A high-precision inertial navigation initial alignment method based on virtual IMU prediction

Abstract

The application provides a high-precision inertial navigation initial alignment method based on virtual IMU prediction, comprising: performing navigation alignment of multiple positions based on a double-axis rotating inertial navigation system, acquiring alignment reference data, IMU data and code disc rotation angle; acquiring virtual IMU gyro angular velocity and acceleration according to the alignment reference data; pre-processing the IMU data to acquire gyro angular rate increment and accelerometer acceleration increment; setting model training input and output; performing model training by using a deep learning LSTM model according to the model training input and output, acquiring a virtual IMU prediction training model; acquiring real-time predicted virtual IMU gyro angular velocity and acceleration based on the virtual IMU prediction training model, according to the gyro angular rate increment and accelerometer acceleration increment calculated according to real-time IMU data, and real-time code disc rotation angle; and performing navigation alignment according to the real-time predicted virtual IMU gyro angular velocity and acceleration. The application can solve the technical problem of insufficient inertial navigation initial alignment precision in the prior art.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of inertial navigation systems, in particular to a high-precision inertial navigation initial alignment method based on virtual IMU prediction. BACKGROUND

[0002] Inertial navigation technology is constantly applied to various weapon equipment on sea, land and air due to its good concealment and strong autonomy. The alignment research on the platform inertial navigation system is quite mature. Usually, Kalman filtering is used to solve the initial alignment problem. However, due to the poor observability of the inertial navigation system, the convergence speed and estimation accuracy of the state estimation of the linear filter are affected, and the alignment accuracy and rapidity are affected, and there is a problem of mutual contradiction between the alignment accuracy and rapidity, so it is still a very valuable problem to study how to improve the speed and accuracy of the initial alignment of the inertial navigation system.

[0003] The double-axis rotating inertial navigation system can modulate out the zero position error of the gyroscope and the accelerometer through double-axis rotation, and improve the alignment and navigation accuracy. However, since the IMU needs to be rotated, the coordinate conversion calculation is needed to obtain the final navigation system attitude result, so during the IMU rotation process, the code disc error will also cause the attitude conversion matrix to have an error, and finally introduce the alignment attitude angle of the navigation system. Therefore, in order to further improve the alignment accuracy of the high-precision inertial navigation system, an improved method needs to be proposed on the basis of the original rotation modulation alignment to meet the current demand of long-range strategic weapon equipment. SUMMARY

[0004] The present application aims at at least solving one of the problems in the prior art.

[0005] The application provides a high-precision inertial navigation initial alignment method based on virtual IMU prediction, which comprises: performing navigation alignment at multiple positions based on a double-axis rotating inertial navigation system, obtaining alignment reference data, IMU data and code disc rotation angle; obtaining virtual IMU gyro angular velocity and acceleration according to the alignment reference data; pre-processing the IMU data to obtain gyro angular rate increment and accelerometer acceleration increment; setting model training input and model training output according to the virtual IMU gyro angular velocity and acceleration, the gyro angular rate increment and the accelerometer acceleration increment, and the code disc rotation angle; performing model training by using a deep learning LSTM model according to the model training input and the model training output, to obtain a virtual IMU prediction training model; obtaining real-time predicted virtual IMU gyro angular velocity and acceleration based on the virtual IMU prediction training model, according to the gyro angular rate increment and the accelerometer acceleration increment calculated from real-time IMU data, and the real-time code disc rotation angle; and performing navigation alignment according to the real-time predicted virtual IMU gyro angular velocity and acceleration.

[0006] Further, according to The virtual IMU gyro angular velocity is obtained, wherein, The virtual IMU gyro angular velocity is obtained, wherein, is the projection of the angle change of the carrier coordinate system b relative to the navigation coordinate system n in the b system, Y is the IMU sampling period, γ, θ, ψ are the reference roll angle, pitch angle, and heading angle, respectively; is the projection of the earth rotation angular velocity in the navigation system n; L is the latitude, ω ie is the earth rotation angular velocity; is the eastward projection of the velocity of the navigation coordinate system n relative to the earth coordinate system e, is the northward projection of the velocity of the navigation coordinate system n relative to the earth coordinate system e, R n is the meridian curvature radius, H is the navigation height, R m is the meridian curvature radius.

[0007] Further, according to The virtual IMU acceleration is obtained, wherein, The virtual IMU acceleration is obtained, wherein, is the projection of the velocity change of the carrier system b relative to the inertial system i in the n system, is the gravitational acceleration, is the velocity of the navigation coordinate system n relative to the earth coordinate system e,

[0008] Further, according to The gyro angular rate increment is obtained, wherein, is the gyro measured angular rate increment in the ΔT time period, ΔT is the data preprocessing interval time period, is the gyro measured angular velocity, T is the IMU sampling period.

[0009] Further, according to The accelerometer acceleration increment is obtained, wherein, is the accelerometer acceleration increment in the ΔT time period, ΔT is the data preprocessing interval time period, is the accelerometer measured specific force, T is the IMU sampling period.

[0010] Further, the model training input is set as wherein Z(·) represents data normalization processing using z-score, and satisfies μ represents the mean of the original data x, and σ represents the standard deviation of the original data x; is the gyro measured angular rate increment components along IMU x-axis, y-axis, z-axis, α is the inner ring angle of the code disc, and β is the outer ring angle of the code disc. respectively, and Δa is the acceleration increment of the accelerometer. components along IMU x-axis, y-axis, z-axis, α is the inner ring angle of the code disc, and β is the outer ring angle of the code disc.

[0011] Further, the model training output is set as wherein, and respectively, and Δa is the acceleration increment of the accelerometer.

[0012] Further, the loss function of the virtual IMU prediction training model is wherein, Y OUT is the model output, Y pred is the prediction value in the model training process, and a is the number of all data pieces for training.

[0013] Further, after the gyro angular rate increment and the accelerometer acceleration increment obtained by real-time calculation and the real-time code disc rotation angle are normalized, the virtual IMU prediction training model is input to perform real-time prediction output of the virtual IMU gyro angular velocity and acceleration.

[0014] The technical scheme of the present application provides a high-precision inertial navigation initial alignment method based on virtual IMU prediction. According to the alignment reference data, the virtual IMU gyro angular velocity and acceleration are obtained. In combination with the IMU data and the code disc rotation angle, a deep learning LSTM model (Long short-term memory, long short-term memory network model) is used for model training to obtain a virtual IMU prediction training model. According to the gyro angular rate increment and the accelerometer acceleration increment obtained by real-time IMU data calculation and the real-time code disc rotation angle, the real-time predicted virtual IMU gyro angular velocity and acceleration are obtained, and navigation alignment is performed. The inertial navigation initial alignment method of the present application not only ensures the effect of rotation modulation for eliminating zero error, but also can estimate and compensate the code disc error, so as to achieve the purpose of fast and high-precision alignment. Compared with the prior art, the technical scheme of the present application can solve the technical problem of insufficient initial alignment precision of inertial navigation in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0015] The accompanying drawings included to provide a further understanding of the embodiments of the present application and constitute a part of the specification, illustrate embodiments of the present application and together with the text description serve to explain the principles of the present application. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0016] Figure 1 A flowchart of a high-precision inertial navigation initial alignment method based on virtual IMU prediction according to a specific embodiment of the present application is shown. DETAILED DESCRIPTION

[0017] It should be noted that the embodiments and features of the embodiments in the present application can be combined with each other without conflict. The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. The description of the at least one exemplary embodiment is actually only illustrative, but not intended to limit the present application and its application or use in any way. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0018] It should be noted that the terms used herein are only intended to describe specific embodiments, and are not intended to limit exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and it should also be understood that when the terms "comprise" and / or "include" are used in the specification, there is a presence of a feature, step, operation, device, component and / or combinations thereof.

[0019] Unless specifically stated otherwise, the relative arrangement of the components and steps illustrated in these embodiments and the numerical expressions and values set forth herein are not limiting of the scope of the present application. It should be understood that the various parts of the drawings are not necessarily drawn to scale, and that, for the purpose of convenience and clarity, not all components can be shown in a given figure. Techniques, methods, and devices known to those of ordinary skill can not be discussed in detail, but rather can be summarily described in order not to unnecessarily obscure aspects of the present application. In all examples shown and discussed herein, any specific values should be interpreted as merely illustrative of the examples, and not as a limitation thereon. Thus, other examples of the exemplary embodiments can have different values. It is noted that like references and labels can be used to denote like items throughout the drawings, and thus, once any part is defined in one drawing, it should not be necessary to further discuss it in subsequent drawings.

[0020] As Figure 1 According to a specific embodiment of the present application, a high-precision inertial navigation initial alignment method based on virtual IMU prediction is provided, which comprises:

[0021] The dual-axis rotary inertial navigation system is used for navigation alignment at multiple positions to obtain alignment reference data, IMU data and code disc rotation angle; virtual IMU gyro angular velocity and acceleration are obtained according to the alignment reference data;

[0022] The IMU data is preprocessed to obtain gyro angular rate increment and accelerometer acceleration increment;

[0023] The virtual IMU gyro angular velocity and acceleration, the gyro angular rate increment and the accelerometer acceleration increment, and the code disc rotation angle are used to set model training input and model training output;

[0024] The deep learning LSTM model is used to train the model according to the model training input and the model training output to obtain a virtual IMU prediction training model;

[0025] Based on the virtual IMU prediction training model, the real-time predicted virtual IMU gyro angular velocity and acceleration are obtained according to the gyro angular rate increment and the accelerometer acceleration increment calculated from the real-time IMU data and the real-time code disc rotation angle; and the navigation alignment is performed according to the real-time predicted virtual IMU gyro angular velocity and acceleration.

[0026] By using the configuration mode, a high-precision inertial navigation initial alignment method based on virtual IMU prediction is provided. The high-precision inertial navigation initial alignment method based on virtual IMU prediction obtains virtual IMU gyro angular velocity and acceleration according to alignment reference data, combines IMU data and code disc rotation angle, uses a deep learning LSTM model to train a model to obtain a virtual IMU prediction training model, obtains real-time predicted virtual IMU gyro angular velocity and acceleration according to gyro angular rate increment and accelerometer acceleration increment calculated from real-time IMU data and real-time code disc rotation angle, and performs navigation alignment. The inertial navigation initial alignment method of the present application not only ensures the effect of eliminating zero error by rotary modulation, but also estimates and compensates the code disc error, so as to achieve the purpose of fast and high-precision alignment.

[0027] Further, in the present application, in order to realize high-precision inertial navigation initial alignment based on virtual IMU prediction, first, the dual-axis rotary inertial navigation system is used for navigation alignment at multiple positions to obtain alignment reference data, IMU data and code disc rotation angle; and virtual IMU gyro angular velocity and acceleration are obtained according to the alignment reference data.

[0028] As a specific embodiment of the present application, the virtual IMU gyro angular velocity and acceleration values can be calculated reversely based on the reference attitude result. Specifically, the virtual IMU gyro angular velocity is obtained according to The virtual IMU gyro angular velocity is obtained, wherein The virtual IMU gyro angular velocity is obtained, wherein is the projection of the angle change of the carrier coordinate system b relative to the navigation coordinate system n in the b system, T is the IMU sampling period, γ, θ, and ψ are the reference roll angle, pitch angle, and heading angle, respectively; is the projection of the Earth's rotation angular velocity in the navigation system n; L is latitude, ω ie is the Earth's rotation angular rate; is the eastward projection of the velocity of the navigation coordinate system n relative to the earth coordinate system e, is the projection of the velocity of the navigation coordinate system n relative to the earth coordinate system e along the north direction, R n is the radius of curvature of the y-axis circle, H is the navigation altitude, R m is the meridian curvature radius.

[0029] according to Get the virtual IMU acceleration, where is the virtual IMU acceleration, is the projection of the velocity change of the carrier system b relative to the inertial system i in system n, is the acceleration due to gravity, is the velocity of the navigation coordinate system n relative to the earth coordinate system e,

[0030] Furthermore, in the present invention, after obtaining the virtual IMU gyro angular velocity and acceleration, the IMU data is preprocessed to obtain the gyro angular rate increment and the added acceleration increment.

[0031] As a specific embodiment of the present invention, and Get the gyro angular rate increment and the additive acceleration increment, where: is the gyro measurement angular rate increment during the ΔT period, is the acceleration increment within the ΔT time period, ΔT is the data preprocessing interval period, Measure the angular velocity of the gyroscope, To add a table to measure specific force, T is the IMU sampling period. When t = n × ΔT, the interval period needs to restart the data and perform IMU incremental initialization.

[0032] Furthermore, in the present invention, after obtaining the gyro angular rate increment and the table acceleration increment, the model training input and model training output are set according to the virtual IMU gyro angular rate and acceleration, the gyro angular rate increment and the table acceleration increment, and the code disk rotation angle.

[0033] As a specific embodiment of the present application, the gyro angular rate increment and the accelerometer acceleration increment, and the code disc rotation angle are normalized respectively, and the normalized data are used as the model training input, and the virtual IMU gyro angular velocity and acceleration are used as the model output.

[0034] Specifically, the model training input can be set as wherein, X IN-z is the model training input, Z(·) represents data normalization processing by using z-score, and satisfies μ represents the mean of the original data x, and σ represents the standard deviation of the original data x. is the gyro measured angular rate increment along the IMU x-axis, y-axis and z-axis, is the accelerometer acceleration increment along the IMU x-axis, y-axis and z-axis, and α is the inner ring angle of the code disc, and β is the outer ring angle of the code disc.

[0035] The model training output is set as wherein, and are the virtual IMU gyro angular velocity and acceleration values respectively.

[0036] Further, in the present application, after the model training input and output are set, the deep learning LSTM model is used for model training according to the model training input and output, and the virtual IMU prediction training model is obtained.

[0037] As a specific embodiment of the present application, the review value of the virtual IMU prediction training model is set as l, and the loss function is wherein, Y pred is the predicted value in the model training process, and a is the number of all data pieces for training.

[0038] The square loss function is used in this embodiment, in addition, the training number of the model training is set as epoch, and after the training number is satisfied, the virtual IMU prediction training model F(p1, p2, p3…ps) can be obtained, wherein p1, p2, p3…ps are the parameters obtained by training, and s is an integer.

[0039] Further, in the present application, after the virtual IMU prediction training model is obtained, the real-time predicted virtual IMU gyro angular velocity and acceleration are obtained based on the virtual IMU prediction training model according to the gyro angular rate increment and the accelerometer acceleration increment calculated based on the real-time IMU data, and the real-time code disc rotation angle; and the navigation alignment is performed according to the real-time predicted virtual IMU gyro angular velocity and acceleration.

[0040] As a specific embodiment of the present application, the trained virtual IMU prediction training model F(p1, p2, p3...ps) is introduced into the alignment application software for application. In the alignment process, the gyro angular rate increment and the accelerometer acceleration increment are obtained according to the real-time IMU data, and the gyro angular rate increment and the accelerometer acceleration increment obtained by real-time calculation and the real-time code disc rotation angle are normalized and used as model input. The prediction period is set to be ΔT, when the time is greater than or equal to l×ΔT, the real-time prediction is carried out according to the virtual IMU prediction training model, and the real-time prediction output wherein, and are the virtual IMU gyro angular velocity and acceleration of the real-time prediction output respectively.

[0041] Further, on the basis of coarse alignment, navigation calculation and Kalman filter alignment are carried out according to the real-time predicted virtual IMU gyro angular velocity and acceleration, and the obtained attitude angle can be used as the final alignment result.

[0042] The high-precision inertial navigation initial alignment method based on virtual IMU prediction of the present application can be applied to the field of platform-type two-axis rotation inertial navigation system initial alignment. According to the alignment reference, the virtual IMU data in the rotation process is reversely calculated, and then a deep learning method based on LSTM is adopted to train the virtual IMU prediction model based on the IMU data in the rotation process and the inner and outer ring code disc rotation angle. In the alignment process, the actual IMU and the code disc inner and outer ring angle information are used to predict the virtual IMU information without error, so as to eliminate the measurement errors of the IMU. The real-time alignment based on the predicted virtual IMU information of the present application not only ensures the effect of rotation modulation for eliminating the zero error, but also can estimate and compensate the code disc error, so as to achieve the purpose of fast and high-precision alignment.

[0043] In order to have a further understanding of the present application, the following will be combined with Figure 1 The high-precision inertial navigation initial alignment method based on virtual IMU prediction of the present application will be described in detail.

[0044] As Figure 1 shown, the high-precision inertial navigation initial alignment method based on virtual IMU prediction according to the specific embodiment of the present application comprises the following steps.

[0045] Step one, navigation alignment at multiple positions is carried out based on the two-axis rotation inertial navigation system, and the alignment reference data, the IMU data and the code disc rotation angle are obtained.

[0046] According to the virtual IMU gyro angular velocity is obtained; and according to the virtual IMU acceleration is obtained.

[0047] Step two, according to and The gyro angular rate increment and the accelerometer acceleration increment are obtained.

[0048] Step three, set the model training input as Set the model training output as

[0049] Step four, according to the model training input and output, a deep learning LSTM model is used for model training, the review value of the virtual IMU prediction training model is l, and the loss function is

[0050] Step five, the trained virtual IMU prediction training model F (p1, p2, p3…ps) is imported into the alignment application software for application. In the alignment process, the gyro angular rate increment and the accelerometer acceleration increment are obtained according to the real-time IMU data solving, and the gyro angular rate increment and the accelerometer acceleration increment obtained by real-time solving and the real-time code disc rotation angle are normalized and used as model input. The prediction period is set as ΔT, when the time is greater than or equal to l×ΔT, real-time prediction is carried out according to the virtual IMU prediction training model, and the real-time prediction output is

[0051] Step six, on the basis of coarse alignment, navigation solving and Kalman filtering alignment are carried out according to the real-time predicted virtual IMU gyro angular velocity and acceleration, and the obtained attitude angle can be used as the final alignment result.

[0052] In summary, the application provides a high-precision inertial navigation initial alignment method based on virtual IMU prediction. The high-precision inertial navigation initial alignment method based on virtual IMU prediction obtains virtual IMU gyro angular velocity and acceleration according to alignment reference data, combines IMU data and code disc rotation angle, uses a deep learning LSTM model for model training to obtain a virtual IMU prediction training model, obtains real-time predicted virtual IMU gyro angular velocity and acceleration according to the gyro angular rate increment and the accelerometer acceleration increment obtained by real-time IMU data solving and the real-time code disc rotation angle, and carries out navigation alignment. The inertial navigation initial alignment method of the application not only ensures the effect of rotation modulation for eliminating zero error, but also can estimate and compensate code disc error, so as to achieve the purpose of fast and high-precision alignment.

[0053] The above only describes the preferred embodiments of the application and is not used to limit the application. For those skilled in the art, the application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A high-precision inertial navigation initial alignment method based on virtual IMU prediction, characterized in that: The high-precision inertial navigation initial alignment method based on virtual IMU prediction includes: Perform navigation alignment at multiple locations based on a dual-axis rotating inertial navigation system to obtain alignment reference data, IMU data, and code disk rotation angle; obtain virtual IMU gyro angular velocity and acceleration based on the alignment reference data; Preprocessing the IMU data to obtain gyro angular rate increment and gimbal acceleration increment; Setting a model training input and a model training output according to the virtual IMU gyro angular velocity and acceleration, the gyro angular rate increment and the added acceleration increment, and the code disk rotation angle; Performing model training using a deep learning LSTM model according to the model training input and model training output to obtain a virtual IMU prediction training model; Based on the virtual IMU prediction training model, the real-time predicted virtual IMU gyro angular velocity and acceleration are obtained according to the gyro angular rate increment and the added acceleration increment calculated according to the real-time IMU data, and the real-time code disk rotation angle: navigation alignment is performed according to the real-time predicted virtual IMU gyro angular velocity and acceleration.

2. The high-precision inertial navigation initial alignment method based on virtual IMU prediction according to claim 1 is characterized in that: according to Get the virtual IMU gyro angular velocity, where is the virtual IMU gyro angular velocity, is the projection of the angle change of the carrier coordinate system b relative to the navigation coordinate system n in the b system, T is the IMU sampling period, , γ, θ, and Ψ are the reference roll angle, pitch angle, and heading angle, respectively; is the projection of the Earth's rotation angular velocity in the navigation system n; , L is latitude, is the Earth's rotation angular rate; = , is the eastward projection of the velocity of the navigation coordinate system n relative to the earth coordinate system e, R n为 The radius of curvature of the Maoyou circle, is the projection of the velocity of the navigation coordinate system n relative to the earth coordinate system e along the north direction, H is the navigation altitude, R m is the meridian curvature radius.

3. The high-precision inertial navigation initial alignment method based on virtual IMU prediction according to claim 2, characterized in that: according to Get the virtual IMU acceleration, where is the virtual IMU acceleration, is the projection of the velocity change of the carrier system b relative to the inertial system i in system n, =0, is the acceleration due to gravity, is the velocity of the navigation coordinate system n relative to the earth coordinate system e, =0 .

4. The high-precision inertial navigation initial alignment method based on virtual IMU prediction according to claim 1, characterized in that: according to Get the gyro angular rate increment, where for The gyro angular rate increment within the time period, is the data preprocessing interval period, is the angular velocity measured by the gyroscope, and T is the IMU sampling period.

5. The high-precision inertial navigation initial alignment method based on virtual IMU prediction according to claim 1, characterized in that: according to Get the acceleration increment of the table, where for The acceleration increment within the time period, is the data preprocessing interval period, is the specific force measured by the meter, and T is the IMU sampling period.

6. The high-precision inertial navigation initial alignment method based on virtual IMU prediction according to claim 1, characterized in that: Set the model training input to X IN-z= ，其中， Indicates that z-score is used for data normalization, satisfying , μ represents the mean of the original data x, σ represents the standard deviation of the original data x; 、 、 are the gyro angular rate increments Components along the IMU x-axis, y-axis, and z-axis, 、 、 They are acceleration increments The components along the IMU x-axis, y-axis, and z-axis, α is the inner ring angle of the code disk, and β is the outer ring angle of the code disk.

7. The high-precision inertial navigation initial alignment method based on virtual IMU prediction according to claim 1, characterized in that: Set the model training output to ,in, and are the virtual IMU gyro angular velocity and acceleration values ​​respectively.

8. The high-precision inertial navigation initial alignment method based on virtual IMU prediction according to claim 1 is characterized in that The loss function of the virtual IMU prediction training model is ,in, is the model output, is the predicted value during model training, and a is the number of all data items for training.

9. The high-precision inertial navigation initial alignment method based on virtual IMU prediction according to any one of claims 1 to 8, characterized in that: After normalizing the gyro angular rate increment and the added acceleration increment obtained by real-time solution and the real-time code disk rotation angle, they are input into the virtual IMU prediction training model to obtain the real-time prediction output of the virtual IMU gyro angular rate and acceleration.

Images