Online self-calibration filtering method for inertial navigation

Abstract

The invention discloses an online self-calibration filtering method for inertial navigation, and relates to the technical field of inertial navigation, and the method comprises the following steps: obtaining multi-modal data, and establishing an inertial device error mathematical model and an inertial navigation system error mathematical model; carrying out recursive estimation on the state variables by the extended Kalman filtering model to obtain error parameters of the inertial device and the navigation system; constructing a fuzzy adaptive model based on the error parameters, generating an observation noise correction factor through a fuzzy logic rule, and correcting an observation noise covariance matrix in the extended Kalman filtering model in real time; carrying out online estimation and compensation on error parameters of the inertial navigation system by utilizing an extended Kalman filtering model, and outputting compensation parameters containing zero offset and scale factors in real time; correcting the original data based on the compensation parameters, injecting the original data into the navigation solution, and updating the fuzzy adaptive model according to the compensated error data.

Description

Technical Field

[0001] This invention relates to the field of inertial navigation technology, and more particularly to an online self-calibration filtering method for inertial navigation. Background Technology

[0002] Inertial navigation is an autonomous navigation technology with good stealth capabilities. It is not affected by external environmental or equipment interference during navigation and continuously provides the vehicle with navigation data such as position, velocity, and attitude. After decades of research and development, inertial navigation systems in my country have been widely applied in civilian and defense sectors.

[0003] With the significant investment of national resources in inertial sensor research, the accuracy of inertial sensors has continuously improved. However, once a certain level of accuracy is reached, further enhancement through external hardware optimization becomes extremely difficult. Therefore, relying solely on assembly and debugging methods to improve inertial sensor accuracy is insufficient. The most effective way to enhance navigation accuracy is through error compensation techniques in software algorithms. Error compensation involves establishing error models for inertial devices and the inertial navigation system within the inertial navigation solution algorithm. These models estimate the correlation coefficients of various errors in real time, and then compensate for these errors using appropriate algorithms. This reduces the cumulative error of the inertial navigation system during the solution process, thereby improving the system's accuracy.

[0004] Online calibration technology for inertial navigation systems (INS) is based on error models of inertial devices and the INS itself. It then uses appropriate filtering algorithms to calibrate various error values ​​of the system. These calibration values ​​are then used to correct the accumulated errors of the navigation system, achieving compensation and improving the accuracy of the INS. Online calibration typically involves establishing state equations using the solution results of the INS algorithm, establishing observation equations by collecting information from a high-precision external measurement system, and then using the Kalman filter algorithm to estimate the calibration compensation parameters of the INS online. However, due to the inherent bias instability of inertial sensors and the influence of the external environment, conventional Kalman filtering cannot accurately estimate the calibration compensation parameters of the INS online, resulting in navigation accuracy far below requirements. Therefore, it is necessary to perform nonlinear processing on the random walk coefficients in the raw data of the inertial sensor and use adaptive filtering to correct the uncertainty of state noise in the filtering process, enabling rapid data convergence during filtering calculations and improving the stability of Kalman filter estimation. Summary of the Invention

[0005] This application provides an online self-calibration filtering method for inertial navigation to address the issues of low calibration accuracy and errors in navigation data calculated by inertial sensors due to the zero-bias instability of inertial sensors and the influence of the external environment during online calibration of inertial navigation systems. This method improves the convergence speed and stability of filtering calculations and enhances the navigation performance of the product.

[0006] This application provides an online self-calibration filtering method for inertial navigation, including:

[0007] S1: Acquire multimodal data and establish mathematical models of inertial device errors, including zero bias error, scaling factor error, random error and random walk, as well as mathematical models of inertial navigation system errors, including velocity error, position error and attitude error.

[0008] S2: Based on the mathematical model of inertial device error and the mathematical model of inertial navigation system error, an extended Kalman filter model is constructed to recursively estimate the state variables and obtain the error parameters of inertial device and navigation system;

[0009] S3: Construct a fuzzy adaptive model based on error parameters, generate observation noise correction factors through fuzzy logic rules, and correct the observation noise covariance matrix in the extended Kalman filter model in real time;

[0010] S4: The extended Kalman filter model is used to estimate and compensate the error parameters of the inertial navigation system online, and the compensation parameters including zero bias and scaling factor are output in real time. The original data is corrected based on the compensation parameters and injected into the navigation solution. The fuzzy adaptive model is updated according to the compensated error data.

[0011] Furthermore, the construction of the extended Kalman filter model includes: a state equation with inertial device error and navigation system error as state variables, and an observation equation with velocity and position information provided by external observation equipment as observation quantities;

[0012] Discretize the state equation and the observation equation to form an extended Kalman filter model for recursive calculation.

[0013] Furthermore, the state variables include: the zero bias error of the gyroscope and accelerometer, the scale factor error, the installation error angle, the random walk error, as well as the velocity error, position error, and attitude error.

[0014] Furthermore, the error parameters include: static error parameters of the inertial device, dynamic error state of the navigation system, and time-varying random error parameters;

[0015] The static error parameters of the inertial devices include gyroscope zero bias error, accelerometer zero bias error, scale factor error, and installation error angle;

[0016] The dynamic error states of the navigation system include velocity error, position error, and attitude error;

[0017] The time-varying random error parameters include random walk error.

[0018] Furthermore, the random walk error includes: modeling the random walk error and defining it as the drift amount accumulated over time, subtracting the drift amount from the attitude error equation as a compensation term, and directly offsetting the negative impact of gyroscope zero-bias drift on attitude angle calculation through the compensation term.

[0019] Furthermore, the construction of the fuzzy adaptive model includes: calculating the ratio of the theoretical variance of attitude velocity to the trace of the measured variance as the first input variable, and calculating the ratio of the theoretical variance of position to the trace of the measured variance as the second input variable; establishing fuzzy logic rules with the first and second input variables as inputs and the observation noise correction factor as the output; and adjusting the observation noise covariance matrix in real time through the observation noise correction factor.

[0020] Furthermore, the fuzzy logic rules include: when 1.5 < input variable ≤ 2, the output correction factor increases; when 0.5 < input variable ≤ 1.5, the output correction factor remains unchanged; when 0 < input variable ≤ 0.5, the output correction factor decreases.

[0021] Furthermore, the estimation and compensation include: performing optimal estimation of the state variables through time update and measurement update loops of the extended Kalman filter model; using the estimated error parameters as compensation quantities; correcting the original output data using a data compensation algorithm to generate compensated inertial data; and injecting the compensated inertial data into the inertial navigation solution algorithm for navigation solution to output corrected attitude, velocity, and position information.

[0022] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0023] By employing an extended Kalman filter algorithm that integrates fuzzy adaptive rules and combining it with comprehensive error modeling of inertial devices and navigation systems, the estimation bias problem caused by the zero-bias instability of inertial sensors and external environmental interference in traditional methods is effectively overcome. This enables high-precision online real-time estimation and compensation of inertial navigation system errors in complex application scenarios, significantly improving the accuracy of navigation solution, the system's adaptive anti-interference capability, and the convergence speed and stability of filtering calculation. Attached Figure Description

[0024] Figure 1 This is a flowchart of an online self-calibration filtering method for inertial navigation according to an embodiment of the present invention. Detailed Implementation

[0025] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0027] Example 1: As Figure 1 As shown, an online self-calibration filtering method for inertial navigation is presented.

[0028] S1: Acquire multimodal data and establish mathematical models of inertial device errors, including zero bias error, scaling factor error, random error and random walk, as well as mathematical models of inertial navigation system errors, including velocity error, position error and attitude error.

[0029] Specifically, the inertial measurement unit (IMU) acquires raw output data from the gyroscope and accelerometer in real time, including angular rate and specific force information; it synchronously receives reference velocity and position information provided by external observation equipment; it obtains time synchronization signals to ensure the consistency of multi-source data timing; and it records environmental parameters to assist in error modeling, establishing the state equation in the filtering model to achieve online self-calibration. The mathematical model of inertial device error is as follows:

[0030] Zero bias error:

[0031] ε=[ε x ε y ε z ] T ,

[0032]

[0033] Where, ε x ε y ε z This represents the zero-bias error of the gyroscope across three axes. This represents the zero bias error of the accelerometers along the three axes.

[0034] Scale factor:

[0035] K G =K Gr (I+δKG ),

[0036] K A =K Ar (I+δK A ),

[0037] Among them, K G K represents the scaling factor of the gyroscope under ideal conditions. Gr δK represents the actual scaling factor of the gyroscope. G K represents the scaling factor error of the gyroscope. A K represents the scale factor of the accelerometer under ideal conditions. Ar δK represents the actual scale factor of the accelerometer. A The scale factor error of the accelerometer is represented by I, which represents the 3×3 identity matrix.

[0038] Random error:

[0039]

[0040] Where, ε b e represents the random error of the gyroscope. t1 This represents a new random number generated by the gyroscope's error after each time update. e represents the random error of the gyroscope. t2 I represents a new random number generated by the accelerometer error after each time update. 3×1 This represents a 3×1 identity matrix.

[0041] Random walk:

[0042]

[0043] Where, σ θ(t) N represents the random walk error of the gyroscope device. s This represents the angular random walk coefficient of the gyroscope, and t represents time.

[0044] The mathematical model for inertial navigation system error is as follows:

[0045] Speed ​​error:

[0046]

[0047] Among them, V n This represents the velocity error of the inertial navigation system. V represents n The first-order differential, f represents the attitude matrix at the current moment. b This indicates the specific force output by the accelerometer. This represents the component of the Earth's rotation angular rate in the n-frame. This represents the angular velocity component in the n-frame caused by the carrier's motion around the Earth's surface, g. n This represents the component of gravitational acceleration in the n-system.

[0048] Position error:

[0049]

[0050] in, The first derivative of the system's longitude error, δV N R represents the northbound velocity error of the system. M V represents the radius of curvature of the meridional circle, h represents the system height, δh represents the error in the system height, and V N Indicates the northward velocity of the system. The first derivative of the system's latitude error, δV E The system's eastward velocity error is represented by δL, where L represents the system's geographical latitude, and V represents the error in the system's geographical latitude. E R represents the eastward velocity of the system. N Indicates the radius of curvature of the zonal loop. The first derivative of the system height error, δV U This indicates the system's celestial velocity error.

[0051] The random walk error includes: modeling the random walk error and defining it as the drift amount accumulated over time; subtracting the drift amount from the attitude error equation as a compensation term; and directly offsetting the negative impact of gyroscope zero-bias drift on attitude angle calculation through the compensation term.

[0052] Specifically, attitude error:

[0053]

[0054] in, Let φ represent the first derivative of the system attitude error. This represents the angular velocity of the n-frame relative to the i-frame. express The first differential, δK G The accelerometer's scale coefficient error is represented by δG, which represents the accelerometer's installation error angle. ε represents the angular velocity of the b-frame relative to the i-frame. n σ represents zero bias. θ(t) This represents the random walk error of the gyroscope device.

[0055] S2: Based on the mathematical model of inertial device error and the mathematical model of inertial navigation system error, an extended Kalman filter model is constructed to recursively estimate the state variables and obtain the error parameters of inertial device and navigation system;

[0056] Discretize the state equation and the observation equation to form an extended Kalman filter model for recursive calculation.

[0057] The error parameters include: static error parameters of inertial devices, dynamic error state of navigation system, and time-varying random error parameters;

[0058] The static error parameters of the inertial devices include gyroscope zero bias error, accelerometer zero bias error, scale factor error, and installation error angle;

[0059] The dynamic error states of the navigation system include velocity error, position error, and attitude error;

[0060] The time-varying random error parameters include random walk error.

[0061] The construction of the extended Kalman filter model includes: a state equation with inertial device error and navigation system error as state variables, and an observation equation with velocity and position information provided by external observation equipment as observation quantities;

[0062] The state variables include: the zero bias error of the gyroscope and accelerometer, the scale factor error, the installation error angle, the random walk error, as well as the velocity error, position error, and attitude error.

[0063] Specifically, an extended Kalman filter model is constructed. In a Kalman filter, accurate reference information is needed as the observed variable to make the estimated state value closer and closer to the true value. This is because the accuracy of the state variable is affected by external random noise, random walks, and other errors that cannot be fundamentally eliminated. Prediction and estimation using both state and observed variables are necessary to make the filtered estimation result closer and closer to the true value. Therefore, a state equation needs to be established, and an observation equation needs to be established in conjunction with external observations. The established model is as follows:

[0064]

[0065] Y(t)=H(t)X(t)+V(t),

[0066] in, Let X(t) represent the first derivative of the state variable, F(t) represent the state variable, G(t) represent the noise interference matrix, W(t) represent the random noise that interferes with the state variable, Y(t) represent the observed variable in the linear system, H(t) represent the observation matrix, and V(t) represent the observation noise vector.

[0067] By establishing the state equation and observation equation, and deriving the root mean square error of the state noise and observation noise, a one-step state prediction is performed, resulting in the following recursive Kalman filter:

[0068]

[0069] Among them, X k / k This is the state estimate at time k obtained by Kalman filtering, which is achieved by using the values ​​of the state variables at time k-1 and the observed values ​​through Kalman filtering. It is using X k / k F represents the state matrix at the current moment, based on the obtained state prediction values; the state is updated using the prediction values ​​and the new information:

[0070]

[0071] Among them, K k+1 Let ε represent the filter gain matrix. k+1 Y represents the measurement prediction error. k+1 Let H represent the external observations, and let H represent the observation matrix. The values ​​of the state variables at the next time step are calculated based on the Kalman gain obtained at the current time step and the observed values ​​of the external variables. That is, the state at that moment is updated, and this process is repeated until the updated state variable converges to the estimated value.

[0072] Balance the confidence of predicted values ​​with observed values ​​using the gain matrix:

[0073] K k+1 =P k+1 / k H T HP k+1 / k H T +R] -1 ,

[0074] Among them, P k+1 / k H represents the one-step prediction variance matrix of the state. T R represents the transpose of the observation matrix, and R represents the measurement noise covariance matrix.

[0075] The optimal estimate for the Kalman filter is calculated using innovation and gain; therefore, calculating the Kalman filter gain is crucial in the filtering process. The magnitude of the Kalman filter gain depends on multiple factors, including external observations and noise effects. During the gain matrix calculation, when R is large, K... k+1 The value of K is relatively small, meaning that the observed values ​​account for a small proportion of the estimated state values, and the predicted values ​​account for a small proportion as well. When the covariance value is relatively large, K... k+1 The larger the value, the greater the dependence of the estimated value on the observed value and the greater its proportion in the calculation. Calculating covariance prediction:

[0076]

[0077] Among them, P k / k The variance matrix represents the state estimation error, and F represents the state transition matrix. Γ represents the transpose of the state transition matrix, Γ represents the system noise distribution matrix, and Q represents the process noise covariance matrix.

[0078] The observation information Y obtained from the external measurement system was improved by one-step state prediction, state update, and filter gain. k+1 Let's consider the state variable X at time k. k / k Update the state variables at time k+1 The basic steps. During the state update process, each iteration also requires a covariance update operation:

[0079] P k+1 / k+1 =[I n -K k+1 H]P k+1 / k ,

[0080] Among them, P k+1 / k+1 I represents the value of the covariance at the next time step. n Let represent the n×1 identity matrix.

[0081] The Kalman filter calculation process includes state update and covariance update. Through real-time iterative update of the covariance, the Kalman filter gain matrix is ​​calculated in real time, and its size also affects the gain.

[0082] S3: Construct a fuzzy adaptive model based on error parameters, generate observation noise correction factors through fuzzy logic rules, and correct the observation noise covariance matrix in the extended Kalman filter model in real time;

[0083] The construction of the fuzzy adaptive model includes: calculating the ratio of the theoretical variance of attitude velocity to the trace of the measured variance as the first input variable, and calculating the ratio of the theoretical variance of position to the trace of the measured variance as the second input variable; establishing fuzzy logic rules with the first and second input variables as inputs and the observation noise correction factor as the output; and adjusting the observation noise covariance matrix in real time through the observation noise correction factor.

[0084] Specifically, let the first input variable (G1) be the ratio of the theoretical variance to the measured variance trace of attitude velocity, and the second input variable (G2) be the ratio of the theoretical variance to the measured variance trace of position. Using these two variables as input terms in the fuzzy logic, the correction factor q for attitude velocity observation noise is... ψV Correction factor q for position observation noise pFor the output term, when G1,G2≈1, the input fuzzy set M is q. ψV q p An invariant fuzzy set; when G1, G2 > 1, the input fuzzy set MA and the output q ψV q p The fuzzy set is increased; when G1, G2 < 1, the input fuzzy set is MI, and the output is q. ψV q p Reduced fuzzy set.

[0085] The basic input and output of fuzzy sets are based on a unit of 1, which represents the ideal state under this input condition, requiring no correction. However, when the input deviates from this ideal state, the magnitude of the correction factor needs to be adjusted appropriately to change the output, making the output closer and closer to the state we need.

[0086] Fuzzy inference is also an input-output system. The basic representation of fuzzy rules derived from fuzzy inference is as follows:

[0087] If x1 is and x2 is ...and x n is then yis B,

[0088] Where x1 represents the first input variable, x1 represents the first input fuzzy set, and x2 represents the second input variable. Let x represent the second input fuzzy set. n This represents the nth input variable. Let represent the nth input fuzzy set, y represent the output variable, and B represent the output fuzzy set.

[0089] The fuzzy logic rules include: when 1.5 < input variable ≤ 2, the output correction factor increases; when 0.5 < input variable ≤ 1.5, the output correction factor remains unchanged; when 0 < input variable ≤ 0.5, the output correction factor decreases.

[0090] Specifically, G1 and G2 use the same input G, q ψV q p Using the same output q, the fuzzy inference rules are established as follows:

[0091] if 1.5 <G≤2,then′G∈MA′,

[0092] if 0.5 <G≤1.5,then′G∈M′,

[0093] if0 <G≤0.5,then′G∈MI′,

[0094] Where G represents the ratio of theoretical variance to measured variance trace, and q represents the correction factor for observation noise.

[0095] The correction factor is obtained as follows:

[0096] If G ∈ MA, then q = 1 + 0.5 N.

[0097] if′G∈M′,q=1+0.3·(1-N)·sign(G-1),

[0098] If G ∈ MI, q = 1 - 0.5 N,

[0099] The N in the correction factor is the membership fuzziness obtained from the input membership function, and its value ranges from [0,1]. The closer N is to 1, the higher the dependence of the input element on the fuzzy set.

[0100] The technical solutions described in the embodiments of this application above have at least the following technical effects or advantages:

[0101] This application employs a collaborative mechanism that integrates fuzzy adaptive rules and extended Kalman filtering. By constructing a comprehensive error model covering inertial devices and the navigation system, it achieves high-precision online estimation and compensation for key errors such as zero bias, scaling factor error, and random walk. By dynamically correcting the observation noise covariance using fuzzy logic, it effectively overcomes the filtering divergence problem caused by sensor zero bias instability and external interference in traditional methods. This significantly improves the accuracy of navigation parameter calculation, the system's anti-interference adaptability, and the filtering convergence speed and stability, thereby ensuring the reliability and accuracy of the inertial navigation system in complex application scenarios over the long term.

[0102] Example 2: In Example 1, high-precision online estimation of key error parameters of the inertial navigation system was achieved, but there was still a problem of insufficient closed-loop connection between compensation operation and navigation solution. This example further supplements the content of Example 1.

[0103] S4: The extended Kalman filter model is used to estimate and compensate the error parameters of the inertial navigation system online, and the compensation parameters including zero bias and scaling factor are output in real time. The original data is corrected based on the compensation parameters and injected into the navigation solution. The fuzzy adaptive model is updated according to the compensated error data.

[0104] The estimation and compensation include: performing optimal estimation of state variables through time update and measurement update loops of the extended Kalman filter model; using the estimated error parameters as compensation quantities; correcting the original output data using a data compensation algorithm to generate compensated inertial data; and injecting the compensated inertial data into the inertial navigation solution algorithm for navigation solution to output corrected attitude, velocity, and position information.

[0105] Specifically, the raw output of the inertial device is corrected in real time, including compensation for the gyroscope angular rate measurement values:

[0106]

[0107] in, The measured gyroscope angular rate is the compensated value. ε is the measured angular rate value of the gyroscope's raw output. b To estimate the obtained gyroscope zero bias error, δK G Let I be the gyroscope scaling factor error matrix, and let I be a 3×3 identity matrix. (I+δK) G ) -1 This is the inverse of the scaling factor error compensation matrix, used to eliminate the effects of scaling factor errors.

[0108] Compensation for accelerometer specific force measurements:

[0109]

[0110] in, The compensated accelerometer specific force measurement value, To estimate the obtained accelerometer zero bias error, δK A Let (I+δK) be the accelerometer scale factor error matrix. A ) -1 This is the inverse of the accelerometer scaling factor error compensation matrix.

[0111] The compensated inertial data and Inject inertial navigation calculation algorithms to perform navigation calculations;

[0112] Attitude calculation is performed using the quaternion method:

[0113]

[0114] Where, q n+1 Let q be the attitude quaternion for the next moment. n Δq() is the attitude quaternion (4×1 vector) at the current moment, representing the rotation relationship from the vehicle coordinate system to the navigation coordinate system, Δt is the angular increment to quaternion transformation function, and Δt is the sampling time interval.

[0115] Velocity calculation under the navigation system:

[0116]

[0117] in, The velocity vector in the navigation frame at the current moment. Let this be the velocity vector in the navigation frame at the next moment. Let be the attitude matrix. Let n be the component of the Earth's rotational angular rate in the n-system. For the angular velocity component in the n-frame caused by the carrier's motion around the Earth's surface, g n Let be the component of gravitational acceleration in the n-system.

[0118] Location updated using latitude, longitude, and altitude:

[0119]

[0120] h k+1 =h k +V U Δt,

[0121] Among them, L k L is the current latitude. k+1 Let λ be the latitude at the next moment. k λ represents the current longitude. k+1 h represents the longitude at the next moment. k h represents the height. k+1 For the altitude at the next moment, R M R is the radius of curvature of the meridian. N Let be the radius of curvature of the y-axis.

[0122] Based on the compensated navigation solution, the ratios G1 and G2 of the traces of the theoretical variance and the measured variance of the attitude velocity are recalculated, and the fuzzy adaptive model is updated based on the following rules:

[0123] if 1.5 <G≤2,then q new =q old ·(1+0.5·N),

[0124] if 0.5 <G≤1.5,then q new =q old ,

[0125] if0 <G≤0.5,then q new =q old ·(1-0.5·N),

[0126] Where, q new q is the correction factor for the observed noise after compensation. old This is the correction factor for the original observation noise.

[0127] The technical solutions described in the embodiments of this application above have at least the following technical effects or advantages:

[0128] This application employs a collaborative mechanism that integrates fuzzy adaptive rules and extended Kalman filtering. By constructing a comprehensive error model covering inertial devices and the navigation system, it establishes a complete technical chain from online estimation of error parameters and real-time data compensation to navigation solution. This enables high-precision online real-time estimation and compensation of key errors such as zero bias, scaling factor error, and random walk. It effectively overcomes the estimation bias and filter divergence problems caused by the instability of inertial sensor zero bias and external environmental interference in traditional methods. This allows for autonomous adaptation and intelligent suppression of inertial navigation system errors in complex application scenarios. It significantly improves the accuracy of navigation parameter solution, the system's anti-interference adaptability, and the convergence speed and stability of filter calculation, thereby ensuring the reliability and accuracy of the inertial navigation system in long-term operation.

[0129] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An online self-calibration filtering method for inertial navigation, characterized in that, include: S1: Acquire multimodal data and establish mathematical models of inertial device errors, including zero bias error, scaling factor error, random error and random walk, as well as mathematical models of inertial navigation system errors, including velocity error, position error and attitude error. S2: Based on the mathematical model of inertial device error and the mathematical model of inertial navigation system error, an extended Kalman filter model is constructed to recursively estimate the state variables and obtain the error parameters of inertial device and navigation system; S3: Construct a fuzzy adaptive model based on error parameters, generate observation noise correction factors through fuzzy logic rules, and correct the observation noise covariance matrix in the extended Kalman filter model in real time; S4: The extended Kalman filter model is used to estimate and compensate the error parameters of the inertial navigation system online, and the compensation parameters including the zero bias and scaling factor are output in real time. The original data is corrected based on the compensation parameters and injected into the navigation solution. The fuzzy adaptive model is then updated based on the compensated error data.

2. The online self-calibration filtering method for inertial navigation as described in claim 1, characterized in that, The construction of the extended Kalman filter model includes: a state equation with inertial device error and navigation system error as state variables, and an observation equation with velocity and position information provided by external observation equipment as observation quantities; Discretize the state equation and the observation equation to form an extended Kalman filter model for recursive calculation.

3. The online self-calibration filtering method for inertial navigation as described in claim 1, characterized in that, The state variables include: the zero bias error of the gyroscope and accelerometer, the scale factor error, the installation error angle, the random walk error, as well as the velocity error, position error, and attitude error.

4. The online self-calibration filtering method for inertial navigation as described in claim 1, characterized in that, The error parameters include: static error parameters of inertial devices, dynamic error state of navigation system, and time-varying random error parameters; The static error parameters of the inertial devices include gyroscope zero bias error, accelerometer zero bias error, scale factor error, and installation error angle; The dynamic error states of the navigation system include velocity error, position error, and attitude error; The time-varying random error parameters include random walk error.

5. The online self-calibration filtering method for inertial navigation as described in claim 3, characterized in that, The random walk error includes: modeling the random walk error and defining it as the drift amount accumulated over time; subtracting the drift amount from the attitude error equation as a compensation term; and directly offsetting the negative impact of gyroscope zero-bias drift on attitude angle calculation through the compensation term.

6. The online self-calibration filtering method for inertial navigation as described in claim 1, characterized in that, The construction of the fuzzy adaptive model includes: calculating the ratio of the theoretical variance of attitude velocity to the trace of the measured variance as the first input variable, and calculating the ratio of the theoretical variance of position to the trace of the measured variance as the second input variable; establishing fuzzy logic rules with the first and second input variables as inputs and the observation noise correction factor as the output; and adjusting the observation noise covariance matrix in real time through the observation noise correction factor.

7. The online self-calibration filtering method for inertial navigation as described in claim 5, characterized in that, The fuzzy logic rules include: when 1.5 < input variable ≤ 2, the output correction factor increases; when 0.5 < input variable ≤ 1.5, the output correction factor remains unchanged; when 0 < input variable ≤ 0.5, the output correction factor decreases.

8. The online self-calibration filtering method for inertial navigation as described in claim 1, characterized in that, The estimation and compensation include: performing optimal estimation of state variables through time update and measurement update loops of the extended Kalman filter model; using the estimated error parameters as compensation quantities; correcting the original output data using a data compensation algorithm to generate compensated inertial data; and injecting the compensated inertial data into the inertial navigation solution algorithm for navigation solution to output corrected attitude, velocity, and position information.

Images