A UAV anti-wind disturbance controller and control method

Abstract

The present invention belongs to the technical field of rotor UAV control, and provides a UAV anti-wind disturbance controller and control method. The design method of the controller includes: establishing a rotor UAV rigid body model and a rotor UAV aerodynamic model; establishing a wind field model; constructing a first controller based on linear adaptive dynamic repetitive control according to the rotor UAV rigid body model, the rotor UAV aerodynamic model and the wind field model; combining the rotor UAV rigid body model and the rotor aerodynamic model, using the hybrid particle swarm gray wolf optimization algorithm to perform parameter setting on the first controller to obtain a second controller; optimizing the second controller using a fuzzy control method to obtain a UAV anti-wind disturbance controller. Using the controller to perform anti-wind disturbance control on the UAV can effectively improve the adaptive ability of the rotor UAV controller, so that the rotor UAV can complete the inspection task with high quality even when it is disturbed by wind.

Description

Technical Field

[0001] The present invention relates to the technical field of rotor UAV control, and in particular to an anti-wind disturbance controller and a control method for a UAV. Background Art

[0002] It is difficult for existing control algorithms to quickly predict and perceive the disturbance caused by natural wind on rotorcraft UAVs and to accurately calculate the corresponding compensation amount, resulting in vibration, deviation and even loss of control of rotorcraft UAVs under the interference of turbulent wind. Especially when the UAV is taking pictures of key damaged locations on the bridge, the unstable photography platform and the offset photography position will affect the quality of the pictures taken. At this time, the UAV control algorithm is particularly important.

[0003] Existing UAV control algorithms mainly include linear, nonlinear control algorithms and intelligent control algorithms. Linear control algorithms mainly include proportional integral differential control (PID) and linear quadratic regulator (LQR). Nonlinear control methods include backstepping, active disturbance rejection control (ADRC), etc. Intelligent control methods mainly include neural networks (NN) and fuzzy logic systems (FLS).

[0004] Among the existing control algorithm studies, the PID control algorithm is simple to implement but has weak robustness; the LQR control algorithm has good control tracking but has strict requirements on the control system; the backstepping control algorithm has strong robustness and good control effect, but has problems such as differential explosion and complex calculation; the neural network control algorithm has strong adaptability, but has problems such as long training time, large amount of calculation and poor interpretability. Summary of the invention

[0005] In view of the defects in the prior art, the present invention provides an unmanned aerial vehicle (UAV) anti-wind disturbance controller and a control method, which improve the self-adaptation and anti-wind disturbance capabilities of the UAV controller.

[0006] An anti-wind disturbance controller for a UAV is designed by the following method:

[0007] Establish the rigid body model and aerodynamic model of the rotary-wing UAV;

[0008] Establish wind field model;

[0009] According to the rigid body model of the rotor UAV, the aerodynamic model of the rotor UAV and the wind field model, a first controller based on linear adaptive dynamic repetitive control is constructed;

[0010] Combining the rigid body model of the rotor UAV and the rotor aerodynamic model, the hybrid particle swarm grey wolf optimization algorithm is used to tune the parameters of the first controller to obtain the second controller;

[0011] The second controller is optimized using the fuzzy control method to obtain the UAV anti-wind disturbance controller.

[0012] Preferably, before establishing the rotor UAV rigid body model, the rotor UAV aerodynamic model and the wind field model, it also includes defining a coordinate system, the coordinate system includes a geodetic coordinate system and a body coordinate system, wherein the transformation matrix of the "ground-body" dual coordinate system is:

[0013]

[0014] in, is the transformation matrix of the “earth-body” dual coordinate system, φ, θ, ψ are the yaw angle, roll angle, and pitch angle of the UAV in the earth coordinate system, respectively.

[0015] Preferably, the rigid body model of the rotorcraft drone is as follows:

[0016]

[0017] in, is the acceleration of the rotorcraft in three directions of the geodetic coordinate system, is the angular velocity of the UAV’s yaw, roll and pitch angles, is the angular acceleration of the yaw angle, roll angle and pitch angle of the UAV, F is the sum of the rotor lift of the UAV, m is the mass of the UAV, g is the acceleration of gravity, J xx , J yy , J zz is the moment of inertia of the X-axis (pitch angle), Y-axis (roll angle) and Z-axis (yaw angle), τ x , τ y , τ z It is the control torque of X-axis (pitch angle), Y-axis (roll angle) and Z-axis (yaw angle).

[0018] The aerodynamic model of the rotorcraft is as follows:

[0019]

[0020] Among them, T wx , T wy , T wz is the component force of the total disturbance force of wind on the rotor UAV in the three-axis direction, M wφ 、M wθ 、M wψ is the torque generated by the total disturbance force on the yaw angle, roll angle and pitch angle.

[0021] Preferably, the wind field model is a combination of constant wind, gust, rapid wind and gradually changing wind:

[0022] V=V c+V z +V r +V j

[0023] Among them, V c is the constant wind, V z For gusts, V r is random wind, V j For the gradual wind;

[0024] The constant wind expression is: V C =k, k is a constant value;

[0025] The wind gust expression is:

[0026] V zmax is the peak wind speed in the gust, T1 is the start time of the gust, T g is the duration of the gust;

[0027] The random wind expression is V smax is the peak value of random wind speed,

[0028] R am is the coefficient of random wind, which is a random value in [-1,1], and ω is a fixed value in [0,2π], is a random value in [0,2π];

[0029] The gradient wind expression is V jmax is the peak wind speed of the gradually changing wind, T1' is the time when the gradually changing wind starts to change, T2' is the time when the gradually changing wind ends, and T is the duration of the gradually changing wind.

[0030] Preferably, the first controller system is a second-order control system including disturbance, and the expression is:

[0031]

[0032] w(t) is the external disturbance of the controlled object, y and u are the output and input items of the system, t is time, b is the controller gain, represents the first-order derivative of the system output, Represents the second-order derivative of the system output.

[0033] Preferably, the linear extended state observer expression in the first controller system is:

[0034]

[0035] z1 and z2 are the observed values ​​of the system state variables, z3 is the total disturbance observed value of the system, is the estimated output value of the system, k1, k2, k3 are the parameters of the linear state observer.

[0036] Preferably, the linear error feedback controller in the first controller system adopts a PD controller, and the expression is:

[0037] u0=k p (ν-z1)+k d z2

[0038] u0 is the intermediate control quantity, k p is the proportional parameter of the PD controller, k d is the differential parameter of the PD controller.

[0039] Preferably, the hybrid particle swarm grey wolf optimization algorithm is expressed as

[0040]

[0041] in, represents the speed of the gray wolf individual ij in the gray wolf pack in the k+1th generation; H is the inertia weight; c1 and c2 represent the individual and group learning factors; r1 and r2 are random factors between [0,1]; represents the position of the gray wolf individual ij at the kth iteration; represents the best position found by the gray wolf individual ij before the kth generation; It represents the best position of all the individual gray wolves in the entire gray wolf pack after k iterations.

[0042] Preferably, the second controller is optimized using a fuzzy control method to obtain an anti-wind disturbance controller for the UAV, including:

[0043] The quantitative deviation e and the deviation change rate e c As the fuzzy input, the fuzzy logic control rules are used to adjust the linear error feedback controller parameters online, and the fuzzy controller output is w c The adjustment values ​​of the three parameters w0, b0 are denoted as Vw0, Vw c , Vb0;

[0044] Specify two fuzzy input variables e and e c With three fuzzy output variables Vw0, Vw c The seven fuzzy control subsets of Vb0 are {NB, NM, NS, ZO, PS, PM, PB}, and the fuzzy rule table of the above five variables is designed by adopting the triangular membership function and applying the Mamdani type fuzzy inference rule.

[0045] A method for controlling an unmanned aerial vehicle against wind disturbances uses the unmanned aerial vehicle against wind disturbances controller described in any one of the above items to perform anti-wind disturbance flight control on the unmanned aerial vehicle.

[0046] The present invention proposes a hybrid particle swarm grey wolf optimization algorithm. At the same time, fuzzy control is added to the controller parameter setting based on linear adaptive dynamic repetitive control. Fuzzy control rules are designed to further enhance the adaptive ability of the controller, so that the rotor UAV can complete the inspection task with high quality even when disturbed by wind. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the specific embodiments of the present invention, the following will briefly introduce the drawings required for the specific embodiments or the prior art description. In all the drawings, each element or part is not necessarily drawn according to the actual scale.

[0048] Figure 1 A design flow chart of anti-wind disturbance control for a UAV provided by one embodiment of the present invention;

[0049] Figure 2 A geodetic coordinate system and a body coordinate system provided by an embodiment of the present invention;

[0050] Figure 3 A schematic diagram of the forces acting on the rotor of a rotary-wing UAV under wind field interference provided by an embodiment of the present invention;

[0051] Figure 4 A fuzzy control flow chart provided for an embodiment of the present invention;

[0052] Figure 5 This is a structural diagram of a UAV anti-wind disturbance controller provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

[0054] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by those skilled in the art to which the present invention belongs.

[0055] In one embodiment, Figure 1 As shown, a UAV anti-wind disturbance controller is provided, which is designed by the following method:

[0056] 1. Establish the rigid body model and aerodynamic model of the rotorcraft UAV;

[0057] In one embodiment, before establishing the rigid body model of the rotary-wing UAV and the aerodynamic model of the rotary-wing UAV, it also includes defining a coordinate system, which includes a geodetic coordinate system and a body coordinate system.

[0058] Furthermore, the specific steps for establishing the rotor UAV motion coordinate system are as follows:

[0059] Step 1: Coordinate system definition. Selecting an appropriate coordinate system can more clearly and accurately show the position and attitude change information of the rotary-wing UAV, thereby reducing the difficulty of modeling the rigid body model of the rotary-wing UAV. The present invention will use the following two coordinate systems to describe the motion state of the UAV:

[0060] (1) Geodetic coordinate system E(O e ,X e ,Y e ,Z e )

[0061] like Figure 2 As shown, the origin O e is the starting point for the UAV to perform flight operations. If the take-off position of the UAV is set on the ground plane, then X e The axis is in the ground plane, with the east direction defined as the positive direction, Z e The axis is perpendicular to the ground plane, and the positive direction is vertically upward Y e Perpendicular to O e X e Z e The plane formed by three points has its positive direction determined by the right-hand rule.

[0062] (2) Body coordinate system B(O b ,X b ,Y b ,Z b )

[0063] Take a quad-rotor drone as an example. Figure 2 As shown, the body coordinate system is fixed to the rotor UAV, with the origin O b is the center of gravity of the rotary wing drone, X b The axis is in the direction of the nose, parallel to the bisector of the angle formed by rotors 1 and 4 and the center of gravity of the drone, and the positive direction is defined as the direction pointing to the side of rotors 1 and 4; Z b Perpendicular to X b Axis, define vertical upward as positive, Y b The axes are oriented according to the right-hand rule.

[0064] The two coordinate systems and their corresponding relationships are shown in the attached Figure 2 shown.

[0065] Step 2: Attitude conversion. The rotation angles between the axes of the "earth-body" coordinate system are called Euler angles, among which the roll angle, pitch angle, and yaw angle correspond to the rotations of the X, Y, and Z axes of the "earth-body" coordinate system, respectively. The conversion relationship of the "earth-body" dual coordinate system is achieved by rotating the three axes in sequence. The current study of the relationship between the two is generally carried out in the order of Z axis (yaw angle) first, then Y axis (roll angle), and finally X axis (pitch angle).

[0066] In one embodiment, the rotation matrix of three rotations around the axis in the coordinate transformation is as follows:

[0067]

[0068] By first converting the yaw angle, then converting the roll angle, and finally converting the pitch angle, the transformation matrix of the "earth-body" dual coordinate system can be obtained by multiplying the three matrices in the above formula:

[0069]

[0070] in, is the transformation matrix of the “earth-body” dual coordinate system, φ, θ, ψ are the yaw angle, roll angle, and pitch angle of the UAV in the earth coordinate system, respectively.

[0071] In one embodiment, the specific steps of establishing the rigid body model of the rotary-wing UAV are as follows: Step 1: Establishing the position motion model of the rotary-wing UAV. According to the relationship between force and mass acceleration according to Newton's second law, and the defined conversion relationship between the earth coordinate system and the body coordinate system, the position motion model of the rotary-wing UAV can be expressed as

[0072]

[0073] All letters of the present invention, with one dot or two dots added to the letters, represent the first-order derivative or the second-order derivative of the variable.

[0074] Where, P is the position of the rotorcraft in the geodetic coordinate system [xyz] T ; V——the speed of the rotorcraft in the geodetic coordinate system [v x v y v z ] T ; m——mass of rotary wing UAV; ——Transformation matrix from the body coordinate system to the earth coordinate system; g——gravitational acceleration; T F ——The sum of the lift of the rotors of a rotary-wing UAV;

[0075]

[0076] F i =λw i2 (i=1,2,3,4......n)(7)

[0077] In the formula, F i ——lift provided by the rotation of a single rotor; λ——lift coefficient; w i ——The rotation speed of a single rotor. According to formulas (5)-(7), the position motion model of the rotor UAV can be expressed as:

[0078]

[0079] Step 2: Establish a rigid body model of the rotorcraft UAV. According to the Newton-Euler equation, the force equation of the rotorcraft UAV in the geodetic coordinate system and the torque in the body coordinate system are as follows:

[0080]

[0081] Where, ω is the angular velocity of the rotor UAV; M is the attitude channel control torque; J is the moment of inertia matrix; ——Angular acceleration of rotorcraft (the first derivative of angular velocity is angular acceleration).

[0082] Taking a quad-rotor drone as an example, the resultant torque M applied to the quad-rotor drone causes the drone to move at an angular velocity ω and an angular acceleration Rotation. Expand the matrix and vector in the above formula, substitute the quadcopter's rotational inertia matrix, the attitude control torque under three axes, and the three Euler angle angular velocities and angular accelerations to get the following formula:

[0083]

[0084] In the formula

[0085] τ x , τ y , τ z ——X-axis (pitch angle), Y-axis (roll angle), and Z-axis (yaw angle) control torque;

[0086] J xx , J yy , J zz ——Moment of inertia of X-axis (pitch angle), Y-axis (roll angle), and Z-axis (yaw angle);

[0087] Rewriting the above formula into an expression of angular acceleration, we can get the following formula:

[0088]

[0089] In summary, the rigid body model of the rotorcraft drone is as follows:

[0090]

[0091] In the formula, is the acceleration of the rotorcraft in three directions of the geodetic coordinate system, is the angular velocity of the UAV’s yaw, roll and pitch angles, is the angular acceleration of the yaw angle, roll angle and pitch angle of the UAV, F is the sum of the rotor lift of the UAV, m is the mass of the UAV, g is the acceleration of gravity, J xx , J yy , J zz is the moment of inertia of the X-axis (pitch angle), Y-axis (roll angle) and Z-axis (yaw angle), τ x , τ y , τ z is the control torque of the X-axis (pitch angle), Y-axis (roll angle) and Z-axis (yaw angle).

[0092] In one embodiment, the aerodynamic model of the rotary wing UAV is established, and the specific steps are as follows:

[0093] Taking the "X"-shaped quad-rotor drone as an example, the expression for the torque generated by its four propellers is:

[0094]

[0095] Where, w1, w2, w3, w4 are the rotation speeds of the four rotors; c T ——Indicates the thrust coefficient of the propeller; c M ——torque coefficient of propeller; d——radius of the quadrotor drone fuselage; f i ——The lift of each propeller of a quadcopter drone.

[0096] For the above formula (13), when the number of rotors of the UAV is more than 4, f1-f4 can be changed to f1-f n , n is an even number among natural numbers. For example, when the number of rotors of a drone is eight, the expression for the torque generated by its eight propellers is:

[0097]

[0098] The above expression of eight rotors is only used as a demonstration of the deduction of the multi-rotor formula of the present invention. The following series of multi-rotor formulas can be derived in the same way, and the present invention will not be elaborated here.

[0099] The quad-rotor drone controls the flight state of the drone by controlling the speed of four motors. However, the speed change of each rotor will cause many changes in the position angle. In order to design the controller more conveniently, the rotor speed change is expressed as four virtual control quantities to achieve decoupling between the four virtual control quantities. The four control quantities are the total lift control quantity U1, the pitch angle control quantity U2, the roll angle control quantity U3 and the yaw angle control quantity U4. The virtual control quantity expression of the "X"-shaped quad-rotor drone is as follows:

[0100]

[0101] Using rotor aerodynamic analysis, under the influence of wind field interference, the lift generated by the rotor rotation is F t , the wind disturbance force on the rotor in the interfering wind field is F w , the wind speed of the interfering wind field is V w The angle between the wind direction in the interfering wind field and the OXY plane in the geodetic coordinate system is α. The forces on the rotor of the quadcopter under wind field interference are shown in the following figure. Figure 3 shown.

[0102] The high-speed rotation of the rotor generates upward lift and downward airflow. The airflow speed is the induced speed V of the rotor. t At this time, the total induced velocity generated by the interfering wind field is V d The relationship between the total induced speed, the rotor induced speed and the interfering wind field speed can be expressed as:

[0103]

[0104] According to the blade element momentum principle and the mechanical analysis of the rotor, the induced speed of the rotor can be expressed as:

[0105]

[0106] Where, ρ is the air density of the rotor; A is the frontal area of ​​the rotor;

[0107] Rotor lift force F t It can be obtained from the rotor speed, see formula (7), and replace F in the above formula (16) t It can be rewritten by the rotor speed expression as:

[0108]

[0109] The relationship between the four virtual control quantities and the rotational speed can be obtained. Then, the rotor rotational speed can be expressed by the virtual control formula as follows:

[0110]

[0111] The induced speed V of each rotor can be obtainedt When the UAV is disturbed by the interfering wind field, the lift generated by each rotor is the sum of the lift generated by the rotor speed and the wind force generated by the interfering wind field disturbance, and its expression is:

[0112] F d =F t +F w =2ρAV t V d (19)

[0113] In the above formula (19), the total wind force F d The direction and total induced velocity V d In the opposite direction, F t Rotor lift and V t The rotor induced speed is opposite, F w The direction of wind disturbance force and V w The wind disturbance direction is the same.

[0114] From formula (19), when there is no wind field interference, V d =V t , that is, the total lift of the rotor is the lift generated by the rotation of the rotor. It can be deduced that:

[0115] F w =2ρAV t V w (20)

[0116] When the wind direction is known, combined with the attitude angle information of the rotor UAV, the wind direction of the wind field and the angle of the rotor blade of the rotor UAV can be obtained. The interference force of the wind on the rotor UAV under the wind field interference can be calculated as:

[0117]

[0118] Where, T w is the total disturbance force of wind on the rotor UAV; M w is the torque generated by the total interference force; F di is the lift of the four rotors.

[0119] Combined with the formula in this section, the aerodynamic model of the rotorcraft taking into account wind disturbance can be deduced as follows:

[0120]

[0121] Where, T wx ,T wy ,T wz is the component force of the total disturbance force of wind on the rotor UAV in the three-axis direction, M wφ 、M wθ 、M wψIt is the torque generated by the total disturbance force on the three Euler angles (yaw angle, roll angle and pitch angle).

[0122] 2. Establish a wind field model; the specific steps are as follows:

[0123] In one embodiment, the natural wind is reconstructed into the following four wind speed models: constant wind V c , Rafale V z , Random Wind V r , gradient style V j The simulated natural wind speed is a combination of the above four winds:

[0124] V=V c +V z +V r +V j (twenty four)

[0125] The normal wind is the main body of natural wind, representing the overall level of wind force, and can be expressed by a constant value:

[0126] V C =k(25)

[0127] Gusts represent the sudden change of natural wind. In natural wind, there are some wind forces that suddenly appear and disappear due to interference from natural or human factors. The expression is:

[0128]

[0129] Where V zmax ——peak wind speed in gust; T1——starting time of gust; T g - duration of gusts;

[0130] Random wind reflects the random change characteristics of natural wind. The formation and propagation of natural wind are complicatedly affected by the objects it contacts. Therefore, there is a part of unpredictable random wind in natural wind, which can be expressed as:

[0131]

[0132] Where V smax ——Random wind peak value; R am ——random wind coefficient, which is a random value in [-1,1]; ω——fixed value in [0,2π]; ——A random value in [0,2π];

[0133] The gradual wind model represents the slow-changing characteristics of natural wind. As the natural wind moves, it will be blocked by various objects, and its wind force will decrease slowly over time. Its expression is:

[0134]

[0135] Where V jmax is the peak value of the gradually changing wind speed; T1' is the time when the gradually changing wind speed starts to change; T2' is the time when the gradually changing wind speed ends; T is the duration of the gradually changing wind;

[0136] 3. According to the rigid body model of the rotor UAV, the aerodynamic model of the rotor UAV and the wind field model, a first controller based on linear adaptive dynamic repetitive control is constructed;

[0137] In one embodiment, a first controller based on linear adaptive dynamic repetitive control is constructed, and the specific steps are as follows:

[0138] In one embodiment, the first controller is a second order control system including disturbances:

[0139]

[0140] Where, w(t) is the external disturbance of the controlled object, t is time; y, u are the output and input items of the system; b is the controller gain; ——the first derivative of the system output; ——The second-order derivative of the system output.

[0141] Formula (29) can also be rewritten as:

[0142]

[0143] In the formula, Where f is the total disturbance of the system. Select the state variable (the variable describing the current state of the dynamic system) x1 = y, The formula is organized as:

[0144]

[0145] In one embodiment, the linear extended state observer in the first controller is designed, and the specific steps are as follows:

[0146] The Linear Extended State Observer (LESO) in the first controller expands the total disturbance f of the system into a new state variable, thereby realizing the observation and elimination of the disturbance. The total disturbance f includes the internal uncertainty of the system (including steady and time-varying, linear and nonlinear disturbances) and external uncertainty (external disturbances). The total disturbance of the system is estimated in real time by constructing a linear extended state observer, and measures are taken in the first controller to compensate for the disturbance. Therefore, the state variables x1 and x2 are constructed, and the extended state variable x3 = f is constructed. The z1, z2, and z3 in formula (37) are approximately equal to x1, x2, and x3. x1, x2, and x3 are the state variables of the derived formula, and z1, z2, and z3 are the system state variable observation values ​​and the total disturbance observation values ​​of the system written according to the derived formula.

[0147] Then formula (31) can be organized as:

[0148]

[0149] The above equation can be rewritten into a state space equation as follows:

[0150]

[0151] In the formula, u and b have the same meaning as in formula (29). Formula (33) can be rewritten as follows:

[0152]

[0153] Formula (34) is a simplified abbreviation of Formula (33), where the coefficient matrix of Formula (33) is replaced by the letters of Formula (34);

[0154] in:

[0155]

[0156] For the system shown in the above equation, a linear extended state observer is designed. The designed linear extended state observer is shown as follows:

[0157]

[0158] Where Z is the observed value; K is the linear state observer; is the estimated output value of the system. The above formula can be rewritten as:

[0159]

[0160] Where, z1, z2 are the observed values ​​of system state variables (referring to the actual values ​​of the internal state of the system obtained by sensors); z3 is the total disturbance observed value of the system; k1, k2, k3 are the parameters of the linear state observer;

[0161] By finding the total disturbance observation value of the system, the disturbance can be eliminated in the first controller system, that is:

[0162]

[0163] In the formula, u0 is the intermediate control quantity; we can get:

[0164]

[0165] In one embodiment, the linear error feedback controller in the first controller is designed, and the specific steps are as follows:

[0166] From formula (39), it can be obtained that the first controller system is a dual integrator control system. According to the characteristics of the system, a PD controller can be used. The form of the PD controller, i.e., the linear error feedback control rate (Linear State Error Feedback, LSEF), is shown in the following formula:

[0167] u0=k p (ν-z1)+k d z2(40)

[0168] In the formula, k p ——PD controller proportional parameter; k d ——PD controller differential parameter;

[0169] Finally, the total disturbance estimate z3 is compensated and the control quantity u is obtained, that is, u=(u0-z3) / b.

[0170] LESO can solve the state observations z1 and z2 of the controlled object, and can also estimate the total disturbance observation value z3 acting on the controlled system. The LESO expression is as follows:

[0171]

[0172] LSEF calculates the feedback control quantity u0 through the input state observation and the expanded state, and then obtains the control output. After that, the state observation of the output value is subtracted from the output in LESO to obtain the observation error of the system. Finally, the total disturbance value is compensated to obtain the control quantity. The whole control algorithm is as follows:

[0173]

[0174] Introducing the linear extended state observer bandwidth w0 and the first controller bandwidth w c The above parameters can be simplified to: k1 = 3w0, k d =2w cAt the same time, b cannot obtain the true value and can be represented by b0, which represents the specific value of the controller gain b. Since b cannot obtain the true value, the approximate value b0 is used to replace b. Then the LADRC control algorithm parameters are finally set to w c , w0, and b0.

[0175] 4. Combining the rigid body model of the rotor UAV and the rotor aerodynamic model, the hybrid particle swarm grey wolf optimization algorithm is used to tune the parameters of the first controller to obtain the second controller;

[0176] The hybrid particle swarm grey wolf optimization algorithm is used to tune the parameters of the linear active disturbance rejection controller. The specific steps are as follows: In the particle swarm optimization algorithm, the method of updating the individual position and velocity of particles according to the individual historical optimal value and the group historical optimal value can be expressed by the formula:

[0177]

[0178] In the formula, represents the speed of particle id in the particle swarm in the kth generation; H represents the inertia weight; c1 and c2 represent individual and group learning factors; r1 and r2 represent random factors between [0,1]; Indicates the position of particle id at the kth iteration; represents the best position found by particle id before the kth generation; It represents the best position of all particles in the whole particle swarm after k iterations;

[0179] In the above parameters, c1 and c2 represent the feedback and communication ability between particles and the group, which is manifested as the particle's desire to learn the individual optimal value and the group optimal value. The larger the c1 and c2 are, the greater the amplitude of the particle's movement to the group's historical optimal position and the individual's historical optimal position. The random values ​​of r1 and r2 in the range of [0,1] are intended to increase the randomness of the particle's optimization process and improve the particle's fault tolerance and optimization ability. The optimal position that the population has reached in each iteration is represented by the matrix P. g =(p g1 ,p g2 ,L,p gD ) represents the optimal position of the group. The optimal value iteration formula of the population is shown in formula (45). H represents the emphasis of the particle swarm on the foraging strategy. When the value is relatively large, the group pays more attention to finding the optimal value in the global range. When the value is relatively small, it focuses on the local optimal search.

[0180]

[0181] In the formula, m is the fitness function, that is, when the fitness of the next iteration is greater than or equal to the previous one, the optimal position is recorded as X i(k+1), when the fitness of the next iteration is less than that of the previous one, the optimal position remains unchanged and is P i (k) P i (k+1) indicates the optimal position of the i-th particle in the k+1th generation; X i (k+1) indicates the optimal position of the i-th particle in the k+1th generation; P i (k) represents the optimal position of the i-th particle in the k-th generation; m(x i (k+1)) represents the optimal position fitness of the i-th particle in the k+1th generation, x i (k+1) represents the corresponding objective function value; m(x i (k)) represents the optimal position fitness of the i-th particle in the k-th generation, x i (k) represents the corresponding objective function value. Specifically, if the objective function value x corresponding to the new position of a particle i (k+1) is greater than the objective function value x at its old position i If (k) is greater, update its position to the new position. Otherwise, keep its position unchanged.

[0182] The trade-off between global search and local search in intelligent optimization algorithms affects the algorithm's ability to find the best solution. Global search tends to find the global optimum in the entire search space, while local search tends to find better individuals around existing individuals. When the gray wolf optimization algorithm iterates, it only records the three historical optimal positions in the wolf pack and the contemporary position information of the gray wolf. The position of the gray wolf is mainly iterated based on the position information of the head wolf. Its position update strategy is single and it is easy to fall into the local optimal solution. The particle swarm optimization algorithm records the historical optimal value of each particle, and its parameters are simple and easy to adjust.

[0183] Therefore, the particle swarm optimization algorithm and the gray wolf optimization algorithm are combined to improve the algorithm's optimization ability and convergence speed. The particle position update replaces the gray wolf individual position update, so that it can remember the best position information in its own iteration process. Even if the position information of the gray wolf individual considers not only the position of the alpha wolf, but also the position of the gray wolf individual's own historical optimal fitness value. The fused speed position update method is shown in the following formula:

[0184]

[0185] in, ——the speed of the individual ij in the wolf pack in the k+1th generation; H——inertia weight; c1, c2——individual and group learning factors; r1, r2——random factors between [0,1]; ——The position of the gray wolf individual ij at the kth iteration; ——The best position found by the gray wolf individual ij before the kth generation; ——The best position among all the individual gray wolves in the entire gray wolf pack after k iterations;

[0186] The hybrid particle swarm gray wolf optimization algorithm will initialize the position of each individual, and the position of each individual is the w0 and w0 in the first controller to be optimized. c , b0, and input each individual position in the population as a parameter into the Simulink model, find the error between the expected height and the actual height calculated by the model, and then integrate the absolute value of the error and the product of time to get the fitness value, which is introduced into the main program of the hybrid particle swarm gray wolf optimization algorithm to calculate the global and individual optimal values, and based on this, complete the position update of each individual in the population, and input the updated position as a parameter into the Simulink model again to calculate the fitness until the population iteration is completed. The final optimization result is the optimal value of this optimization, and the position of the optimal value is the best parameter.

[0187] 5. The second controller is optimized using the fuzzy control method to obtain the UAV anti-wind disturbance controller.

[0188] In one embodiment, Figure 5 As shown, the quantitative deviation e and the deviation change rate e c As the fuzzy input, the fuzzy logic control rules are used to adjust the linear error feedback controller parameters online, and the fuzzy controller output is w c The adjustment values ​​of the three parameters w0, b0 are denoted as Vw0, Vw c 、Vb0. ;

[0189] Specify two fuzzy input variables e and e c With three fuzzy output variables Vw0, Vw c The seven fuzzy control subsets of Vb0 are {NB, NM, NS, ZO, PS, PM, PB}, and the fuzzy rule table of the above five variables is designed by using the triangular membership function and the Mamdani type fuzzy inference rule. The fuzzy controller mainly completes the control function by the three core links of fuzzification, fuzzy inference and defuzzification. The process is as follows: Figure 4 shown.

[0190] The specific steps of formulating fuzzy inference rules are as follows: Step 1: fuzzifier dimension selection, the present invention selects a two-dimensional fuzzy controller; Step 2: fuzzy set, fuzzy domain quantization factor selection, the present invention selects the input variable quantitative deviation e and deviation change rate e cThe fuzzy sets are divided into 7 levels. The fuzzy set of quantitative deviation e is {NB, NM, NS, ZO, PS, PM, PB}. Among them, NB, NM, NS represent that the input deviation is negative, that is, the actual value of the system is higher than the expected value. Their meanings are negative large, negative medium and negative small, which respectively mean that the actual value is much higher, higher and a little higher than the expected value. ZO means that the actual value is equal to the expected value. PS, PM, PB represent that the input deviation is positive, that is, the actual value is lower than the expected value. The three fuzzy language meanings are positive small, positive medium and positive large, which respectively mean that the actual value is a little lower, lower and much lower than the expected value. The fuzzy set of the input deviation change rate is {NB, NM, NS, ZO, PS, PM, PB}. Among them, NB, NM, NS represent that the actual value has a tendency to increase compared with the expected value, which means fast quantitative increase speed, moderate increase speed and slow increase speed respectively; ZO means that the actual value changes synchronously with the expected value or does not change; PS, PM, PB represent that the actual value has a tendency to decrease compared with the expected value, which means that the input value decreases slowly, moderately and fast. The output variable fuzzy set is divided into 7 levels, among which {NB, NM, NS, ZO, PS, PM, PB} represent the output parameter adjustment state of negative large, negative medium, negative small, zero, positive small, positive medium and positive large respectively. Step 3: Membership function selection. Combined with the characteristics of LADRC controller, this paper selects triangular membership functions with fast operation speed and suitable for continuous input to ensure the reasoning efficiency and accuracy of fuzzy controller. At the same time, the division of fuzzy domain will also affect the performance of the controller. If the fuzzy domain is small, there will be parts that are not covered, affecting the control ability; if it is too large, overlap will occur, affecting the reasoning speed. This paper divides the fuzzy domain evenly to cover more areas and minimize the overlap area. The fuzzy rule table is shown in the following table:

[0191] Table 1 Vw0 fuzzy rule table

[0192]

[0193] Table 2Vw c Fuzzy rule table

[0194]

[0195]

[0196] Table 3 Vb0 fuzzy rule table

[0197]

[0198] Furthermore, the detailed steps of building a simulink model in matlab for simulation experiments are as follows:

[0199] Step 1: Build a rotor UAV attitude and position controller in matlab / simulink; Step 2: Build a wind field model simulation module in matlab / simulink; Step 3: Verify the performance of the controller proposed in the present invention under gusts; Step 4: Verify the performance of the controller proposed in the present invention under gradually changing wind; Step 5: Verify the performance of the controller proposed in the present invention under combined wind.

[0200] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. An anti-wind disturbance controller for a UAV, characterized in that: The following design method is used: Define the coordinate system, which includes the earth coordinate system and the body coordinate system. The transformation matrix of the "earth-body" dual coordinate system is: in, is the transformation matrix of the "earth-body" dual coordinate system, φ, θ, ψ are the yaw angle, roll angle and pitch angle of the UAV in the earth coordinate system respectively; Establish the rigid body model and aerodynamic model of the rotary-wing UAV; A wind field model is established, wherein the wind field model is a combination of constant wind, gust, rapid wind and gradually changing wind: V=V c +V z +V r +V j Where V is the wind speed of the wind field model, V c is the constant wind, V z For gusts, V r is random wind, V j For the gradual wind; The constant wind expression is: V C =k, k is a constant value; The wind gust expression is: Among them, V zmax is the peak wind speed in the gust, T1 is the start time of the gust, T g is the duration of the gust; The random wind expression is: Among them, V smax is the peak value of random wind force, R am is the coefficient of random wind, which is a random value in [-1,1], and ω is a fixed value in [0,2π], is a random value in [0,2π]; The gradient wind expression is: Among them, V jmax is the peak wind speed of the gradually changing wind, T1' is the time when the gradually changing wind starts to change, T2' is the time when the gradually changing wind ends, and T is the duration of the gradually changing wind; According to the rigid body model, aerodynamic model and wind field model of the rotor UAV, a first controller based on linear adaptive dynamic repetitive control is constructed. The linear error feedback controller in the first controller system adopts a PD controller, and the expression is: u0=k p (n-z1)+k d z2 Among them, V is the wind speed of the wind field model, z1 and z2 are the state variable observation values ​​of the first controller system, u0 is the intermediate control quantity, k p is the proportional parameter of the PD controller, k d is the differential parameter of the PD controller; Combining the rigid body model and the rotor aerodynamic model of the rotor UAV, the hybrid particle swarm grey wolf optimization algorithm is used to perform parameter tuning on the first controller to obtain a second controller; The hybrid particle swarm grey wolf optimization algorithm is expressed as in, represents the speed of the gray wolf individual ij in the gray wolf pack in the k+1th generation; H is the inertia weight; c1 and c2 represent the individual and group learning factors; r1 and r2 are random factors between [0,1]; represents the position of the gray wolf individual ij at the kth iteration; represents the best position found by the gray wolf individual ij before the kth generation; It represents the best position of all the individual gray wolves in the entire gray wolf pack after k iterations; The second controller is optimized using a fuzzy control method to obtain a UAV anti-wind disturbance controller, including: The quantitative deviation e and the deviation change rate e c As the fuzzy input, the fuzzy logic control rules are used to adjust the linear error feedback controller parameters online, and the fuzzy controller output is w c The adjustment values ​​of the three parameters w0, b0 are denoted as Vw0, Vw c , Vb0; Specify two fuzzy input variables e and e c With three fuzzy output variables Vw0, Vw c The seven fuzzy control subsets of Vb0 are {NB, NM, NS, ZO, PS, PM, PB}, and the fuzzy rule table of the above five variables is designed by adopting the triangular membership function and applying the Mamdani type fuzzy inference rule.

2. The anti-wind disturbance controller for a UAV as claimed in claim 1, characterized in that: The rigid body model of the rotorcraft UAV is as follows: in, is the acceleration of the rotorcraft in three directions of the geodetic coordinate system, is the angular velocity of the UAV’s yaw, roll and pitch angles, is the angular acceleration of the yaw angle, roll angle and pitch angle of the UAV, F is the sum of the rotor lift of the UAV, m is the mass of the UAV, g is the acceleration of gravity, J xx , J yy , J zz is the moment of inertia of the X-axis (pitch angle), Y-axis (roll angle) and Z-axis (yaw angle), τ x , τ y , τ z It is the control torque of X-axis (pitch angle), Y-axis (roll angle) and Z-axis (yaw angle). The aerodynamic model of the rotor UAV is as follows: Among them, T wx 、T wy 、T wz is the component force of the total disturbance force of wind on the rotor UAV in the three-axis direction, M wφ 、M wθ 、M wψ is the torque generated by the total disturbance force on the yaw angle, roll angle and pitch angle.

3. A method for controlling an unmanned aerial vehicle against wind disturbance, characterized in that: The UAV anti-wind disturbance controller according to any one of claims 1 to 2 is used to perform anti-wind disturbance flight control on the UAV.

Images