Error Calibration Method, Device, Equipment and Medium of Inertial Measurement Unit

Abstract

The present application discloses an error calibration method, device, equipment and medium for an inertial measurement unit, relating to the technical field of inertial navigation. The method includes: obtaining an output sequence of the inertial measurement unit; according to the output sequence, using an error prediction model to determine an error sequence of the inertial measurement unit; wherein, the error prediction model includes an encoder and a decoder; the encoder is used to encode the output sequence by using a Transformer network and a long short-term memory network to obtain a hidden state sequence; the decoder is used to decode the hidden state sequence by using a Kalman filter to obtain an error sequence of the inertial measurement unit; according to the error sequence of the inertial measurement unit, correcting the output sequence of the inertial measurement unit to calibrate the error of the inertial measurement unit. The present application improves the calibration accuracy of the inertial measurement unit, and further improves the measurement accuracy of the inertial measurement unit.

Description

Technical Field

[0001] The present application relates to the technical field of inertial navigation, and particularly to an error calibration method, device, equipment and medium for an inertial measurement unit. Background Art

[0002] With the vigorous development of navigation technology and the advantages of high precision, low cost and small size of the inertial measurement unit (IMU), it is widely used in the military and civilian fields. However, due to the diverse error sources of the IMU, including manufacturing defects of the sensors themselves, interference from environmental factors, and improper operations during installation and use. Over time, the errors of inertial devices will gradually accumulate, significantly affecting the performance of the inertial navigation system. Therefore, calibrating the systematic errors of the IMU is an essential step in using the IMU.

[0003] According to different measurement factors and calibration levels, the related inertial measurement unit calibration methods can be divided into two categories: discrete calibration and system-level calibration. The discrete calibration uses a high-precision turntable to excite error parameters, but the calibration accuracy is restricted by the turntable and affected by the calibration time. To improve the accuracy, a more efficient system-level calibration method has emerged. However, the traditional multi-position IMU system-level calibration method is difficult to adjust different parameters, and there will be problems such as improper model parameter settings leading to poor filtering effects, reducing the calibration accuracy of the IMU and directly affecting the accuracy of the navigation system. Summary of the Invention

[0004] The purpose of the present application is to provide an error calibration method, device, equipment and medium for an inertial measurement unit, which can improve the calibration accuracy of the inertial measurement unit error.

[0005] To achieve the above purpose, the present application provides the following solutions:

[0006] In the first aspect, the present application provides an error calibration method for an inertial measurement unit, including:

[0007] Obtain the output sequence of the inertial measurement unit;

[0008] According to the output sequence, use an error prediction model to determine the error sequence of the inertial measurement unit;

[0009] Among them, the error prediction model includes an encoder and a decoder; the encoder is used to encode the output sequence by using a Transformer network and a long short-term memory network to obtain a hidden state sequence; the decoder is used to decode the hidden state sequence by using a Kalman filter to obtain the error sequence of the inertial measurement unit;

[0010] According to the error sequence of the inertial measurement unit, correct the output sequence of the inertial measurement unit to calibrate the errors of the inertial measurement unit.

[0011] In a second aspect, the present application provides an error calibration device for an inertial measurement unit, including:

[0012] An output sequence acquisition module, configured to acquire the output sequence of the inertial measurement unit;

[0013] An error sequence prediction module, configured to determine the error sequence of the inertial measurement unit according to the output sequence by using an error prediction model;

[0014] Wherein, the error prediction model includes an encoder and a decoder; the encoder is configured to encode the output sequence by using a Transformer network and a long short-term memory network to obtain a hidden state sequence; the decoder is configured to decode the hidden state sequence by using a Kalman filter to obtain the error sequence of the inertial measurement unit;

[0015] An error calibration module, configured to correct the output sequence of the inertial measurement unit according to the error sequence of the inertial measurement unit to calibrate the errors of the inertial measurement unit.

[0016] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the above-mentioned error calibration method for an inertial measurement unit.

[0017] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned error calibration method for an inertial measurement unit is implemented.

[0018] According to the specific embodiments provided by the present application, the present application has the following technical effects:

[0019] The present application provides an error calibration method, device, equipment and medium for an inertial measurement unit. By combining a Transformer network and a long short-term memory network to encode the output sequence of the inertial measurement unit, the characteristics of the extremely strong non-linear mapping ability and generalization ability of the neural network are utilized, and the error characteristics of the inertial measurement unit can be better learned, thereby improving the estimation accuracy of the deterministic error of the inertial measurement unit. Further, a Kalman filter is used to decode the hidden state sequence to obtain an error sequence, and the error sequence is used to correct the output sequence of the inertial measurement unit, improving the calibration accuracy of the inertial measurement unit, and further improving the measurement accuracy of the inertial measurement unit. Description of the Drawings

[0020] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0021] Figure 1 It is an application environment diagram of an error calibration method for an inertial measurement unit in an embodiment of the present application.

[0022] Figure 2 It is a schematic flowchart of an error calibration method for an inertial measurement unit provided in an embodiment of the present application.

[0023] Figure 3 It is a schematic structural diagram of an error prediction model in an embodiment of the present application.

[0024] Figure 4 It is a schematic diagram of the functional modules of an error calibration device for an inertial measurement unit provided in an embodiment of the present application.

[0025] Figure 5 It is a schematic structural diagram of a computer device provided in an embodiment of the present application. Detailed implementation manners

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0027] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the drawings and specific implementation manners.

[0028] The error calibration of the inertial measurement unit provided in the embodiments of the present application can be applied to, for example Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set separately, integrated on the server 104, placed on the cloud or other servers. The terminal 102 can send the output sequence of the inertial measurement unit to the server 104. After receiving the output sequence of the inertial measurement unit, the server 104 uses an error prediction model to determine the error sequence of the inertial measurement unit to calibrate the error of the inertial measurement unit. The server 104 can feedback the obtained error sequence of the inertial measurement unit to the terminal 102. In addition, in some embodiments, the error calibration method of the inertial measurement unit can also be implemented separately by the server 104 or the terminal 102.

[0029] Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0030] In an exemplary embodiment, as Figure 2 shown, a method for calibrating the error of an inertial measurement unit is provided. This method is executed by a computer device, and can specifically be executed separately by a computer device such as a terminal or a server, or jointly executed by a terminal and a server. In the embodiments of the present application, taking this method applied to Figure 1 the server 104 in as an example for illustration, it includes the following steps 201 to step 203.

[0031] Step 201, obtain the output sequence of the inertial measurement unit. The output sequence includes the measurement values at each moment within a set time period. The measurement values include angular velocity measurement values and acceleration measurement values. Among them, the angular velocity measurement value is the output data of the gyroscope, and the acceleration measurement value is the output data of the accelerometer.

[0032] In a specific application example, the IMU is installed on a three-axis turntable with a tooling, and the output data of the IMU within a period of time at a set rotation speed is obtained.

[0033] Step 202, according to the output sequence, use an error prediction model to determine the error sequence of the inertial measurement unit.

[0034] Among them, the error prediction model includes an encoder and a decoder. The encoder takes the output sequence of the IMU as input, and uses Transformer + Long Short-Term Memory (LSTM) to obtain the hidden state of the encoded sequence, that is, the hidden state sequence. The hidden state sequence obtained by the encoder is used as the initial state and input into the decoder to predict the error of the IMU. As Figure 3 shown, the encoder includes multiple Transformer + LSTM, and the decoder includes multiple Kalman filters, which process the data at each moment respectively, K indicating the total number of moments within a set time period.

[0035] (1) The encoder is used to encode the output sequence by using a Transformer network and a long short-term memory network to obtain a hidden state sequence. The hidden state sequence includes the hidden states at each moment within a set time period.

[0036] Inputting the output sequence of the IMU into the LSTM can output a new sequence that more effectively describes the state, but the LSTM cannot capture long-term dependencies. Therefore, this application adopts the Transformer + LSTM structure because attention usually precedes memory, and the Transformer network can capture long-term dependencies, enhancing the robustness to noisy observations. Moreover, the Transformer network draws on the design of convolutional neural networks. Therefore, the Transformer network can capture significant features that convolutional neural networks cannot capture. Combining the LSTM with the Transformer network in this application can enhance the structural advantages and time series modeling capabilities of the Transformer network. Encoding the measured values of the IMU by combining the Transformer network and the LSTM before filtering makes it easier for the Kalman filter to estimate the error sequence.

[0037] In a specific application example, the encoder is pre-trained using a training sample set. Each training sample in the training sample set includes a historical output sequence and the corresponding true hidden state sequence. During training, the ratio of the training set to the test set is set to 9:1.

[0038] Among them, first set the structure of the encoder. The hidden layer includes 3 layers of neural networks, and the number of hidden neurons is 10. During training, the Adam optimizer is used, and the supervised learning method is adopted. The mean square error is used as the evaluation index to construct the loss function of the encoder: .

[0039] Among them, is the loss function value, N is the number of training samples, is then The true hidden state sequence of a training sample is the n predicted hidden state sequence of the nth training sample.

[0040] In an exemplary embodiment, the encoder uses the following formula to determine the hidden state at a certain time period k moment.

[0041] .

[0042] .

[0043] .

[0044] Wherein, is the k hidden state at the kth moment, is the k system state covariance matrix at the kth moment, is the k system state covariance matrix at the (k - 1)th moment, is the k measurement noise covariance matrix at the kth moment, is the k measurement noise covariance matrix at the (k - 1)th moment, is the sigmoid activation function, is the weight matrix for transferring from the system state covariance matrix to the measurement value, is the weight matrix for transferring between system state covariance matrices, is the weight matrix for transferring from the measurement noise covariance matrix to the measurement value, is the weight matrix for transferring between measurement noise covariance matrices, is the k measurement value at the kth moment.

[0045] (2) The decoder is used to decode the hidden state sequence by using a Kalman filter to obtain the error sequence of the inertial measurement unit. The error sequence includes error values at each moment within a set time period. The error values include the zero biases of the three axes of the gyroscope, the scale factor error matrix in the three-axis directions of the gyroscope, the installation error matrix in the three-axis directions of the gyroscope, the zero biases of the three axes of the accelerometer, the scale factor error matrix in the three-axis directions of the accelerometer, and the installation error matrix in the three-axis directions of the accelerometer.

[0046] In an exemplary embodiment, the process of decoding the hidden state sequence by using a Kalman filter specifically includes the following steps 21 to 24.

[0047] Step 21: Design the state equation and measurement equation of the Kalman filter.

[0048] 。

[0049] 。

[0050] Among them, is k the state variable at is k the state variable at is k from k to is k the system noise matrix at is k the system process noise sequence at is k the measurement value of the sensor at is k the measurement matrix at is k the measurement noise sequence at

[0051] The state equation and measurement equation use the hidden state sequence output by the encoder for the recursion of the Kalman filter. The five recursion formulas for the Kalman filter of discrete systems are as follows.

[0052] 。

[0053] 。

[0054] 。

[0055] 。

[0056] 。

[0057] Among them, is the predicted state value, is the state transition matrix, describing the state change of the system from k to k time, is k the estimated value of the state quantity corrected by the measurement equation at is k the estimated value of the state quantity corrected by the measurement equation at is the prior estimation error covariance matrix, which describes the correlation between state variables. is k the posterior estimation error covariance matrix at time is k the posterior estimation error covariance matrix at time is k the Kalman gain at time is the identity matrix consistent with the state space dimension. The superscript T represents the transpose operation of the matrix, and the superscript -1 represents the inverse matrix of the matrix.

[0058] Since most inertial navigation systems can meet the application requirements of the Kalman filter, the corresponding navigation parameters can be obtained using the above formulas. Parameters such as the statistical characteristics of the state noise, measurement noise, and initial state of the Kalman filter usually cannot be obtained in advance and can only be set and adjusted based on experience. This application uses an encoder to obtain two parameters: the system state variance matrix and the measurement noise covariance matrix of the Kalman filter.

[0059] Step 22: Establish the Kalman filter state update equation and measurement equation for the gyroscope.

[0060] .

[0061] .

[0062] Among them, is the state variable of the gyroscope, is the system state matrix of the gyroscope, is the system state variable, is the system noise drive matrix, W is the system noise, , is x the white noise of the angular velocity measurement on the is y the white noise of the angular velocity measurement on the is z the white noise of the angular velocity measurement on the is x the zero-bias random walk noise on the is y the zero-bias random walk noise on the is z the zero-bias random walk noise on the is the observation error of the gyroscope, , is the eastward error, is the northward error, is the vertical upward direction error, is the measurement matrix of the gyroscope, is the 3x3 identity matrix, is the 3x12 all-zero matrix, is the system observation white noise.

[0063] Step 23, establish the Kalman filter state update equation and measurement equation for the accelerometer.

[0064] .

[0065] .

[0066] Among them, is the state variable of the accelerometer, is the system state matrix of the accelerometer, is the velocity error and attitude error of the accelerometer, is the measurement matrix of the accelerometer, is the 6x6 identity matrix, is the 6x12 all-zero matrix.

[0067] Step 24, use the hidden state sequence output by the encoder as the input, and online estimate and generate the error sequence of the IMU: ; among them, is k the error value at time is the zero bias of the three axes of the gyroscope, is the zero bias of the three axes of the accelerometer, is the scale factor and installation error matrix in the three-axis directions of the gyroscope, is the scale factor and installation error matrix in the three-axis directions of the accelerometer.

[0068] Step 203, correct the output sequence of the inertial measurement unit according to the error sequence of the inertial measurement unit to calibrate the errors of the inertial measurement unit.

[0069] In an exemplary embodiment, correct the angular velocity measurement value according to the zero bias of the three axes of the gyroscope, the scale factor error matrix in the three-axis directions of the gyroscope, and the installation error matrix in the three-axis directions of the gyroscope. Correct the acceleration measurement value according to the zero bias of the three axes of the accelerometer, the scale factor error matrix in the three-axis directions of the accelerometer, and the installation error matrix in the three-axis directions of the accelerometer.

[0070] Among them, the correction value = measured value - zero bias - scale factor error - installation error. The zero bias indicates that the outputs of the gyroscope and accelerometer are not zero when stationary, and the error is a fixed offset. The scale factor error represents the deviation of the ratio relationship between the output of the IMU and the actual value. The installation error is the directional deviation caused by inaccuracies or mechanical errors during the installation process when the IMU is installed on the carrier or device.

[0071] This application further establishes a random error model for the IMU, including a gyroscope random error model and an accelerometer random error model.

[0072] The gyroscope random error model is as follows.

[0073] 。

[0074] Where, is the output data of the gyroscope, is x the measured angular velocity value of the axis, y is the measured angular velocity value of the z axis, is the zero bias of the three axes of the gyroscope, is the x axis zero bias of the gyroscope, is the y axis zero bias of the gyroscope, is the z axis zero bias of the gyroscope, is the scale factor error matrix in the three-axis direction of the gyroscope, is the x axis scale factor error of the gyroscope, is the y axis scale factor error of the gyroscope, is the z axis scale factor error of the gyroscope, is the installation error matrix in the three-axis direction of the gyroscope, is the installation error from the x axis to the y axis of the gyroscope, is the installation error from the x axis to the z axis of the gyroscope, is the installation error from the y axis to the x axis of the gyroscope, is the installation error from the y axis to the z axis of the gyroscope, is the installation error from the z axis to the xThe installation error of the axis, is the installation error from the z axis to the y axis of the gyroscope, is the actual angular velocity, is the x actual angular velocity of the axis, y is the actual angular velocity of the z axis, is the random noise of the gyroscope, is the x axis random noise of the gyroscope, is the y axis random noise of the gyroscope, is the z axis random noise of the gyroscope.

[0075] The random error model of the accelerometer is as follows.

[0076] .

[0077] Wherein, is the output data of the accelerometer, is the x axis acceleration measurement value of the accelerometer, is the y axis acceleration measurement value of the accelerometer, is the z axis acceleration measurement value of the accelerometer, is the three-axis zero bias of the accelerometer, is the x axis zero bias of the accelerometer, is the y axis zero bias of the accelerometer, is the z axis zero bias of the accelerometer, is the scale factor error matrix in the three-axis direction of the accelerometer, is the x axis direction scale factor error of the accelerometer, is the y axis direction scale factor error of the accelerometer, is the z axis direction scale factor error of the accelerometer, is the installation error matrix in the three-axis direction of the accelerometer, is the x axis to y axis installation error of the accelerometer, is the x axis to z axis installation error of the accelerometer, is the accelerometer y axis to x axis installation error is the accelerometer y axis to z axis installation error is the accelerometer z axis to x axis installation error is the accelerometer z axis to y axis installation error is the actual acceleration is x actual acceleration of the is y actual acceleration of the is z actual acceleration of the is the random noise of the accelerometer is the x axis random noise of the accelerometer is the y axis random noise of the accelerometer is the z axis random noise of the accelerometer

[0078] This application combines the Transformer network and the LSTM network, utilizes the strong non - linear mapping ability and generalization ability of the neural network, can better learn the error characteristics of the IMU, improve the estimation accuracy of the deterministic error of the IMU, has a simple algorithm structure, a fast calculation speed, and uses the error sequence estimation and compensation of various errors in the IMU to improve the calibration accuracy of the IMU, and further improve the measurement accuracy of the IMU.

[0079] Based on the same inventive concept, the embodiment of this application also provides an error calibration device for implementing the above - mentioned error calibration method. The implementation solution provided by this device to solve the problem is similar to the implementation solution described in the above method. Therefore, the specific limitations in one or more embodiments of the following error calibration device can refer to the limitations on the error calibration method in the above text, and will not be repeated here.

[0080] In an exemplary embodiment, as Figure 4 shown, an error calibration device for an inertial measurement unit is provided, including: an output sequence acquisition module 401, an error sequence prediction module 402, and an error calibration module 403.

[0081] The output sequence acquisition module 401 is used to acquire the output sequence of the inertial measurement unit.

[0082] The error sequence prediction module 402 is used to determine the error sequence of the inertial measurement unit according to the output sequence by using an error prediction model.

[0083] Wherein, the error prediction model includes an encoder and a decoder. The encoder is used to encode the output sequence by using a Transformer network and a long short-term memory network to obtain a hidden state sequence. The decoder is used to decode the hidden state sequence by using a Kalman filter to obtain the error sequence of the inertial measurement unit.

[0084] The error calibration module 403 is used to correct the output sequence of the inertial measurement unit according to the error sequence of the inertial measurement unit, so as to calibrate the error of the inertial measurement unit.

[0085] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal, and its internal structure diagram can be as Figure 5 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the output sequence of the inertial measurement unit. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements an error calibration method for an inertial measurement unit.

[0086] Those skilled in the art can understand that Figure 5 the structure shown in

[0087] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0088] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0089] In this application, all actions of obtaining signals, information, or data are carried out on the premise of complying with the corresponding data protection regulations and policies of the location and obtaining authorization from the owner of the corresponding device.

[0090] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random access memories (ReRAM), magnetoresistive random access memories (MRAM), ferroelectric random access memories (FRAM), phase change memories (PCM), graphene memories, etc. Volatile memories can include random access memory (RAM) or external cache memories, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0091] The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0092] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0093] In this text, specific examples are used to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. An error calibration method for an inertial measurement unit, characterized in that The error calibration method of the inertial measurement unit includes: Obtaining the output sequence of the inertial measurement unit; According to the output sequence, using an error prediction model to determine the error sequence of the inertial measurement unit; Wherein, the error prediction model includes an encoder and a decoder; the encoder is used to encode the output sequence by using a Transformer network and a long short-term memory network to obtain a hidden state sequence; the decoder is used to decode the hidden state sequence by using a Kalman filter to obtain the error sequence of the inertial measurement unit; According to the error sequence of the inertial measurement unit, correcting the output sequence of the inertial measurement unit to calibrate the error of the inertial measurement unit; The process of decoding the hidden state sequence by using a Kalman filter includes: Designing the state equation and measurement equation of the Kalman filter: ； ； wherein, is k the state variable at time , including position, velocity and attitude, where the position and velocity are related to acceleration and the attitude is related to angular velocity, is k the state variable at time - 1, is k the one - step transition matrix from time - 1 to k time , is k the system noise matrix at time - 1, is k the system process noise sequence at time - 1, is k the measurement value of the sensor at time , is k the measurement matrix at time , is k the measurement noise sequence at time ; The state equation and measurement equation use the hidden state sequence output by the encoder for Kalman filter recursion; the five recursion formulas of discrete system Kalman filter are as follows: ； ； ； ； ； Among them, is the predicted state value, is the state transition matrix, which describes the state change of the system from k -1 to k time instant, is k the estimated state quantity value corrected by the measurement equation at is k the estimated state quantity value corrected by the measurement equation at is the prior estimation error covariance matrix, which describes the correlation between state quantities, is k the posterior estimation error covariance matrix at is k the posterior estimation error covariance matrix at is k the Kalman gain at is the identity matrix consistent with the state space dimension. The superscript T represents the transpose operation of the matrix, and the superscript -1 represents the inverse matrix of the matrix; Establishing the Kalman filter state update equation and measurement equation of the gyroscope: ； ； Among them, is the state variable of the gyroscope, is the system state matrix of the gyroscope, is the system state variable, is the system noise driving matrix, W is the system noise, , is x the white noise of the angular velocity measurement on the axis, y is the white noise of the angular velocity measurement on the z axis, is x the white noise of the angular velocity measurement on the axis, y is the random walk noise of the zero bias on the z axis, is the observation error of the gyroscope, , is the eastward error, is the northward error, is the vertical upward direction error, is the measurement matrix of the gyroscope, is the 3×3 identity matrix, is the 3×12 all-zero matrix, is the system observation white noise; Establishing the Kalman filter state update equation and measurement equation of the accelerometer: ； ； Among them, is the state variable of the accelerometer, is the system state matrix of the accelerometer, is the velocity error and attitude error of the accelerometer, is the measurement matrix of the accelerometer, is the 6×6 identity matrix, is the 6×12 zero matrix; Taking the hidden state sequence output by the encoder as the input, online estimate and generate the error sequence of the inertial measurement unit: ; where is k the error value at time the zero biases of the three axes of the gyroscope, the zero biases of the three axes of the accelerometer, the scale factor and installation error matrix in the three-axis directions of the gyroscope, the scale factor and installation error matrix in the three-axis directions of the accelerometer.

2. The error calibration method of the inertial measurement unit according to claim 1, characterized in that The encoder is pre-trained by using a training sample set; each training sample in the training sample set includes a historical output sequence and a corresponding true hidden state sequence.

3. The error calibration method of the inertial measurement unit according to claim 1, characterized in that The output sequence includes measurement values at each moment within a set time period; the hidden state sequence includes hidden states at each moment within a set time period; the error sequence includes error values at each moment within a set time period.

4. The error calibration method of the inertial measurement unit according to claim 3, characterized in that The encoder determines the hidden state at a moment within a set period by using the following formula: k ​ ； ； ； Among them, is k the hidden state at time is k the system state covariance matrix at time is k the system state covariance matrix at time -1 is k the measurement noise covariance matrix at time is k the measurement noise covariance matrix at time -1 is the sigmoid activation function, is the weight matrix for transmitting from the system state covariance matrix to the measurement value, is the weight matrix for transmitting between system state covariance matrices, is the weight matrix for transmitting from the measurement noise covariance matrix to the measurement value, is the weight matrix for transmitting between measurement noise covariance matrices, is k the measurement value at time 5. The error calibration method of the inertial measurement unit according to claim 3, characterized in that, The measurement values include angular velocity measurement values and acceleration measurement values.

6. The error calibration method of the inertial measurement unit according to claim 5, characterized in that According to the error sequence of the inertial measurement unit, correcting the output sequence of the inertial measurement unit specifically includes: Correcting the angular velocity measurement value according to the zero bias of the three axes of the gyroscope, the scale factor error matrix in the three-axis directions of the gyroscope, and the installation error matrix in the three-axis directions of the gyroscope; Correcting the acceleration measurement value according to the zero bias of the three axes of the accelerometer, the scale factor error matrix in the three-axis directions of the accelerometer, and the installation error matrix in the three-axis directions of the accelerometer.

7. An error calibration device for an inertial measurement unit, applied to the error calibration method of the inertial measurement unit according to any one of claims 1-6, characterized in that, The error calibration device of the inertial measurement unit includes: An output sequence acquisition module for acquiring the output sequence of the inertial measurement unit; An error sequence prediction module for determining the error sequence of the inertial measurement unit according to the output sequence by using an error prediction model; Wherein, the error prediction model includes an encoder and a decoder; the encoder is used to encode the output sequence by using a Transformer network and a long short-term memory network to obtain a hidden state sequence; the decoder is used to decode the hidden state sequence by using a Kalman filter to obtain the error sequence of the inertial measurement unit; An error calibration module for correcting the output sequence of the inertial measurement unit according to the error sequence of the inertial measurement unit to calibrate the error of the inertial measurement unit.

8. A computer device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the error calibration method of the inertial measurement unit according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the error calibration method of the inertial measurement unit according to any one of claims 1-6.

Images