A fixed-wing unmanned aerial vehicle falling body motion horizontal attitude estimation method

Abstract

The application discloses a fixed-wing unmanned aerial vehicle falling body motion horizontal attitude estimation method, which comprises the following steps: step one, pre-processing the gyroscope and accelerometer data, converting the sensor data to the front-right-down body coordinate system and normalizing; step two, preliminarily judging the belly orientation according to the skyward accelerometer; step three, after starting the engine, using a maneuvering strategy to estimate the preliminary constant zero offset of the gyroscope and the accelerometer; and step four, according to the gyroscope and accelerometer data, using an improved adaptive extended Kalman filtering method based on quaternions to fuse the sensor data, further estimating the zero offset of the gyroscope and the accelerometer, and simultaneously obtaining the real-time estimation of the horizontal attitude angle. The application overcomes the shortcoming of the traditional transfer alignment method which has high requirements for the delivery carrier, can determine the opening time of the wings and the starting time of the engine during the delivery process of the fixed-wing unmanned aerial vehicle, and realizes the estimation task of the horizontal attitude and the online estimation of the sensor error.

Description

Technical Field

[0001] This invention belongs to the field of inertial navigation and sensor data fusion, specifically a method for estimating the horizontal attitude of a fixed-wing unmanned aerial vehicle in freefall motion. Background Technology

[0002] Acquiring the attitude of the carrier is a crucial prerequisite for manipulating and controlling the carrier and realizing other functions. In recent years, microelectromechanical systems (MEMS) sensors have developed rapidly, gaining widespread use due to their advantages of low cost, small size, and flexible, unrestricted application. Fixed-wing unmanned aerial vehicles (UAVs) are advanced UAV systems, characterized by high maneuverability and rapid response, making them suitable for attacking various moving targets and possessing significant military value. Estimating the horizontal attitude (roll and pitch angles) of a fixed-wing UAV and determining the timing of wing deployment and engine start-up are key research focuses and challenges in controlling fixed-wing UAVs. Traditional initial alignment methods place high demands on the delivery carrier and are susceptible to satellite rejection during fire strikes. Therefore, this patent delves into the research of using a fixed-wing UAV equipped with a six-DOF MEMS inertial sensor (including a 3-axis gyroscope and a 3-axis accelerometer) and an airspeed indicator to select the timing of wing deployment and engine start-up and to estimate the horizontal attitude during freefall. Summary of the Invention

[0003] To address the technical problems mentioned in the background section, this invention proposes a method for estimating the horizontal attitude of a fixed-wing unmanned aerial vehicle during free fall.

[0004] To achieve the above-mentioned technical objectives, the technical solution of the present invention is as follows:

[0005] A method for estimating the horizontal attitude of a fixed-wing unmanned aerial vehicle (UAV) during freefall includes the following steps:

[0006] (1) After the UAV has been in free fall for a certain period of time, it acquires data from the airborne MEMS gyroscope, MEMS accelerometer and airspeed indicator. The gyroscope and accelerometer data are preprocessed and the sensor data are converted to the front right lower body coordinate system and normalized.

[0007] (2) Based on the accelerometer, the orientation of the fuselage is initially determined. If the fuselage is facing the ground, the UAV can directly open its wings and start the engine. Otherwise, the UAV can adjust its attitude angle by the flight control and then open its wings and start the engine.

[0008] (3) After starting the engine, the initial constant zero bias of the gyroscope and accelerometer is estimated using the maneuver strategy;

[0009] (4) Based on the gyroscope and accelerometer data, an improved adaptive extended Kalman filtering method based on quaternions is used to fuse the sensor data, further estimate the biases of the gyroscope and accelerometer, and obtain a real-time estimate of the horizontal attitude angle at the same time.

[0010] Further, in step (1), the north-east-down geographic coordinate system is selected as the navigation coordinate system, and the gyroscope and accelerometer data are correspondingly converted to the front-right-down body coordinate system and normalized. The formula for obtaining the forward drag according to the airspeed indicator is as follows: where ρ is the air density, v is the speed of the vehicle relative to the air, obtained from the airspeed indicator, C is the air drag coefficient, and A is the cross-sectional area of the UAV.

[0011] The formula for obtaining the forward acceleration caused by the forward drag is as follows: where m is the mass of the UAV, and taking a m as the compensation formula for the forward acceleration is as follows: a x补 = a x - a m , where a x补 is the compensated forward accelerometer data, and a x is the forward accelerometer data before compensation.

[0012] Further, in step (2), the belly orientation is judged according to the vertical accelerometer a z . The specific process is as follows: Set the vertical accelerometer threshold a (0.6g < a < 0.8g). If a z is negative and its absolute value is greater than a, it can be judged that the belly is facing down and the attitude is not much different from the horizontal, then the wings can be opened and the engine can be started to enter the next stage. If a z is positive and its absolute value is greater than a, then after rotating 180° through the flight control, the wings are opened and the engine is started to enter the next stage. If the absolute value of a z is less than a, then it can be rotated 90° and then the foregoing judgment is made again.

[0013] Further, in step (3), the method for determining that the UAV is approaching horizontal using the maneuvering strategy is as follows: Let the attitude at the start of the engine be P1. After starting the engine, first let the UAV rotate a certain angle around the forward axis, i.e., roll, and then the attitude is P2. Detect the output of the vertical accelerometer a z during the time from P1 to P2. When a z is the maximum value, the attitude is P3. After making the UAV return to the attitude P3 through the flight control, let the UAV rotate a certain angle around the lateral axis, i.e., pitch, and then the attitude is P4. Detect the output of the vertical accelerometer a z during the time from P3 to P4. When a z is the maximum value, the attitude is P5. After making the UAV return to the attitude P5 through the flight control, it is considered that the UAV attitude is close to horizontal at this time.

[0014] Furthermore, in step (3), the method for estimating the constant zero bias of the gyroscope using the maneuvering strategy is as follows: Based on the gyroscope data during the time period from position P3 to P2 and back to P3, since the angular increment error of the two processes is the angular increment generated by the constant zero bias throughout the process, the constant zero bias of the gyroscope forward axis is estimated based on this angular increment, as shown in the following formula: Where ε bx Let ω1 be the constant zero bias of the forward axis of the gyroscope, ω2 be the total data of the forward axis gyroscope during the time interval from P3 to P2, and T be the sampling period. Similarly, the constant zero bias of the lateral axis of the gyroscope can be estimated based on the gyroscope data during the time interval from P5 to P4 and back to P5.

[0015] Furthermore, in step (3), the method for estimating the accelerometer constant zero bias using the maneuver strategy is as follows: The formula for calculating the roll angle at time P2 based on the gyroscope data is as follows: γ P2 =Σ((ω1+ε bx )T), where γ P2 The roll angle at time P2 is obtained from the gyroscope. The formula for calculating the roll angle at time P2 using the accelerometer is as follows: Where γ P2 'The roll angle at time P2 is obtained from the accelerometer, a' y a z These are the accelerometer's lateral and longitudinal axis data, respectively. To achieve zero bias for the accelerometer, the formula for zero bias can be derived from the equality of roll angles as follows:

[0016] Furthermore, in step (4), the specific process is as follows:

[0017] (401) Construct error models for MEMS gyroscopes and accelerometers, selecting quaternions q0, q1, q2, and q3, and the three-axis constant zero bias error ε of the gyroscope. bx ε by ε bz Three-axis random walk error ε rx ε ry ε rz Accelerometer triaxial constant zero bias error A total of 13 state variables;

[0018] (402) Determine the initial values ​​of each state quantity, construct the nonlinear continuous state equation and discretize and discretize the measurement equation;

[0019] (403) Begin the extended Kalman filter process;

[0020] (404) A dual-input single-output fuzzy controller is used to achieve adaptive adjustment of the parameters of the system measurement variance matrix R;

[0021] (405) Output the quaternion after EKF update and convert it into horizontal attitude roll and pitch angles.

[0022] Furthermore, in step (401), the error models for the gyroscope and accelerometer are established as follows:

[0023]

[0024]

[0025] Where, ω x ω y ω z It is the three-axis measurement value of the gyroscope, ω tx ω ty ω tz It is the true value of the gyroscope's three axes, ε bx ε by ε bz It is a gyroscope with constant zero bias across three axes, η ωx η ωy η ωz It is three-axis Gaussian white noise from the gyroscope. x a y a z It is a triaxial measurement value from the accelerometer. tx a ty a tz These are the true values ​​of the accelerometer across all three axes. It is the triaxial constant zero bias of the accelerometer, η ax η ay η az It is three-axis Gaussian white noise from the accelerometer.

[0026] Furthermore, in step (402), the specific method for constructing the state equation and measurement equation is as follows: based on the quaternion differential equation: And the mathematical model formula for noise: The continuous nonlinear state equations are constructed as follows: Find the Jacobian matrix for F(x(t), t): in Therefore, the state equation is as follows: The discretization formula is as follows: Where T is the iteration period. The discretized state equation is as follows: X k+1 =F k X k +Gk W k W k The system white noise at time k is the 6-dimensional value. The triaxial accelerometer measurement value a is selected. k =[a x补 a y a z ] T For measurement, according to

[0027] in For the attitude transformation matrix between the navigation coordinate system and the body coordinate system, Find the Jacobian matrix:

[0028] The measurement equations were then constructed as follows:

[0029] a k =H k X k +V k V k The 3D measurement noise at time k is given.

[0030] Furthermore, in step (403), the state one-step prediction formula is as follows: Where X k|k-1 To predict the state matrix in one step, Let X be the state transition matrix. k-1 Let be the state matrix at time k-1. The formula for the mean square error of the one-step prediction is as follows: Where P k|k-1 For the one-step prediction mean square error matrix, P k-1 Let G be the mean square error matrix at time k-1. k Let Q be the system noise matrix. k-1 Let K be the variance matrix of the system noise at time k-1. The filter gain formula is as follows: K k =P k|k-1 H k T (H k P k|k-1 H k T +R k ) -1 , where R k Let be the system measurement variance matrix at time k. The state estimation formula is as follows: The formula for estimating the mean square error is as follows: P k|k =(IK k H k )P k|k-1 , where I is the identity matrix.

[0031] Furthermore, in step (404), the motion acceleration error is utilized. and its forward difference a s As two inputs to the fuzzy controller, the measurement noise variance is used as the output. Both the input and output are divided into five fuzzy subsets: {PB, PS, ZO, NS, NB}, where PB represents very small, PS represents small, ZO represents medium, NS represents large, and NB represents very large. The membership function is a uniformly distributed triangular function, and the fuzzy rule is a. s a s The larger the value, the larger the measurement noise variance, and the Centroid algorithm is used to achieve defuzzification.

[0032] Furthermore, in step (405), the formulas for obtaining the roll angle γ and pitch angle θ from the updated quaternions are as follows:

[0033] The beneficial effects of adopting the above technical solution are as follows:

[0034] (1) The present invention adopts a fixed-wing UAV maneuver strategy to achieve coarse alignment, which overcomes the shortcomings of traditional transfer alignment methods that have high requirements for UAV carriers. It can be independent of the launch platform and carrier after launch, thus expanding the application scope and adaptability.

[0035] (2) The present invention can fully consider the common errors contained in MEMS sensors and make estimations. The error values ​​can be compared with the parameter values ​​of the selected sensor to verify the accuracy of the model. At the same time, the parameters can be adaptively adjusted, thereby effectively improving the attitude estimation accuracy.

[0036] (3) The method implemented by the present invention has low computational load and high accuracy. It can determine the time of opening the wings and starting the engine during the deployment of fixed-wing UAVs and realize the task of estimating the horizontal attitude. It has broad market prospects and application value. Attached Figure Description

[0037] Figure 1 This is a basic flowchart of the present invention; Detailed Implementation

[0038] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0039] This invention designs a method for estimating the horizontal attitude of a fixed-wing unmanned aerial vehicle (UAV) during freefall motion, such as... Figure 1 As shown, the steps are as follows:

[0040] Step 1: After the drone freely falls for a certain period of time, obtain the data of the onboard MEMS gyroscope, MEMS accelerometer, and airspeed indicator. Preprocess the gyroscope and accelerometer data, and convert the sensor data to the front-right-down body coordinate system and normalize it;

[0041] Step 2: Initially judge the belly orientation based on the vertical accelerometer. If the belly is facing downwards, the drone can directly open the wings and start the engine. Otherwise, adjust a certain attitude angle through the drone flight control and then open the wings and start the engine;

[0042] Step 3: After starting the engine, use a maneuvering strategy to estimate the initial constant zero biases of the gyroscope and accelerometer;

[0043] Step 4: According to the gyroscope and accelerometer data, use an improved adaptive extended Kalman filtering method based on quaternions to fuse the sensor data, further estimate the zero biases of the gyroscope and accelerometer, and simultaneously obtain the real-time estimation of the horizontal attitude angle.

[0044] In this embodiment, the specific process of preprocessing the sensor data in Step 1 above is as follows:

[0045] In Step (1), select the north-east-down geographic coordinate system as the navigation coordinate system, and correspondingly convert the gyroscope and accelerometer data to the front-right-down body coordinate system and normalize it. The formula for obtaining the forward drag based on the airspeed indicator is as follows: where ρ is the air density, v is the speed of the vehicle relative to the air, obtained from the airspeed indicator, C is the air drag coefficient, and A is the cross-sectional area of the drone.

[0046] The formula for obtaining the forward motion acceleration caused by the forward drag is as follows: where m is the mass of the drone, and take a m as the compensation formula for the forward motion acceleration: a x补 = a x - a m where a x补 is the compensated forward accelerometer data, and a x is the forward accelerometer data before compensation.

[0047] In this embodiment, the specific process of judging the belly orientation in Step 2 above is as follows:

[0048] Set the vertical accelerometer data as a z and set the threshold a (0.6g < a < 0.8g). If a z is negative and its absolute value is greater than a, it can be judged that the belly is facing downwards and the attitude is not much different from the horizontal, then the wings can be opened and the engine can be started to enter the next stage. If a zIf the value is positive and its absolute value is greater than a, then the wings will open and the engines will start after the flight control rotates 180° to enter the next stage. If a z If the absolute value of is less than a, then it can be rotated 90° before the aforementioned judgment is made.

[0049] In this embodiment, the specific process of the maneuver strategy finding the UAV's near-horizontal position and estimating the gyroscope constant zero bias order of magnitude in step 3 above is as follows:

[0050] (1) Let the attitude be P1 when the engine is started. After starting the engine, let the UAV roll around the forward axis for a certain angle, and the attitude be P2. Detect the accelerometer a during the time from P1 to P2. z Output when a z When the value is at its maximum, the attitude is P3. After the drone returns to attitude P3 via flight control, it is rotated around the lateral axis (pitch) by a certain angle to attitude P4. The axial accelerometer a is measured during the time from P3 to P4. z Output when a z When the value is at its maximum, the attitude is P5. The drone is brought back to attitude P5 by the flight control system, and it is considered that the drone's attitude is close to horizontal at this time.

[0051] (2) Based on the gyroscope data from position P3 to P2 and back to P3, since the angular increment error of the two processes is a constant zero bias generated during the entire process, the constant zero bias of the gyroscope's forward axis can be estimated based on this angular increment, as shown in the following formula: Where ε bx Let ω1 be the constant zero bias of the forward axis of the gyroscope, ω2 be the total data of the forward axis gyroscope during the time interval from P3 to P2, and T be the sampling period. Similarly, the constant zero bias of the lateral axis of the gyroscope can be estimated based on the gyroscope data during the time interval from P5 to P4 and back to P5.

[0052] (3) The formula for calculating the roll angle at time P2 based on gyroscope data is as follows: β P2 =∑((ω1+ε bx )T), where γ P2 The roll angle at time P2 is obtained from the gyroscope. The formula for calculating the roll angle at time P2 using the accelerometer is as follows: Where γ P2 'The roll angle at time P2 is obtained from the accelerometer, a' y a z These are the accelerometer's lateral and longitudinal axis data, respectively. To achieve zero bias for the accelerometer, the formula for zero bias can be derived from the equality of roll angles as follows:

[0053] In this example, the specific implementation process of step 4 is as follows:

[0054] (1) The error models for the gyroscope and accelerometer are established as follows:

[0055]

[0056]

[0057] Where, ω x ω y ω z It is the three-axis measurement value of the gyroscope, ω tx ω ty ω tz It is the true value of the gyroscope's three axes, ε bx ε by ε bz It is a gyroscope with constant zero bias across three axes, η ωx η ωy η ωz It is three-axis Gaussian white noise from the gyroscope. x a y a z It is a triaxial measurement value from the accelerometer. tx a ty a tz These are the true values ​​of the accelerometer across all three axes. It is the triaxial constant zero bias of the accelerometer, η ax η ay η az It is three-axis Gaussian white noise from the accelerometer.

[0058] (2) The specific method for constructing the state equation and measurement equation is as follows: Based on the quaternion differential equation: And the mathematical model formula for noise: The continuous nonlinear state equations are constructed as follows: Find the Jacobian matrix for F(x(t), t): in Therefore, the state equation is as follows: The discretization formula is as follows: Where T is the iteration period. The discretized state equation is as follows: X k+1 =F k X k +G k W k W k The system white noise at time k is the 6-dimensional value. The triaxial accelerometer measurement value a is selected. k =[a x补 a y a z ] TFor measurement, according to in For the attitude transformation matrix between the navigation coordinate system and the body coordinate system, Find the Jacobian matrix: The measurement equations were then constructed as follows:

[0059] a k =H k X k +V k V k The 3D measurement noise at time k is given.

[0060] (3) The one-step state prediction formula is as follows: Where X k|k-1 To predict the state matrix in one step, Let X be the state transition matrix. k-1 Let be the state matrix at time k-1. The formula for the mean square error of the one-step prediction is as follows: Where P k|k-1 For the one-step prediction mean square error matrix, P k-1 Let G be the mean square error matrix at time k-1. k Let Q be the system noise matrix. k-1 Let K be the variance matrix of the system noise at time k-1. The filter gain formula is as follows: K k =P k|k- 1H k T (H k P k|k-1 H k T +R k ) -1 , where R k Let be the system measurement variance matrix at time k. The state estimation formula is as follows: The formula for estimating the mean square error is as follows: P k|k =(IK k H k )P k|k-1 , where I is the identity matrix.

[0061] (4) Utilizing motion acceleration error and its forward difference a s As two inputs to the fuzzy controller, the measurement noise variance is used as the output. Both the input and output are divided into five fuzzy subsets: {PB, PS, ZO, NS, NB}, where PB represents very small, PS represents small, ZO represents medium, NS represents large, and NB represents very large. The membership function is a uniformly distributed triangular function, and the fuzzy rule is a. sa s The larger the value, the larger the measurement noise variance, and the Centroid algorithm is used to achieve defuzzification.

[0062] (5) The formulas for obtaining the roll angle γ and pitch angle θ from the updated quaternions are as follows:

[0063] The above description is merely an illustration of the technical concept of the present invention in conjunction with specific preferred embodiments, and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A method for estimating the horizontal attitude of a fixed-wing unmanned aerial vehicle during freefall motion, characterized in that, Includes the following steps: 1) The drone is equipped with a MEMS gyroscope, MEMS accelerometer, and airspeed indicator. The MEMS gyroscope and MEMS accelerometer each contain three-axis data: forward x-axis, lateral y-axis, and upward z-axis. Data from the MEMS gyroscope, MEMS accelerometer, and airspeed indicator are acquired after the drone has been in free fall for a certain period of time. The MEMS gyroscope and MEMS accelerometer data are then preprocessed. 2) The orientation of the fuselage is initially determined by the z-axis data of the MEMS accelerometer. If the fuselage is facing the ground, the UAV can directly open its wings and start the engine. Otherwise, the attitude angle is adjusted by the UAV flight control before the wings are opened and the engine is started. 3) Estimate the initial zero bias of the MEMS gyroscope and MEMS accelerometer after starting the engine; 4) Based on the data from the MEMS gyroscope and MEMS accelerometer, the zero bias of the MEMS gyroscope and MEMS accelerometer is further estimated through the quaternion extended Kalman filter (EKF) attitude update algorithm, and the real-time estimate of the UAV's horizontal attitude angle is obtained.

2. The method for estimating the horizontal attitude of a fixed-wing UAV during freefall as described in claim 1, characterized in that, Step 1) The preprocessing process is as follows: 1.1) Select the northeast geographic coordinate system as the navigation coordinate system, and convert the MEMS gyroscope and MEMS accelerometer data to the front right lower body coordinate system and unify them to the International System of Units (SI) to obtain the three-axis data of the MEMS gyroscope. , , MEMS accelerometer triaxial data , , ;Calculate the forward drag based on airspeed data The formula is as follows: ,in air density, It is the speed of the carrier relative to the air, obtained from the airspeed indicator; C is the air resistance coefficient; and A is the cross-sectional area of ​​the UAV. 1.2) Solve for the forward acceleration. The formula is as follows: Where m is the mass of the UAV; solve for the compensation of the forward acceleration. The formula is as follows: .

3. The method for estimating the horizontal attitude of a fixed-wing UAV during freefall as described in claim 2, characterized in that, Step 2) Determining the orientation of the fuselage involves three scenarios: First, set the MEMS accelerometer's axial z-axis data to... And set a threshold a, which satisfies: 0.6g <a<0.8g; A) If If the value is negative and its absolute value is greater than the threshold a, then it is determined that the fuselage is facing down and the drone's attitude is near the horizontal line. The drone then opens its wings and starts its engine. B) If If the value is positive and its absolute value is greater than the threshold a, then the wings will open and the engine will start after the flight control rotates 180°. C) If If the absolute value is less than the threshold a, then the flight control will rotate 90° and then the judgment will be made according to cases A) and B).

4. The method for estimating the horizontal attitude of a fixed-wing UAV during freefall as described in claim 3, characterized in that, The implementation process of step 3) is as follows: 3.1) Let the attitude of the UAV when it starts its engine be P1. After starting the engine, let the UAV roll around the forward axis x by a certain angle, and the attitude at this time is P2. Detect the MEMS accelerometer z-axis data of the vertical axis during the time taken from attitude P1 to P2. When the MEMS accelerometer z-axis data of the vertical axis is at its maximum value, the attitude is P3. After the UAV returns to attitude P3 through the flight control, let the UAV roll around the lateral axis x by a certain angle, and the attitude at this time is P4. Detect the MEMS accelerometer z-axis data of the vertical axis during the time taken from attitude P3 to P4. When the MEMS accelerometer z-axis data of the vertical axis is at its maximum value, the attitude is P5. Detect the UAV returns to attitude P5 through the flight control. It is considered that the attitude of the UAV at this time is close to horizontal. 3.2) Obtain MEMS gyroscope data from the two time periods in step 3.1) when the UAV changes attitude from P3 to P2 and then back to P3. Since the angular increment error during the two transformation processes is zero, the angular increment generated throughout the entire process is estimated based on this angular increment. The formula is as follows: ,in This contains all data from the MEMS gyroscope during the time it takes for the attitude to change from attitude P3 to attitude P2. The MEMS gyroscope data during the time interval from attitude P2 back to attitude P3 is used, where T is the sampling period. Similarly, the MEMS gyroscope lateral axis y-bias is estimated based on the MEMS gyroscope data during the time interval from P5 to P4 and back to P5. ; 3.3) Calculate the roll angle at time P2 based on MEMS gyroscope data. The formula is as follows: Calculate the roll angle at time P2 based on the MEMS accelerometer. It is believed that the three axes of the MEMS accelerometer have the same zero bias. ,Right now =∇ x =∇ y =∇ z The formula is as follows: , , These are the lateral and longitudinal axis data from the MEMS accelerometer, respectively. Solving for the zero bias of the MEMS accelerometer The solution formula is as follows: .

5. The method for estimating the horizontal attitude of a fixed-wing unmanned aerial vehicle in freefall motion as described in claim 4, characterized in that, The implementation process of step 4) is as follows: 4.1) Select quaternions q0, q1, q2, and q3 for the MEMS gyroscope's three-axis zero bias. , ,in Initial value set to 0, three-axis random walk error , Equal and initial values ​​are all set to 0; MEMS accelerometer triaxial zero bias , There are a total of 13 state variables; 4.2) Determine the initial values ​​of each state variable, construct the nonlinear continuous state equation and discretize it, and construct the discrete measurement equation; 4.3) Begin the extended Kalman filtering process; 4.4) Output the quaternion updated by the Extended Kalman Filter (EKF) and convert it to the roll angle. Pitch angle .

6. The method for estimating the horizontal attitude of a fixed-wing unmanned aerial vehicle in freefall motion as described in claim 5, characterized in that, In step 4.3), the magnitude of the motion acceleration error during the filtering process is utilized in the extended Kalman filtering process. and its forward difference The filter measurement variance matrix at time k serves as one of the two inputs to the fuzzy controller. The noise variance measured by the diagonal elements is used as the output, and both the input and output are divided into 5 fuzzy subsets: The subsets PB through NB increase sequentially, and each subset is numbered 1, 2, 3, 4, 5. This represents the time step at each moment during the filtering process. and The values ​​are evenly distributed into 5 input subsets in ascending order, with subset numbers N1, N2, and N2 respectively. The formula for calculating the output subset number N is: The process involves rounding N up, assigning the output to the corresponding fuzzy subset based on the index of N, and then using the centroid tracking algorithm to perform defuzzification to obtain the output value. The output value is then used as a matrix. The diagonal elements are used to implement the matrix during the filtering process. Adaptive updates.

7. The method for estimating the horizontal attitude of a fixed-wing unmanned aerial vehicle in freefall motion as described in claim 6, characterized in that, In step 4.4), the roll angle is obtained from the updated quaternion. Pitch angle The formula is as follows: .

Images