest singular value All others are 0, constructing a truncated singular value matrix.

[0035] Step 5: Using the formula Calculate the reconstruction matrix Step 6: Reconstruct the matrix The arithmetic mean of the 64 elements is calculated to obtain a high-precision angular velocity fusion result: It achieves the mapping output of 64-dimensional data to 1-dimensional high-precision data.

[0036] Experimental verification results The X, Y, and Z axes of the four prototypes were calculated using the SVD fusion algorithm of this invention and the traditional arithmetic mean method, respectively. The results are shown in the table below: Table 1 Comparison of Experimental Results Beneficial effects: 1) Strong anti-interference capability: Through the principal component extraction characteristics of the SVD algorithm, it can accurately shield the sinusoidal disturbances and noise anomalies of individual sensors, solving the core pain point of traditional methods being affected by abnormal sensors; 2) Improved accuracy: The zero-bias stability index is more than twice as high as that of the traditional averaging algorithm, and the noise level is reduced by more than 50%; 3) High engineering feasibility: No need to modify the SVD algorithm kernel, mature numerical computing libraries can be directly called; 4) High degree of standardization: The integration process is fixed at 7 steps, and the parameter configuration is clearly defined (only requiring compliance with...). This facilitates industrial production and mass application.

[0037] The above specific embodiments further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for data fusion of MEMS array gyroscopes based on singular value decomposition, characterized in that, The data fusion method includes: Step S1: MEMS gyroscope array configuration; Step S2: Collect data; Step S3: Construct a fusion matrix and arrange the data according to the principle of sensor number priority and row priority filling; Step S4: Perform standard singular value decomposition on the fusion matrix; Step S5: Perform the first principal component screening; Step S6: Reconstruct the fusion matrix to obtain the reconstructed fusion matrix; Step S7: Average the reconstructed fusion matrix.

2. The MEMS array gyroscope data fusion method based on singular value decomposition according to claim 1, characterized in that, Step S1: MEMS gyroscope array configuration specifically includes: use Only MEMS gyroscope chips of the same model are used to build a redundant array, and the sensitive axes of all gyroscopes are arranged in parallel to achieve multi-source synchronous sampling with consistent frequency.

3. The MEMS array gyroscope data fusion method based on singular value decomposition according to claim 2, characterized in that, Step S2: Data collection specifically includes: Each collection Angular velocity data from each gyroscope sensor; each gyroscope sensor collects... Each point is fused once, where , are positive integers, and satisfy... , It is a positive integer.

4. The MEMS array gyroscope data fusion method based on singular value decomposition according to claim 1, characterized in that, Step S3: Constructing a fusion matrix, arranging the data according to the sensor number priority and row priority principle, specifically includes: Data is arranged according to sensor number priority and row priority filling principle: Extract the first sensor sequentially Second sampling data, from the second sensor Sub-sampled data...until all All sampling data from each sensor is retrieved, and the collected data is sequentially filled into the matrix according to the order of "filling the first row of the matrix first, then the second row, and so on..." OK The data matrix in the column forms the fusion matrix. .

5. The MEMS array gyroscope data fusion method based on singular value decomposition according to claim 1, characterized in that, Step S4: Performing standard singular value decomposition on the fusion matrix specifically includes: The numerical linear algebra library is used to perform standard singular value decomposition on the fusion matrix A. The decomposition is as follows: in, Let be a left singular vector matrix, satisfying Orthogonality of columns; For a diagonal singular value matrix, the diagonal elements ,and ; Let be a right singular vector matrix, satisfying Orthogonality of columns This is a transpose.

6. The MEMS array gyroscope data fusion method based on singular value decomposition according to claim 1, characterized in that, Step S5: Performing the first principal component screening specifically includes: Preserve the maximum singular value Construct the truncated singular value matrix in the corresponding principal component subspace. 。 7. The MEMS array gyroscope data fusion method based on singular value decomposition according to claim 1, characterized in that, Step S6: Reconstructing the fusion matrix to obtain the reconstructed fusion matrix specifically includes: The formula for calculating the reconstructed matrix based on the truncated singular value matrix is ​​as follows: Reconstructed Noise and abnormal data interference have been filtered out, while retaining the core common characteristics and timing correlation information of each gyroscope channel.

8. The MEMS array gyroscope data fusion method based on singular value decomposition according to claim 1, characterized in that, Step S7: Calculating the average of the reconstructed fusion matrix specifically includes: The reconstructed matrix Calculate the arithmetic mean of all elements; Output fused angular velocity value ， accomplish Mapping 3D sensor data to 1D high-precision angular velocity.

9. A MEMS array gyroscope data fusion system based on singular value decomposition, employing the MEMS array gyroscope data fusion method based on singular value decomposition as described in any one of claims 1-8, characterized in that, The data fusion system includes: Redundant gyroscope array module for MEMS gyroscope array configuration; The synchronous acquisition module is used for data acquisition; The matrix construction module is used to construct a fusion matrix, arranging the data according to the principle of sensor number priority and row priority. The SVD decomposition module is used to perform standard singular value decomposition on the fusion matrix; The principal component screening module is used to perform the first principal component screening; The reconstruction module is used to reconstruct the fusion matrix and obtain the reconstructed fusion matrix. The fusion output module is used to average the reconstructed fusion matrix.

Images