CN113961010B - Tracking control method for quadrotor plant protection UAV - Google Patents

Abstract

The invention relates to a four-rotor plant protection unmanned aerial vehicle tracking control method based on an anti-saturation finite time self-adaptive neural network fault tolerance technology, which solves the defect that the four-rotor plant protection unmanned aerial vehicle is difficult to realize high-precision tracking control compared with the prior art. The invention comprises the following steps: setting and storing expected track data; collecting and storing real-time track data; building a composite mathematical model of the four-rotor plant protection unmanned aerial vehicle; establishing a four-rotor plant protection unmanned aerial vehicle flight error mathematical model; design and data storage of a saturation compensation system; designing and storing data of parameters of the self-adaptive neural network; the design of the fault-tolerant tracking controller and the storage of control signals based on the anti-saturation finite time adaptive neural network; updating real-time track data; and adjusting parameter values in the position system and the attitude system. The method can ensure that the track tracking error of the four-rotor plant protection unmanned aerial vehicle converges within a limited time range.

Description

Tracking control method of quadrotor plant protection UAV

Technical Field

The present invention relates to the technical field of control of quad-rotor plant protection UAVs, and in particular to a tracking control method of quad-rotor plant protection UAVs based on anti-saturation finite-time adaptive neural network fault-tolerant technology.

Background Art

In recent years, drones have been widely used in modern agriculture due to their advantages of flexible mobility, convenient operation, simple structure, and strong terrain adaptability. Compared with traditional manual fertilization, quad-rotor plant protection drones avoid the impact of improper manual operation on health through remote control. At the same time, quad-rotor plant protection drones are suitable for large-scale crop spraying, which improves work efficiency, reduces labor intensity, and is conducive to the precise prevention and control of pests and diseases, and promotes the rapid growth of the agricultural economy.

In order to ensure the efficient and accurate operation of the quad-rotor agricultural drone, one of the main difficulties is how to achieve high-precision trajectory tracking of the drone. However, there are still many practical difficulties in designing a high-performance trajectory tracking controller: First, the drone platform has the characteristics of under-actuation, strong coupling and high nonlinearity, which requires the designed controller to have good decoupling and strong robustness; secondly, since the application scenarios of drones are mostly outdoor environments, they are easily affected by external time-varying wind disturbances, which requires the designed controller to have strong anti-interference ability; in addition, when the drone is flying for a long time, the actuator is prone to failure and input saturation problems, which requires the designed controller to have good fault tolerance and anti-saturation capabilities. However, considering the above problems at the same time will increase the difficulty and complexity of the controller design.

At present, many researchers have carried out a lot of research on the trajectory tracking control problem of quadrotor agricultural drones, mainly including linear control methods and nonlinear control methods. Among them, the most common linear control methods are Proportional integral derivative (PID), linear quadratic regulator (LQR) and H infinite control. The basic idea is to first linearize the drone model at the equilibrium point, and then design a linear controller based on the linear model. Therefore, the linear control method has low dependence on the model, simple design and strong practicality. However, when the drone is performing large-scale flight missions, the control parameters of the linear control method are prone to sudden changes, resulting in reduced control accuracy.

In recent years, with the development of computer technology and control theory, some advanced nonlinear control methods have been proposed to improve the tracking control performance of UAVs, mainly including backstepping, sliding film control, neural network control and adaptive control. The backstepping method can effectively suppress disturbances or uncertainties in nonlinear systems and has strong robustness. However, in the backstepping process, the virtual control input needs to be continuously differentiated, which is easy to cause the problem of "exponential explosion" and increase the computational burden. Sliding film control is a nonlinear control method with strong robustness, fast response speed and simple design. However, the actual output signal is prone to jitter, which reduces the trajectory tracking quality.

In order to solve this problem, some researchers have proposed using bounded layer technology to replace discontinuous terms in the controller. However, these methods reduce the tracking control accuracy of the system. With its powerful online approximation capability, neural network control is particularly suitable for systems that need to consider multiple uncertainties or complex nonlinear systems at the same time. It usually requires an adaptive algorithm to adjust the neural network weights online, which greatly increases the online calculation time and calculation pressure of the control algorithm, which makes it difficult to meet the fast maneuvering flight mission requirements of quad-rotor agricultural drones. In addition, in practical applications, it is very important to achieve convergence of the tracking error within a finite time range. The convergence time is only determined by the control parameters and has nothing to do with the initial state of the tracking. Therefore, it is particularly important to design a finite time tracking controller.

Summary of the invention

The purpose of the present invention is to solve the defect that the tracking control of the quad-rotor plant protection UAV in the prior art is difficult to achieve, and to provide a quad-rotor plant protection UAV tracking control method based on anti-saturation finite time adaptive neural network fault tolerance technology to solve the above problem.

In order to achieve the above object, the technical solution of the present invention is as follows:

A tracking control method for a quadrotor plant protection UAV based on an anti-saturation finite-time adaptive neural network fault-tolerant technology comprises the following steps:

Setting and storing the expected trajectory data: according to the flight mission requirements performed by the quad-rotor plant protection UAV, the expected trajectory data is input through the ground terminal; the expected trajectory data input by the ground terminal is received through the airborne network module and saved in the flight data storage device I;

Collection and storage of real-time trajectory data: The real-time trajectory data is collected through the position sensor and attitude sensor carried by the quad-rotor agricultural drone, and the collected real-time trajectory data is stored in the flight data storage device II;

Establishment of a composite mathematical model of a quadrotor agricultural drone: Based on the inherent mechanical characteristics of the quadrotor agricultural drone and the interference factors of actuator failure, input saturation and time-varying wind disturbance during flight, a complete composite mathematical model of the quadrotor agricultural drone is established;

Establishment of the flight error mathematical model of quadrotor plant protection UAV: Based on the composite mathematical model of quadrotor plant protection UAV, the flight error mathematical model of quadrotor plant protection UAV is established;

Design and data storage of saturation compensation system: Design a saturation compensation system based on the flight error mathematical model of the quad-rotor agricultural drone, and update and save the saturation compensation system signal data into the flight data storage device III;

Design and data storage of adaptive neural network parameters: Design the adaptive neural network parameters based on the flight error mathematical model of the quadrotor plant protection UAV, and update and save the data of the adaptive neural network parameters into the flight data storage IV;

Design of fault-tolerant tracking controller based on anti-saturation finite-time adaptive neural network and storage of control signals: Based on the flight error mathematical model and saturation compensation system of the quadrotor plant protection UAV, a fault-tolerant tracking controller based on anti-saturation finite-time adaptive neural network is designed, and the fault-tolerant tracking control signal data based on anti-saturation finite-time adaptive neural network is updated and saved in the flight data storage V;

Update of real-time trajectory data: input the fault-tolerant tracking control signal based on the anti-saturation finite-time adaptive neural network into the complete composite mathematical model of the quadrotor plant protection UAV, output the real-time trajectory data and save it in the flight data memory II;

Adjustment of the numerical values of the position system parameters: By monitoring the data changes of the saturation compensation signal, the data changes of the adaptive neural network parameters, and the difference changes between the expected trajectory data and the actual trajectory data, the design parameters and control parameters in the position system are adjusted to achieve tracking control of the quadrotor plant protection UAV.

The establishment of the composite mathematical model of the quadrotor plant protection UAV includes the following steps:

Based on the coordinate system transformation method, the position kinematics mathematical model and attitude kinematics mathematical model of the quadrotor plant protection UAV are established. The specific expressions are as follows:

Where P = [x, y, z] ^T and represents the Euclidean position vector and Euler angle vector of the quadrotor agricultural drone in the earth coordinate system, respectively, where x, y, and z represent the position coordinates on the x _a- axis, y _a- axis, and z _a- axis, respectively, and φ, θ, and Respectively represent the roll angle around the x _a axis, the pitch angle around the y _a axis, and the yaw angle around the z _a axis, V = [u, v, w] ^T and Ω = [p, q, r] ^T respectively represent the linear velocity vector and angular velocity vector of the quadrotor agricultural drone in the body coordinate system, where u, v, and w represent the linear velocities on the x _b axis, y _b axis, and z _b axis, respectively, and p, q, and r represent the roll angular velocity around the x _b axis, the pitch angular velocity around the y _b axis, and the yaw angular velocity around the z _b axis, respectively. and They represent the linear velocity vector and angular velocity vector of the quadrotor agricultural drone in the earth coordinate system, respectively. and They represent the linear velocities on the x _a axis, y _a axis and z _a axis respectively, and They represent the roll angular velocity around the x _a axis, the pitch angular velocity around the y _a axis, and the yaw angular velocity around the z _a axis, respectively. R _t and R _s represent the orthogonal matrix and the Euler matrix, respectively. The specific expressions are as follows:

Using the Euler Lagrangian modeling method, considering the mechanical structure characteristics of the quadrotor agricultural drone and the influence of external time-varying wind disturbance during flight, the position dynamics mathematical model and attitude dynamics mathematical model of the quadrotor agricultural drone are established. The specific expressions are as follows:

Wherein, I _r = diag(I _x ,I _y ,I _z ) represents the positive definite moment of inertia matrix, where I _x , I _y and I _z represent the moment of inertia coefficients around the x-axis, y-axis and z-axis respectively, m represents the mass of the quadrotor plant protection drone, and They represent the linear acceleration vector and angular acceleration vector of the quadrotor agricultural drone in the body coordinate system, respectively. and denote the linear acceleration on the x _b axis, y _b axis and z _b axis respectively, and They represent the roll angular acceleration around the x _b axis, the pitch angular acceleration around the y _b axis, and the yaw angular acceleration around the z _b axis respectively;

_Fs = [0,0,u _o,1 ] ^T and _Ts = [u _o,2 ,u _o,3 ,u _o,4 ] ^T represent lift and control torque respectively.

Among them, the specific calculation expression of u _o,i ,(i＝1,2,3,4) is as follows:

Among them, w _i ,(i＝1,2,3,4) represents the rotation speed of the i-th motor rotor, d represents the distance between the motor and the mass center of the quadrotor agricultural drone, c ₁ and c ₂ represent the propeller thrust coefficient and torque coefficient respectively, F _a and T _a represent the air resistance in the attitude power system and position power system respectively. The specific expressions are as follows:

Wherein, _Kf = diag(Kf _,1 , _Kf,2 , Kf _,3 ) and _Kt = diag(Kt _,1 , _Kt,2 , _Kt,3 ) represent the drag coefficient matrix of the attitude system and the drag coefficient matrix of the position system respectively;

F _g represents the system gravity, and its specific expression is as follows:

Where, E = [0,0,1] ^T , m represents the mass of the quadrotor agricultural drone, g represents the acceleration of gravity, is the inverse matrix of the orthogonal matrix R _t , T _g represents the gyro torque, and its specific expression is as follows:

Where J represents the inertia coefficient of each rotor,

The symbol (Ω) _× represents a skew-symmetric matrix of the Ω vector, which satisfies the following form:

Based on the kinematic mathematical model, dynamic mathematical model, position dynamic mathematical model and attitude dynamic mathematical model of the quadrotor plant protection UAV, a non-complete composite mathematical model that does not consider actuator errors and input saturation is established. The specific expressions are as follows:

in, Represents the virtual input vector of the quadrotor agricultural drone;

as well as and

denote the nonlinearity in the position system and attitude system respectively, _da = [ _dx , _dy , _dz ] ^T and denote the lumped disturbances in the position system and attitude system respectively;

Considering the impact of actuator errors, the specific mathematical expression is as follows:

u _o,i ＝ρ _i u _i +r _i ,i＝1,2,3,4,

Where, u _o,i and u _i represent the actual control signal and the expected control signal respectively, ρ _i and _ri represent the effective coefficient and the additional fault respectively;

Considering the influence of actuator input saturation, the specific mathematical expression is as follows:

sat(u _i )=sign(u _i )min{|u _i |,u _max,i },i=1,2,3,4,

Where u _max,i represents the upper limit of the control signal u _i , and the sign function sign(u _i ) is defined as

Based on the incomplete composite mathematical model, the mathematical expression of actuator error and the mathematical expression of input saturation, a complete composite mathematical model considering actuator error and input saturation is established. and The specific expression is as follows:

Among them, ρ _b = [ρ ₁ , ρ ₂ , ρ ₃ ] ^T , r _b = [r ₁ , r ₂ , r ₃ ] ^T

and

The establishment of the flight error mathematical model of the quadrotor plant protection UAV includes the following steps:

Define position error e ₁ , attitude error e ₂ , linear velocity error And the angular velocity error The specific mathematical expressions are as follows:

e ₁ =PP _d , e ₂ =Θ-Θ _d ,

Among them, P _d =[x _d ,y _d ,z _d ] ^T and Respectively represent the expected position signal and the expected attitude signal in the earth coordinate system;

Based on the defined position error e ₁ , attitude error e ₂ , and linear velocity error And the angular velocity error Design the filter tracking error ξ ₁ of the position system and the filter tracking error ξ ₂ of the attitude system. The specific mathematical expressions are as follows:

Wherein, γ ₁ >0 and γ ₂ >0 represent filter coefficients. By increasing γ ₁ and γ ₂ , the convergence speed of the tracking error can be improved.

Based on the position-based filtered tracking error ξ ₁ , the attitude-based filtered tracking error ξ ₂ , and the complete composite mathematical model and Establishing a mathematical model of flight error for a quadrotor agricultural drone and The specific mathematical expression is as follows:

in, and Represent the complex nonlinear variables in the position system and attitude system respectively.

The design and data storage of the saturation compensation system include the following steps:

Based on the filtered tracking error ξ ₁ of the position system and the filtered tracking error ξ ₂ of the attitude system, a saturation compensation system is established and Its specific mathematical expression is as follows:

in, ΔU ₂ =T _t -sat(T _t ), K ₁ >0 and K ₂ >0 represent control parameters of the input compensation auxiliary system, υ ₁ and υ ₂ represent compensation auxiliary variables input in the position system and attitude system, respectively, ρ ₁ and ρ ₂ represent positive odd numbers, and satisfy the condition ρ ₁ <ρ ₂ ;

The compensated auxiliary variable data υ ₁ and υ ₂ are updated and saved in the flight data memory III.

The design and data storage of the adaptive neural network parameters include the following steps:

According to the definitions of s ₁ and s ₂ , we get the following inequality:

Then, based on the strong approximation ability of radial basis function neural network to nonlinear functions, radial basis function neural network is introduced, and its specific mathematical expression is as follows:

h(Z)＝W ^* TΞ(Z)+δ(Z),

in, and W ^*T Ξ(Z) respectively represent the input and output of the radial basis function neural network, n represents the number of inputs, h(Z) represents the nonlinear function, δ(Z) represents the approximation error, and W ^* represents the optimal weight vector, which is calculated according to the following formula:

in, represents the weight vector of the radial basis function neural network, represents the Gaussian basis function, and its specific mathematical expression is as follows:

Where, m = 1, 2, ..., k _sum , k _sum represents the total number of neurons in the hidden layer; and μ represent the center and radius of the radial basis function neural network, respectively;

The radial basis function neural network is used to approximate the nonlinear functions η ₁ (Z ₁ ) and η ₂ (Z ₂ ), and its specific mathematical expression is as follows:

Further:

in, and Ψ _i (Z _i )＝1+Ξ _i (Z _i ), (i＝1,2) represent unknown imaginary parameters and known computable positive scalar parameters respectively;

Designed adaptive neural network parameters and As shown below:

Wherein, _bi and c _i (i＝1,2) both represent positive design parameters; represents the upper bound estimate of β _i ;

The adaptive neural network parameter data and Update and save to flight data memory IV.

The design of the anti-saturation finite-time adaptive neural network fault-tolerant tracking controller and the storage of control signals include the following steps:

Based on filtering tracking errors ξ ₁ and ξ ₂ , saturation compensation system and Adaptive neural network parameters and And a complete composite mathematical model and The design is based on an anti-saturation finite-time adaptive neural network fault-tolerant tracking controller, and its specific mathematical expression is as follows:

Where, _ki and _ai , (i = 1, 2) represent positive design parameters,

Since the quadrotor agricultural drone has four inputs (u _o,1 ,u _o,2 ,u _o,3 ,u _o,4 ) and six outputs For an under-actuated system, the actual control input signal u _o,1 is calculated using three virtual control input signals (q ₁ ,q ₂ ,q ₃ ), namely:

In addition, the calculation formulas for the desired pitch angle φ _d and the desired yaw angle θ _d are respectively as follows:

The anti-saturation finite time adaptive neural network fault-tolerant tracking control signal data and T _t are updated and saved in the flight data memory V.

The updating of the real-time trajectory data comprises the following steps:

The fault-tolerant tracking control signal based on the anti-saturation finite-time adaptive neural network is input into the complete composite mathematical model of the quadrotor plant protection UAV, and the second-order derivative of the real-time trajectory data is output, namely: three linear accelerations and three angular accelerations And save it to the flight data memory II;

For three linear accelerations and three angular accelerations Perform secondary integration to obtain real-time trajectory data.

The adjustment of the parameter values in the position system and the attitude system comprises the following steps:

Inputting the expected trajectory data, real-time trajectory data, real-time adaptive neural network parameters, and complex nonlinear variables stored in flight data memories I, II, and III into the saturation compensation system and the radial basis function neural network, and outputting new saturation compensation signals and new adaptive neural network parameters;

The real-time trajectory data and expected trajectory data stored in flight data memory I and II, the new saturation compensation signal and the new adaptive neural network parameters are input into the anti-saturation finite-time adaptive neural network fault-tolerant tracking controller, and the control signal for adjusting the trajectory of the quadrotor plant protection UAV is output;

The control signal used to adjust the trajectory of the quadrotor plant protection drone is input into the complete composite mathematical model of the quadrotor plant protection drone, and the second-order derivative of the real-time trajectory data is output, namely: three linear accelerations and three angular accelerations Then perform a second integration on the second-order derivative of the real- _time trajectory data to obtain the new real-time trajectory data, namely: the position coordinates x, y and z on the xa axis, _ya axis and _za axis, as well as the roll angle φ around the _xa axis, the pitch angle θ around the _ya axis and the yaw angle around the _za axis.

The new real-time trajectory data, the new saturation compensation signal data, the new adaptive neural network parameter data and the input control signal for adjusting the trajectory of the quad-rotor plant protection UAV are updated and stored in the flight data memories II, III, IV and V respectively;

Changes in saturation compensation signal data in Observation Flight Database III:

Change adjustment of design parameters _k1 and _a1 in the position system: if the absolute value of the saturation compensation signal in the position system changes in a range greater than or equal to 0.2, the design parameter _k1 is increased by 0.5, and the value of the design parameter _a1 is increased by 0.2, until the absolute value of the saturation compensation signal in the position system changes in a range less than or equal to 0.2; if the absolute value of the saturation compensation signal in the position system changes in a range less than or equal to 0.2, the design parameter _k1 is increased by 0.3, and the value of the design parameter _a1 is increased by 0.1, until the absolute value of the saturation compensation signal in the position system changes in a range less than or equal to 0.02; if the absolute value of the saturation compensation signal in the position system changes in a range less than or equal to 0.02, the values of the design parameters _k1 and _a1 do not change, which meets the performance requirements of input signal compensation in the position system of the quadrotor plant protection UAV;

Change adjustment of design parameters _k2 and _a2 in the attitude system: if the absolute value of the saturated compensation signal in the attitude system changes in the range of greater than or equal to 0.2, the design parameter _k2 is increased by 0.5, and the value of the design parameter _a2 is increased by 0.2, until the absolute value of the saturated compensation signal in the attitude system changes in the range of less than or equal to 0.2; if the absolute value of the saturated compensation signal in the attitude system changes in the range of less than or equal to 0.2, the design parameter _k2 is increased by 0.3, and the value of the design parameter _a2 is increased by 0.1, until the absolute value of the saturated compensation signal in the attitude system changes in the range of less than or equal to 0.02; if the absolute value of the saturated compensation signal in the attitude system changes in the range of less than or equal to 0.02, the values of design parameters _k2 and _a2 do not change, which meets the performance requirements of input signal compensation in the attitude system of the quadrotor plant protection UAV.

Changes in adaptive neural network parameter data in observed flight database IV:

Change adjustment of design parameters _b1 and _c1 in the position system: if the adaptive neural network parameter value in the position system increases with time, the value of the design parameter _b1 decreases by 0.2, and the value of the design parameter _c1 increases by 0.25, until the adaptive neural network parameter value in the position system decreases monotonically with time; if the adaptive neural network parameter value in the position system needs more than 25 seconds to converge to zero, the value of the design parameter _b1 decreases by 0.08, and the value of the design parameter _c1 increases by 0.12, until the adaptive neural network parameter value in the position system needs less than 25 seconds to converge to zero; if the adaptive neural network parameter value in the position system needs less than 25 seconds to converge to zero, the value of the design parameter _b1 decreases by 0.04, and the value of the design parameter _c1 increases by 0.06, until the adaptive neural network parameter value in the position system needs to converge to zero within 10 seconds to 25 seconds; if the adaptive neural network parameter value in the position system needs to converge to zero within 10 seconds to 25 seconds, the design parameter b The value of ₁ does not change, and the value of the design parameter c ₁ increases by 0.04 until the adaptive neural network parameter value in the position system can converge to near zero within 10 seconds; if the adaptive neural network parameter value in the position system converges to zero within 10 seconds, the values of the design parameters b ₁ and c ₁ do not change, which meets the performance requirements of the convergence of the adaptive neural network parameters in the position system of the quadrotor plant protection drone;

Change adjustment of design parameters _b2 and _c2 in the attitude system: if the adaptive neural network parameter value in the attitude system increases with time, the value of the design parameter _b2 decreases by 0.2, and the value of the design parameter _c2 increases by 0.25, until the adaptive neural network parameter value in the attitude system decreases monotonically with time; if the adaptive neural network parameter value in the attitude system needs more than 25 seconds to converge to zero, the value of the design parameter _b2 decreases by 0.08, and the value of the design parameter _c2 increases by 0.12, until the adaptive neural network parameter value in the attitude system needs less than 25 seconds to converge to zero; if the adaptive neural network parameter value in the attitude system needs less than 25 seconds to converge to zero, the value of the design parameter _b2 decreases by 0.04, and the value of the design parameter _c2 increases by 0.06, until the adaptive neural network parameter value in the attitude system needs 10 seconds to 25 seconds to converge to zero; if the adaptive neural network parameter value in the attitude system needs 10 seconds to 25 seconds to converge to zero, the design parameter b The value of ₂ does not change, and the value of design parameter c ₂ increases by 0.04 until the adaptive neural network parameter value in the attitude system converges to zero within 10 seconds; if the adaptive neural network parameter value in the attitude system converges to zero within 10 seconds, the values of design parameters b ₂ and c ₂ do not change, which meets the performance requirements of the convergence of the adaptive neural network parameters in the attitude system of the quadrotor plant protection UAV.

Compare the difference between the expected trajectory data and the actual trajectory data in flight databases I and II:

Change adjustment of control parameter γ ₁ in the position system: if the sum of the absolute values of the differences between the three expected positions (x _d , y _d , z _d ) and the corresponding real-time positions (x, y, z) is greater than 1.5, the value of control parameter γ ₁ is increased by 0.18 until the sum of the absolute values of the differences is less than or equal to 1.5; if the sum of the absolute values of the differences between the three expected positions (x _d , y _d , z _d ) and the corresponding real-time positions (x, y, z) is less than or equal to 1.5, the value of control parameter γ ₁ is increased by 0.1 until the sum of the absolute values of the differences is less than or equal to 0.1; if the sum of the absolute values of the differences between the three expected positions (x _d , y _d , z _d ) and the corresponding real-time positions (x, y, z) is less than or equal to 0.1, the value of control parameter γ ₁ is increased by 0.06 until the sum of the absolute values of the differences is less than or equal to 0.01; if the sum of the absolute values of the differences between the three expected positions (x _d , y _d , z _d ) and the corresponding real-time position (x, y, z) is less than or equal to 0.01, and the value of the control parameter γ ₁ does not change, which meets the performance requirements of the trajectory tracking accuracy of the quadrotor plant protection UAV in the position system;

If the three desired attitude angles And the corresponding real-time attitude angle If the sum of the absolute values of the differences is greater than 1, the value of the control parameter _γ2 is increased by 0.15 until the sum of the absolute values of the differences is less than or equal to 1; if the three desired attitude angles are And the corresponding real-time attitude angle If the sum of the absolute values of the differences is less than or equal to 1, the value of the control parameter _γ2 is increased by 0.08 until the sum of the absolute values of the differences is less than or equal to 0.1; if the three desired attitude angles are And the corresponding real-time attitude angle If the sum of the absolute values of the differences is less than or equal to 0.1, the value of the control parameter _γ2 is increased by 0.03 until the sum of the absolute values of the differences is less than or equal to 0.01; if the three desired attitude angles are And the corresponding real-time attitude angle If the sum of the absolute values of the differences is less than or equal to 0.01, the value of the control parameter γ ₂ does not change, which meets the performance requirements of the trajectory tracking accuracy of the quadrotor plant protection UAV in the attitude system.

Beneficial Effects

The tracking and control method of the quadrotor plant protection UAV based on the anti-saturation finite-time adaptive neural network fault-tolerant technology of the present invention can ensure that the trajectory tracking error of the quadrotor plant protection UAV converges within a limited time range compared with the prior art; secondly, the method has strong robustness, good anti-saturation and actuator fault tolerance, and can realize high-performance trajectory tracking control of the quadrotor plant protection UAV in the presence of external time-varying interference, input saturation and actuator errors; at the same time, the method can effectively reduce the number of online calculations of the adaptive neural network parameters, thereby reducing the online calculation burden.

The present invention can ensure that the trajectory tracking error of the quad-rotor plant protection UAV can still converge to a bounded range within a limited time under the influence of external disturbances, input saturation and actuator errors, thereby improving the robustness, anti-saturation and actuator fault tolerance of the flight control system, while reducing the number of adaptive parameters in the controller and alleviating the burden of online parameter calculation.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1 is a method sequence diagram of the present invention;

FIG2 is a schematic diagram of a model of a four-rotor plant protection UAV in the present invention;

FIG3 is a real-time three-dimensional trajectory tracking response curve diagram of the present invention;

FIG4 is a trajectory tracking response curve diagram of the present invention in the x-axis direction;

FIG5 is a trajectory tracking response curve diagram of the present invention in the y-axis direction;

FIG6 is a trajectory tracking response curve diagram of the present invention in the z-axis direction;

FIG7 is a roll angle trajectory tracking response curve diagram of the present invention;

FIG8 is a pitch angle trajectory tracking response curve diagram of the present invention;

FIG9 is a yaw angle trajectory tracking response curve diagram of the present invention;

FIG10 is a graph showing an actual control input response of the present invention;

FIG. 11 is a graph showing the parameter response of the adaptive neural network of the present invention.

DETAILED DESCRIPTION

In order to have a further understanding and recognition of the structural features and the effects achieved by the present invention, a preferred embodiment and accompanying drawings are used for detailed description as follows:

As shown in FIG1 , a schematic diagram of a quad-rotor agricultural drone model of the present invention is shown, wherein four identical motors (numbered as motor 1, motor 2, motor 3 and motor 4) are mounted at symmetrical positions of the drone, and the position and attitude motion of the drone are controlled by adjusting the speed of the four motors. {O _a , x _a , y _a , z _a } is the earth coordinate system, {O _b , x _b , y _b , z _b } is the body coordinate system, d, m and g represent the distance between the motor and the center of mass of the drone, the mass of the quad-rotor agricultural drone and the gravity coefficient, respectively.

As shown in FIG2 , the tracking control method of the quadrotor plant protection UAV based on the anti-saturation finite time adaptive neural network fault tolerance technology of the present invention comprises the following steps:

The first step is to set and store the expected trajectory data: according to the flight mission requirements of the quadrotor plant protection UAV, the expected trajectory data is input through the ground terminal; the expected trajectory data input by the ground terminal is received through the airborne network module and saved in the flight data storage device I. The expected trajectory data includes: expected attitude angles (expected roll angles, expected pitch angles, and expected yaw angles) and expected position coordinates (expected x-axis position coordinates, expected y-axis position coordinates, and expected z-axis position coordinates).

The second step is to collect and store real-time trajectory data: collect real-time trajectory data through the position sensor and attitude sensor carried by the quadrotor agricultural drone, and save the collected real-time trajectory data in the flight data storage device II. The real-time trajectory data includes: real-time attitude angles (real-time roll angles, real-time pitch angles, and real-time yaw angles) and real-time position coordinates (real-time x-axis position coordinates, real-time y-axis position coordinates, and real-time z-axis position coordinates).

The third step is to establish a composite mathematical model of the four-rotor plant protection UAV. A complete composite mathematical model of the four-rotor plant protection UAV is established based on the inherent mechanical characteristics of the four-rotor plant protection UAV and the interference factors of actuator failure, input saturation and time-varying wind disturbance during flight. It should be noted that the present invention takes into account three factors: external disturbance, input saturation and actuator error, and considers the coupling between the attitude dynamic system and the position dynamic system, which can not only better reflect the flight characteristics of the four-rotor plant protection UAV under actual conditions, but also better deal with emergencies encountered during actual flight. However, these interference factors greatly increase the difficulty and complexity of the controller design. The specific steps are as follows:

(1) Based on the coordinate system transformation method, the position kinematics mathematical model and attitude kinematics mathematical model of the quadrotor agricultural drone are established. The specific expressions are as follows:

Where P = [x, y, z] ^T and represents the Euclidean position vector and Euler angle vector of the quadrotor agricultural drone in the earth coordinate system, respectively, where x, y, and z represent the position coordinates on the x _a- axis, y _a- axis, and z _a- axis, respectively, and φ, θ, and Respectively represent the roll angle around the x _a axis, the pitch angle around the y _a axis, and the yaw angle around the z _a axis, V = [u, v, w] ^T and Ω = [p, q, r] ^T respectively represent the linear velocity vector and angular velocity vector of the quadrotor agricultural drone in the body coordinate system, where u, v, and w represent the linear velocities on the x _b axis, y _b axis, and z _b axis, respectively, and p, q, and r represent the roll angular velocity around the x _b axis, the pitch angular velocity around the y _b axis, and the yaw angular velocity around the z _b axis, respectively. and They represent the linear velocity vector and angular velocity vector of the quadrotor agricultural drone in the earth coordinate system, respectively. and They represent the linear velocities on the x _a axis, y _a axis and z _a axis respectively, and They represent the roll angular velocity around the x _a axis, the pitch angular velocity around the y _a axis, and the yaw angular velocity around the z _a axis, respectively. R _t and R _s represent the orthogonal matrix and the Euler matrix, respectively. The specific expressions are as follows:

(2) Using the Euler Lagrangian modeling method, the position dynamics mathematical model and attitude dynamics mathematical model of the quadrotor agricultural drone are established by considering the mechanical structure characteristics of the quadrotor agricultural drone and the influence of external time-varying wind disturbance during flight. The specific expressions are as follows:

Wherein, I _r = diag(I _x ,I _y ,I _z ) represents the positive definite moment of inertia matrix, where I _x , I _y and I _z represent the moment of inertia coefficients around the x-axis, y-axis and z-axis respectively, m represents the mass of the quadrotor plant protection drone, and They represent the linear acceleration vector and angular acceleration vector of the quadrotor agricultural drone in the body coordinate system, respectively. and denote the linear acceleration on the x _b axis, y _b axis and z _b axis respectively, and They represent the roll angular acceleration around the x _b axis, the pitch angular acceleration around the y _b axis, and the yaw angular acceleration around the z _b axis respectively;

_Fs = [0,0,u _o,1 ] ^T and _Ts = [u _o,2 ,u _o,3 ,u _o,4 ] ^T represent lift and control torque respectively.

Among them, the specific calculation expression of u _o,i ,(i＝1,2,3,4) is as follows:

Among them, w _i ,(i＝1,2,3,4) represents the rotation speed of the i-th motor rotor, d represents the distance between the motor and the mass center of the quadrotor agricultural drone, c ₁ and c ₂ represent the propeller thrust coefficient and torque coefficient respectively, F _a and T _a represent the air resistance in the attitude power system and position power system respectively. The specific expressions are as follows:

Wherein, _Kf = diag(Kf _,1 , _Kf,2 , Kf _,3 ) and _Kt = diag(Kt _,1 , _Kt,2 , _Kt,3 ) represent the drag coefficient matrix of the attitude system and the drag coefficient matrix of the position system respectively;

F _g represents the system gravity, and its specific expression is as follows:

Where, E = [0,0,1] ^T , m represents the mass of the quadrotor agricultural drone, g represents the acceleration of gravity, is the inverse matrix of the orthogonal matrix R _t , T _g represents the gyro torque, and its specific expression is as follows:

Where J represents the inertia coefficient of each rotor,

The symbol (Ω) _× represents a skew-symmetric matrix of the Ω vector, which satisfies the following form:

(3) Based on the kinematic mathematical model, dynamic mathematical model, position dynamic mathematical model and attitude dynamic mathematical model of the quadrotor agricultural drone, a non-complete composite mathematical model that does not consider actuator errors and input saturation is established. The specific expressions are as follows:

in, Represents the virtual input vector of the quadrotor agricultural drone;

as well as and denote the nonlinearity in the position system and attitude system respectively, _da = [ _dx , _dy , _dz ] ^T and denote the lumped disturbances in the position system and attitude system, respectively.

(4) Considering the influence of actuator error, the specific mathematical expression is as follows:

u _o,i ＝ρ _i u _i +r _i ,i＝1,2,3,4,

Among them, u _o,i and _ui represent the actual control signal and the expected control signal respectively, ρ _i and _ri represent the effective coefficient and the additional fault respectively.

(5) Considering the influence of actuator input saturation, the specific mathematical expression is as follows:

sat(u _i )=sign(u _i )min{|u _i |,u _max,i },i=1,2,3,4,

Where u _max,i represents the upper limit of the control signal u _i , and the sign function sign(u _i ) is defined as

(6) Based on the incomplete composite mathematical model, the mathematical expression of actuator error, and the mathematical expression of input saturation, a complete composite mathematical model considering actuator error and input saturation is established and The specific expression is as follows:

Among them, ρ _b = [ρ ₁ , ρ ₂ , ρ ₃ ] ^T , r _b = [r ₁ , r ₂ , r ₃ ] ^T

and

The fourth step is to establish a mathematical model of the flight error of the quadrotor plant protection drone: Based on the composite mathematical model of the quadrotor plant protection drone, a mathematical model of the flight error of the quadrotor plant protection drone is established. It should be noted that the flight error model of the quadrotor plant protection drone established in the present invention is to introduce filter coefficients γ ₁ and γ ₂ so as to more directly adjust the convergence speed of the position error and the convergence speed of the attitude error.

The establishment of the flight error mathematical model of the quadrotor plant protection UAV includes the following steps:

(1) Define position error e ₁ , attitude error e ₂ , and linear velocity error And the angular velocity error The specific mathematical expressions are as follows:

e ₁ =PP _d , e ₂ =Θ-Θ _d ,

Among them, P _d =[x _d ,y _d ,z _d ] ^T and They represent the expected position signal and expected attitude signal in the earth coordinate system respectively.

(2) Based on the defined position error e ₁ , attitude error e ₂ , and linear velocity error And the angular velocity error Design the filter tracking error ξ ₁ of the position system and the filter tracking error ξ ₂ of the attitude system. The specific mathematical expressions are as follows:

Here, γ ₁ >0 and γ ₂ >0 represent filter coefficients, and the convergence speed of the tracking error can be improved by increasing γ ₁ and γ ₂ .

(3) Based on the position-based filtering tracking error ξ ₁ , the attitude-based filtering tracking error ξ ₂ , and the complete composite mathematical model and Establishing a mathematical model of flight error for a quadrotor agricultural drone and The specific mathematical expression is as follows:

in, and Represent the complex nonlinear variables in the position system and attitude system respectively.

Step 5, design and data storage of saturation compensation system: design a saturation compensation system based on the flight error mathematical model of the quadrotor plant protection UAV, and update and save the saturation compensation system signal data into the flight data memory III. The saturation compensation system designed by the present invention does not need to assume that the size of the expected control input is bounded, and can also ensure that the saturation compensation system converges in a finite time.

The design and data storage of the saturation compensation system include the following steps:

(1) Based on the filter tracking error ξ ₁ of the position system and the filter tracking error ξ ₂ of the attitude system, a saturation compensation system is established and Its specific mathematical expression is as follows:

in, ΔU ₂ =T _t -sat(T _t ), K ₁ >0 and K ₂ >0 represent control parameters of the input compensation auxiliary system, υ ₁ and υ ₂ represent compensation auxiliary variables input in the position system and attitude system, respectively, ρ ₁ and ρ ₂ represent positive odd numbers, and satisfy the condition ρ ₁ <ρ ₂ ;

(2) Update and save the compensation auxiliary variable data _υ1 and _υ2 into the flight data memory III.

Step 6, design and data storage of adaptive neural network parameters: Adaptive neural network parameters are designed based on the flight error mathematical model of the quad-rotor plant protection UAV, and the data of the adaptive neural network parameters are updated and saved in the flight data memory IV. The present invention uses the designed adaptive algorithm to update two lumped parameters online instead of updating two vectors or two matrices. Therefore, this greatly reduces the number of adaptive parameters in the controller, thereby effectively reducing the computational burden. In addition, the adaptive algorithm of the present invention is designed based on δ-modification technology, thereby effectively avoiding the problem of parameter drift. The design and data storage of adaptive neural network parameters include the following steps:

(1) According to the definitions of s ₁ and s ₂ , we get the following inequality:

Then, based on the strong approximation ability of radial basis function neural network to nonlinear functions, radial basis function neural network is introduced, and its specific mathematical expression is as follows:

h(Z)＝W ^* TΞ(Z)+δ(Z),

in, and W ^*T Ξ(Z) respectively represent the input and output of the radial basis function neural network, n represents the number of inputs, h(Z) represents the nonlinear function, δ(Z) represents the approximation error, and W ^* represents the optimal weight vector, which is calculated according to the following formula:

in, represents the weight vector of the radial basis function neural network, represents the Gaussian basis function, and its specific mathematical expression is as follows:

Where, m = 1, 2, ..., k _sum , k _sum represents the total number of neurons in the hidden layer; and μ represent the center and radius of the radial basis function neural network, respectively.

(2) The radial basis function neural network is used to approximate the nonlinear functions η ₁ (Z ₁ ) and η ₂ (Z ₂ ). The specific mathematical expressions are as follows:

Further:

in, and Ψ _i (Z _i ) = 1 + Ξ _i (Z _i ), (i = 1, 2) represent the unknown imaginary parameters and the known computable positive scalar parameters respectively.

(3) Designed adaptive neural network parameters and As shown below:

Wherein, _bi and c _i (i＝1,2) both represent positive design parameters; represents the upper bound estimate of β _i ;

(4) Adaptive neural network parameter data and Update and save to flight data memory IV.

The seventh step is the design of the anti-saturation finite time adaptive neural network fault-tolerant tracking controller and the storage of control signals: based on the flight error mathematical model and saturation compensation system of the four-rotor plant protection UAV, a finite time adaptive neural network fault-tolerant tracking controller based on anti-saturation is designed, and the control signal data based on the anti-saturation finite time adaptive neural network fault-tolerant tracking is updated and saved in the flight data storage device V. The anti-saturation finite time adaptive neural network fault-tolerant tracking controller designed by the present invention can simultaneously handle the problems of external disturbances, actuator errors and input saturation, which greatly increase the difficulty of controller design. In addition, this greatly reduces the number of adaptive parameters in the controller, simplifies the design structure, and reduces the burden of online parameter calculation. Therefore, the designed controller is more economical and reliable, and can better cope with emergency situations encountered during actual flight and meet the requirements of safe flight. The design of the anti-saturation finite time adaptive neural network fault-tolerant tracking controller and the storage of control signals include the following steps:

(1) Based on the filtering tracking errors ξ ₁ and ξ ₂ and the saturation compensation system and Adaptive neural network parameters and And a complete composite mathematical model and The design is based on an anti-saturation finite-time adaptive neural network fault-tolerant tracking controller, and its specific mathematical expression is as follows:

Among them, _ki and _ai , (i = 1, 2) represent positive design parameters,

Since the quadrotor agricultural drone has four inputs (u _o,1 ,u _o,2 ,u _o,3 ,u _o,4 ) and six outputs For an under-actuated system, the actual control input signal u _o,1 is calculated using three virtual control input signals (q ₁ ,q ₂ ,q ₃ ), namely:

In addition, the calculation formulas for the desired pitch angle φ _d and the desired yaw angle θ _d are respectively as follows:

(2) The fault-tolerant tracking control signal data based on the anti-saturation finite-time adaptive neural network and T _t are updated and saved in the flight data memory V.

Step 8. Update of real-time trajectory data: Input the fault-tolerant tracking control signal based on the anti-saturation finite-time adaptive neural network into the complete composite mathematical model of the quadrotor plant protection UAV, output the real-time trajectory data and save it in the flight data storage device II.

(1) Input the fault-tolerant tracking control signal based on the anti-saturation finite-time adaptive neural network into the complete composite mathematical model of the quadrotor plant protection UAV, and output the second-order derivative of the real-time trajectory data, namely: three linear accelerations and three angular accelerations And saved to the flight data memory II.

(2) For three linear accelerations and three angular accelerations Perform secondary integration to obtain real-time trajectory data.

The ninth step is to adjust the parameter values in the position system and attitude system: by monitoring the data changes of the saturation compensation signal, the data changes of the adaptive neural network parameters, and the difference changes between the expected trajectory data and the actual trajectory data, the design parameters and control parameters in the position system are adjusted to achieve tracking control of the quadrotor plant protection UAV.

The specific steps are as follows:

(1) The desired trajectory data, real-time trajectory data, real-time adaptive neural network parameters, and complex nonlinear variables stored in flight data memories I, II, and III are input into the saturation compensation system and radial basis function neural network, and new saturation compensation signals and new adaptive neural network parameters are output.

(2) The real-time trajectory data and expected trajectory data stored in flight data memories I and II, the new saturation compensation signal, and the new adaptive neural network parameters are input into the anti-saturation finite-time adaptive neural network fault-tolerant tracking controller, and the control signal for adjusting the trajectory of the quadrotor plant protection UAV is output.

(3) The control signal used to adjust the trajectory of the quadrotor agricultural drone is input into the complete composite mathematical model of the quadrotor agricultural drone, and the second-order derivative of the real-time trajectory data is output, namely: three linear accelerations and three angular accelerations Then perform a second integration on the second-order derivative of the real- _time trajectory data to obtain the new real-time trajectory data, namely: the position coordinates x, y and z on the xa axis, _ya axis and _za axis, as well as the roll angle φ around the _xa axis, the pitch angle θ around the _ya axis and the yaw angle around the _za axis.

(4) The new real-time trajectory data, the new saturation compensation signal data, the new adaptive neural network parameter data and the input control signal for adjusting the trajectory of the quad-rotor plant protection UAV are updated and stored in the flight data memories II, III, IV and V respectively.

(5) Changes in saturation compensation signal data in the observation flight database III:

A1) Change adjustment of design parameters _k1 and _a1 in the position system: if the absolute value of the saturation compensation signal in the position system changes within a range greater than or equal to 0.2, the design parameter _k1 is increased by 0.5, and the value of the design parameter _a1 is increased by 0.2, until the absolute value of the saturation compensation signal in the position system changes within a range less than or equal to 0.2; if the absolute value of the saturation compensation signal in the position system changes within a range less than or equal to 0.2, the design parameter _k1 is increased by 0.3, and the value of the design parameter _a1 is increased by 0.1, until the absolute value of the saturation compensation signal in the position system changes within a range less than or equal to 0.02; if the absolute value of the saturation compensation signal in the position system changes within a range less than or equal to 0.02, the values of the design parameters _k1 and _a1 do not change, which meets the performance requirements of input signal compensation in the position system of the quad-rotor plant protection UAV;

A2) Change adjustment of design parameters _k2 and _a2 in the attitude system: If the absolute value of the saturated compensation signal in the attitude system changes in a range greater than or equal to 0.2, the design parameter _k2 is increased by 0.5, and the value of the design parameter _a2 is increased by 0.2, until the absolute value of the saturated compensation signal in the attitude system changes in a range less than or equal to 0.2; if the absolute value of the saturated compensation signal in the attitude system changes in a range less than or equal to 0.2, the design parameter _k2 is increased by 0.3, and the value of the design parameter _a2 is increased by 0.1, until the absolute value of the saturated compensation signal in the attitude system changes in a range less than or equal to 0.02; if the absolute value of the saturated compensation signal in the attitude system changes in a range less than or equal to 0.02, the values of the design parameters _k2 and _a2 do not change, which meets the performance requirements of input signal compensation in the attitude system of the quadrotor plant protection UAV.

(6) Changes in the adaptive neural network parameter data in the observed flight database IV:

B1) Change adjustment of design parameters _b1 and _c1 in the position system: If the adaptive neural network parameter value in the position system increases with time, the value of the design parameter _b1 decreases by 0.2, and the value of the design parameter _c1 increases by 0.25, until the adaptive neural network parameter value in the position system decreases monotonically with time; if the adaptive neural network parameter value in the position system needs more than 25 seconds to converge to zero, the value of the design parameter _b1 decreases by 0.08, and the value of the design parameter _c1 increases by 0.12, until the adaptive neural network parameter value in the position system needs less than 25 seconds to converge to zero; if the adaptive neural network parameter value in the position system needs less than 25 seconds to converge to zero, the value of the design parameter _b1 decreases by 0.04, and the value of the design parameter c1 increases by 0.12. The value of ₁ is increased by 0.06 until the adaptive neural network parameter value in the position system needs 10 to 25 seconds to converge to zero; if the adaptive neural network parameter value in the position system needs 10 to 25 seconds to converge to zero, the value of the design parameter b ₁ does not change, and the value of the design parameter c ₁ is increased by 0.04 until the adaptive neural network parameter value in the position system can converge to zero within 10 seconds; if the adaptive neural network parameter value in the position system converges to zero within 10 seconds, the values of the design parameters b ₁ and c ₁ do not change, which meets the performance requirements of the adaptive neural network parameter convergence in the position system of the quadrotor plant protection drone;

B2) Change adjustment of design parameters _b2 and _c2 in the attitude system: If the adaptive neural network parameter value in the attitude system increases with time, the value of the design parameter _b2 decreases by 0.2, and the value of the design parameter _c2 increases by 0.25, until the adaptive neural network parameter value in the attitude system decreases monotonically with time; if the adaptive neural network parameter value in the attitude system needs more than 25 seconds to converge to zero, the value of the design parameter _b2 decreases by 0.08, and the value of the design parameter _c2 increases by 0.12, until the adaptive neural network parameter value in the attitude system needs less than 25 seconds to converge to zero; if the adaptive neural network parameter value in the attitude system needs less than 25 seconds to converge to zero, the value of the design parameter _b2 decreases by 0.04, and the value of the design parameter c2 increases by 0.12. The value of ₂ is increased by 0.06 until the adaptive neural network parameter value in the attitude system converges to near zero within 10 seconds to 25 seconds; if the adaptive neural network parameter value in the attitude system converges to near zero within 10 seconds to 25 seconds, the value of the design parameter _b2 does not change, and the value of the design parameter _c2 is increased by 0.04 until the adaptive neural network parameter value in the attitude system converges to zero within 10 seconds; if the adaptive neural network parameter value in the attitude system converges to zero within 10 seconds, the values of the design parameters _b2 and _c2 do not change, which meets the performance requirements of the adaptive neural network parameter convergence in the attitude system of the quadrotor plant protection UAV.

(7) Compare the difference between the expected trajectory data and the actual trajectory data in flight databases I and II:

C1) Change adjustment of control parameter γ ₁ in the position system: If the sum of the absolute values of the differences between the three expected positions (x _d , y _d , z _d ) and the corresponding real-time positions (x, y, z) is greater than 1.5, the value of control parameter γ ₁ is increased by 0.18 until the sum of the absolute values of the differences is less than or equal to 1.5; if the sum of the absolute values of the differences between the three expected positions (x _d , y _d , z _d ) and the corresponding real-time positions (x, y, z) is less than or equal to 1.5, the value of control parameter γ ₁ is increased by 0.1 until the sum of the absolute values of the differences is less than or equal to 0.1; if the sum of the absolute values of the differences between the three expected positions (x _d , y _d , z _d ) and the corresponding real-time positions (x, y, z) is less than or equal to 0.1, the value of control parameter γ ₁ is increased by 0.06 until the sum of the absolute values of the differences is less than or equal to 0.01; if the sum of the absolute values of the differences between the three expected positions (x _d , y _d , z _{d )} and the corresponding real-time positions (x, y, z) is less than or equal to 0.1 ) and the corresponding real-time position (x, y, z) is less than or equal to 0.01, and the value of the control parameter γ ₁ does not change, which meets the performance requirements of the trajectory tracking accuracy of the quadrotor plant protection UAV in the position system;

C2) Change adjustment of control parameter γ ₂ in attitude system: If the three desired attitude angles And the corresponding real-time attitude angle If the sum of the absolute values of the differences is greater than 1, the value of the control parameter _γ2 is increased by 0.15 until the sum of the absolute values of the differences is less than or equal to 1; if the three desired attitude angles are And the corresponding real-time attitude angle If the sum of the absolute values of the differences is less than or equal to 1, the value of the control parameter _γ2 is increased by 0.08 until the sum of the absolute values of the differences is less than or equal to 0.1; if the three desired attitude angles are And the corresponding real-time attitude angle If the sum of the absolute values of the differences is less than or equal to 0.1, the value of the control parameter _γ2 is increased by 0.03 until the sum of the absolute values of the differences is less than or equal to 0.01; if the three desired attitude angles And the corresponding real-time attitude angle If the sum of the absolute values of the differences is less than or equal to 0.01, the value of the control parameter γ ₂ does not change, which meets the performance requirements of the trajectory tracking accuracy of the quadrotor plant protection UAV in the attitude system.

In order to prove that the tracking error signals _e1 and _e2 of the quadrotor plant protection drone converge to a bounded region within a finite time range, consider the following composite Lyapunov function V:

in, and represents the estimation error, and denote the estimated values of _β1 and _β2 , respectively.

Substituting the anti-saturation finite-time adaptive neural network fault-tolerant tracking controller, the flight error mathematical model of the quadrotor plant protection UAV, the saturation compensation system and the adaptive neural network parameters into the first-order derivative of the Lyapunov function V, we can obtain:

in,

as well as

According to the finite time stability theorem, as long as and K ₂ ＞0, the tracking signals ξ ₁ ,ξ ₂ , υ ₁ ,υ ₂ converge to a bounded region Ω ₁ near the origin in finite time T ₁ .

Specifically, the calculation formula for the finite time T ₁ is:

, where: 0＜ε ₁ ＜1, T ₀ represents the initial time.

Specifically, the bounded region Ω ₁ is calculated as

Therefore, based on the designed position filter tracking error and attitude filter tracking error, it can be concluded that the position tracking error and attitude tracking error will converge to the following bounded areas respectively:

In summary, the present invention, when the quad-rotor plant protection UAV is affected by external time-varying wind disturbance, input saturation and actuator errors, the designed anti-saturation finite time adaptive neural network fault-tolerant tracking controller can still ensure that all closed-loop system signals and tracking errors converge to a bounded area within a finite time, thereby enhancing the robustness of the system, the actuator anti-saturation performance and the actuator fault tolerance, and helping to ensure the high-performance, safe and autonomous flight of the quad-rotor plant protection UAV. At the same time, the number of online calculations of the adaptive neural network parameters is reduced, effectively reducing the computational burden of the airborne control center.

In order to verify the superiority of the controller proposed in the present invention, in a specific embodiment, a trajectory tracking control system of a four-rotor plant protection UAV is constructed based on a MATLAB simulation platform. In the embodiment of the present invention, the physical parameters of the four-rotor plant protection UAV are as follows: m = 2 [kg], g = 9.8 [m/s ² ], d = 0.2 [m], _Jr = 0.002 [kg·m ² ], _Io,x = _Io,y = 1.2416 [N·m·s ² /rad], Io _,z = 2.4832 [N·m·s ² /rad], _Kf,1 = Kf _,2 = _Kf,3 = 0.01 [N·s/m] and _Kt,1 = _Kt,2 = _Kt,3 = 0.001 [N·m·s/rad].

The time-varying disturbances to the quadrotor plant protection drone are as follows:

and The actuator failures of the quadrotor agricultural drone are as follows: [r ₁ ,r ₂ ,r ₃ ,r ₄ ] ^T ＝[0.1,0.02,0.1sin(0.2t),0] ^T . The input saturation limits of the quadrotor agricultural drone are as follows: u _max,1 ＝35[N] and u _max,2 ＝u _max,3 ＝u _max,4 ＝40[N·m]. The quadrotor agricultural drone changes from the initial trajectory

Takeoff tracking desired trajectory

The selected control parameters are as follows:

γ ₁ =10, γ ₂ =6, k ₁ =10, k ₂ =2, a ₁ =a ₂ =1, b ₁ =0.003, b ₂ =0.002, c ₁ =c ₂ =0.6, β ₁ (0 )=β ₂ (0)=0, σ=0, μ=5, k _sum =200, ρ ₁ =15, ρ ₂ =17, K ₁ =2.5 and K ₂ =3.5.

As shown in Figures 3 to 10, it can be seen from Figure 3 that the actual trajectory of the quadrotor agricultural drone can track the expected trajectory well. From Figures 4, 5, and 6, it can be seen that under the influence of external wind disturbance, actuator error, and input saturation, the real-time trajectory signals of the quadrotor agricultural drone on the x, y, and z axes can accurately track the corresponding expected position signals respectively. From Figures 7, 8, and 9, it can be seen that under the influence of external wind disturbance, actuator error, and input saturation, the real-time roll angle signal, real-time pitch angle signal, and real-time yaw angle signal of the quadrotor agricultural drone can accurately track the corresponding expected attitude signals respectively. From Figure 10, it can be seen that the four output signals of the quadrotor agricultural drone meet the input saturation constraint conditions, effectively solving the problem of input saturation. From Figure 11, it can be seen that the values of the two adaptive parameters are bounded and eventually converge to near 0, avoiding the problem of parameter drift.

The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims (7)

1. The four-rotor plant protection unmanned aerial vehicle tracking control method is characterized by comprising the following steps of:

11 Setting and storing of desired trajectory data: according to the flight task requirement executed by the four-rotor plant protection unmanned aerial vehicle, inputting expected track data through a ground terminal; receiving expected track data input by a ground terminal through an airborne network module, and storing the expected track data into a flight data memory I;

12 Acquisition and storage of real-time trajectory data: collecting real-time track data through a position sensor and a posture sensor carried by the four-rotor plant protection unmanned aerial vehicle, and storing the collected real-time track data in a flight data storage II;

13 Four rotor plant protection unmanned aerial vehicle composite mathematical model establishment: according to the inherent mechanical characteristics of the four-rotor plant protection unmanned aerial vehicle and the interference factors of actuator faults, input saturation and time-varying wind disturbance during flight, a complete composite mathematical model of the four-rotor plant protection unmanned aerial vehicle is built;

the building of the composite mathematical model of the four-rotor plant protection unmanned aerial vehicle comprises the following steps:

131 Based on a coordinate system conversion method, a position kinematic mathematical model and an attitude kinematic mathematical model of the four-rotor plant protection unmanned aerial vehicle are established, and specific expressions are respectively shown as follows:

wherein, P= [ x, y, z] ^T Andrespectively representing Euclidean position vector and Euler angle vector of the four-rotor plant protection unmanned aerial vehicle in an earth coordinate system, wherein x, y and z respectively represent x _a Axis, y _a Axis and z _a Position coordinates on the axis, phi, theta and +.>Respectively denote the winding x _a The degree of roll angle of the shaft, y _a Pitch angle degree and z-axis _a Yaw angle degree of shaft, v= [ u, V, w] ^T And Ω= [ p, q, r ]] ^T Respectively representing a linear velocity vector and an angular velocity vector of the four-rotor plant protection unmanned aerial vehicle in a machine body coordinate system, wherein u, v and w respectively represent x _b Axis, y _b Axis and z _b The linear velocity on the axis, p, q and r respectively denote the velocity around x _b Roll angular velocity of shaft, around y _b Pitch rate and z-axis of rotation _b Yaw rate of the shaft, +.>And->Respectively representing a linear velocity vector and an angular velocity vector of the four-rotor plant protection unmanned aerial vehicle in an earth coordinate system, wherein +.>And->Respectively expressed in x _a Axis, y _a Axis and z _a Linear speed on shaft, +.>And->Respectively denote the winding x _a Roll angular velocity of shaft, around y _a Pitch rate and z-axis of rotation _a Yaw rate of shaft, R _t And R is _s Respectively representing an orthogonal matrix and an Euler matrix, wherein the specific expressions are respectively as follows:

132 By utilizing an Euler Lagrange modeling method, the position dynamics mathematical model and the attitude dynamics mathematical model of the four-rotor plant protection unmanned aerial vehicle are established by considering the self mechanical structure characteristics of the four-rotor plant protection unmanned aerial vehicle and the influence of external time-varying wind disturbance received during flight, and the specific expressions are respectively shown as follows:

wherein I is _r ＝diag(I _x ,I _y ,I _z ) Representing a positive fixed moment of inertia matrix, where I _x 、I _y And I _z Respectively represent the rotational inertia coefficients around the x axis, the y axis and the z axis, m represents the self mass of the four-rotor plant protection unmanned aerial vehicle,and->Respectively representing a linear acceleration vector and an angular acceleration vector of the four-rotor plant protection unmanned aerial vehicle in a machine body coordinate system, wherein +.>Andrespectively expressed in x _b Axis, y _b Axis and z _b Linear acceleration on axis, +.>And->Respectively denote the winding x _b Roll angular acceleration of axis, around y _b Pitch acceleration and z-axis _b Yaw angular acceleration of the shaft;

F _s ＝[0,0,u _o,1 ] ^T and T _s ＝[u _o,2 ,u _o,3 ,u _o,4 ] ^T Respectively representing the lifting force and the control moment,

wherein u is _o,i The specific calculation expression of (i=1, 2,3, 4) is as follows:

wherein w is _i (i=1, 2,3, 4) represents the rotation speed of the ith motor rotor, d represents the distance between the motor and the mass center of the four-rotor plant protection unmanned aerial vehicle, c ₁ And c ₂ Respectively represent the thrust coefficient and the torque coefficient of the propeller, F _a And T _a The air resistance in the attitude power system and the position power system are respectively represented, and the specific expressions are as follows:

wherein K is _f ＝diag(K _f,1 ,K _f,2 ,K _f,3 ) And K _t ＝diag(K _t,1 ,K _t,2 ,K _t,3 ) Respectively representing a resistance coefficient matrix of the attitude system and a resistance coefficient matrix of the position system;

F _g representing the gravity of the system, the specific expression is as follows:

wherein e= [0, 1 ]] ^T M represents the self mass of the four-rotor plant protection unmanned aerial vehicle, g represents the gravitational acceleration,is an orthogonal matrix R _t Inverse matrix of T _g The specific expression of the gyro moment is as follows:

where J represents the coefficient of inertia of each rotor,

sign (omega) _× Representing the omega vectorA skewed symmetry matrix satisfying the form:

133 Based on the kinematic mathematical model, the dynamic mathematical model, the position dynamic mathematical model and the attitude dynamic mathematical model of the four-rotor plant protection unmanned aerial vehicle, an incomplete composite mathematical model without considering the error of an actuator and input saturation is established, and the specific expressions are respectively as follows:

Wherein,,a virtual input vector representing a quad-rotor plant protection drone;

and +.>Andrepresenting non-linearities in a position system and a gesture system, respectively,/->d _a ＝[d _x ,d _y ,d _z ] ^T Andrepresenting lumped disturbances in the position system and the attitude system, respectively;

134 Considering the effect of actuator errors, the specific mathematical expression is as follows:

u _o,i ＝ρ _i u _i +r _i ,i＝1,2,3,4，

wherein u is _o,i And u _i Representing the actual control signal and the desired control signal, ρ, respectively _i And r _i Representing the significant coefficient and the additional fault, respectively;

135 Considering the effect of actuator input saturation, a specific mathematical expression is as follows:

sat(u _i )＝sign(u _i )min{|u _i |,u _max,i },i＝1,2,3,4，

wherein u is _max,i Representing the control signal u _i Upper bound of sign (u) _i ) Is defined as

136 Based on the incomplete complex mathematical model, the mathematical expression of the actuator error, and the mathematical expression of the input saturation, a complete complex mathematical model is built that accounts for the actuator error and the input saturationAnd->The specific expression is as follows:

wherein ρ is _b ＝[ρ ₁ ,ρ ₂ ,ρ ₃ ] ^T ，r _b ＝[r ₁ ,r ₂ ,r ₃ ] ^T

And

14 Four rotor plant protection unmanned aerial vehicle flight error mathematical model establishment: based on the composite mathematical model of the four-rotor plant protection unmanned aerial vehicle, establishing a flight error mathematical model of the four-rotor plant protection unmanned aerial vehicle;

15 Design and data storage of the saturation compensation system: designing a saturation compensation system based on a flight error mathematical model of the four-rotor plant protection unmanned aerial vehicle, and updating and storing signal data of the saturation compensation system into a flight data memory III;

16 Design and data storage of adaptive neural network parameters: designing a self-adaptive neural network parameter based on a flight error mathematical model of the four-rotor plant protection unmanned aerial vehicle, and updating and storing data of the self-adaptive neural network parameter into a flight data memory IV;

17 Design of fault-tolerant tracking controller based on anti-saturation finite time adaptive neural network and storage of control signals: based on a four-rotor plant protection unmanned aerial vehicle flight error mathematical model and a saturation compensation system, designing an anti-saturation limited time self-adaptive neural network fault-tolerant tracking controller, and updating and storing data of the anti-saturation limited time self-adaptive neural network fault-tolerant tracking control signal into a flight data storage V;

18 Update of real-time trajectory data): inputting fault-tolerant tracking control signals based on the anti-saturation finite time self-adaptive neural network into a complete composite mathematical model of the four-rotor plant protection unmanned aerial vehicle, outputting real-time track data and storing the real-time track data into a flight data storage II;

19 Parameter values in the position and attitude systems: the tracking control of the four-rotor plant protection unmanned aerial vehicle is realized by monitoring the data change of the saturation compensation signal, the data change of the self-adaptive neural network parameter and the difference change of the expected track data and the actual track data and adjusting the design parameters and the control parameters in the positioning system.

2. The four-rotor plant protection unmanned aerial vehicle tracking control method according to claim 1, wherein the building of the four-rotor plant protection unmanned aerial vehicle flight error mathematical model comprises the following steps:

21 Defining a position error e ₁ Attitude error e ₂ Error in linear velocityAngular velocity error +.>The specific mathematical expressions are shown below, respectively:

e ₁ ＝P-P _d ，e ₂ ＝Θ-Θ _d ，

wherein P is _d ＝[x _d ,y _d ,z _d ] ^T Andrepresenting a desired position signal and a desired attitude signal in an earth coordinate system, respectively;

22 Based on the defined position error e) ₁ Attitude error e ₂ Error in linear velocityAngular velocity error +.>Filtering tracking error ζ of a design position system ₁ And the filtered tracking error ζ of the attitude system ₂ The specific mathematical expression is as follows:

wherein, gamma ₁ > 0 and gamma ₂ > 0 represents the filter coefficient by increasing γ ₁ And gamma ₂ Convergence of tracking error can be improvedA speed;

23 Position-based filtered tracking error ζ ₁ Filtered tracking error ζ of gesture ₂ Complete composite mathematical modelAndestablishing a flight error mathematical model of a four-rotor plant protection unmanned aerial vehicle>And->The specific mathematical expression is as follows:

wherein,,and->Representing complex nonlinear variables in the position and attitude systems, respectively.

3. The four rotor plant protection drone tracking control method of claim 2, wherein the design and data storage of the saturation compensation system comprises the steps of:

31 Position system based filtered tracking error ζ ₁ And the filtered tracking error ζ of the attitude system ₂ Establishing a saturation compensation systemAnd->The specific mathematical expression is as follows:

wherein,,ΔU ₂ ＝T _t -sat(T _t )，K ₁ > 0 and K ₂ > 0 represents the control parameter of the input compensation auxiliary system, v ₁ And v ₂ Representing the compensation auxiliary variables entered in the position and attitude systems, respectively ρ ₁ And ρ ₂ Represents a positive odd number and satisfies the condition ρ ₁ ＜ρ ₂ ；

32 To compensate the auxiliary variable data v ₁ And v ₂ Updated and saved to the flight data memory III.

4. The four-rotor plant protection unmanned aerial vehicle tracking control method of claim 3, wherein the design and data storage of the adaptive neural network parameters comprises the steps of:

41 According to s) ₁ Sum s ₂ Is defined by the following inequality:

then, based on the strong approximation capability of the radial basis function neural network to the nonlinear function, the radial basis function neural network is introduced, and the specific mathematical expression is as follows:

h(Z)＝W ^*T Ξ(Z)+δ(Z)，

wherein,,and W is ^*T Xi (Z) represents the input and output of the radial basis function neural network, n represents the number of inputs, h (Z) represents the nonlinear function, delta (Z) represents the approximation error, W ^* Represents an optimal weight vector, which is calculated according to the following formula:

Wherein,,a weight vector representing a radial basis function neural network,the specific mathematical expression of the Gaussian basis function is as follows:

where m=1, 2,..k _sum ，k _sum Representing the total neurons in the hidden layer;and μ represents the center and radius of the radial basis function neural network, respectively;

42 Approximation of a nonlinear function eta using a radial basis function neural network ₁ (Z ₁ ) And eta ₂ (Z ₂ ) Specific mathematical table thereofThe expression is as follows:

η ₁ (Z ₁ )＝W ₁ ^*T Ξ ₁ (Z ₁ )+δ ₁ ，η ₂ (Z ₂ )＝W ₂ ^*T Ξ ₂ (Z ₂ )+δ ₂ ，

further obtain:

wherein,,and psi is _i (Z _i )＝1+Ξ _i (Z _i ) (i=1, 2) represents an unknown virtual parameter and a known computable positive scalar parameter, respectively;

43 Adaptive neural network parameters for designAnd->The following is shown:

wherein b _i And c _i (i=1, 2) each represents a positive design parameter;representing beta _i Upper bound estimate of (c);

44 To adapt the neural network parameter dataAnd->Updated and saved to the flight data memory IV.

5. The four-rotor plant protection unmanned aerial vehicle tracking control method of claim 4, wherein the design and storage of control signals based on the anti-saturation finite time adaptive neural network fault tolerant tracking controller comprises the following steps:

51 Based on the filtered tracking error ζ ₁ And xi ₂ Saturation compensation systemAnd->Adaptive neural network parameters- >And->Complete composite mathematical model->And->The fault-tolerant tracking controller based on the anti-saturation finite time self-adaptive neural network is designed, and a specific mathematical expression is as follows:

wherein k is _i And a _i (i=1, 2) represents a positive design parameter,

because the quad-rotor plant protection drone has four inputs (u _o,1 ,u _o,2 ,u _o,3 ,u _o,4 ) 6 outputsUtilizes three virtual control input signals (q ₁ ,q ₂ ,q ₃ ) Calculating the actual control input signal u _o,1 The method comprises the following steps:

in addition, a pitch angle phi is desired _d And a desired yaw angle theta _d The calculation formulas of (a) are respectively as follows:

52 Fault-tolerant tracking control signal data based on anti-saturation finite time adaptive neural networkAnd T _t Updated and saved to the flight data memory V.

6. The four-rotor plant protection unmanned aerial vehicle tracking control method of claim 5, wherein the updating of the real-time trajectory data comprises the steps of:

61 Inputting fault-tolerant tracking control signals based on anti-saturation finite time adaptive neural network to a complete composite mathematical model of the four-rotor plant protection unmanned aerial vehicle, and outputting real timeThe second derivative of the trajectory data, namely: three linear accelerationsAnd three angular accelerations->And storing the data into a flight data memory II;

62 For three linear accelerationsAnd three angular accelerations->And performing secondary integration to obtain real-time track data.

7. The four-rotor plant protection unmanned aerial vehicle tracking control method according to claim 6, wherein the adjustment of the parameter values in the position system and the attitude system comprises the steps of:

71 Inputting the desired trajectory data, the real-time trajectory data, and the real-time adaptive neural network parameters, and the complex nonlinear variables stored in the flight data memories I, II and III into the saturation compensation system and the radial basis function neural network, and outputting new saturation compensation signals and new adaptive neural network parameters;

72 Inputting the real-time track data and expected track data stored in the flight data memories I and II, the new saturation compensation signals and the new self-adaptive neural network parameters into an anti-saturation finite time self-adaptive neural network fault-tolerant tracking controller, and outputting control signals for adjusting the track of the four-rotor plant protection unmanned aerial vehicle;

73 The control signal for adjusting the track of the four-rotor plant protection unmanned aerial vehicle is input into a complete composite mathematical model of the four-rotor plant protection unmanned aerial vehicle, and the second derivative of real-time track data is output, namely: three linear accelerations And three angular accelerationsAnd then carrying out secondary integration on the second derivative of the real-time track data to obtain new real-time track data, namely: at x _a Axis, y _a Axis and z _a Position coordinates x, y and z on axis and around x _a Angle of roll of axis phi, y _a Pitch angle degree θ and roll z of axis _a Yaw angle degree +.>

74 Updating the new real-time track data, the new saturation compensation signal data, the new self-adaptive neural network parameter data and the input control signals for adjusting the track of the four-rotor plant protection unmanned aerial vehicle, and storing the updated data in the flight data memories II, III, IV and V respectively;

75 Observing changes in saturation compensation signal data in flight database III:

751 Design parameter k in a position system ₁ And a ₁ Is adjusted by the change of (a):

if the absolute value of the saturation compensation signal in the position system varies in a range of 0.2 or more, the design parameter k ₁ Increase by 0.5 and design parameter a ₁ The value of (2) is increased by 0.2 until the absolute value of the saturation compensation signal in the position system changes within a range of 0.2 or less;

if the absolute value of the saturation compensation signal in the position system varies within a range of 0.2 or less, the design parameter k ₁ Increase in size by 0.3 and design parameter a ₁ The value of (2) is increased by 0.1 until the absolute value of the saturation compensation signal in the position system changes within a range of 0.02 or less;

if the absolute value of the saturation compensation signal in the position system varies within a range of 0.02 or less, the design parameter k ₁ And a ₁ Is unchanged in valueThe performance requirement of input signal compensation of the four-rotor plant protection unmanned aerial vehicle in a position system is met;

752 Design parameter k in gesture System ₂ And a ₂ Is adjusted by the change of (a):

if the absolute value of the saturation compensation signal in the attitude system varies within a range of 0.2 or more, the design parameter k ₂ Increase by 0.5 and design parameter a ₂ The value of (2) is increased by 0.2 until the absolute value of the saturation compensation signal in the attitude system changes within the range of less than or equal to 0.2;

if the absolute value of the saturation compensation signal in the attitude system varies within a range of 0.2 or less, the design parameter k ₂ Increase in size by 0.3 and design parameter a ₂ The value of (2) is increased by 0.1 until the absolute value of the saturation compensation signal in the attitude system changes within the range of less than or equal to 0.02;

if the absolute value of the saturation compensation signal in the attitude system varies within a range of 0.02 or less, the design parameter k ₂ And a ₂ The values of the input signals are unchanged, so that the performance requirement of the four-rotor plant protection unmanned aerial vehicle on input signal compensation in a gesture system is met;

76 Observing changes in the adaptive neural network parameter data in the flight database IV:

761 Design parameter b) in a position system ₁ And c ₁ Is adjusted by the change of (a):

if the value of the adaptive neural network parameter in the location system is incrementally changed over time, then the design parameter b ₁ The value of (c) is decreased by 0.2 and the design parameter c ₁ The value of (2) is increased by 0.25 until the parameter value of the self-adaptive neural network in the position system is monotonically decreased along with time;

if the adaptive neural network parameter values in the location system take more than 25 seconds to converge to zero, then design parameter b ₁ The value of (c) is decreased by 0.08 and the design parameter c ₁ Increasing by 0.12 until the adaptive neural network parameter values in the location system need to converge to around zero in less than 25 seconds;

if in a location systemThe adaptive neural network parameter value of (2) needs to converge to around zero in less than 25 seconds, then the design parameter b ₁ The value of (c) is decreased by 0.04 and the design parameter c ₁ The value of (c) increases by 0.06 until the adaptive neural network parameter value in the position system takes 10 seconds to 25 seconds to converge to around zero, then the parameter b is designed ₁ Not changing the value of (c) while designing the parameter c ₁ The value of (2) is increased by 0.04 until the adaptive neural network parameter value in the position system can be converged to be near zero in 10 seconds;

if the adaptive neural network parameter values in the location system converge to zero within 10 seconds, then the design parameter b ₁ And c ₁ The values of the self-adaptive neural network parameters are unchanged, so that the performance requirement of the self-adaptive neural network parameter convergence of the four-rotor plant protection unmanned aerial vehicle in a position system is met;

762 Design parameter b in gesture System ₂ And c ₂ Is adjusted by the change of (a):

if the adaptive neural network parameter values in the gesture system change incrementally over time, then design parameter b ₂ The value of (c) is decreased by 0.2 and the design parameter c ₂ The value of (2) is increased by 0.25 until the parameter value of the self-adaptive neural network in the gesture system is monotonically decreased along with time;

if the adaptive neural network parameter values in the attitude system take more than 25 seconds to converge to zero, then design parameter b ₂ The value of (c) is decreased by 0.08 and the design parameter c ₂ The value of (2) is increased by 0.12 until the adaptive neural network parameter value in the attitude system needs to converge to near zero in less than 25 seconds;

if the adaptive neural network parameter values in the attitude system require less than 25 seconds to converge to around zero, then the design parameter b ₂ The value of (c) is decreased by 0.04 and the design parameter c ₂ The value of (2) increases by 0.06 until the adaptive neural network parameter values in the attitude system need to converge to around zero in the range of 10 seconds to 25 seconds;

if the adaptive neural network parameter values in the attitude system require convergence to near zero in the range of 10 seconds to 25 seconds, then design parameter b ₂ Is not of the value of (a)Varying, while designing parameter c ₂ The value of (2) is increased by 0.04 until the adaptive neural network parameter value in the gesture system converges to zero within 10 seconds;

if the adaptive neural network parameter values in the attitude system converge to zero within 10 seconds, then the design parameter b ₂ And c ₂ The values of the self-adaptive neural network parameters are unchanged, so that the performance requirement of the self-adaptive neural network parameter convergence of the four-rotor plant protection unmanned aerial vehicle in a gesture system is met;

77 By difference comparison of the expected trajectory data with the actual trajectory data in the flight databases I and II):

771 Control parameter gamma in a position system ₁ Is adjusted by the change of (a):

if three desired positions (x _d ,y _d ,z _d ) The sum of absolute values of the differences with the corresponding real-time positions (x, y, z) is greater than 1.5, the control parameter gamma ₁ Increasing the value of (2) by 0.18 until the sum of absolute values of the differences is less than or equal to 1.5;

If three desired positions (x _d ,y _d ,z _d ) The sum of absolute values of the difference values with the corresponding real-time positions (x, y, z) is less than or equal to 1.5, and then the control parameter gamma ₁ The value of (2) is increased by 0.1 until the sum of absolute values of the difference values is less than or equal to 0.1;

if three desired positions (x _d ,y _d ,z _d ) The sum of absolute values of the difference values with the corresponding real-time positions (x, y, z) is less than or equal to 0.1, and then the control parameter gamma ₁ The value of (2) is increased by 0.06 until the sum of absolute values of the difference values is less than or equal to 0.01;

if three desired positions (x _d ,y _d ,z _d ) The sum of absolute values of the difference values with the corresponding real-time positions (x, y, z) is less than or equal to 0.01, and the control parameter gamma ₁ The value of (2) is not changed, so that the performance requirement of track tracking precision of the four-rotor plant protection unmanned aerial vehicle in a position system is met;

772 Control parameter gamma in a gesture system ₂ Is adjusted by the change of (a):

if three desired attitude angle numbersCorresponding real-time attitude angle degree->The sum of the absolute values of the differences is greater than 1, the control parameter gamma ₂ The value of (2) is increased by 0.15 until the sum of absolute values of the difference values is less than or equal to 1;

if three desired attitude angle numbersCorresponding real-time attitude angle degree->The sum of the absolute values of the differences is 1 or less, the control parameter gamma ₂ The value of (2) is increased by 0.08 until the sum of absolute values of the difference values is less than or equal to 0.1;

If three desired attitude angle numbersCorresponding real-time attitude angle degree->The sum of the absolute values of the differences is less than or equal to 0.1, and the control parameter gamma ₂ The value of (2) is increased by 0.03 until the sum of absolute values of the difference values is less than or equal to 0.01;

if three desired attitude angle numbersCorresponding real-time attitude angle degree->The sum of the absolute values of the differences is less than or equal to 0.01, the control parameter gamma ₂ The value of (2) is not changed, so that the performance of track tracking precision of the four-rotor plant protection unmanned aerial vehicle in a gesture system is metCan be required.