Calibration system of motion reference unit and error correction method

Through the calibration system and error correction method of the motion reference unit, six-degrees of freedom motion and multi-source data fusion are used, combined with support vector classification model and Bayesian optimization, the problem of drift and deviation of the output data of the motion reference unit is solved, and the measurement is improved Accuracy and stability.

CN118896629APending Publication Date: 2024-11-05GUANGDONG UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410950380.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-05

AI Technical Summary

Technical Problem

During long-term use of the motion reference unit, due to inherent errors and environmental factors, the output data may drift and deviation, resulting in serious challenges in high-precision application scenarios.

Method used

A calibration system using a motion reference unit, including a motion platform, a motion reference unit, a dynamic capture device and calibration and error correction module, calibrating and error correction is performed through six-degree of freedom motion and multi-source data fusion, using the support vector classification model and Bayesian optimization to identify and correct errors.

Benefits of technology

A comprehensive calibration and error correction of the motion reference unit is achieved, improving its measurement accuracy and stability in various application environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118896629A_ABST
    Figure CN118896629A_ABST
Patent Text Reader

Abstract

The invention discloses a calibration system of a motion reference unit and an error correction method, and the system comprises a motion platform which is used for carrying out six-degree-of-freedom motion; the motion reference unit is fixed on the motion platform and is used for generating six-degree-of-freedom motion information according to input excitation provided by the motion platform; the dynamic capture device is fixed above the motion platform and is used for capturing motion information of the motion reference unit; and the calibration and correction module is used for processing the output data of the dynamic capture device and the output data of the motion reference unit so as to carry out calibration and error correction on the motion reference unit. According to the calibration system of the motion reference unit and the error correction method, six-degree-of-freedom output of the motion reference unit can be comprehensively calibrated, and error correction is carried out on the output result, so that the measurement precision and stability of the motion reference unit in various application environments are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of motion reference unit calibration, and in particular to a motion reference unit calibration system and an error correction method. Background Art

[0002] The Inertial Measurement Unit (IMU) is an important sensor device that is widely used in aerospace, robotics, heave compensation, motion tracking and other fields. IMU can provide six degrees of freedom (6DOF) motion information (displacement, velocity, acceleration, etc.), including heave, lateral movement, longitudinal movement, roll, pitch, and pitch, so as to accurately measure the position, attitude and motion state of the object. However, due to the inherent errors of the sensor itself and the influence of environmental factors, the output data of the IMU may drift and deviate in long-term use, which poses a serious challenge to high-precision application scenarios. Summary of the invention

[0003] The purpose of the present invention is to provide a motion reference unit calibration system and an error correction method to solve one or more technical problems existing in the above-mentioned background technology.

[0004] To achieve this object, the present invention adopts the following technical solutions:

[0005] A motion reference unit calibration system, comprising:

[0006] A motion platform for making six degrees of freedom movements;

[0007] A motion reference unit, fixed on the motion platform, is used to generate six-degree-of-freedom motion information according to input excitation provided by the motion platform;

[0008] A dynamic capture device, fixed above the motion platform, is used to capture the motion information of the motion reference unit;

[0009] The calibration and error correction module is used to process the output data of the motion capture device and the output data of the motion reference unit to calibrate and correct the errors of the motion reference unit.

[0010] Preferably, the motion capture device comprises a mounting frame and a plurality of cameras, the mounting frame is fixed above the motion platform, and the plurality of cameras are evenly arranged on the mounting frame.

[0011] Preferably, the motion capture device further comprises three reflective balls, and the three reflective balls are pasted on three different planes of the motion reference unit.

[0012] The present invention also provides a motion reference unit error correction method, the method comprising the following steps:

[0013] S100: running the calibration system to calibrate the motion reference unit;

[0014] S200: The platform is made to perform six-degree-of-freedom motion, and the six-degree-of-freedom outputs of the motion reference unit and the dynamic capture device are collected through the calibration system, and the motion information of the motion platform measured by the dynamic capture device is used as the reference true value to obtain the measurement error of the motion reference unit as the total training data set, and the collection time is set;

[0015] S300: Perform error analysis on the total training data set, divide the error into six error characteristic intervals, and establish SVC_model_1, SVC_model_2, SVC_model_3, SVC_model_4, and SVC_model_5 models, where model_1 distinguishes (min, 0) and (0, max), model_2 distinguishes [min / 4, 0) from other intervals, model_3 distinguishes [min / 2, min / 4) from other intervals, model_4 distinguishes (0, max / 4] from other intervals, and model_5 distinguishes (max / 4, max / 2] from other intervals;

[0016] S400: Use the collected data to complete model training, complete the training of models SVC_model_1, SVC_model_2, SVC_model_3, SVC_model_4, SVC_model_5, and obtain models model_1, model_2, model_3, model_4, and model_5 for error correction;

[0017] S500: Use the trained model to correct the error of the motion reference unit.

[0018] Preferably, step S400 specifically includes the following steps:

[0019] S410: Establish a support vector classification model SVC_model_1 to be trained, set the number of iterations and the accuracy of exiting the iteration;

[0020] S420: Input the training data set and labels collected by the calibration system into the SVC_model_1 model to be trained;

[0021] S430: Set the initial value range of c and g parameter values, use Bayesian optimization to select the best c and g values ​​through the probability distribution of the accuracy of the SVC_model_1 model output, and input them into the SVC_model model to be trained;

[0022] S440: Determine whether the set number of iterations or accuracy is reached;

[0023] If the set number of iterations or accuracy is not reached, the highest accuracy and its corresponding c and g parameter values ​​are recorded, and steps S420-S430 are repeated;

[0024] If the set number of iterations or accuracy has been reached, the highest accuracy recorded in the iteration process and its corresponding c and g parameter values ​​are output to complete the training of the SVC_model_1 model and obtain the model model_1 for error correction;

[0025] Repeat steps S410-S440 to obtain models model_2, model_3, model_4, and model_5.

[0026] Preferably, S500 specifically includes the following steps:

[0027] S510: using the acceleration, velocity and displacement information output by the motion reference unit as the input of the model, and inputting it into the model_1 model;

[0028] S520: model_1 model identifies whether the error range corresponding to the group of data is within (min, 0);

[0029] If yes, the data will be passed to model_2 and model_3. The two models will identify whether the data meets their own error characteristics. If yes, they will output 1, if not, they will output -1. The output results of the two models will be output to logic processor 1.

[0030] If not, the data will be passed to model_4 and model_5. The two models will identify whether the data meets their own error characteristics. If yes, they will output 1, if not, they will output -1, and the output results of the two models will be output to logic processor 2.

[0031] S530: After processing the received result, the logic processor 1 or the logic processor 2 outputs a corresponding correction value.

[0032] Preferably, step S530 includes:

[0033] If the results output to the logic processor 1 are 1 and -1 respectively, the logic processor 1 outputs the first correction value;

[0034] If the results output to the logic processor 1 are -1 and 1 respectively, the logic processor 1 outputs the second correction value

[0035] If the results output to the logic processor 1 are all -1, the logic processor 1 outputs a third correction value;

[0036] If the results output to the logic processor 1 are all 1, then the correction is made according to the correction value of the previous data, and the corresponding value before the correction is output;

[0037] If the results output to the logic processor 2 are 1 and -1 respectively, the logic processor 2 outputs a fourth correction value;

[0038] If the results output to the logic processor 2 are respectively -1 and 1, the logic processor 2 outputs a fifth correction value;

[0039] If the results output to the logic processor 2 are all -1, the logic processor 2 outputs the sixth correction value;

[0040] If the results output to the logic processor 2 are all 1, they are corrected according to the correction value of the previous data, and the value before correction is output accordingly.

[0041] Preferably, in step S530, the first correction value is min / 8, the second correction value is 3*min / 8, the third correction value is 6*min / 8, the fourth correction value is max / 8, the fifth correction value is 3*max / 8, and the sixth correction value is 6*max / 8.

[0042] Preferably, S100 specifically includes the following steps:

[0043] S110: Fix the motion reference unit on the motion platform, and stick three reflective balls on three different surfaces of the motion reference unit respectively;

[0044] S120: Start the motion platform, and place the motion platform in the middle of the travel through the control software, and wait for 5 minutes;

[0045] S130: Convert the origin of the coordinate system of the dynamic capture device to the initial coordinates of the motion reference unit through calibration software;

[0046] S140: The motion platform is made to move in six degrees of freedom, and calibrated by combining the output data of the motion capture device and the output data of the motion reference unit itself.

[0047] The beneficial effects of the present invention are as follows: the calibration system and error correction method of the motion reference unit of the present invention can comprehensively calibrate the six degrees of freedom output of the motion reference unit and perform error correction on its output results, thereby improving its measurement accuracy and stability in various application environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings further illustrate the present invention, but the contents in the accompanying drawings do not constitute any limitation to the present invention.

[0049] Figure 1 is a schematic structural diagram of a calibration system according to one embodiment of the present invention;

[0050] Figure 2 is an overall flow chart of an error correction method according to one embodiment of the present invention;

[0051] Figure 3 is a model training flow chart of one embodiment of the present invention;

[0052] Figure 4 FIG. 4 is a flowchart of an error correction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.

[0054] Example 1

[0055] A motion reference unit calibration system of this embodiment is shown in the attached Figure 1 include:

[0056] A motion platform 1, used for making motions with six degrees of freedom;

[0057] The motion reference unit 2 is fixed on the motion platform 1 and is used to generate six-degree-of-freedom motion information according to the input excitation provided by the motion platform 1;

[0058] A dynamic capture device 3, fixed above the motion platform 1, is used to capture the motion information of the motion reference unit 2;

[0059] The calibration module is used to process the output data of the motion capture device 3 and the output data of the motion reference unit 2 to calibrate the motion reference unit 2.

[0060] By combining the six degrees of freedom provided by the motion platform, the motion information generated by the motion reference unit, and the captured data of the dynamic capture device, the system can achieve high-precision calibration of the motion reference unit. This multi-source data fusion calibration method helps to improve the accuracy and reliability of the calibration. The use of a motion platform that can make six degrees of freedom allows the system to fully cover all possible motion states, can truly reflect the performance of the motion reference unit in actual use, and has higher calibration accuracy.

[0061] Preferably, the motion capture device 3 includes a mounting frame and a plurality of cameras. The mounting frame is fixed above the motion platform 1, and the plurality of cameras are evenly arranged on the mounting frame.

[0062] Preferably, the motion capture device 3 further includes three reflective balls, which are pasted on three different planes of the motion reference unit 2 .

[0063] By sticking a reflective ball on the motion reference unit, it is ensured that the dynamic capture device can capture its motion information, thereby improving the capture effect in high-speed or complex motion states.

[0064] Example 2

[0065] A motion reference unit error correction method of this embodiment is shown in the attached Figure 2 , the method comprises the following steps:

[0066] S100: running the calibration system to calibrate the motion reference unit;

[0067] S200: The platform is made to perform six-degree-of-freedom motion, and the six-degree-of-freedom outputs of the motion reference unit and the dynamic capture device are collected through the calibration system, and the motion information of the motion platform measured by the dynamic capture device is used as the reference true value to obtain the measurement error of the motion reference unit as the total training data set, and the collection time is set;

[0068] S300: Perform error analysis on the total training data set, divide the error into six error feature intervals, and establish SVC_model_1, SVC_model_2, SVC_model_3, SVC_model_4, and SVC_model_5 models, where model_1 distinguishes between (min, 0) and (0, max), model_2 distinguishes between [min / 4, 0) and other intervals, model_3 distinguishes between [min / 2, min / 4) and other intervals, model_4 distinguishes between (0, max / 4] and other intervals, and model_5 distinguishes between (max / 4, max / 2] and other intervals; the above 'min' and 'max' are the minimum and maximum error values ​​in the training data, respectively.

[0069] S400: Use the collected data to complete model training and obtain model_1, model_2, model_3, model_4 and model_5 models;

[0070] S500: Use the trained model to correct the error of the motion reference unit.

[0071] Therefore, the motion reference unit obtained by the calibration system and the reference value of the dynamic capture device are used to train the SVC model, and the error correction is performed on its output results, thereby improving its measurement accuracy and stability in various application environments.

[0072] Preferably, refer to the attached Figure 3 Step S400 specifically includes the following steps:

[0073] S410: Establish a support vector classification model SVC_model_1 to be trained, set the number of iterations and the accuracy of exiting the iteration;

[0074] S420: Input the training data set and labels collected by the calibration system into the SVC_model_1 model to be trained;

[0075] S430: setting the initial value range of the c and g parameter values, using Bayesian optimization to select the best c and g values ​​through the probability distribution of the SVC_model_1 model output accuracy, and inputting them into the SVC_model_1 model to be trained;

[0076] S440: Determine whether the set number of iterations or accuracy is reached;

[0077] If the set number of iterations or accuracy is not reached, the highest accuracy and its corresponding c and g parameter values ​​are recorded, and steps S420-S430 are repeated;

[0078] If the set number of iterations has been reached, the highest accuracy recorded in the iteration process and its corresponding c and g parameter values ​​are output to complete the training of the SVC_model_1 model and obtain the model_1 model;

[0079] Repeat steps S410-S440 to obtain models model_2, model_3, model_4 and model_5.

[0080] Thus, the training data set is collected by the calibration system. SVC_model is a mathematical model to be trained. Training the model requires inputting features, labels, c, and g. Among them, c and g are parameters that need to be optimized using Bayesian optimization. Given their initial value ranges, the initial value range is used to find the optimal range through experiments. In this embodiment, c∈(0, 1000), g∈(0, 10240), and then Bayesian optimization is used to select the optimal c and g values ​​through the probability distribution of the accuracy output by SVC_model. The selection basis is the improvement of the accuracy rate, and the highest accuracy rate and its corresponding c and g parameter values ​​are output through multiple iterations. Models model_1, model_2, model_3, model_4 and model_5 are obtained for error correction.

[0081] Preferably, refer to the attached Figure 4 , S500 specifically includes the following steps:

[0082] S510: using the acceleration, velocity and displacement information output by the motion reference unit as the input of the model, and inputting it into the model_1 model;

[0083] S520: model_1 model identifies whether the error range corresponding to the group of data is within (min, 0];

[0084] If yes, the data will be passed to model_2 and model_3. The two models will identify whether the data meets their own error characteristics. If yes, they will output 1, if not, they will output -1. The output results of the two models will be output to logic processor 1.

[0085] If not, the data will be passed to model_4 and model_5. The two models will identify whether the data meets their own error characteristics. If yes, they will output 1, if not, they will output -1, and the output results of the two models will be output to logic processor 2.

[0086] S530: After processing the received result, the logic processor 1 or the logic processor 2 outputs a corresponding correction value.

[0087] Further, step S530 includes:

[0088] If the results output to the logic processor 1 are 1 and -1 respectively, the logic processor 1 outputs the first correction value;

[0089] If the results output to the logic processor 1 are -1 and 1 respectively, the logic processor 1 outputs the second correction value

[0090] If the results output to the logic processor 1 are all -1, the logic processor 1 outputs a third correction value;

[0091] If the results output to the logic processor 1 are all 1, then the correction is made according to the correction value of the previous data, and the corresponding value before the correction is output;

[0092] If the results output to the logic processor 2 are 1 and -1 respectively, the logic processor 2 outputs a fourth correction value;

[0093] If the results output to the logic processor 2 are respectively -1 and 1, the logic processor 2 outputs a fifth correction value;

[0094] If the results output to the logic processor 2 are all -1, the logic processor 2 outputs the sixth correction value;

[0095] If the results output to the logic processor 2 are all 1, they are corrected according to the correction value of the previous data, and the value before correction is output accordingly.

[0096] Further, in step S530, the first correction value is min / 8, the second correction value is 3*min / 8, the third correction value is 6*min / 8, the fourth correction value is max / 8, the fifth correction value is 3*max / 8, and the sixth correction value is 6*max / 8.

[0097] The acceleration, velocity and displacement information output by the motion reference unit are used as the input of the model and passed to the model_1 model. The model_1 model will identify the error range corresponding to the group of data. Take "yes" as an example; if it is identified as "yes", the group of data will be passed to the model_2 model and the model_3 model. The two models will identify whether the data meets their respective error characteristics and pass the identification results to the logic processor 1, where the results are voted. The voting principle of the logic processor 1 is as follows: if the outputs of the two models are 1 and -1 respectively, then output 1, that is, output min / 8 correction positive; if they are -1 and 1 respectively, then output 2, that is, output 3*min / 8 correction value; if both are -1, then the corresponding output is 3, that is, output 6*min / 8 correction value; if both are 1, then the correction is made according to the correction value of the previous data, corresponding to output 4. The correction method on the other side is similar.

[0098] Therefore, by inputting the output data of the motion reference unit into multiple models in sequence, performing logical processing according to the output results of the models, and finally outputting the corresponding correction value, the error correction is achieved accurately. By making targeted corrections to different error characteristics, the accuracy and effectiveness of the correction are improved.

[0099] Preferably, S100 specifically includes the following steps:

[0100] S110: Fix the motion reference unit on the motion platform, and stick three reflective balls on three different surfaces of the motion reference unit respectively;

[0101] S120: Start the motion platform, and place the motion platform in the middle of the travel through the control software, and wait for 5 minutes; ensure that the motion platform reaches a stable state before the calibration begins, reducing the error caused by the instability of the initial position of the platform;

[0102] S130: Convert the origin of the coordinate system of the motion capture device to the initial coordinates of the motion reference unit through calibration software; thereby ensuring the consistency of the coordinate systems between the motion capture device and the motion reference unit, providing an accurate basis for subsequent data fusion and calibration.

[0103] S140: The motion platform is made to move in six degrees of freedom, and the output data of the motion capture device and the output data of the motion reference unit itself are combined for calibration and error correction.

[0104] The technical principle of the present invention is described above in conjunction with specific embodiments. These descriptions are only for explaining the principle of the present invention and cannot be interpreted as limiting the scope of protection of the present invention in any way. Based on the explanations herein, those skilled in the art can associate other specific implementations of the present invention without paying creative labor, and these methods will fall within the scope of protection of the present invention.

Claims

1. A calibration system for a motion reference unit, It is characterized in that include: A motion platform for making six degrees of freedom movements; A motion reference unit, fixed on the motion platform, is used to generate six-degree-of-freedom motion information according to input excitation provided by the motion platform; A dynamic capture device, fixed above the motion platform, is used to capture the motion information of the motion reference unit; The calibration and correction module is used to process the output data of the motion capture device and the output data of the motion reference unit to calibrate and correct the error of the motion reference unit.

2. A motion reference unit calibration system according to claim 1, It is characterized in that The dynamic capture device comprises a mounting frame and a plurality of cameras. The mounting frame is fixed above the motion platform, and the plurality of cameras are evenly arranged on the mounting frame.

3. A motion reference unit calibration system according to claim 1, It is characterized in that The dynamic capture device also includes a plurality of reflective balls, which are pasted on three different planes of the motion reference unit.

4. A motion reference unit error correction method, It is characterized in that The method comprises the following steps: S100: running the calibration system to calibrate the motion reference unit; S200: The platform is made to perform six-degree-of-freedom motion, and the six-degree-of-freedom outputs of the motion reference unit and the dynamic capture device are collected through the calibration system, and the motion information of the motion platform measured by the dynamic capture device is used as the reference true value to obtain the measurement error of the motion reference unit as the total training data set, and the collection time is set; S300: Perform error analysis on the total training data set, and divide the error into six error feature intervals, namely (min, 0), (0, max), [min / 4, 0), [min / 2, min / 4), (0, max / 4], (max / 4, max / 2]; establish SVC_model_1, SVC_model_2, SVC_model_3, SVC_model_4, SVC_model_5 models, where SVC_model_1 distinguishes between (min, 0) and (0, max), SVC_model_2 distinguishes between [min / 4, 0) and other intervals, SVC_model_3 distinguishes between [min / 2, min / 4) and other intervals, SVC_model_4 distinguishes between (0, max / 4] and other intervals, and SVC_model_5 distinguishes between (max / 4, max / 2] and other intervals; S400: Use the collected data to complete model training, complete the training of models SVC_model_1, SVC_model_2, SVC_model_3, SVC_model_4, and SVC_model_5, and obtain models model_1, model_2, model_3, model_4, and model_5 for error correction; S500: Use the trained model to correct the error of the motion reference unit.

5. A motion reference unit error correction method according to claim 4, It is characterized in that Step S400 specifically includes the following steps: S410: Establish a support vector classification model SVC_model_1 to be trained, set the number of iterations and the accuracy rate for exiting the iteration; S420: Input the training data set and labels collected by the calibration system into the SVC_model_1 model to be trained; S430: setting the initial value range of the c and g parameter values, using Bayesian optimization to select the best c and g values ​​through the probability distribution of the SVC_model_1 model output accuracy, and inputting them into the SVC_model_1 model to be trained; S440: Determine whether the set number of iterations or accuracy is reached; If the set number of iterations or accuracy is not reached, the highest accuracy and its corresponding c and g parameter values ​​are recorded, and steps S420-S430 are repeated; If the set number of iterations or accuracy has been reached, the highest accuracy recorded in the iteration process and its corresponding c and g parameter values ​​are output to complete the training of the SVC_model_1 model and obtain the model_1 for error correction; Repeat steps S410-S440 to complete the training of SVC_model_2, SVC_model_3, SVC_model_4, and SVC_model_5 models to obtain models model_2, model_3, model_4, and model_5 for error correction.

6. A motion reference unit error correction method according to claim 4, It is characterized in that S500 specifically includes the following steps: S510: using the acceleration, velocity and displacement information output by the motion reference unit as the input of the model, and inputting it into the model_1 model; S520: model_1 model identifies whether the error range corresponding to the group of data is within (min, 0); If yes, the data will be passed to model_2 and model_3. The two models will identify whether the data meets their own error characteristics. If yes, they will output 1, if not, they will output -1. The output results of the two models will be output to logic processor 1. If not, the data will be passed to model_4 and model_5. The two models will identify whether the data meets their own error characteristics. If yes, they will output 1, if not, they will output -1, and the output results of the two models will be output to logic processor 2. S530: After processing the received result, the logical processor 1 or the logical processor 2 outputs a corresponding correction value.

7. A motion reference unit error correction method according to claim 6, It is characterized in that Step S530 includes: If the results output to the logic processor 1 are 1 and -1 respectively, the logic processor 1 outputs the first correction value; If the results output to the logic processor 1 are -1 and 1 respectively, the logic processor 1 outputs the second correction value If the results output to the logic processor 1 are all -1, the logic processor 1 outputs the third correction value; If the results output to the logic processor 1 are all 1, then the correction is made according to the correction value of the previous data, and the corresponding value before the correction is output; If the results output to the logic processor 2 are 1 and -1 respectively, the logic processor 2 outputs a fourth correction value; If the results output to the logic processor 2 are respectively -1 and 1, the logic processor 2 outputs a fifth correction value; If the results output to the logic processor 2 are all -1, the logic processor 2 outputs the sixth correction value; If the results output to the logic processor 2 are all 1, they are corrected according to the correction value of the previous data, and the value before correction is output accordingly.

8. A motion reference unit error correction method according to claim 7, It is characterized in that In step S530, the first correction value is min / 8, the second correction value is 3*min / 8, the third correction value is 6*min / 8, the fourth correction value is max / 8, the fifth correction value is 3*max / 8, and the sixth correction value is 6*max / 8.

9. A motion reference unit error correction method according to claim 4, It is characterized in that S100 specifically includes the following steps: S110: Fix the motion reference unit on the motion platform, and stick three reflective balls on three different surfaces of the motion reference unit respectively; S120: Start the motion platform, and place the motion platform in the middle of the travel through the control software, and wait for 5 minutes; S130: Convert the origin of the coordinate system of the dynamic capture device to the initial coordinates of the motion reference unit through calibration software; S140: The motion platform is made to move in six degrees of freedom, and calibrated by combining the output data of the motion capture device and the output data of the motion reference unit itself.