Periodic attitude tracking task-oriented unmanned aerial vehicle predetermined time anti-interference learning control method

Abstract

A periodic attitude tracking task-oriented unmanned aerial vehicle predetermined time anti-interference learning control method comprises the following steps: establishing a kinetic model of a four-rotor aircraft system, dividing the uncertainty of the system into a periodic uncertainty part and a non-periodic uncertainty part, and initializing the state and control parameters of the system; designing a predetermined time disturbance observer, and quickly estimating and compensating a non-periodic uncertain part; designing a predetermined time sliding mode variable; designing a full-saturation repetitive learning law, and accurately estimating and compensating the uncertain part of the period; and designing a repetitive learning controller based on the predetermined time disturbance observer, the predetermined time sliding mode variable and the full-saturation repetitive learning law. According to the method, the influence of the system uncertainty on the control precision of the unmanned aerial vehicle can be effectively suppressed, and it is ensured that the upper bound of the minimum convergence time of the observation error can be preset by adjusting an adjustable parameter, so that the tracking error is converged to the neighborhood near the original point within the preset time.

Description

Technical Field

[0001] The invention belongs to the technical field of attitude control of unmanned aerial vehicles. Aiming at the uncertainty in the unmanned aerial vehicle system and the system performing periodic attitude tracking tasks, a predetermined time anti-interference learning control method of the unmanned aerial vehicle for periodic attitude tracking tasks is proposed. Background Art

[0002] In recent years, unmanned aerial vehicles have attracted widespread attention from researchers. Quadrotors are a typical type of unmanned aerial vehicle that has the characteristics of low cost, small size, precise hovering, vertical take-off and landing, and has been widely used in various fields.

[0003] The control system of a quadrotor includes attitude, speed and position control, among which attitude control is the core inner loop of the controller. Its stability and control performance are crucial to the flight performance of the entire quadrotor system. In actual work, quadrotors are usually involved in periodic attitude tracking tasks, such as aerial photography, search and rescue, agricultural inspection, etc. However, existing nonlinear control technologies rarely use the periodic feature, making it difficult to achieve high-precision control. Repeated learning control is suitable for controlled objects that operate periodically. By continuously optimizing the control performance using previous operation data, the tracking accuracy of the system for periodic tasks can be effectively improved. Therefore, it is very meaningful to conduct in-depth research on unmanned aerial vehicle control methods for periodic attitude tracking tasks.

[0004] Accurate, fast and reliable aircraft attitude control has always been one of the hot issues in the research of unmanned aerial vehicle systems. In order to improve the convergence speed of the system, researchers have proposed finite time and fixed time controllers. However, the upper bound of its convergence time requires more complex calculations and cannot be obtained by only one scalar. Scheduled time control is a control method for time optimization. Compared with finite time control and fixed time control, scheduled time control is less dependent on the initial value of the system and does not require complex system parameters. At the same time, the minimum upper bound of the system state convergence time can be explicitly contained in a single system parameter. Therefore, the scheduled time control method can pre-set the minimum upper bound of the system convergence time by setting this parameter in advance.

[0005] In practical applications, quadrotors face challenges such as multiple external disturbances (such as gusts, load changes) and system uncertainties (such as parameter disturbances, unmodeled dynamics), which may lead to significant control deviations. Therefore, it is necessary to adopt corresponding anti-disturbance strategies for quadrotors. At present, some work has been done to estimate and compensate for disturbances by designing finite-time disturbance observers and fixed-time disturbance observers. However, these disturbance observers contain discontinuous sign functions, which may cause system jitter, affect system stability and cause mechanical wear. In addition, in these methods, the upper bound of the convergence time of the observation error is difficult to give accurately and cannot be given in advance according to actual needs. Therefore, in order to further improve the rapidity of disturbance estimation and compensation, and to enable the minimum upper bound of the convergence time of the observation error to be determined in advance by an adjustable parameter, it is necessary to design a predetermined time disturbance observer and ensure the continuity of the observer.

[0006] Therefore, for unmanned aerial vehicles with periodic attitude tracking tasks and system uncertainties, it is of great research significance to design a control method that can quickly estimate and compensate for system uncertainties, make the minimum convergence time upper bound of the observation error only related to an adjustable parameter, and enable the system to converge to a neighborhood near the origin within a predetermined time. Summary of the invention

[0007] In order to reduce the influence of system uncertainty on the periodic attitude tracking task of the unmanned aerial vehicle, and at the same time improve the accuracy of the minimum convergence time upper bound of the attitude tracking error, so that the tracking error converges to the neighborhood near the origin within a preset time, the present invention proposes a predetermined time anti-interference learning control method for unmanned aerial vehicles for periodic attitude tracking tasks. The proposed method first divides the uncertainty in the system into two parts: periodic uncertainty and non-periodic uncertainty. Secondly, a predetermined time disturbance observer is constructed to estimate the non-periodic uncertainty in the system, and the continuity of the observer is guaranteed. A fully saturated repetitive learning law is designed to estimate the periodic uncertainty in the system. On this basis, a fully saturated repetitive learning controller is designed. This design can effectively suppress the influence of system uncertainty on the control accuracy of the unmanned aerial vehicle, ensure that the minimum convergence time upper bound of the observation error can be preset by adjusting an adjustable parameter, and at the same time ensure that the tracking error converges to the neighborhood near the origin within a predetermined time.

[0008] The technical solutions proposed to solve the above technical problems are as follows:

[0009] A method for predictive time anti-interference learning control of an unmanned aerial vehicle for a periodic attitude tracking task comprises the following steps:

[0010] Step 1: Establish the dynamic model of the quadrotor system and initialize the state and control parameters of the system. The process is as follows:

[0011] 1.1 The dynamics of a quadrotor is expressed as:

[0012]

[0013] Among them, τ x ,τ y ,τ z are the control moments acting on the roll axis, pitch axis and yaw axis of the quadrotor in the body coordinate system; I xx ,I yy ,I zz are the moments of inertia of each axis of the quadrotor in the body coordinate system; ω x ,ω y ,ω z are the angular velocities of each axis of the quadrotor in the body coordinate system; They are ω x ,ω y ,ω z The first-order derivative of represents the angular acceleration of each axis of the quadrotor in the body coordinate system; × represents the cross product; · represents the dot product;

[0014] 1.2 Through formula (1), the expression of the attitude angular acceleration of each axis of the quadrotor aircraft during motion is obtained as follows:

[0015]

[0016] 1.3 Define the following state variables:

[0017]

[0018] Among them, φ, θ, ψ are the roll axis attitude angle, pitch axis attitude angle and yaw axis attitude angle of the quadrotor aircraft in the body coordinate system respectively; x1 represents the attitude angle state quantity, and x2 represents the attitude angular velocity state quantity;

[0019] 1.4 In actual work, considering that the quadcopter may be subject to external interference, combining equations (2)-(3), the following state equation is constructed:

[0020]

[0021] in, Represent the first-order derivatives of x1 and x2 respectively; represents the control torque;

[0022]

[0023] Among them, d x ,d y ,d zThey are respectively represented as external disturbances acting on the control channels of each axis of the quadrotor in the body coordinate system;

[0024] Step 2, taking the yaw axis of a quadrotor as an example, design a predetermined time disturbance observer. The process is as follows:

[0025] 2.1 Combined with formula (4), the following quadrotor yaw axis attitude angle system model is obtained:

[0026]

[0027] Among them, x ψ1 ,x ψ2 are the yaw-axis attitude angle and yaw-axis attitude angular velocity of the quadrotor in the body coordinate system, x ψ1 ,x ψ2 The first derivative of ; They are the yaw axis control torque gain, yaw axis control torque, yaw axis system term and external disturbance on the yaw axis control channel of the quadrotor aircraft in the body coordinate system;

[0028] 2.2 In the actual flight of the quadrotor, considering the error between the model parameters and the true values, the yaw axis system term f ψ and the external disturbance d on the yaw axis control channel ψ The sum of ′ is regarded as the lumped uncertainty Δ ψ , that is, Δ ψ =f ψ +d ψ ′, the aggregate uncertainty Δ ψ Divided into periodic terms f ψr and the non-periodic term d ψ , combined with formula (5), the system model is transformed into the following form:

[0029]

[0030] Among them, the non-periodic term d ψ =f ψ -f ψr +d ψ ′;

[0031] 2.3 Assuming the aggregate uncertainty Δ ψ and the first derivative of the lumped uncertainty There is an upper bound, let D be Δ ψ The upper limit value of L is The upper bound value of

[0032] 2.4 The following observation errors are defined:

[0033]

[0034] Among them, η1, η2 represent the observation error, z ψ1 Represents state x ψ2 The estimated value of z ψ2 represents the non-periodic term d ψ An estimated value of

[0035] 2.5 For any constant scalars λ and δ, define sig λ The (δ) function is:

[0036] sig λ (δ)=|δ| λ sign(δ) (8)

[0037] Among them, sign(·) represents the sign function;

[0038] 2.6 For the yaw axis attitude angle system model (6), design a predetermined time disturbance observer:

[0039]

[0040] in, Represents the estimated value z ψ1 The first derivative of; 0<k1<1,k2>D,k>0,T s represents a predetermined time constant, tanh(·) represents a hyperbolic tangent function;

[0041] Step 3, design the sliding mode variable with a predetermined time. The process is as follows:

[0042] 3.1 Define the following attitude angle tracking error:

[0043] e1=x ψ1 -x ψr (10)

[0044] Among them, e1 represents the yaw axis attitude angle tracking error of the quadrotor aircraft, x ψr Indicates the desired yaw axis attitude angle of the quadrotor;

[0045] Derivation of formula (10) yields:

[0046]

[0047] in, Represent e1,x respectively ψr The first derivative of ;

[0048] 3.2 Design the following scheduled time sliding mode variables:

[0049]

[0050] in,

[0051]

[0052] Where s represents the predetermined time sliding mode variable, and the predetermined time constant T c >0,0<γ<1, l1=(1+γ)ε -γ ,l2=(-γ)ε -γ-1 ,ε>0;

[0053] Step 4: Design a fully saturated repetitive learning law to compensate for the periodic term in the quadrotor system. The process is as follows:

[0054] 4.1 Design the fully saturated repeated learning law as follows:

[0055]

[0056] in, f ψr (t), The estimated value of for The value at time tT; μ0>0 is the repeated learning gain coefficient; φ(t) is the transition function, and the expression is:

[0057]

[0058] sat(·) is the saturation function, and its expression is:

[0059]

[0060] in, is the limit value of Υ,

[0061] Step 5: Design a preset time sliding mode repetitive learning controller.

[0062] Further, the process of step 5 is as follows:

[0063] 5.1 Taking the derivative of equation (12), we get:

[0064]

[0065] in,

[0066]

[0067] in, Respectively The derivative of

[0068] 5.2 Combining equations (9), (14) and (17), the preset time sliding mode controller is designed as:

[0069]

[0070] in, express The first derivative of , constant n ≥ 0.5;

[0071] 5.3 Choose the following Lyapunov function:

[0072]

[0073] Derivative of formula (20) yields:

[0074]

[0075] in, represents the first-order derivative of V1(t);

[0076] Substituting equation (17) and equation (19) into equation (21), we get That is, the system is judged to be stable.

[0077] The present invention designs a predetermined time anti-interference learning control method for an unmanned aerial vehicle for periodic attitude tracking tasks. The proposed predetermined time disturbance observer can quickly estimate and compensate for the non-periodic uncertainty in the system and ensure the continuity of the observer. The designed fully saturated repetitive learning law can well estimate and compensate for the periodic uncertainty in the system. On this basis, a predetermined time sliding mode controller is designed to improve the anti-interference performance of the system and the convergence speed of the tracking error, so as to realize the accurate tracking of the periodic desired attitude angle of the unmanned aerial vehicle within the preset parameters.

[0078] The technical concept of the present invention is: for unmanned aerial vehicles with periodic attitude tracking tasks and system uncertainty, the system uncertainty is first divided into two parts: periodic uncertainty and non-periodic uncertainty, and a predetermined time disturbance observer is designed to quickly estimate the unknown state of the system and compensate for the non-periodic uncertainty, thereby improving the anti-interference ability of the system, and ensuring that the upper limit of the minimum convergence time of the observation error can be explicitly contained in an adjustable parameter, which can be set in advance according to actual needs. In addition, the designed fully saturated repetitive learning law estimates and compensates for the periodic uncertainty in the system, and a predetermined time sliding mode controller is designed to achieve high-precision tracking of the periodic desired attitude angle of the unmanned aerial vehicle within a predetermined time.

[0079] The effective effects of the present invention are as follows: the unmanned aerial vehicle is able to track a periodic desired attitude angle within a preset time, and the influence of system uncertainty on the attitude tracking accuracy of the unmanned aerial vehicle is suppressed. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 is a control flow chart of the present invention;

[0081] Figure 2 It is the yaw angle tracking effect;

[0082] Figure 3 is the yaw angle tracking error signal;

[0083] Figure 4 is the sliding mode variable signal;

[0084] Figure 5 is the control torque signal;

[0085] Figure 6 Estimating curves for repeated learning;

[0086] Figure 7 is the observation effect of the disturbance observer. DETAILED DESCRIPTION

[0087] The present invention will be further described below in conjunction with the accompanying drawings.

[0088] Reference Figure 1-Figure 7 , a quadrotor aircraft attitude repetitive learning control method based on a predetermined time disturbance observer, comprising the following steps:

[0089] Step 1: Establish the dynamic model of the quadrotor system and initialize the state and control parameters of the system. The process is as follows:

[0090] 1.1 The dynamics of a quadrotor is expressed as:

[0091]

[0092] Among them, τ x ,τ y ,τ z are the control moments acting on the roll axis, pitch axis and yaw axis of the quadrotor in the body coordinate system respectively; I xx ,I yy ,I zz are the moments of inertia of each axis of the quadrotor in the body coordinate system; ω x ,ω y ,ω z are the angular velocities of each axis of the quadrotor in the body coordinate system; They are ω x ,ω y ,ω z The first-order derivative of represents the angular acceleration of each axis of the quadrotor in the body coordinate system; × represents the cross product; · represents the dot product;

[0093] 1.2 Through formula (1), the expression of the attitude angular acceleration of each axis of the quadrotor aircraft during motion is obtained as follows:

[0094]

[0095] 1.3 Define the following state variables:

[0096]

[0097] Among them, φ, θ, ψ are the roll axis attitude angle, pitch axis attitude angle and yaw axis attitude angle of the quadrotor aircraft in the body coordinate system respectively; x1 represents the attitude angle state quantity, and x2 represents the attitude angular velocity state quantity;

[0098] 1.4 In actual work, considering that the quadcopter may be subject to external interference, combining equations (2)-(3), the following state equation is constructed:

[0099]

[0100] in, Represent the first-order derivatives of x1 and x2 respectively; represents the control torque;

[0101]

[0102] Among them, d x ,d y ,d z They are respectively represented as external disturbances acting on the control channels of each axis of the quadrotor in the body coordinate system;

[0103] Step 2, taking the yaw axis of a quadrotor as an example, design a predetermined time disturbance observer. The process is as follows:

[0104] 2.1 Combined with formula (4), the following quadrotor yaw axis attitude angle system model is obtained:

[0105]

[0106] Among them, x ψ1 ,x ψ2 are the yaw-axis attitude angle and yaw-axis attitude angular velocity of the quadrotor in the body coordinate system, x ψ1 ,x ψ2 The first derivative of ; They are the yaw axis control torque gain, yaw axis control torque, yaw axis system term and external disturbance on the yaw axis control channel of the quadrotor aircraft in the body coordinate system;

[0107] 2.2 In the actual flight of the quadrotor, considering the error between the model parameters and the true values, the yaw axis system term f ψ and the external disturbance d on the yaw axis control channel ψ The sum of ′ is regarded as the lumped uncertainty Δ ψ , that is, Δ ψ =f ψ +d ψ ′, the aggregate uncertainty Δ ψ Divided into periodic terms f ψr and the non-periodic term d ψ , combined with formula (5), the system model is transformed into the following form:

[0108]

[0109] Among them, the non-periodic term d ψ =f ψ -f ψr +d ψ ′;

[0110] 2.3 Assuming the aggregate uncertainty Δ ψ and the first derivative of the lumped uncertainty There is an upper bound, let D be Δ ψ The upper limit value of L is The upper bound value of

[0111] 2.4 The following observation errors are defined:

[0112]

[0113] Among them, η1, η2 represent the observation error, z ψ1 Represents state x ψ2 The estimated value of z ψ2 represents the non-periodic term d ψ An estimated value of

[0114] 2.5 For any constant scalars λ and δ, define sig λ The (δ) function is:

[0115] sig λ (δ)=|δ λ sign(δ) (29)

[0116] Among them, sign(·) represents the sign function;

[0117] 2.6 For the yaw axis attitude angle system model (6), design a predetermined time disturbance observer:

[0118]

[0119] in, Represents the estimated value z ψ1 The first derivative of; 0<k1<1,k2>D,k>0,T s represents a predetermined time constant, tanh(·) represents a hyperbolic tangent function;

[0120] Step 3, design the sliding mode variable with a predetermined time. The process is as follows:

[0121] 3.1 Define the following attitude angle tracking error:

[0122] e1=x ψ1 -x ψr (31)

[0123] Among them, e1 represents the yaw axis attitude angle tracking error of the quadrotor aircraft, x ψr Indicates the desired yaw axis attitude angle of the quadrotor;

[0124] Derivation of formula (10) yields:

[0125]

[0126] in, Represent e1,x respectively ψr The first derivative of ;

[0127] 3.2 Design the following scheduled time sliding mode variables:

[0128]

[0129] in,

[0130]

[0131] Where s represents the predetermined time sliding mode variable, and the predetermined time constant T c >0,0<γ<1, l1=(1+γ)ε -γ ,l2=(-γ)ε -γ-1 ,ε>0;

[0132] Step 4: Design a fully saturated repetitive learning law to compensate for the periodic term in the quadrotor system. The process is as follows:

[0133] 4.1 Design the fully saturated repeated learning law as follows:

[0134]

[0135] in, They are The estimated value of for The value at time tT; μ0>0 is the repeated learning gain coefficient; φ(t) is the transition function, and the expression is:

[0136]

[0137] sat(·) is the saturation function, and its expression is:

[0138]

[0139] in, is the limit value of Υ,

[0140] Step 5: Design a preset time sliding mode repetitive learning controller. The process is as follows:

[0141] 5.1 Taking the derivative of equation (12), we get:

[0142]

[0143] in,

[0144]

[0145] in, Respectively The derivative of

[0146] 5.2 Combining equations (9), (14) and (17), the preset time sliding mode controller is designed as:

[0147]

[0148] in, express The first derivative of , constant n ≥ 0.5;

[0149] 5.3 Choose the following Lyapunov function:

[0150]

[0151] Derivative of formula (20) yields:

[0152]

[0153] in, represents the first-order derivative of V1(t);

[0154] Substituting equation (17) and equation (19) into equation (21), we get That is, the system is judged to be stable.

[0155] In order to verify the effectiveness of the proposed method, the present invention conducts a simulation experiment on the control effect shown in formula (19). The initial conditions and control parameters in the experiment are set as follows: the sampling step is 0.001s; the system parameters are selected as I xx =0.0552,I yy =0.0551,I zz =0.0555; the parameters of the preset time disturbance observer are selected as k1=0.5, k2=0.1, k=0.1, T s =5s; repeated learning parameters are The controller parameters are γ=0.3, ε=0.0001, T c =5s,n=1; the expected attitude angle is x φr =5sin(0.2πt),x θr =5sin(0.2πt),x ψr =5sin(0.2πt); the non-periodic uncertainty term is set to d=0.01rand; the initial state of the disturbance observer is set to (z φ1 ,z θ1 ,z ψ1 ,z φ2 ,z θ2 ,z ψ2 )=(0,0,0,0,0,0); the initial state of the system is set to (x φ1 ,x θ1 ,x ψ1 ,x φ2 ,x θ2 ,x ψ2 )=(2,2,2,0,0,0).

[0156] Figure 2-Figure 7 The simulation results of the proposed unmanned aerial vehicle scheduled time anti-interference learning control method for periodic attitude tracking tasks on system (6) are shown in FIG. Figure 2 This is the yaw angle tracking effect diagram. It can be seen from the figure that the yaw angle tracks the expected signal in about 2s, and then the expected yaw angle signal is maintained with good consistency; Figure 3 is the yaw angle tracking error signal. It can be seen from the figure that the yaw angle tracking error converges to near 0 in about 2s, and the steady-state error is ±5×10 -5 rad within the bounds; Figure 4 It is the sliding mode variable signal. It can be seen from the figure that the sliding mode variable converges to near 0 at around 0.05s; Figure 5 It is the control torque signal. It can be seen from the figure that the control signal is periodic and relatively smooth. Figure 6 is the repeated learning estimation curve. It can be seen from the figure that the designed repeated learning law can achieve accurate estimation of the system periodic term; Figure 7It is the observation effect of the disturbance observer. It can be seen from the figure that the disturbance observer can estimate the disturbance and system state very well. In summary, the scheduled time anti-interference learning control method for unmanned aerial vehicle for periodic attitude tracking tasks can effectively realize the scheduled time tracking of the attitude angle of the unmanned aerial vehicle, suppress the influence of system uncertainty on the accuracy of the attitude control of the unmanned aerial vehicle, and realize accurate estimation of the system state and high-precision tracking of the periodic expected attitude angle.

[0157] The above is a simulation experiment given by the present invention to show the superiority of the designed method. Obviously, the present invention is not limited to the above examples. It can be implemented in various ways without departing from the basic spirit of the present invention and not exceeding the scope of the essential content of the present invention. The control method designed by the present invention has a good control effect on the periodic expected attitude angle of the unmanned aerial vehicle tracking period with system uncertainty, and can effectively improve the tracking accuracy and rapidity of the unmanned aerial vehicle.

Claims

1. A method for predictive time anti-interference learning control of an unmanned aerial vehicle for a periodic attitude tracking task, characterized in that the method comprises the following steps: Step 1: Establish the dynamic model of the quadrotor system and initialize the state and control parameters of the system. The process is as follows: 1.1 The dynamics of a quadrotor is expressed as: Among them, τ x ,τ y ,τ z are the control moments acting on the roll axis, pitch axis and yaw axis of the quadrotor in the body coordinate system; I xx ,I yy ,I zz are the moments of inertia of each axis of the quadrotor in the body coordinate system; ω x ,ω y ,ω z are the angular velocities of each axis of the quadrotor in the body coordinate system; They are ω x ,ω y ,ω z The first-order derivative of represents the angular acceleration of each axis of the quadrotor in the body coordinate system; × represents the cross product; · represents the dot product; 1.2 Through formula (1), the expression of the attitude angular acceleration of each axis of the quadrotor aircraft during motion is obtained as follows: 1.3 Define the following state variables: in, φ, θ, ψ are the roll axis attitude angle, pitch axis attitude angle and yaw axis attitude angle of the quadrotor aircraft in the body coordinate system respectively; x1 represents the attitude angle state quantity, and x2 represents the attitude angular velocity state quantity; 1.4 Considering that the quadrotor aircraft may be subject to external interference, combining equations (2)-(3), the following state equation is constructed: in, Represent the first-order derivatives of x1 and x2 respectively; represents the control torque; Among them, d x ,d y ,d z They are respectively represented as external disturbances acting on the control channels of each axis of the quadrotor in the body coordinate system; Step 2, taking the yaw axis of a quadrotor as an example, design a predetermined time disturbance observer. The process is as follows: 2.1 Combined with formula (4), the following quadrotor yaw axis attitude angle system model is obtained: in, x ψ1 ,x ψ2 are the yaw-axis attitude angle and yaw-axis attitude angular velocity of the quadrotor in the body coordinate system, x ψ1 ,x ψ2 The first derivative of ; They are the yaw axis control torque gain, yaw axis control torque, yaw axis system term and external disturbance on the yaw axis control channel of the quadrotor aircraft in the body coordinate system; 2.2 In the actual flight of the quadrotor, considering the error between the model parameters and the true values, the yaw axis system term f ψ and the external disturbance d on the yaw axis control channel ψ The sum of ′ is regarded as the lumped uncertainty Δ ψ , that is, Δ ψ =f ψ +d ψ ′, the aggregate uncertainty Δ ψ Divided into periodic terms f ψr and the non-periodic term d ψ , combined with formula (5), the system model is transformed into the following form: Among them, the non-periodic term d ψ =f ψ -f ψr +d ψ ′; 2.3 Assuming the aggregate uncertainty Δ ψ and the first derivative of the lumped uncertainty There is an upper bound, let D be Δ ψ The upper limit value of L is The upper bound of |Δ ψ |≤D, 2.4 The following observation errors are defined: in, η1,η2 represent the observation error, z ψ1 Represents state x ψ2 The estimated value of z ψ2 represents the non-periodic term of the system d ψ An estimated value of 2.5 For any constant scalars λ and δ, define sig λ The (δ) function is: sig λ (d)=|d λ sign(δ) (8) Among them, sign(·) represents the sign function; 2.6 For the yaw axis attitude angle system model (6), design a predetermined time disturbance observer: in, Represents the estimated value z ψ1 The first derivative of; 0<k1<1,k2>D,k>0,T s represents a predetermined time constant, tanh(·) represents a hyperbolic tangent function; Step 3, design the sliding mode variable with a predetermined time. The process is as follows: 3.1 Define the following attitude angle tracking error: e1=x ψ1 -x ψr (10) Among them, e1 represents the yaw axis attitude angle tracking error of the quadrotor aircraft, x ψr Indicates the desired yaw axis attitude angle of the quadrotor; Derivation of formula (10) yields: in, Represent e1,x respectively ψr The first derivative of ; 3.2 Design the following scheduled time sliding mode variables: in, Where s represents the predetermined time sliding mode variable, and the predetermined time constant T c >0,0<γ<1, l1=(1+γ)ε -γ ,l2=(-γ)ε -γ-1 ,ε>0; Step 4: Design a fully saturated repetitive learning law to compensate for the periodic terms in the quadrotor system.

2. The method for predictive anti-interference learning control of an unmanned aerial vehicle for periodic attitude tracking tasks as claimed in claim 1, characterized in that: The process of step 4 is as follows: 4.1 Design the fully saturated repeated learning law as follows: in, f ψr (t), The estimated value of for The value at time tT; μ0>0 is the repeated learning gain coefficient; φ(t) is the transition function, and the expression is: sat(·) is the saturation function, and its expression is: in, is the limit value of Υ, Step 5: Design a preset time sliding mode repetitive learning controller. The process is as follows: 5.1 Taking the derivative of equation (12), we get: in, in, They represent β(e1), s, The derivative of 5.2 Combining equations (9), (14) and (17), the preset time sliding mode controller is designed as: in, express The first derivative of , constant n ≥ 0.5; 5.3 Choose the following Lyapunov function: Derivative of formula (20) yields: in, represents the first-order derivative of V1(t); Substituting equation (17) and equation (19) into equation (21), we get That is, the system is judged to be stable.

Images