An imu fault adaptive estimation method
By constructing a nonlinear system filtering model and using Kalman filtering decoupling technology, adaptive estimation of IMU faults was achieved, solving the fault diagnosis problem of MEMS-IMU under environmental factors and improving the accuracy and reliability of unmanned aerial vehicle navigation systems.
Patent Information
- Application Number
- CN202310253627.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-16
- Publication Date
- 2026-02-06
- Estimated Expiration
- 2043-03-16
AI Technical Summary
In existing technologies, the accuracy and reliability of MEMS-IMUs are greatly affected by the environment. Factors such as high temperature, vibration, and electromagnetic interference can easily cause IMU failures, leading to the failure of the unmanned aerial vehicle navigation system. Existing fault diagnosis methods cannot accurately isolate IMU faults.
An adaptive estimation method for IMU faults is adopted. By periodically collecting IMU and GNSS output information, a nonlinear system filtering model is constructed. The Kalman filter is used to decouple the model into a reduced-order filter, and the noise covariance matrix is adaptively updated to achieve online estimation of IMU fault types and aircraft information.
It improves the estimation accuracy and algorithm robustness of gyroscope and accelerometer fault information, ensuring accurate estimation of aircraft attitude, velocity, and position information in the event of IMU failure, thereby enhancing the reliability of the navigation system.
Smart Images
Figure CN116337110B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the field of sensor fault diagnosis, and particularly relates to an IMU fault adaptive estimation method. BACKGROUND
[0002] An unmanned aerial vehicle has the advantages of small size, light weight and strong flexibility, and is widely applied to fields of military reconnaissance, power inspection, and agricultural and forestry monitoring. A reliable navigation system is a key to guaranteeing stable flight of the aerial vehicle. Currently, commonly used navigation sensors include an inertial measurement unit (IMU), a global satellite navigation system (GNSS), a barometric altimeter, and a visual sensor. The IMU is composed of a three-axis gyroscope and a three-axis accelerometer, and measures angular velocity and linear acceleration of a carrier relative to an inertial space. The IMU can obtain attitude, speed and position information of the aerial vehicle through integral recursion of itself, and is a basic but key sensor in a multi-sensor fusion navigation system.
[0003] Limited by factors of load, power consumption and size, a small unmanned aerial vehicle usually adopts a low-cost micro-electromechanical system inertial measurement unit (MEMS-IMU). The precision and reliability of the MEMS-IMU are greatly affected by the environment. High temperature, vibration and electromagnetic interference may cause faults of the IMU and lead to failure of the navigation system, which seriously affects safe and stable flight of the aerial vehicle.
[0004] To guarantee the reliability of the navigation system, it is necessary to monitor the health status of the sensor in real time. Current sensor fault diagnosis methods are mainly divided into hardware redundancy and analytical redundancy. Hardware redundancy refers to carrying multiple same sensor hardware devices, and isolating a faulty sensor by comparing and voting multiple output signals. This method is commonly used in civil aviation passenger planes and advanced military aircrafts which have high requirements for flight safety. However, for unmanned aerial vehicles, additional hardware devices increase the size, cost and power consumption, thereby bringing more flight challenges. Analytical redundancy refers to constructing a mathematical model between system parameters and sensor component parameters to generate a redundant signal. Currently, federated filtering architecture is usually used to monitor the residual error of the filter in real time to realize fault detection and isolation of the navigation sensor. This method usually takes the inertial navigation system as a reference system and assumes that the inertial measurement unit is fault-free. When the IMU fails, the sensor fault cannot be accurately isolated. SUMMARY
[0005] The purpose of the present application is to provide an IMU fault adaptive estimation method to solve the problems existing in the prior art.
[0006] To achieve the above purpose, the present application provides an IMU fault adaptive estimation method, comprising:
[0007] periodically collect output information of the IMU and output information of the GNSS, and construct a nonlinear system filtering model based on a kinematic relationship of the output information of the IMU and the output information of the GNSS;
[0008] decouple the nonlinear system filtering model into a reduced-order filter based on Kalman filtering, and simultaneously perform filtering update to obtain a state estimation value and a state estimation error covariance matrix;
[0009] perform adaptive update on the state estimation error covariance matrix based on a covariance matching technique to obtain an updated noise covariance matrix;
[0010] estimate and output a fault type of the IMU and aircraft information based on the updated noise covariance matrix.
[0011] Preferably, the output information of the IMU includes projections of angular velocity information of the aircraft relative to an inertial system output by a gyroscope on three axes of an on-board system and projections of non-gravitational acceleration of the aircraft relative to the inertial system output by an accelerometer on the three axes of the on-board system.
[0012] The output information of the GNSS includes attitude information, velocity information and position information of the aircraft in a navigation system.
[0013] Preferably, the process of constructing the nonlinear system filtering model includes:
[0014] constructing a state equation of the system based on the output information of the IMU;
[0015] processing the output information of the GNSS to generate an observation of the system, and obtaining a system measurement equation based on the observation of the system;
[0016] obtaining the nonlinear system filtering model based on the state equation of the system and the system measurement equation.
[0017] Preferably, the process of decoupling the nonlinear system filtering model into the reduced-order filter includes:
[0018] decoupling the nonlinear system filtering model into a fault-free filter and a fault filter based on optimal two-step extended Kalman filtering;
[0019] the system matrix of the fault-free filter after discretization is obtained based on a filtering sampling period, a state estimation value at k-1 time of the fault-free filter, a state transition matrix of the fault-free filter and a system noise matrix of the fault-free filter;
[0020] the discrete state equation of the fault filter is obtained based on random walk noise of unknown faults of the gyroscope and the accelerometer.
[0021] Preferably, the process of obtaining the state estimation value and the state estimation error covariance matrix comprises:
[0022] Filtering and updating the reduced order filter by using the extended Kalman filter to obtain an updated filter;
[0023] Obtaining the state estimation value and the state estimation error covariance matrix based on the updated filter.
[0024] Preferably, the process of obtaining the updated noise covariance matrix comprises:
[0025] Obtaining an actual innovation covariance matrix based on a sliding window;
[0026] Obtaining a theoretical innovation covariance matrix based on the dimension of the fault vector and the diagonal elements;
[0027] Matching the theoretical innovation covariance and the actual innovation covariance, and updating in real time based on the main diagonal elements to obtain the updated noise covariance matrix.
[0028] Preferably, the theoretical innovation covariance matrix is obtained based on the measurement matrix of the system, the radius of the earth, the filter sampling period, and the system noise matrix of the fault-free filter;
[0029] The actual innovation covariance matrix is obtained based on the sliding window size, the innovation of the system filter model, and the filter sampling period;
[0030] Preferably, the process of outputting the fault type of the IMU and the aircraft information comprises:
[0031] Coupling the state estimation value of the fault-free filter and the state estimation value of the fault filter to obtain the fault type of the IMU and the aircraft information;
[0032] The aircraft information comprises the aircraft attitude, speed, and position.
[0033] The technical effect of the present application is:
[0034] The present application provides an IMU fault adaptive estimation method, compared with the prior art, the method can adaptively update the fault random walk noise covariance matrix according to the filter innovation, so as to accurately match the dynamic characteristics of various types of faults, and improve the estimation accuracy of the gyroscope and accelerometer fault information and the robustness of the algorithm. At the same time, the present application decouples the IMU fault and the navigation state quantity, realizes accurate estimation of the aircraft attitude, speed, and position information under the condition of IMU fault, and further improves the reliability of the navigation system. BRIEF DESCRIPTION OF DRAWINGS
[0035] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The illustrations, together with their description, serve to explain the application without unduly
[0036] Figure 1 Flow chart of adaptive estimation method in embodiments of the application;
[0037] Figure 2 Gyroscope fault estimation result chart in embodiments of the application;
[0038] Figure 3 Accelerometer fault estimation result chart in embodiments of the application;
[0039] Figure 4 Attitude estimation error chart in embodiments of the application;
[0040] Figure 5 Velocity estimation error chart in embodiments of the application;
[0041] Figure 6 Position estimation error chart in embodiments of the application. DETAILED DESCRIPTION
[0042] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0043] It should be noted that the steps shown in the flow chart of the drawings can be executed in a computer system such as a group of computer executable instructions, and although the logical order is shown in the flow chart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.
[0044] Embodiment one
[0045] As shown in Figure 1 The present embodiment provides an IMU fault adaptive estimation method, which comprises:
[0046] Step 1: periodically collect the output information of IMU and GNSS at time k;
[0047] Step 2: according to the kinematic relationship between the collected sensor output information, a nonlinear system filtering model containing IMU fault is constructed;
[0048] Step 3: the system filtering model is decoupled into two parallel reduced order filters for filtering update by using optimal two-step extended Kalman filter;
[0049] Step 4: the fault random walk noise covariance matrix in step 3 is adaptively updated by using covariance matching technology;
[0050] Step 5: estimating the fault information of the gyroscope and the accelerometer at the current time and the attitude, speed and position of the aircraft online;
[0051] Step 6: repeating steps 1-5.
[0052] Further, the IMU data collected in step 1 includes: the projection of the angular velocity information of the aircraft relative to the inertial system output by the gyroscope on the three axes of the body system the projection of the non-gravitational acceleration of the aircraft relative to the inertial system output by the accelerometer on the three axes of the body system A xm ,A ym ,A zm The collected GNSS output information includes: attitude information (roll angle φ m , pitch angle θ m , heading angle ψ m ) of the aircraft in the navigation system; speed information (northward speed v nm , eastward speed v em , downward speed v dm ); position information (longitude λ m , latitude L m , height h m ).
[0053] Wherein, the related coordinate systems are defined as follows:
[0054] The inertial coordinate system (i system) is the geocentric inertial coordinate system, the origin of which is located at the center of the earth, the z axis of which is along the rotation axis of the earth, the x and y axes of which are in the equatorial plane, and which satisfies the right-handed orthogonal coordinate system; the body coordinate system (b system) is a front-right-down orthogonal coordinate system, the origin of which is located at the center of mass of the aircraft, the x, y and z axes of which respectively point to the head direction, the right direction and the downward direction of the aircraft; the navigation coordinate system (n system) is a north-east-ground orthogonal coordinate system, the origin of which is located at the center of mass of the aircraft, the x axis of which is along the tangent of the local meridian (northward), the y axis of which is along the tangent of the local latitude (eastward), and the z axis of which is along the local geographic vertical.
[0055] Further, in step 2, a nonlinear system filtering model containing IMU faults is constructed according to the kinematic relationship between the collected sensor output information, including the following steps:
[0056] Step 2.1, constructing the state equation of the system
[0057] The aircraft satisfies the following angular motion differential equation, linear motion differential equation and position differential equation:
[0058]
[0059] In the formula, φ, θ and ψ are the roll angle, pitch angle and heading angle of the aircraft respectively; v nv e v d v e v ie v
[0060] Substitute the IMU output information collected in step 1 into the above differential equations, and the continuous system state equation containing IMU faults can be obtained, described as follows:
[0061]
[0062] Wherein, each variable is defined as follows:
[0063] x = [φ θ ψ v n v e v d λ L h] T
[0064]
[0065] f = [f ax f ay f az f wx f wy f wz ] T
[0066] w = [w ax w ay w az w wx w wy w wz ] T
[0067] In the formula, x is the state quantity of the system; u m is the output information of the IMU in step 1; f ax , f ay , f az and f wx , f wy , f wz are faults of the three-axis accelerometer and the three-axis gyroscope respectively; w ax , w ay , w az and w wx , w wy , w wz are measurement noises of the three-axis accelerometer and the three-axis gyroscope respectively; G(x(t)) is the system noise matrix, calculated as follows:
[0068]
[0069]
[0070]
[0071] Step 2.2, Constructing the measurement equation of the system
[0072] Taking the aircraft attitude, velocity, position information output by GNSS in step 1 as the observation of the system, the measurement equation of the system can be obtained as follows:
[0073] y m = H k x + v
[0074] Wherein, each variable is defined as follows:
[0075] y m = [φ m θ m ψ m v nm v em v dm λ m L m h m ] T
[0076]
[0077] In the formula, y m is the output information of GNSS in step 1; v is the measurement noise of GNSS; H k is the measurement matrix of the system, H k = I 9×9 .
[0078] Further, the optimal two-step extended Kalman filter is used in step 3 to decouple the system filter model into two parallel reduced-order filters for filter updating, including the following steps:
[0079] Step 3.1, Discretization and linearization of system state equation
[0080] The two parallel reduced-order filters are respectively the fault-free filter and the fault filter. Among them, the fault-free filter ignores the fault term (f = 0) in the system state equation, and the system matrix after discretization is calculated as follows:
[0081]
[0082]
[0083] In the formula, T is the filter sampling period; is the state estimation of the fault-free filter at time k-1; is the state transition matrix of the fault-free filter; is the system noise matrix of the fault-free filter.
[0084] The state quantity of the fault filter is the fault f of the gyroscope and accelerometer, and the random fault that can occur is constructed as a random walk model. The discrete state equation of the fault filter is as follows:
[0085]
[0086] In the formula, is the random walk noise of the unknown fault of the gyroscope and accelerometer.
[0087] Step 3.2, calculate the coupling matrix of the fault-free filter and the fault filter
[0088] S k = H k U k
[0089]
[0090]
[0091] In the formula, S k is the observation matrix of the fault filter; is the compensation input of the fault-free filter; is the compensation covariance matrix of the fault-free filter; Q k-1 = E{ww T} is the system noise variance matrix; is the fault random walk noise covariance matrix; is the covariance matrix between the fault noise and the system noise; U k , V k are derived by two-step U-V variation, and the calculation formula is as follows:
[0092]
[0093]
[0094]
[0095] In the formula, is the gain matrix of the fault-free filter; is the one-step prediction error covariance matrix of the fault filter.
[0096] Step 3.3, filter update of the fault-free filter by using the extended Kalman filter
[0097] Calculate the one-step prediction value of the state variables of the fault-free filter and the one-step prediction error covariance matrix The time update equation is as follows:
[0098]
[0099]
[0100] In the formula, It is the state estimation error covariance matrix of the fault-free filter at time k-1.
[0101] Calculate the gain matrix of the fault-free filter
[0102]
[0103] Using the observations y provided by GNSS k Calculate the state estimate of the fault-free filter and state estimation error covariance matrix The measurement update equation is as follows:
[0104]
[0105]
[0106] In the formula, R k =E{vv T} is the measurement noise variance matrix.
[0107] Step 3.4: Update the fault filter using an extended Kalman filter.
[0108] Calculate the one-step prediction value of the fault filter state variables and the one-step prediction error covariance matrix The time update equation is as follows:
[0109]
[0110]
[0111] Calculate the gain matrix of the fault filter
[0112]
[0113] Will As observations of the fault filter, calculate the state estimate of the fault filter. and state estimation covariance matrix The measurement update equation is as follows:
[0114]
[0115]
[0116] Further, the fault random walk noise covariance matrix in step 3 is updated adaptively using the covariance matching technique in step 4, and the specific process is as follows:
[0117] The innovation γ of the system filtering model at time m m As follows:
[0118]
[0119] The actual innovation covariance matrix C of the system filtering model at time k is obtained using the sliding window k :
[0120]
[0121] In the formula, N is the size of the window.
[0122] The theoretical innovation covariance matrix of the system filtering model at time k is calculated as follows:
[0123]
[0124] Definition:
[0125]
[0126]
[0127] In the formula, l is the dimension of the fault vector; is the jth diagonal element of
[0128] The theoretical innovation covariance and the actual innovation covariance are matched, and the noise covariance matrix of the unknown random fault can be updated in real time based on the main diagonal elements of the following formula:
[0129]
[0130] Further, the fault information of the gyroscope and the accelerometer at the current time and the attitude, speed, and position of the aircraft are estimated online in step 5, and the specific process is as follows:
[0131] The state estimation value of the fault-free filter in step 3 and the state estimation value of the fault filter are coupled to obtain the optimal estimation of the attitude, speed, and position of the aircraft, and the calculation is as follows:
[0132]
[0133] Step 3 state estimation value of fault filter That is, the estimated gyroscope and accelerometer fault information.
[0134] Example two
[0135] Step 1, in this embodiment, the output information of the IMU and GNSS at time k is obtained through a simulation experiment period, and the specific process is as follows:
[0136] This embodiment first sets a flight path including different motion states such as rising, turning, descending, accelerating, and decelerating, which is used as a real data source for sensor simulation. On this basis, the IMU and GNSS output information is simulated. The simulation duration is 800s, and the initial attitude, speed, and position information of the aircraft is: [0° 0° 180° 0m / s 0m / s 0m / s 90.906802° 29.296007° 3600m].
[0137] The simulation parameters of the IMU and GNSS are as follows: the gyroscope drift is 0.1° / h; the accelerometer drift is 0.001m / s 2 ; the attitude measurement error of the GNSS is 0.2°; the speed measurement error is 0.1m / s; the horizontal position measurement error is 10m; and the height measurement error is 15m. The sampling frequency of the IMU and GNSS is 50Hz.
[0138] The collected information includes: the angular velocity information of the aircraft relative to the inertial system projected on the three axes of the body system output by the gyroscope The non-gravitational acceleration of the aircraft relative to the inertial system projected on the three axes of the body system output by the accelerometer A xm ,A ym ,A zm The attitude information (roll angle φ m , pitch angle θ m , heading angle ψ m ) of the aircraft in the navigation system output by the GNSS; the speed information (northward speed v nm , eastward speed v em , and downward speed v dm ); and the position information (longitude λ m , latitude L m , and height h m ).
[0139] To verify the identification effect of the technical scheme of the present application on the IMU fault, different types of faults are injected into the gyroscope and accelerometer at different time periods in this embodiment, and the fault scenarios are shown in Table 1.
[0140] Table 1
[0141]
[0142]
[0143] Step 2, according to the kinematics relationship between the collected sensor output information, a nonlinear system filtering model containing IMU fault is constructed, including the following steps:
[0144] Step 2.1, the state equation of the system is constructed
[0145] The angular motion differential equation, linear motion differential equation and position differential equation of the aircraft satisfy the following formula:
[0146]
[0147] In the formula, φ, θ, ψ are the roll angle, pitch angle and heading angle of the aircraft respectively; v n ,v e ,v d are the velocity components of the aircraft in the three axes of the navigation system; λ, L, h are the longitude, latitude and altitude of the aircraft respectively; R e is the earth radius; w ie is the earth rotation angular velocity.
[0148] Substitute the IMU output information collected in step 1 into the above differential equation, then the continuous system state equation containing IMU fault can be obtained, which is described as follows:
[0149]
[0150] Wherein, each variable is defined as follows:
[0151] x=[φ θ ψ v n v e v d λ L h] T
[0152]
[0153] f=[f ax f ay f az f wx f wy f wz ] T
[0154] w=[w ax w ay w az w wx w wy wwz ] T
[0155] In the formula, x is the state variable of the system; u m This refers to the IMU output information in step 1; f ax ,f ay ,f az and f wx ,f wy ,f wz The faults are respectively the three-axis accelerometer and the three-axis gyroscope; w ax ,w ay ,w az and w wx ,w wy ,w wz These are the measurement noises from the triaxial accelerometer and triaxial gyroscope, respectively; G(x(t)) is the system noise matrix, calculated as follows:
[0156]
[0157]
[0158]
[0159] Step 2.2, Construct the measurement equations of the system.
[0160] Using the aircraft attitude, velocity, and position information output by GNSS in step 1 as the system's observations, the system's measurement equations can be obtained as follows:
[0161] y m =H k x+v
[0162] The variables are defined as follows:
[0163] y m =[φ m θ m ψ m v nm v em v dm λ m L m h m ] T
[0164]
[0165] In the formula, y m This refers to the GNSS output information in step 1; v is the GNSS measurement noise; H k It is the system's measurement matrix, H k =I9×9 .
[0166] Step 3, the system filter model is decoupled into two parallel reduced-order filters for filter update by using the optimal two-step extended Kalman filter, including the following steps:
[0167] Step 3.1, discretization and linearization of the system state equation
[0168] The two parallel reduced-order filters are respectively the fault-free filter and the fault filter. Among them, the fault-free filter ignores the fault term (f = 0) in the system state equation, and the discretized system matrix is calculated as follows:
[0169]
[0170]
[0171] In the formula, T is the filter sampling period; is the state estimation value of the fault-free filter at time k-1; is the state transition matrix of the fault-free filter; is the system noise matrix of the fault-free filter.
[0172] The state quantity of the fault filter is the fault f of the gyroscope and accelerometer, and the random fault that may occur is constructed as a random walk model, and the discretized state equation of the fault filter is as follows:
[0173]
[0174] In the formula, is the random walk noise of the unknown fault of the gyroscope and accelerometer.
[0175] Step 3.2, calculation of the coupling matrix of the fault-free filter and the fault filter
[0176] S k =H k U k
[0177]
[0178]
[0179] In the formula, S k is the observation matrix of the fault filter; is the compensation input of the fault-free filter; is the compensation covariance matrix of the fault-free filter; Q k-1 =E{ww T} is the system noise variance matrix; is the fault random walk noise covariance matrix; is the covariance matrix between fault noise and system noise; U k , V k is derived from two-step U-V variation, and the calculation formula is as follows:
[0180]
[0181]
[0182]
[0183] wherein, is the gain matrix of the fault-free filter; is the one-step prediction error covariance matrix of the fault filter.
[0184] In this embodiment, the initial parameters are set as follows: Take 0; Q k Corresponding to the measurement noise set by the accelerometer and the gyroscope.
[0185] Step 3.3, using extended Kalman filter to perform filtering update of the fault-free filter
[0186] Calculate the one-step prediction value of the state quantity of the fault-free filter and the one-step prediction error covariance matrix The time update equation is as follows:
[0187]
[0188]
[0189] wherein, is the state estimation error covariance matrix of the fault-free filter at k-1.
[0190] Calculate the gain matrix of the fault-free filter
[0191]
[0192] Using the observation value y provided by GNSS k Calculate the state estimation value of the fault-free filter and the state estimation error covariance matrix The measurement update equation is as follows:
[0193]
[0194]
[0195] where R k = E{vv T} is the measurement noise variance matrix.
[0196] In this embodiment, the initial state value of the fault filter is The initial state estimation error covariance matrix of the fault filter is R k corresponding to the measurement noise set by the GNSS.
[0197] Step 3.4, filtering update of the fault filter using the extended Kalman filter
[0198] The one-step prediction value of the state quantity of the fault filter is calculated and the one-step prediction error covariance matrix is calculated The time update equation is as follows:
[0199]
[0200]
[0201] The gain matrix of the fault filter is calculated
[0202]
[0203] The is taken as the observation value of the fault filter, the state estimation value of the fault filter is calculated and the state estimation covariance matrix is calculated The measurement update equation is as follows:
[0204]
[0205]
[0206] In this embodiment, the initial fault estimation value of the fault filter is set to 0.
[0207] Step 4, adaptive update of the fault random walk noise covariance matrix in step 3 using the covariance matching technique, the specific process is as follows:
[0208] The innovation γ of the system filtering model at time m is as follows: m
[0209]
[0210] The actual innovation covariance matrix C of the system filtering model at time k is obtained using the sliding window: k
[0211]
[0212] In the formula, N is the size of the window, and in the embodiment, the window size N = 10.
[0213] The theoretical innovation covariance matrix of the system filtering model at the k th moment is calculated as follows:
[0214]
[0215] Definition:
[0216]
[0217]
[0218] In the formula, l is the dimension of the fault vector; is the j th diagonal element of .
[0219] The theoretical innovation covariance and the actual innovation covariance are matched, and the noise covariance matrix of the unknown random fault can be updated in real time based on the main diagonal elements of the following formula:
[0220]
[0221] Step 5, online estimation of the fault information of the gyroscope and the accelerometer at the current moment and the attitude, speed and position of the aircraft, the specific process is as follows:
[0222] The state estimation value of the fault-free filter in step 3 and the state estimation value of the fault filter are coupled to obtain the optimal estimation of the attitude, speed and position of the aircraft, which is calculated as follows:
[0223]
[0224] The state estimation value of the fault filter in step 3 is the estimated fault information of the gyroscope and the accelerometer.
[0225] Embodiment three
[0226] Under this embodiment, the fault estimation result of the IMU is as shown in Figure 2 ; the estimation error of the aircraft navigation information is as shown in Figure 3 . 100 Monte Carlo simulation tests are performed, and the root mean square error between the navigation information estimation value and the true navigation information when the IMU fails is counted, and the experimental results are shown in Table 2.
[0227] Table 2
[0228]
[0229] From the above experimental results, it can be seen that the application provides an IMU fault adaptive estimation method, which can adaptively update the fault random walk noise covariance matrix, thereby accurately following the dynamic change characteristics of various unknown faults that may occur in the gyroscope and accelerometer, and accurately estimating the size of the IMU bias, drift and oscillation fault. At the same time, the application decouples the IMU fault and the navigation information, and can still maintain high accuracy of the aircraft attitude, speed and position estimation when the IMU fails, which is of great significance to improve the reliability of the navigation system.
[0230] The above is only the preferred specific embodiment of the application, but the protection scope of the application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the application, which should be covered within the protection scope of the application. Therefore, the protection scope of the application should be subject to the protection scope of the claims.
Claims
1. An IMU fault adaptive estimation method, characterized in that, The method comprises the following steps: periodically collecting output information of the IMU and output information of the GNSS, and constructing a nonlinear system filtering model based on a kinematic relationship between the output information of the IMU and the output information of the GNSS; decoupling the nonlinear system filtering model into a reduced-order filter based on Kalman filtering, and simultaneously performing filtering update to obtain a state estimation value and a state estimation error covariance matrix; performing adaptive update on the state estimation error covariance matrix based on a covariance matching technique to obtain an updated noise covariance matrix; online estimating and outputting a fault type of the IMU and aircraft information based on the updated noise covariance matrix; the process of constructing the nonlinear system filtering model comprises: constructing a state equation of the system based on the output information of the IMU; processing the output information of the GNSS to generate a system observation, and obtaining a system measurement equation based on the system observation; obtaining the nonlinear system filtering model based on the state equation of the system and the system measurement equation; the process of decoupling the nonlinear system filtering model into the reduced-order filter comprises: decoupling the nonlinear system filtering model into a fault-free filter and a fault filter based on optimal two-step extended Kalman filtering; the system matrix of the fault-free filter after discretization is obtained based on a filtering sampling period, a state estimation value at k-1 time of the fault-free filter, a state transition matrix of the fault-free filter, and a system noise matrix of the fault-free filter; the discrete state equation of the fault filter is obtained based on random walk noise of unknown faults of the gyroscope and the accelerometer; the process of obtaining the updated noise covariance matrix specifically comprises: obtaining an actual innovation covariance matrix based on a sliding window; obtaining a theoretical innovation covariance matrix based on dimensions and diagonal elements of a fault vector; matching the theoretical innovation covariance and the actual innovation covariance, and performing real-time update based on main diagonal elements to obtain the updated noise covariance matrix.
2. The IMU fault adaptive estimation method according to claim 1, wherein the output information of the IMU comprises projections of angular velocity information of the aircraft relative to an inertial system output by the gyroscope on three axes of an on-board system, and projections of non-gravitational acceleration of the aircraft relative to the inertial system output by the accelerometer on three axes of the on-board system; the output information of the GNSS comprises attitude information, speed information, and position information of the aircraft in a navigation system.
3. The IMU fault adaptive estimation method of claim 1, wherein, the process of obtaining the state estimation value and the state estimation error covariance matrix comprises: performing filtering update on the reduced-order filter by using extended Kalman filtering to obtain an updated filter; obtaining the state estimation value and the state estimation error covariance matrix based on the updated filter.
4. The IMU fault adaptive estimation method according to claim 1, wherein the theoretical innovation covariance matrix is obtained based on a measurement matrix of the system, an earth radius, a filtering sampling period, and a system noise matrix of the fault-free filter; the actual innovation covariance matrix is obtained based on a sliding window size, innovation of a system filtering model, and a filtering sampling period.
5. The IMU fault adaptive estimation method of claim 1, wherein, the process of outputting the fault type of the IMU and the aircraft information comprises: The state estimation value of the fault-free filter and the state estimation value of the fault filter are coupled to obtain fault type of the IMU and aircraft information; The aircraft information includes aircraft attitude, speed, and position.