Error parameter identification and prediction method for micro-inertial navigation with carrier motion excitation

Abstract

The application provides a carrier motion excitation micro-inertial navigation error parameter identification and prediction method, which effectively combines micro-inertial navigation use data and calibration data, compensates device errors according to use environment and motion conditions, fully utilizes historical data, improves micro-inertial navigation key error parameter compensation precision, and further improves micro-inertial navigation performance; meanwhile, the application also introduces the influence of motion excitation on error parameter compensation precision, avoids data pollution of calibration results caused by failure to excite errors; finally, the learning and prediction method of micro-inertial navigation error parameter change trend provided by the application can reduce device errors caused by use life growth, and prolong the service life of the micro-inertial navigation.

Description

Technical Field

[0001] This invention belongs to the field of navigation, guidance and control technology, and in particular relates to a method for identifying and predicting micro inertial navigation error parameters for carrier motion excitation. Background Technology

[0002] Micro inertial navigation systems (INS) are gradually evolving towards miniaturization, low cost, and low power consumption, and are widely used for navigation of moving objects such as pedestrians, unmanned vehicles, and drones. The autonomous navigation performance of micro INS is greatly affected by the environment, the carrier, and its motion mode. Long-term operation results in significant parameter drift, making error calibration and compensation crucial for ensuring its performance. Traditional calibration and error compensation methods are limited by laboratory-specific equipment and cannot achieve in-situ calibration on the carrier. Rapid on-site calibration methods free micro INS error compensation from site limitations, but single rapid calibrations may suffer from insufficient error excitation, reducing calibration accuracy and affecting the performance of the micro INS. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a method for identifying and predicting micro inertial navigation error parameters based on carrier motion excitation. By establishing a deep learning network model to predict errors, and fully utilizing historical measurement data to excite errors, the method improves the accuracy of micro inertial navigation error compensation and enhances the performance of the micro inertial navigation system.

[0004] A method for identifying and predicting error parameters of a micro inertial navigation system (INS) under carrier motion excitation is provided. The method determines the error parameters used to compensate the INS based on the results of turntable calibration experiments and carrier motion experiments. Specifically, if only turntable calibration experiments are conducted, or if both turntable calibration and carrier motion experiments are conducted simultaneously, the error parameter calibration results from the turntable calibration experiment are directly used as the error parameters for compensating the INS. If only carrier motion experiments are conducted, it is determined whether an error parameter prediction model has been established. If yes, the result with greater reliability between the error parameter identification results from the carrier motion experiment and the error parameter prediction results output by the error parameter prediction model is selected as the error parameter for compensating the INS. If no, the error parameter identification results from the carrier motion experiment are used as the error parameters for compensating the INS.

[0005] Furthermore, the micro-inertial navigation system includes an accelerometer and a gyroscope, and the method for obtaining the calibration results of the error parameters in the turntable calibration experiment is as follows:

[0006] The error parameter models for the gyroscope and accelerometer are constructed as follows:

[0007]

[0008] in, These are the three-axis components of the carrier angular velocity measured by the gyroscope in the carrier coordinate system b; These are the three-axis components of the expected angular velocity of the carrier in the carrier coordinate system b; These are the three-axis components of the gyroscope's zero bias in the carrier coordinate system b; These are the three-axis components of the gyroscope random noise in the carrier coordinate system b; K ωx K ωy K ωz These are the three-axis components of the gyroscope calibration coefficients in the carrier coordinate system b; θ ωx θ ωy θ ωz These are the three-axis components of the gyroscope mounting angle error in the carrier coordinate system b; These are the triaxial components of the carrier acceleration measured by the accelerometer in the carrier coordinate system b; These are the triaxial components of the expected value of the carrier acceleration in the carrier coordinate system b; These are the three-axis components of the accelerometer zero bias in the carrier coordinate system b; These are the three-axis components of the accelerometer random noise in the carrier coordinate system b; K ax K ay K az These are the three-axis components of the accelerometer scale coefficients in the carrier coordinate system b; θ az θ ay θ az These are the three-axis components of the accelerometer mounting angle error in the carrier coordinate system b;

[0009] Using the carrier angular velocity, expected value of carrier angular velocity, and random noise measured by the gyroscope, and the carrier acceleration, expected value of carrier acceleration, and random noise measured by the accelerometer as known quantities, the least squares method is used to calculate the calibration results of the error parameters in the gyroscope error parameter model and the accelerometer error parameter model. The calibration results of the error parameters in the gyroscope error parameter model are gyroscope zero bias, gyroscope scale coefficient, and gyroscope installation angle error; the calibration results of the error parameters in the accelerometer error parameter model are accelerometer zero bias, accelerometer scale coefficient, and accelerometer installation angle error.

[0010] Furthermore, the micro-inertial navigation system includes an accelerometer and a gyroscope, and the method for obtaining the error parameter identification results of the carrier motion experiment is as follows:

[0011] The error parameter model for the gyroscope and accelerometer is constructed as follows:

[0012]

[0013] Z = HX + V

[0014] Among them, the state parameter sequence δVn The triaxial acceleration error of the carrier in the navigation coordinate system n, measured by the accelerometer; φ n μ represents the three-axis attitude error of the carrier in the navigation coordinate system n; b The installation angle error of the gyroscope and accelerometer in the carrier coordinate system b; δK a For the accelerometer's triaxial scale coefficient error; δK ω This refers to the error of the three-axis scale coefficient of the gyroscope. The accelerometer has zero bias along its three axes in the carrier coordinate system b; ε b The gyroscope has zero bias on all three axes in the carrier coordinate system b. The derivative of X is the observed value Z = [δV]. n φ n ] T W and V are uncorrelated Gaussian white noise, F is the state matrix, and H is the observation matrix;

[0015] The Kalman filter micro-inertial navigation error parameter identification algorithm is used to solve the error parameter model of the gyroscope and accelerometer. The error parameter identification result obtained is the three-axis zero bias of the accelerometer. Three-axis zero bias ε of the gyroscope b Accelerometer triaxial scale coefficient error δK a Gyroscope three-axis scale coefficient error δK ω The installation angle error μ of the gyroscope and accelerometer in the carrier coordinate system b b .

[0016] Furthermore, the method for obtaining the error parameter prediction results output by the error parameter prediction model is as follows:

[0017] The environmental parameters and date information of the day of the carrier motion experiment are input into the error parameter prediction model based on Transformer. The error parameter prediction model outputs the error parameter prediction results, which include gyroscope zero bias, gyroscope scale coefficient, gyroscope installation angle error, accelerometer zero bias, accelerometer scale coefficient, and accelerometer installation angle error.

[0018] Furthermore, the method for obtaining the reliability of the error parameter identification results and the error parameter prediction results output by the error parameter prediction model from the carrier motion experiment is as follows:

[0019] The motion feature code ID1 of the carrier's motion trajectory, obtained using a data autoencoding method, is as follows:

[0020] ID1 = AutoEncoder(Body, Mode)

[0021] Among them, Body is the carrier type label, Mode is the motion type label sequence formed by a series of motions performed by the carrier during the motion experiment, and AutoEncoder is the data self-encoding method;

[0022] The environmental parameter feature encoding ID2 of the carrier motion trajectory obtained by the parameter feature extraction algorithm is as follows:

[0023] ID2=Onehot(E)=Onehot(T,RH)

[0024] Where E represents the environmental parameters of the carrier motion experiment, T represents the temperature, RH represents the humidity, and Onehot represents the unique thermal encoding method;

[0025] The following model is constructed based on motion feature encoding ID1 and environmental parameter feature encoding to obtain the confidence level of error parameters in a Gaussian multivariate regression model:

[0026] s = (Y, r, ID1, ID2)

[0027] m(s) = E[μ(s)]

[0028] k(s,s')=E[(f(s)-m(s))(f(s')-m(s'))]

[0029] μ(s)~GP(m(s),k(s,s'))

[0030] Wherein, s is the true value of the state variable, s' is the estimated value of the state variable obtained based on the IMU data Y of the micro-inertial navigation system before compensation and the auxiliary navigation reference information r, Y is the IMU data, and Y includes the three-axis components Gyro_x, Gyro_y, Gyro_z of the carrier angular velocity measured by the gyroscope in the carrier coordinate system b, and the three-axis components Acc_x, Acc_y, Acc_z of the carrier acceleration measured by the accelerometer in the carrier coordinate system b; r is the auxiliary navigation reference information, and r includes the GNSS position information, velocity information, and attitude information; k(s,s') is the covariance function, μ(s) is the credibility of the error parameter identification result or the error parameter prediction result output by the error parameter prediction model of the body motion experiment downloaded to the true value of the state variable s, E[·] represents the expectation function, f(·) represents the regression model function, GP(·) represents the Gaussian process distribution function, and m(·) represents the mean function.

[0031] Beneficial effects:

[0032] This invention provides a method for identifying and predicting error parameters of a micro inertial navigation system (INS) under carrier motion excitation. It effectively combines INS usage data and calibration data, selectively compensates for device errors based on the usage environment and motion conditions, fully utilizes historical data, improves the accuracy of compensation for key INS error parameters, and thus enhances INS performance. Furthermore, this invention introduces the influence of motion excitation on the accuracy of error parameter compensation, avoiding the contamination of calibration results by data that failed to excite errors. Finally, the proposed method for learning and predicting the changing trends of INS error parameters can reduce device errors caused by increased service life and extend the lifespan of the INS. Attached Figure Description

[0033] Figure 1 The flowchart shows a method for identifying and predicting micro-inertial navigation error parameters for carrier motion excitation provided by the present invention. Detailed Implementation

[0034] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0035] First, based on the experimental type of the collected data, this invention divides the historical micro-inertial navigation data into micro-inertial navigation carrier motion experiment data and micro-inertial navigation laboratory turntable calibration experiment data. The historical micro-inertial navigation data is processed in units of these units; that is, micro-inertial navigation motion data is processed as data from a complete motion trajectory, and micro-inertial navigation calibration data is processed as data from a single complete calibration experiment. The storage structures for the two types of micro-inertial navigation experimental data differ, as detailed below:

[0036] (1) Data structure of micro-inertial carrier motion experiment

[0037] The data is divided into units based on each complete motion trajectory, and the data format is shown in Table 1.

[0038] Table 1. Data Structure Design for Carrier Motion Experiment

[0039]

[0040] (2) Data structure for micro inertial navigation turntable calibration experiment

[0041] Each complete calibration data point is used as the unit of division, and the data format is shown in Table 2.

[0042] Table 2. Data Structure Design for Turntable Calibration Experiment

[0043]

[0044] Secondly, this invention uses reference time synchronization information to align the acquired IMU data with reference data. Specifically, it aligns the IMU data, navigation reference auxiliary information data, and environmental parameter information in the micro-inertial navigation motion data, and aligns the IMU data, IMU reference data, and environmental parameter information in the micro-inertial navigation calibration data. Data preprocessing, such as outlier removal, is then applied to the aligned data.

[0045] For the motion data of the micro-inertial navigation system (MIS) carrier, the subsequent carrier / motion type identification is performed as follows: First, the IMU data is divided according to the selected window length, and the sample size is calculated as acquisition time / window duration. The sample data from the first 20 seconds of the carrier startup phase is input into the carrier type classifier to obtain the carrier type label. The data samples are then input into the motion type classifier in chronological order to obtain the motion type label time series for that trajectory.

[0046] Based on this, such as Figure 1 As shown, a method for identifying and predicting error parameters of a micro inertial navigation system (INS) for carrier motion excitation is proposed. The method determines the error parameters used to compensate the INS based on the results of turntable calibration experiments and carrier motion experiments. Specifically, if only turntable calibration experiments are conducted, or if both turntable calibration and carrier motion experiments are conducted simultaneously, the error parameter calibration results from the turntable calibration experiment are directly used as the error parameters for compensating the INS. If only carrier motion experiments are conducted, it is determined whether an error parameter prediction model has been established. If yes, the result with greater reliability between the error parameter identification results from the carrier motion experiment and the error parameter prediction results output by the error parameter prediction model is selected as the error parameter for compensating the INS. If no, the error parameter identification results from the carrier motion experiment are used as the error parameters for compensating the INS.

[0047] The following section details the process of establishing the micro inertial navigation error model and identifying its parameters.

[0048] S2.1: First, establish a micro-inertial navigation error model based on actual application requirements and determine the error parameters to be identified. The micro-inertial navigation error model is constructed as follows:

[0049] Y = Model(k, X) (1)

[0050] Among them, Y, X∈Rn, Y=(y1,y2,…,y n ) T For the micro inertial navigation output value, X = (x1, x2, ... x n ) T The input values ​​for the micro-inertial navigation system are all n-dimensional vectors, k∈R. l ,k=(k1,k2,…k l ) T Let R be the parameter vector of the micro-inertial navigation error model, where Model ∈ R n ×R l →Rn This is the mapping from input to output of the micro inertial navigation system, i.e., the micro inertial navigation error model architecture.

[0051] The micro inertial navigation turntable calibration experiment data execution step S2.2, and the micro inertial navigation motion experiment data execution step S2.3.

[0052] S2.2: Identification of error parameters in turntable calibration experiment. For micro-inertial navigation turntable calibration experiment data, in equation (1) Y=(y1,y2,…,y n ) T Let Y be the IMU data in Table 2, and X = (x1, x2, ... x n ) T The IMU reference data X is shown in Table 2. An error parameter calibration algorithm was designed based on the turntable calibration experiment principle to obtain the micro-inertial navigation system calibration error parameters for each test. The error parameter identification results from turntable calibration experiments at different times constitute the time series data of calibration error parameters.

[0053] S2.3: Identification of Error Parameters in Carrier Motion Experiments. Since the micro-inertial navigation input values ​​cannot be directly obtained from the micro-inertial navigation experiment data, equation (1) cannot be directly applied. Therefore, based on the types of navigation reference auxiliary information in Table 1, a micro-inertial navigation error parameter identification algorithm based on carrier motion experiment data is used to obtain the identification error parameters for each carrier motion experiment. The error parameter identification results of carrier motion experiments at different times constitute the time series data of the identification error parameters.

[0054] It should be noted that the accuracy of error parameters obtained from the motion experimental data of the micro-inertial navigation system (INS) carrier is limited by factors such as the accuracy of navigation reference auxiliary information and the degree of excitation of the error parameters by the motion trajectory. Therefore, the reliability of the INS identification error parameters differs from that of the INS calibration error parameters. A reliability index is added to the time series of INS error parameters, and then the INS identification and calibration error parameter series are fused based on the reliability.

[0055] S3.1: Calculation of the reliability of error parameters. Calculation of the reliability of calibration experiment error parameters: For the time series of micro-inertial navigation system calibration error parameters, it is considered completely reliable, i.e., the reliability is 1. Calculation of the reliability of carrier motion experiment error parameters: For the time series of micro-inertial navigation system identification error parameters, the algorithm design for the reliability of micro-inertial navigation system identification error parameters is as follows:

[0056] (1) The motion feature encoding ID1 of the carrier's motion trajectory is obtained using the data autoencoding method as follows:

[0057] ID1=AutoEncoder(Body,Mode) (2)

[0058] Among them, Body is the carrier type label, Mode is the motion type label sequence formed by a series of motions performed by the carrier during the motion experiment, and AutoEncoder is the data self-encoding method;

[0059] (2) The environmental parameter feature encoding ID2 of the carrier motion trajectory is obtained by using a parameter feature extraction algorithm as follows:

[0060] ID2=Onehot(E)=Onehot(T,RH) (3)

[0061] Where E represents the environmental parameter information of the carrier motion experiment in Table 1, T represents the temperature in Table 1, RH represents the humidity in Table 1, and Onehot represents the one-hot encoding method.

[0062] (3) Design an algorithm to identify the reliability of error parameters. The algorithm takes motion features and environmental features as input and outputs the reliability. The mathematical form is as follows:

[0063] μ=f(Y,r,ID1,ID2) (4)

[0064] Where μ = (μ1, μ2, ..., μ l ) T The confidence level corresponds to l error parameters (range [0, 1]); in equation (2), Y = (y1, y2, ..., y n ) T Y represents the IMU data in Table 1, and r represents the auxiliary navigation reference information in Table 1.

[0065] S3.2: Reliable Fusion of Error Parameter Sequences. There are three cases for error parameter fusion in turntable calibration experiments and carrier motion experiments:

[0066] (1) On the same date, there are turntable calibration experiments and carrier motion experiments. The error parameter calibration result of the turntable calibration experiment is taken as the final error parameter for that day, and the confidence level is recorded as 1.

[0067] (2) Only the turntable calibration experiment was performed on the same date. The error parameter calibration result of the turntable calibration experiment was taken as the final error parameter for that day, and the confidence level was recorded as 1.

[0068] (3) Only the carrier motion experiment is performed on the same date. When an error parameter prediction model has been established, the error parameter calibration result and the error parameter prediction result of the carrier motion experiment are reliably fused as the final error parameter for that day, and a reliable fusion algorithm is designed to provide reliability. When an error parameter prediction model has not been established, the error parameter calibration result of the carrier motion experiment is taken as the final error parameter for that day, and the reliability is provided by the error parameter reliability identification algorithm in S3.1. The threshold-based reliable fusion algorithm is designed as follows:

[0069]

[0070] Where, k i ,i=1,2,...,l represents the i-th error parameter, Let represent the i-th error parameter obtained by prediction using the established error parameter prediction model and identification using the least squares method, respectively. This represents the confidence level corresponding to the i-th error parameter in the prediction and identification.

[0071] Thus, a reliable sequence of error parameters in the following form is obtained.

[0072] Table 3

[0073]

[0074]

[0075] Based on the construction of a micro inertial navigation error parameter prediction model, the model realizes the function of inputting a reliable error parameter sequence, environmental parameter feature encoding and time information, and outputting the predicted value of the error parameter.

[0076] Based on the actual application scenario, environmental parameters E = (T, RH...) that significantly affect the identification accuracy and variation range of micro-inertial navigation error parameters are selected, and the model is established as follows.

[0077] (k,μ)=G(E,Day) (5)

[0078] Where E represents the environmental parameters in Tables 1 and 2, T represents the temperature in Table 1, RH represents the humidity in Table 1, and k∈R l k = (k1, k2, ... k) l ) T is the parameter vector of the micro inertial navigation error model, Day is the date information of the day the data was collected, and μ is the credibility obtained by equation (4) in S3.2 and S3.1.

[0079] During the micro-inertial navigation system (MIS)-borne carrier motion experiment, the predicted values ​​of the MIS error parameters are used to compensate the MIS output. When processing the data, S3.1 is executed to obtain the identified error parameters and their reliability for this carrier motion experiment. Then, the error parameter reliability fusion algorithm in S3.2(3) is used to extract the identified and predicted error parameters with high reliability, thereby improving the accuracy of the error parameters.

[0080] Furthermore, based on motion feature encoding ID1 and environmental parameter feature encoding, the following model is constructed to obtain the reliability of error parameters based on a Gaussian multivariate regression model:

[0081] s = (Y, r, ID1, ID2)

[0082] m(s) = E[μ(s)]

[0083] k(s,s')=E[(f(s)-m(s))(f(s')-m(s'))]

[0084] μ(s)~GP(m(s),k(s,s'))

[0085] Wherein, s is the true value of the state variable, s' is the estimated value of the state variable obtained based on the IMU data Y of the micro-inertial navigation system before compensation and the auxiliary navigation reference information r, Y is the IMU data, and Y includes the three-axis components Gyro_x, Gyro_y, Gyro_z of the carrier angular velocity measured by the gyroscope in the carrier coordinate system b, and the three-axis components Acc_x, Acc_y, Acc_z of the carrier acceleration measured by the accelerometer in the carrier coordinate system b; r is the auxiliary navigation reference information, and r includes the GNSS position information, velocity information, and attitude information; k(s,s') is the covariance function, μ(s) is the credibility of the error parameter identification result or the error parameter prediction result output by the error parameter prediction model of the body motion experiment downloaded to the true value of the state variable s, E[·] represents the expectation function, f(·) represents the regression model function, GP(·) represents the Gaussian process distribution function, and m(·) represents the mean function.

[0086] Furthermore, based on the dataset structure of the micro-inertial navigation carrier motion experiment provided in Tables 4 and 5, the error parameter calibration results of the turntable calibration experiment and the error parameter identification results of the carrier motion experiment are further introduced.

[0087] Table 4

[0088]

[0089]

[0090] Table 5

[0091]

[0092] The time series data in the dataset is aligned according to timestamps and preprocessed. Then, the IMU data is divided into IMU data with a window length of 2 seconds to obtain carrier type / motion type identification samples. The sample data of the carrier startup phase is input into the carrier type classifier to obtain the carrier type label. The data samples are input into the motion type classifier in chronological order to obtain the motion type label time series of the trajectory.

[0093] Based on this, the method for obtaining the calibration results of the error parameters in the turntable calibration experiment is as follows:

[0094] The error parameter models for the gyroscope and accelerometer are constructed as follows:

[0095]

[0096] in, These are the three-axis components of the carrier angular velocity measured by the gyroscope in the carrier coordinate system b; These are the three-axis components of the expected angular velocity of the carrier in the carrier coordinate system b; These are the three-axis components of the gyroscope's zero bias in the carrier coordinate system b; These are the three-axis components of the gyroscope random noise in the carrier coordinate system b; K ωx K ωy K ωz These are the three-axis components of the gyroscope calibration coefficients in the carrier coordinate system b; θ ωx θ ωy θ ωz These are the three-axis components of the gyroscope mounting angle error in the carrier coordinate system b; These are the triaxial components of the carrier acceleration measured by the accelerometer in the carrier coordinate system b; These are the triaxial components of the expected value of the carrier acceleration in the carrier coordinate system b; These are the three-axis components of the accelerometer zero bias in the carrier coordinate system b; These are the three-axis components of the accelerometer random noise in the carrier coordinate system b; K ax K ay K az These are the three-axis components of the accelerometer scale coefficients in the carrier coordinate system b; θ az θ ay θ az These are the three-axis components of the accelerometer mounting angle error in the carrier coordinate system b;

[0097] Using the carrier angular velocity, expected value of carrier angular velocity, and random noise measured by the gyroscope, and the carrier acceleration, expected value of carrier acceleration, and random noise measured by the accelerometer as known quantities, the least squares method is used to calculate the calibration results of the error parameters in the gyroscope error parameter model and the accelerometer error parameter model. The calibration results of the error parameters in the gyroscope error parameter model are gyroscope zero bias, gyroscope scale coefficient, and gyroscope installation angle error; the calibration results of the error parameters in the accelerometer error parameter model are accelerometer zero bias, accelerometer scale coefficient, and accelerometer installation angle error.

[0098] Furthermore, the method for obtaining the error parameter identification results of the carrier motion experiment is as follows:

[0099] The error parameter model for the gyroscope and accelerometer is constructed as follows:

[0100]

[0101] Z = HX + V

[0102] Among them, the state parameter sequence δV n The triaxial acceleration error of the carrier in the navigation coordinate system n, measured by the accelerometer; φ n μ represents the three-axis attitude error of the carrier in the navigation coordinate system n; b The installation angle error of the gyroscope and accelerometer in the carrier coordinate system b; δK a For the accelerometer's triaxial scale coefficient error; δK ω This refers to the error of the three-axis scale coefficient of the gyroscope. The accelerometer has zero bias along its three axes in the carrier coordinate system b; ε b The gyroscope has zero bias on all three axes in the carrier coordinate system b. The derivative of X is the observed value Z = [δV]. n φ n ] T W and V are uncorrelated Gaussian white noise, F is the state matrix, and H is the observation matrix;

[0103] The Kalman filter micro-inertial navigation error parameter identification algorithm is used to solve the error parameter model of the gyroscope and accelerometer. The error parameter identification result obtained is the three-axis zero bias of the accelerometer. Three-axis zero bias ε of the gyroscope b Accelerometer triaxial scale coefficient error δK a Gyroscope three-axis scale coefficient error δK ω The installation angle error μ of the gyroscope and accelerometer in the carrier coordinate system b b .

[0104] Furthermore, the method for obtaining the error parameter prediction results output by the error parameter prediction model is as follows:

[0105] The environmental parameters and date information of the day of the carrier motion experiment are input into the error parameter prediction model based on Transformer. The error parameter prediction model outputs the error parameter prediction results, which include gyroscope zero bias, gyroscope scale coefficient, gyroscope installation angle error, accelerometer zero bias, accelerometer scale coefficient, and accelerometer installation angle error.

[0106] It should be noted that the parameter settings for the Transformer-based error parameter prediction model are as follows:

[0107] Table 5

[0108]

[0109]

[0110] In summary, this invention effectively combines usage data and calibration data of micro inertial navigation systems (INS), selectively compensates for device errors based on the usage environment and motion conditions, fully utilizes historical data, improves the compensation accuracy of key error parameters of INS, and thus enhances the performance of INS. Simultaneously, this invention also introduces the influence of motion excitation on the compensation accuracy of error parameters, avoiding the contamination of calibration results by data that fails to excite errors. Finally, the learning and prediction method for the changing trends of INS error parameters proposed in this invention can reduce device errors caused by the increase in service life and extend the lifespan of the INS.

[0111] Of course, the present invention may have other various embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. A method for identifying and predicting micro-inertial navigation error parameters for carrier motion excitation, characterized in that, The error parameters used to compensate for the micro inertial navigation system are determined based on the results of the turntable calibration experiment and the carrier motion experiment. If only the turntable calibration experiment is conducted, or if both the turntable calibration experiment and the carrier motion experiment are conducted simultaneously, the error parameter calibration result of the turntable calibration experiment is directly used as the error parameter for compensating the micro inertial navigation system. If only the carrier motion experiment is conducted, it is determined whether an error parameter prediction model has been established. If yes, the result with greater confidence between the error parameter identification result of the carrier motion experiment and the error parameter prediction result output by the error parameter prediction model is selected as the error parameter for compensating the micro inertial navigation system. If no, the error parameter identification result of the carrier motion experiment is used as the error parameter for compensating the micro inertial navigation system.

2. The method for identifying and predicting micro-inertial navigation error parameters for carrier motion excitation as described in claim 1, characterized in that, The micro inertial navigation system includes an accelerometer and a gyroscope, and the method for obtaining the calibration results of the error parameters in the turntable calibration experiment is as follows: The error parameter models for the gyroscope and accelerometer are constructed as follows: in, These are the three-axis components of the carrier angular velocity measured by the gyroscope in the carrier coordinate system b; These are the three-axis components of the expected angular velocity of the carrier in the carrier coordinate system b; These are the three-axis components of the gyroscope's zero bias in the carrier coordinate system b; These are the three-axis components of the gyroscope random noise in the carrier coordinate system b; K ωx K ωy K ωz These are the three-axis components of the gyroscope calibration coefficients in the carrier coordinate system b; θ ωx θ ωy θ ωz These are the three-axis components of the gyroscope mounting angle error in the carrier coordinate system b; These are the triaxial components of the carrier acceleration measured by the accelerometer in the carrier coordinate system b; These are the triaxial components of the expected value of the carrier acceleration in the carrier coordinate system b; These are the three-axis components of the accelerometer zero bias in the carrier coordinate system b; These are the three-axis components of the accelerometer random noise in the carrier coordinate system b; K ax K ay K az These are the three-axis components of the accelerometer scale coefficients in the carrier coordinate system b; θ az θ ay θ az These are the three-axis components of the accelerometer mounting angle error in the carrier coordinate system b; Using the carrier angular velocity, expected value of carrier angular velocity, and random noise measured by the gyroscope, and the carrier acceleration, expected value of carrier acceleration, and random noise measured by the accelerometer as known quantities, the least squares method is used to calculate the calibration results of the error parameters in the gyroscope error parameter model and the accelerometer error parameter model. The calibration results of the error parameters in the gyroscope error parameter model are gyroscope zero bias, gyroscope scale coefficient, and gyroscope installation angle error; the calibration results of the error parameters in the accelerometer error parameter model are accelerometer zero bias, accelerometer scale coefficient, and accelerometer installation angle error.

3. The method for identifying and predicting micro-inertial navigation error parameters for carrier motion excitation as described in claim 1, characterized in that, The micro-inertial navigation system includes accelerometers and gyroscopes, and the method for obtaining the error parameter identification results of the carrier motion experiment is as follows: The error parameter model for the gyroscope and accelerometer is constructed as follows: Z = HX + V Among them, the state parameter sequence δV n The triaxial acceleration error of the carrier in the navigation coordinate system n, measured by the accelerometer; φ n μ represents the three-axis attitude error of the carrier in the navigation coordinate system n; b The installation angle error of the gyroscope and accelerometer in the carrier coordinate system b; δK a For the accelerometer's triaxial scale coefficient error; δK ω This refers to the error of the three-axis scale coefficient of the gyroscope. The accelerometer has zero bias along its three axes in the carrier coordinate system b; ε b The gyroscope has zero bias on all three axes in the carrier coordinate system b. The derivative of X is the observed value Z = [δV]. n φ n ] T W and V are uncorrelated Gaussian white noise, F is the state matrix, and H is the observation matrix; The Kalman filter micro-inertial navigation error parameter identification algorithm is used to solve the error parameter model of the gyroscope and accelerometer. The error parameter identification result obtained is the three-axis zero bias of the accelerometer. Three-axis zero bias ε of the gyroscope b Accelerometer triaxial scale coefficient error δK a Gyroscope three-axis scale coefficient error δK ω The installation angle error μ of the gyroscope and accelerometer in the carrier coordinate system b b .

4. The method for identifying and predicting micro-inertial navigation error parameters for carrier motion excitation as described in claim 1, characterized in that, The method for obtaining the error parameter prediction results output by the error parameter prediction model is as follows: The environmental parameters and date information of the day of the carrier motion experiment are input into the error parameter prediction model based on Transformer. The error parameter prediction model outputs the error parameter prediction results, which include gyroscope zero bias, gyroscope scale coefficient, gyroscope installation angle error, accelerometer zero bias, accelerometer scale coefficient, and accelerometer installation angle error.

5. The method for identifying and predicting micro-inertial navigation error parameters for carrier motion excitation as described in claim 1, characterized in that, The method for obtaining the reliability of the error parameter identification results and the error parameter prediction results output by the error parameter prediction model in the carrier motion experiment is as follows: The motion feature code ID1 of the carrier's motion trajectory, obtained using a data autoencoding method, is as follows: ID1 = AutoEncoder(Body, Mode) Among them, Body is the carrier type label, Mode is the motion type label sequence formed by a series of motions performed by the carrier during the motion experiment, and AutoEncoder is the data self-encoding method; The environmental parameter feature encoding ID2 of the carrier motion trajectory obtained by the parameter feature extraction algorithm is as follows: ID2=Onehot(E)=Onehot(T,RH) Where E represents the environmental parameters of the carrier motion experiment, T represents the temperature, RH represents the humidity, and Onehot represents the unique thermal encoding method; The following model is constructed based on motion feature encoding ID1 and environmental parameter feature encoding to obtain the confidence level of error parameters in a Gaussian multivariate regression model: s = (Y, r, ID1, ID2) m(s) = E[μ(s)] k(s,s')=E[(f(s)-m(s))(f(s')-m(s'))] μ(s)~GP(m(s),k(s,s')) Wherein, s is the true value of the state variable, s' is the estimated value of the state variable obtained based on the IMU data Y of the micro-inertial navigation system before compensation and the auxiliary navigation reference information r, Y is the IMU data, and Y includes the three-axis components Gyro_x, Gyro_y, Gyro_z of the carrier angular velocity measured by the gyroscope in the carrier coordinate system b, and the three-axis components Acc_x, Acc_y, Acc_z of the carrier acceleration measured by the accelerometer in the carrier coordinate system b; r is the auxiliary navigation reference information, and r includes the GNSS position information, velocity information, and attitude information; k(s,s') is the covariance function, μ(s) is the credibility of the error parameter identification result or the error parameter prediction result output by the error parameter prediction model of the body motion experiment downloaded to the true value of the state variable s, E[·] represents the expectation function, f(·) represents the regression model function, GP(·) represents the Gaussian process distribution function, and m(·) represents the mean function.

Images