Unmanned aerial vehicle system adaptive control method, system, equipment and medium

Abstract

The invention discloses an unmanned aerial vehicle system adaptive control method, system and device and a medium, and relates to the technical field of unmanned aerial vehicle system control, and the method comprises the steps: dividing an unmanned aerial vehicle system into a position subsystem and an attitude subsystem; for a position subsystem, constructing an asymmetric time-varying obstacle Lyapunov function according to system errors and system constraints of the subsystem, selecting a tubular preset performance boundary, and applying quantitative constraints to overshoot output by the system; an RBF neural network is utilized to approach dx, external environment interference of the unmanned aerial vehicle is estimated, an adaptive law of the neural network is generated, a control law of each subsystem is generated by adopting a backstepping method, and then a position subsystem controller is constructed; constructing an attitude subsystem controller by adopting a backstepping sliding mode control method; the unmanned aerial vehicle system is adaptively controlled through the position subsystem controller and the attitude subsystem controller, and then the purpose of applying system output overshoot quantitative constraint is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle system control, and in particular to an unmanned aerial vehicle system adaptive control method, system, equipment and medium. Background Art

[0002] Quadrotor unmanned aerial vehicles (QUAVs) have broad application prospects in the fields of surveillance, material transportation, and mapping due to their simple structure, easy control, and vertical take-off and landing. They are one of the key research objects in academia and industry. Due to the underactuated, strongly coupled, nonlinear characteristics of quadrotor drones and the superposition of external interference during flight, the flight control system design of quadrotor drones has higher, more complex, and more challenging requirements.

[0003] In recent years, in order to improve the performance of quadrotor UAV systems, domestic and foreign scholars have proposed a series of control strategies. For example, sliding mode control, backstepping control, and anti-disturbance control. These control methods can make the system have good steady-state performance. However, because there is no error constraint mechanism to keep the output error within the predefined constraint range, these control strategies can only obtain better transient performance by adjusting the controller parameters. In order to introduce a constraint mechanism into the system, Tee et al. first defined the barrier Lyapunov function, and proposed logarithmic and integral barrier Lyapunov functions. On this basis, they introduced the switch function and constructed a class of asymmetric barrier Lyapunov functions. Xu et al. considered both constrained and unconstrained conditions and proposed a time-varying tangent Lyapunov function. For the control problem of UAVs under asymmetric time-varying full-state constraints, a new type of integral barrier Lyapunov function was applied to broaden the scope of application of traditional barrier Lyapunov functions. For the high-precision control problem of heterogeneous formations of UAVs, performance constraints on errors were introduced in the time-varying Lyapunov function to ensure the speed and accuracy of error convergence. A Lyapunov function based on asymmetric time-varying integral barrier is proposed to ensure that the system always satisfies the full-state constraints of dynamic changes. However, the constraints on transient performance indicators such as overshoot of the system in these works are not accurate and flexible enough, and due to the introduction of switching functions, the constraints are not continuous.

[0004] Taking into account realistic conditions and safety factors, such as when a drone passes through a narrow passage or when a drone is hanging and carrying cargo, it is often necessary to impose certain constraints on the overshoot of the system; the overshoot of the system is limited through error transfer transformation. The existing improved generalized predictive control algorithm uses the prediction function and information of predictive control to dynamically adjust the weighted parameters of the control quantity online, so as to achieve the purpose of fast response and overshoot suppression; a nonlinear control strategy is designed based on the energy method to achieve precise control of the drone position and overshoot suppression during flight. However, none of the above methods can impose more precise constraints on the system overshoot, resulting in the inability to complete the task of tracking complex trajectories. Summary of the invention

[0005] In view of the deficiency that the prior art cannot impose more precise constraints on system overshoot, the present invention proposes an adaptive control method, system, device and medium for an unmanned aerial vehicle system. By constructing an asymmetric time-varying obstacle Lyapunov function and selecting a tubular preset performance boundary, the difference between the actual trajectory and the expected trajectory of the unmanned aerial vehicle is limited within the tubular preset performance boundary, thereby solving the problems existing in the prior art.

[0006] An adaptive control method for an unmanned aerial vehicle system comprises the following steps:

[0007] Obtain the position and attitude information of the UAV, and divide the UAV system into a position subsystem and an attitude subsystem according to the position and attitude information of the UAV;

[0008] According to the system error and system constraints of the position subsystem, an asymmetric time-varying barrier Lyapunov function is constructed; according to the asymmetric time-varying barrier Lyapunov function, a tubular preset performance boundary is selected, and by limiting the difference between the actual trajectory and the expected trajectory of the UAV within the tubular preset performance boundary, a quantitative constraint is imposed on the overshoot of the system output; at the same time, the RBF neural network is used to approximate the disturbance of each subsystem of the UAV, the external environmental interference of the UAV is estimated, and the adaptive law of the neural network is generated, and then the position subsystem controller is constructed;

[0009] According to the position channel control law generated by the asymmetric time-varying barrier Lyapunov function, the attitude subsystem controller is constructed using the backstepping sliding mode control method.

[0010] The UAV system is adaptively controlled through the position subsystem controller and the attitude subsystem controller.

[0011] Furthermore, the acquisition of the position and attitude information of the drone and the division of the drone system into a position subsystem and an attitude subsystem according to the position and attitude information of the drone specifically include the following steps:

[0012] The dynamic model of the UAV system is established, which is expressed as:

[0013]

[0014] Among them: [x, y, z] represents the position state of the drone, [φ, θ, Ψ] represents the attitude state of the drone, They represent the speed in the x direction, the speed in the y direction, the speed in the z direction, the speed of the roll angle, the speed of the pitch angle, and the speed of the yaw angle respectively; represents the corresponding acceleration; m represents the mass of the drone, l represents the distance from the center of the propeller to the center of gravity, g represents the acceleration due to gravity, and k i represents the air resistance coefficient, I1, I2, I3 represent the moment of inertia of the body around the X, Y, Z axis respectively, u1, u2, u3, u4 represent the control input, d(t) represents the time-varying disturbance, t is the time;

[0015] Define the drone system state variables as:

[0016]

[0017] definition:

[0018]

[0019] Then the position subsystem and attitude subsystem are expressed as:

[0020]

[0021] Furthermore, the generation process of the adaptive law of the neural network specifically includes the following steps:

[0022] For any subsystem, it can be expressed as:

[0023]

[0024] Among them, u1 represents the control input, m represents the mass of the body, and k1 represents the air resistance coefficient;

[0025] Subsystem error e 1x The derivative is:

[0026]

[0027] Among them, x 1d Indicates the desired position;

[0028] Construct the barrier Lyapunov function V x1 , the expression is:

[0029]

[0030] Where Θ is the conversion error, is the upper bound of the constraint, and B is the lower bound of the constraint; the tubular preset performance boundary is selected as remember:

[0031]

[0032] V x1 The derivative is:

[0033]

[0034] Among them, e 2x =x2-α x , a x is the virtual control law, a x It is expressed as:

[0035]

[0036] Among them, γ 1x is a constant greater than zero;

[0037] According to the barrier Lyapunov function V x1 and subsystem error e 1x Derivative formula, we get:

[0038]

[0039] For 2x The derivative is:

[0040]

[0041] Using RBF neural network to approximate d x , the output of the neural network is:

[0042]

[0043] in, is the estimated value of the neural network weight, and h(x) is the Gaussian basis function output of the neural network;

[0044] According to the output expression of the neural network, the expression of the control law is obtained, which is expressed as:

[0045]

[0046] Furthermore, the method of using the RBF neural network to approximate the disturbances to each subsystem of the drone, estimating the external environmental interference of the drone, and generating the adaptive law of the neural network specifically includes the following steps:

[0047] Construct Lyapunov function V x2 , the expression is:

[0048]

[0049] V x2 Taking the derivative, we get:

[0050]

[0051] According to the Lyapunov function V x2 And the output expression of the neural network, we get:

[0052]

[0053] Construct Lyapunov function V x3 , the expression is:

[0054]

[0055] V x3 Taking the derivative, we get:

[0056]

[0057] Among them, ε x is the approximation error of the neural network;

[0058] get The adaptive law is:

[0059]

[0060] Furthermore, the backstepping sliding mode control method is used to construct the attitude subsystem controller, which specifically includes the following steps:

[0061] The attitude subsystem is divided into three subsystems: roll, pitch and yaw. For the roll subsystem, it is expressed as:

[0062]

[0063] Among them, u2 represents the control input;

[0064] For 1φ Taking the derivative, we get:

[0065]

[0066] Construct Lyapunov function V φ1 , the expression is:

[0067]

[0068] Vφ1 The derivative is:

[0069]

[0070] Among them, e φ2 =x8-α φ1 , α φ1 =-c φ1 e φ1 +x 8d , c φ1 is a positive constant;

[0071] Design sliding surfaces φ :

[0072] s φ =λ φ e φ1 +e φ2 ;

[0073] Among them, λ φ is a defined positive constant;

[0074] Derivative of the sliding surface:

[0075]

[0076] Select the Lyapunov function:

[0077]

[0078] V φ2 Taking the derivative, we get:

[0079]

[0080] According to the derivative expression of the sliding surface, the control law of the rolling subsystem is obtained as follows:

[0081]

[0082] Similarly, the control laws of the pitch and yaw subsystems are:

[0083]

[0084] The attitude subsystem controller is constructed according to the control laws of the three subsystems: roll, pitch and yaw.

[0085] Furthermore, the tubular preset performance boundary It is expressed as:

[0086]

[0087] Among them, B a0 , B b0is the preset performance boundary function q a ,q b The initial value of B a∞ , B b∞ When t→∞, q a ,q b The steady-state value of c1 and c2 is q a ,q b The convergence speed, q a >q b ≥0; the initial error value e(0) satisfies B b0 <e(0)<B a0 , then the overshoot of the UAV system satisfies:

[0088]

[0089] Furthermore, the approximation function of the RBF neural network is expressed as:

[0090]

[0091] Among them, h(x) is the approximation error, μ(x) is the neural network approximation error, and W* is the ideal weight;

[0092]

[0093] Where: c j is the center value of the Gaussian function, b j is the width.

[0094] The present invention also includes an adaptive control system for an unmanned aerial vehicle system, comprising:

[0095] The acquisition module is used to obtain the position and attitude information of the UAV, and divide the UAV system into a position subsystem and an attitude subsystem according to the position and attitude information of the UAV;

[0096] The position subsystem controller construction module is used to construct an asymmetric time-varying barrier Lyapunov function based on the system error and system constraints of the position subsystem; select a tubular preset performance boundary based on the asymmetric time-varying barrier Lyapunov function, and impose quantitative constraints on the overshoot of the system output by limiting the difference between the actual trajectory and the expected trajectory of the UAV within the tubular preset performance boundary; at the same time, the RBF neural network is used to approximate the disturbance to each subsystem of the UAV, estimate the external environmental interference of the UAV, generate the adaptive law of the neural network, and then construct the position subsystem controller;

[0097] The attitude subsystem controller construction module is used to construct the attitude subsystem controller using the backstepping sliding mode control method according to the position channel control law generated by the asymmetric time-varying obstacle Lyapunov function;

[0098] The control module is used to adaptively control the UAV system through the position subsystem controller and the attitude subsystem controller.

[0099] The present invention also includes a drone system adaptive control computer device, including: a memory, a processor, and a computer program stored in the memory, and the processor implements the steps of the drone system adaptive control method when executing the computer program.

[0100] The present invention also includes a readable storage medium, which stores a computer program. The computer program includes program instructions. When the program instructions are executed by a processor, they are used to execute the steps of the adaptive control method of the unmanned aerial vehicle system.

[0101] The present invention provides a method, system, device and medium for adaptive control of a drone system, which has the following beneficial effects:

[0102] The present invention constructs an asymmetric time-varying barrier Lyapunov function and selects a tubular preset performance boundary, thereby achieving the purpose of limiting the difference between the actual trajectory and the expected trajectory of the UAV within the tubular preset performance boundary. The feedback control law and the neural network adaptive law are designed by the inversion method to ensure the performance constraint of the system. The control strategy can complete the tracking task of the complex trajectory under the premise of applying quantitative overshoot constraints to the position of the UAV, and the system has more flexible and accurate transient performance constraints and strong robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] Figure 1 This is a schematic diagram of the structure and coordinates of a quad-rotor drone in an embodiment of the present invention;

[0104] Figure 2 Schematic diagram of tubular preset performance boundary in an embodiment of the present invention;

[0105] Figure 3 This is a control block diagram in an embodiment of the present invention;

[0106] Figure 4 A schematic diagram showing a comparison of tracking effects in an embodiment of the present invention;

[0107] Figure 5 It is a schematic diagram of a position tracking curve in an embodiment of the present invention;

[0108] Figure 6 Schematic diagram of a position error convergence curve in an embodiment of the present invention;

[0109] Figure 7 Schematic diagram of a pitch angle and yaw angle tracking curve (ABLF) in an embodiment of the present invention;

[0110] Figure 8 Schematic diagram of a pitch angle and yaw angle tracking curve (TBLF) in an embodiment of the present invention;

[0111] Fig. 9 A schematic diagram of a roll angle tracking curve in an embodiment of the present invention;

[0112] Fig.10 Schematic diagram of control input comparison curve in an embodiment of the present invention;

[0113] Fig.11 A schematic diagram showing a comparison of tracking effects in an embodiment of the present invention;

[0114] Fig.12 It is a schematic diagram of a position tracking curve in an embodiment of the present invention;

[0115] Fig.13 Schematic diagram of a position error convergence curve in an embodiment of the present invention;

[0116] Fig.14 Schematic diagram of a pitch angle and yaw angle tracking curve (ABLF) in an embodiment of the present invention;

[0117] Fig.15 Schematic diagram of a pitch angle and yaw angle tracking curve (TBLF) in an embodiment of the present invention;

[0118] Fig.16 Schematic diagram of a pitch angle and yaw angle tracking curve (SMC) in an embodiment of the present invention;

[0119] Fig.17 A schematic diagram of a roll angle tracking curve in an embodiment of the present invention;

[0120] Fig.18 Schematic diagram of control input comparison curve in an embodiment of the present invention. DETAILED DESCRIPTION

[0121] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0122] The present invention proposes a preset performance neural network adaptive control method based on a new type of time-varying barrier Lyapunov function. First, a new type of time-varying barrier Lyapunov function is proposed, which can impose continuous asymmetric constraints on the system, thereby enriching the setting form of the preset performance boundary. Secondly, a new type of tubular preset performance boundary function is designed, which imposes quantitative constraints on the overshoot of the system output and meets the steady-state performance requirements. On this basis, the feedback control law and the neural network adaptive law are designed by the inversion method to ensure the performance constraints of the system. Finally, based on the Lyapunov function stability theory, the consistent ultimate boundedness of all closed-loop signals is proved, and experimental comparison is carried out through numerical simulation to verify the effectiveness of the proposed method.

[0123] Quadrotor UAV mathematical model: Figure 1 As shown in the figure, the quadrotor unmanned aerial vehicle (QUAVs) consists of four arms with brushless DC motors. The four arms are arranged in a symmetrical structure. Considering the gyroscopic effect of the propeller, the rotation directions of each two adjacent motors are opposite. In order to establish the dynamic model of the quadrotor unmanned aerial vehicle (QUAVs) system, the following reasonable assumptions are made on the basis of ignoring some aerodynamic effects:

[0124] Assumption 1: The body is a rigid body, the center of gravity and the geometric center are completely coincident, the body is completely symmetrical, and the body model parameters are all positive constants. In order to facilitate the description of the mathematical model of QUAVs, an inertial coordinate system (O e X e Y e Z e ) and the body coordinate system (O b X b Y b Z b ),like Figure 1 As shown, the dynamic model of QUAVs is as follows:

[0125]

[0126] Among them, [x, y, z] represents the position state of the drone, [φ, θ, Ψ] represents the attitude state of the drone, They represent the speed in the x direction, the speed in the y direction, the speed in the z direction, the speed of the roll angle, the speed of the pitch angle, and the speed of the yaw angle respectively; represents the corresponding acceleration; m represents the mass of the drone, l represents the distance from the center of the propeller to the center of gravity, g represents the acceleration due to gravity, and k i represents the air resistance coefficient, I1, I2, I3 represent the moment of inertia of the body around the X, Y, Z axis respectively, u1, u2, u3, u4 represent the control input, d(t) represents the time-varying disturbance, and t is the time.

[0127] The desired position and angle of the QUAVs are denoted by x d ,y d 、z d , Φ d ,θ d , d , define the error variable of the system as:

[0128]

[0129] Define the system state variables as:

[0130]

[0131] To simplify the model, define:

[0132]

[0133] The simplified position subsystem and attitude subsystem are expressed as:

[0134]

[0135] To facilitate the subsequent controller design and related proofs, the following assumptions and lemmas are given.

[0136] Assumption 2: In system (1) i Bounded, that is, d i ≤D im And D im Known.

[0137] Lemma 1: For a system with bounded initial conditions, if there exists a C 1 The continuous and positive definite Lyapuno v function V(x) satisfies like in R n →R is K ∞ If the system is a function-like system and ρ and γ are positive constants, then the solution x(t) of the system is uniformly ultimately bounded.

[0138] Asymmetric time-varying barrier Lyapunov function: The traditional tangent time-varying barrier Lyapunov function is shown in formula (9):

[0139]

[0140] Where: tan(·) is the tangent function, F1 constrains the boundary, and its schematic diagram is as follows Figure 2 As shown in (a), symmetry constraints are imposed on the system, but the constraint form is not precise enough, especially for the transient performance of the system such as overshoot, there is a lack of more flexible constraints.

[0141] The traditional asymmetric time-varying barrier Lyapunov function is shown in formula (10):

[0142]

[0143] Where: p is a positive integer, k c To constrain the lower bound, To constrain the upper bound, y d is the expected trajectory, and q(·) is the switching function.

[0144] Due to the introduction of the switching function, the traditional asymmetric time-varying barrier Lyapunov function is relatively complex, the constraints imposed on the system are not continuous, and the derivation process is relatively complicated. In order to impose continuous asymmetric time-varying constraints on the system and simplify the derivation process, the present invention proposes a new type of asymmetric time-varying barrier Lyapunov function, which is expressed as follows:

[0145]

[0146] Where Θ is the conversion error, e 1i is the system tracking error, is the upper bound of the constraint, B is the lower bound of the constraint, and based on the new barrier Lyapunov function, a more accurate and flexible tubular preset performance boundary for transient performance constraints is further designed. Its schematic diagram is as follows Figure 2 As shown:

[0147]

[0148] Among them, B a0 , B b0 is the preset performance boundary function q a ,q b The initial value of B a∞ , B b∞ When t→∞, q a ,q b The steady-state value of c1 and c2 is q a ,q b The convergence speed, q a >q b ≥0; the initial error value e(0) satisfies B b0 <e(0)<B a0 , then the overshoot of the system satisfies:

[0149]

[0150] RBF neural network: Radial basis function neural network (RBFNN) has attracted widespread attention in many fields due to its nonlinear modeling ability, adaptability, real-time and simple structure. In the trajectory tracking control process of quad-rotor UAV, it is often difficult to achieve good control effect due to the influence of factors such as parameter uncertainty, external disturbance, unknown information, etc. on the system. Therefore, the present invention introduces RBFNN as an estimator to approximate the disturbance to the UAV under the premise of ensuring sufficient accuracy. The approximation function expression of RBFNN is:

[0151]

[0152] Where: h(x) is the approximation error, μ(x) is the neural network approximation error, and W* is the ideal weight.

[0153]

[0154] Where: c j is the center value of the Gaussian function, b j is the width.

[0155] A preset performance neural network adaptive control method based on a novel time-varying barrier Lyapunov function, as shown in system (1), the quad-rotor UAV system has 4 control inputs and 6 control outputs, and is a highly coupled, under-driven nonlinear system. In order to facilitate the design of the controller, the present invention divides the UAV system into two parts: a position subsystem (7) and an attitude subsystem (8), and adopts a double closed-loop control structure. In order to impose more accurate and flexible performance constraints on the position loop of the quad-rotor UAV, the novel asymmetric time-varying barrier Lyapunov function proposed in the present invention is used to design the control law of the position loop, and a backstepping sliding mode controller is used for the attitude loop. The overall structure is shown in Figure (3). The control method specifically includes the following steps:

[0156] S1. RBF neural network estimator.

[0157] In order to compensate for the influence of external environmental disturbance of the UAV on the controller performance, the RBF neural network estimator is introduced:

[0158] d p (x p )=W p *T h p (x p )+ε p (18)

[0159] in: p=x,y,z,φ,θ,ψ,h(x p ) is the central function of the neural network adaptive law; εp is the approximation error.

[0160] Right p The estimate is:

[0161]

[0162] S2. Position control subsystem controller design:

[0163] As shown in system (7), the position subsystem can be regarded as consisting of three subsystems: X, Y, and Z. The controller design steps of the position subsystem include: selecting an asymmetric time-varying Lyapunov function and setting a tubular preset performance boundary; using the backstepping method to design the control law u, which specifically includes the following steps:

[0164] Taking the position system controller design as an example, the X subsystem is expressed as:

[0165]

[0166] Step 1: For e 1x The derivative is

[0167]

[0168] e 1x is the difference between the actual trajectory and the expected trajectory.

[0169] Construct the barrier Lyapunov function V x1 , the expression is:

[0170]

[0171] Among them, e 1x is the X subsystem error; the tubular preset performance boundary is selected as Its expression is shown in formula (12). Note:

[0172]

[0173] V x1 The derivative is:

[0174]

[0175] Among them, e 2x =x2-α x , a x is the virtual control law, represents the term in the X channel that makes the mathematical form simple and concise. From formula (23), we can design a x for:

[0176]

[0177] Among them, γ 1x is a constant greater than zero; Substituting formula (24) into formula (23), we can get:

[0178]

[0179] step2: for e 2x The derivative is:

[0180]

[0181] is the difference between x2 and the virtual control law.

[0182] Using RBF neural network to approximate d x , the network output is:

[0183]

[0184] Among them, d x It indicates the disturbance of X channel among the three channels X, Y and Z of the drone.

[0185] Construct Lyapunov function V x2 , the expression is:

[0186]

[0187] V x2 The derivative is:

[0188]

[0189] According to formula (29), the control law is designed:

[0190]

[0191] Among them, γ 2x Represents a positive constant designed in the control law.

[0192] Substituting formula (30) into formula (29) yields:

[0193]

[0194] Construct Lyapunov function V x3 , the expression is:

[0195]

[0196] V x3 Taking the derivative we get:

[0197]

[0198] design The adaptive law is:

[0199]

[0200] To analyze the system stability, substitute equation (34) into equation (33) and we get:

[0201]

[0202] According to Young's inequality, we have:

[0203]

[0204] when When , we have:

[0205]

[0206] remember According to Lemma 1, we get:

[0207]

[0208] Integrating both sides of equation (38) simultaneously, we can obtain:

[0209] 0≤V x3 (t)≤μ(t) (39)

[0210] in:

[0211]

[0212] From equations (32) and (39):

[0213]

[0214] Solving inequality (41) yields:

[0215]

[0216] It can be concluded that Then the system error e 1x Constrained at the Boundary Within.

[0217] Similarly, the control law and neural network adaptive law of Y and Z channels can be obtained as follows:

[0218]

[0219] S3. Design of attitude control subsystem controller: As shown in system (8), the attitude subsystem is considered to consist of three subsystems: roll, pitch and yaw. The backstepping sliding mode method is used to design its controller. Taking the design of the roll subsystem controller as an example, the roll system subsystem is expressed as:

[0220]

[0221] Among them, K φ = lk4 / I1, l represents the distance from the propeller center to the center of gravity, I1 represents the moment of inertia of the body around the X axis, d φ Represents the external disturbance of the φ channel.

[0222] step1: for e 1φ The derivative is:

[0223]

[0224] Construct Lyapunov function V φ1 , the expression is:

[0225]

[0226] V φ1 The derivative is:

[0227]

[0228] Among them, e φ2 =x8-α φ1 , α φ1 =-c φ1 e φ1 +x 8d , c φ1 Is a normal number.

[0229] Step 2: Design the sliding surface:

[0230] s φ =λ φ e φ1 +e φ2 (51)

[0231] Derivative of the sliding surface:

[0232]

[0233] Select the Lyapunov function:

[0234]

[0235] V φ2 The derivative is:

[0236]

[0237] According to formula (54), the control law is designed:

[0238]

[0239] Among them, λ φ is a defined positive constant.

[0240] To analyze the system stability, we can substitute equation (55) into equation (54) to obtain:

[0241]

[0242] Construct a positive definite matrix:

[0243]

[0244] We can get:

[0245]

[0246] Take η φ ≥D φ +η φ0 , η φ0 >0

[0247] Similarly, the control laws of the pitch and yaw channels can be obtained as follows:

[0248]

[0249] S4, finite time convergent third order differentiator:

[0250] The desired pitch angle θ in the present invention is d , yaw angle Through the control law x 、u V 、u z and φ d Calculated, where φ d is a given value. According to formula (6), we can get:

[0251]

[0252] In the control laws (59) and (60), the intermediate signal θ generated by the outer loop needs to be d and Taking into account the differential explosion problem caused by the derivative of discontinuous signals and the noise suppression problem in practical applications, a finite time convergent differentiator is used to obtain and The expression is as follows:

[0253]

[0254] The input signal to be differentiated is v(t), ω=0.04, x a1 is the tracked signal, x a2 is an estimate of the first-order derivative of the signal, x a3 is the estimate of the second derivative of the signal. The initial value of the differentiator is x a1 (0) = x a2 (0) = x a3 (0)=0.

[0255] Simulation results and analysis:

[0256] In order to verify the effectiveness and superiority of the novel asymmetric Lyapunov function control scheme (ABLF) proposed in this paper, two experimental scenarios are designed in this section. The first experimental scenario simulates the flight mode of a quadcopter UAV hovering and ascending under time-varying disturbances, and selects the traditional tangent time-varying barrier Lyapunov function control scheme (TBLF) to verify the effectiveness of the method proposed in this paper in constraining output overshoot; the second experimental scenario sets a more complex flight trajectory, and conducts an ablation experiment while applying complex time-varying disturbances, removes the observer, and selects the traditional tangent time-varying barrier Lyapunov function control scheme (TBLF) and the sliding mode method (SMC) as comparative experiments to verify the robustness of the method proposed in this paper.

[0257] The physical parameters of the quadcopter are set as follows:

[0258] Table 1 Physical parameters of quadrotor drone

[0259]

[0260] Experimental scenario 1:

[0261] Assuming that the drone takes off from a 5m high building roof, the initial conditions of the system are set as:

[0262] [x(0),y(0),z(0)] T =[5,5,5] T ,

[0263] [(0),θ(0),ψ(0)] T =[0,0,0] T

[0264] Set the preset trajectory of the quadrotor drone to:

[0265]

[0266] The external disturbance is d i(t) = 0.1sin(t), where

[0267] The RBF network observer structure takes the structure of the neural network (input layer - hidden layer - output layer), and designs the Gaussian basis function parameters, taking c j =[-1-0.500.51], b j =3, the initial weight of the network is 0.1, and the adaptive law γ 3i =5. The specific parameters of the controller are shown in Table 2 and Table 3, where j = x, y, z, θ,Ψ.

[0268] In order to ensure the fairness of the comparative experiment, the structures and parameters of the neural network observers used in the control scheme (ABLF) and the control scheme (TBLF) proposed in the present invention are kept consistent.

[0269] Table 2 Controller parameters for experimental scenario 1

[0270]

[0271] The simulation results are as follows Figures 4 to 10 As shown in the figure: the ABLF controller can quantitatively constrain the overshoot while ensuring the rapidity of the body's running trajectory, thereby tracking the desired trajectory more smoothly and with better steady-state accuracy; the overshoot of the TBLF controller cannot meet the quantitative constraints, the temporary steady-state performance is poor, and the calculated angle fluctuation is more obvious.

[0272] Figure 6 The purple solid line in the middle is the preset performance boundary of the error. It can be seen that the traditional TBLF controller cannot constrain the error overshoot, while the ABLF controller proposed in the present invention can make the error converge quickly within a preset range, and then quantitatively constrain the error overshoot, verifying the effectiveness and feasibility of the algorithm of the present invention.

[0273] Figure 7-Figure 8 In order to calculate the expected attitude angle and its tracking status according to the expected position target, the attitude loops of all experiments use a backstepping sliding mode controller with consistent parameter settings. It can be shown that the expected angle fluctuation time of the attitude loop calculated by the method of the present invention is shorter and smoother.

[0274] Fig.10 This is the control input curve of the quadrotor drone. It can be seen that the control input of the ABLF controller is relatively smooth and the fluctuation time is small, which is beneficial to reduce the frequent adjustment of the actuator and increase the service life of the equipment.

[0275] Experimental scenario 2:

[0276] The setting of experimental scenario 2 is: the drone takes off from the roof of a building with a height of 5m and tracks a more complex "double-leaf clover" curve. The initial conditions of the system are set as follows:

[0277] [x(0),y(0),z(0)] T =[5,5,5] T ,

[0278] [φ(0),θ(0),ψ(0)] T =[0,0,0] T

[0279] Set the preset trajectory of the quadrotor drone to:

[0280]

[0281] y d =20cos(0.2t)sin(0.2t) t∈[0,30] (65)

[0282]

[0283] The external disturbance is d i (t) = 0.1sin(t), where At the same time, a set of multi-source time-varying disturbances consisting of constant disturbances, periodic disturbances, parameter change periodic disturbances, and divergent disturbances is set, and its form is shown in formula (67):

[0284]

[0285] Experimental scenario 2 conducts an ablation experiment based on setting a more complex desired trajectory and a more complex multi-source time-varying disturbance, removes the observer, and conducts a comparative simulation experiment combining the TBLF and terminal sliding mode (SMC) control schemes. In order to reflect the versatility of the control method proposed in the present invention, the system parameters remain consistent with those in experimental scenario 1.

[0286] The simulation results are as follows Figures 11 to 18 As shown; the experimental results show that when tracking more complex trajectories and being subject to more complex multi-source time-varying disturbances, the method of the present invention still has a faster tracking speed and can quantitatively constrain overshoot. Compared with the TBLF control scheme, the attitude angle fluctuation amplitude solved and tracked by the method of the present invention is smaller and smoother; although the attitude angle fluctuation amplitude solved and tracked by the SMC scheme is not much different from that of the present invention, multiple jitters occurred during the tracking process, and the smoothness is worse than that of the method of the present invention.

[0287] Fig.13The purple solid line in the middle is the preset performance boundary of the error. Since the expected trajectory changes at 16 seconds, the error tracking curve fluctuates. The accuracy of the z channel after recovery is higher than that of the traditional TBLF control scheme and is close to that of the SMC control scheme; however, the X and Y channels recover significantly faster than the other two methods and have higher accuracy.

[0288] Figure 14-15 The expected attitude angle and its tracking status calculated according to the expected position target are shown in Figure 2. Since the attitude loops of all experiments use a backstepping sliding mode controller and the parameter settings are consistent, it can be shown that the attitude angle calculated and tracked by the method of the present invention has a smaller fluctuation amplitude and is smoother; the expected attitude angle calculated by the TBLF control scheme has a large initial fluctuation and a long fluctuation time. Although the attitude angle calculated and tracked by the SMC scheme has a fluctuation amplitude that is not much different from that of the present invention, multiple jitters occur during the tracking process, and the smoothness is worse than that of the method of the present invention.

[0289] Fig.18 It is the control input curve of the quadrotor drone. It can be seen that the control input u1 of the method of the present invention has a smaller fluctuation amplitude when the steady-state size is similar to that of the other two methods. Although the steady-state value of u2 is slightly larger than that of the other two methods, the steady-state values ​​and fluctuation time of u3 and u4 are much smaller than those of the TBLF scheme, and the number of fluctuations is less than that of the SMC scheme.

[0290] The present invention aims at a quad-rotor unmanned aerial vehicle with time-varying disturbances, and proposes a control scheme with overshoot quantitative constraint based on a new obstacle Lyapunov function. The new obstacle Lyapunov function can impose continuous, more flexible and precise performance constraints on the system. Simulation experiments show that the new obstacle Lyapunov function can impose more flexible and precise constraints on the system. Based on the new obstacle Lyapunov function, a set of tubular preset performance boundaries are designed to achieve the purpose of imposing quantitative constraints on system output overshoot. Simulation experiments show that the tubular preset performance function can effectively impose quantitative constraints on system overshoot, thereby improving the transient performance of the system. Theoretical and experimental simulations show that the control strategy designed by the present invention can complete the tracking task of a complex trajectory under the premise of imposing quantitative overshoot constraints on the position of the unmanned aerial vehicle, and the system has more flexible and precise transient performance constraints and has strong robustness.

[0291] Based on the same inventive concept, the present invention also proposes an adaptive control system for a UAV system, comprising:

[0292] The acquisition module is used to obtain the position and attitude information of the drone, and divide the drone system into a position subsystem and an attitude subsystem according to the position and attitude information of the drone.

[0293] The position subsystem controller construction module is used to construct an asymmetric time-varying barrier Lyapunov function based on the system error and system constraints of the position subsystem; a tubular preset performance boundary is selected according to the asymmetric time-varying barrier Lyapunov function, and a quantitative constraint is imposed on the overshoot of the system output by limiting the difference between the actual trajectory and the expected trajectory of the UAV within the tubular preset performance boundary; at the same time, the RBF neural network is used to approximate the disturbance suffered by each subsystem of the UAV, the external environmental interference of the UAV is estimated, and the adaptive law of the neural network is generated, thereby constructing the position subsystem controller.

[0294] The attitude subsystem controller construction module is used to construct the attitude subsystem controller based on the position channel control law generated by the asymmetric time-varying obstacle Lyapunov function and the backstepping sliding mode control method.

[0295] The control module is used to adaptively control the UAV system through the position subsystem controller and the attitude subsystem controller.

[0296] The present invention also proposes a computer device for adaptive control of an unmanned aerial vehicle system, comprising: a memory, a processor, and a computer program stored in the memory, and the processor implements the steps of the method for adaptive control of the unmanned aerial vehicle system when executing the computer program.

[0297] The present invention also proposes a readable storage medium, which stores a computer program. The computer program includes program instructions. When the program instructions are executed by a processor, they are used to execute the steps of the adaptive control method of the unmanned aerial vehicle system.

[0298] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An adaptive control method for an unmanned aerial vehicle system, characterized in that: The following steps are involved: Obtain the position and attitude information of the UAV, and divide the UAV system into a position subsystem and an attitude subsystem according to the position and attitude information of the UAV; According to the system error and system constraints of the position subsystem, an asymmetric time-varying barrier Lyapunov function is constructed; according to the asymmetric time-varying barrier Lyapunov function, a tubular preset performance boundary is selected, and by limiting the difference between the actual trajectory and the expected trajectory of the UAV within the tubular preset performance boundary, a quantitative constraint is imposed on the overshoot of the system output; at the same time, the RBF neural network is used to approximate the disturbance of each subsystem of the UAV, the external environmental interference of the UAV is estimated, and the adaptive law of the neural network is generated, and then the position subsystem controller is constructed; According to the position channel control law generated by the asymmetric time-varying barrier Lyapunov function, the attitude subsystem controller is constructed using the backstepping sliding mode control method. The UAV system is adaptively controlled through the position subsystem controller and the attitude subsystem controller.

2. The method for adaptive control of a drone system according to claim 1, characterized in that: The acquisition of the position and attitude information of the drone, and the division of the drone system into a position subsystem and an attitude subsystem according to the position and attitude information of the drone, specifically includes the following steps: The dynamic model of the UAV system is established, which is expressed as: Among them, [x, y, z] represents the position state of the drone, [φ, θ, Ψ] represents the attitude state of the drone, They represent the speed in the x direction, the speed in the y direction, the speed in the z direction, the speed of the roll angle, the speed of the pitch angle, and the speed of the yaw angle respectively; represents the corresponding acceleration; m represents the mass of the drone, l represents the distance from the center of the propeller to the center of gravity, g represents the acceleration due to gravity, and k i represents the air resistance coefficient, I1, I2, I3 represent the moment of inertia of the body around the X, Y, Z axis respectively, u1, u2, u3, u4 represent the control input, d(t) represents the time-varying disturbance, t is the time; Define the drone system state variables as: definition: Then the position subsystem and attitude subsystem are expressed as:

3. The method for adaptive control of a drone system according to claim 2, characterized in that: The generation process of the adaptive law of the neural network specifically includes the following steps: For any subsystem, it can be expressed as: Among them, u1 represents the control input, m represents the mass of the body, and k1 represents the air resistance coefficient; Subsystem error e 1x The derivative is: Among them, x 1d Indicates the desired position; Construct the barrier Lyapunov function V x1 , the expression is: Where Θ is the conversion error, is the upper bound of the constraint, and B is the lower bound of the constraint; the tubular preset performance boundary is selected as remember: V x1 The derivative is: Among them, e 2x =x2-α x , a x is the virtual control law, a x It is expressed as: Among them, γ 1x is a constant greater than zero; According to the barrier Lyapunov function V x1 and subsystem error e 1x Derivative formula, we get: For 2x The derivative is: Using RBF neural network to approximate d x , the output of the neural network is: in, is the estimated value of the neural network weight, and h(x) is the Gaussian basis function output of the neural network; According to the output expression of the neural network, the expression of the control law is obtained, which is expressed as:

4. The method for adaptive control of a drone system according to claim 3, characterized in that: The method of using the RBF neural network to approximate the disturbances to each subsystem of the drone, estimating the external environmental interference of the drone, and generating the adaptive law of the neural network specifically includes the following steps: Construct Lyapunov function V x2 , the expression is: V x2 Taking the derivative, we get: According to the Lyapunov function V x2 And the output expression of the neural network, we get: Construct Lyapunov function V x3 , the expression is: V x3 Taking the derivative, we get: Among them, ε x is the approximation error of the neural network; get The adaptive law is:

5. The method for adaptive control of a drone system according to claim 1, characterized in that: The backstepping sliding mode control method is used to construct the attitude subsystem controller, which specifically includes the following steps: The attitude subsystem is divided into three subsystems: roll, pitch and yaw. For the roll subsystem, it is expressed as: Among them, u2 represents the control input; For 1φ Taking the derivative, we get: Construct Lyapunov function V φ1 , the expression is: V φ1 The derivative is: Among them, e φ2 =x8-α φ1 , α φ1 =-c φ1 e φ1+ x 8d , c φ1 is a positive constant; Design sliding surfaces φ : s φ =λ φ And φ1 +e φ2 ; Among them, λ φ is a defined positive constant; Derivative of the sliding surface: Select the Lyapunov function: V φ2 Taking the derivative, we get: According to the derivative expression of the sliding surface, the control law of the rolling subsystem is obtained as follows: Similarly, the control laws of the pitch and yaw subsystems are: The attitude subsystem controller is constructed according to the control laws of the three subsystems: roll, pitch and yaw.

6. The method for adaptive control of a drone system according to claim 3, characterized in that: The tubular preset performance boundary It is expressed as: Among them, B a0 , B b0 is the preset performance boundary function q a ,q b The initial value of B a∞ , B b∞ When t→∞, q a ,q b The steady-state value of c1 and c2 is q a ,q b The convergence speed, q a >q b ≥0; the initial error value e(0) satisfies B b0 <e(0)<B a0 , then the overshoot of the UAV system satisfies:

7. The method for adaptive control of a drone system according to claim 3, characterized in that: The approximation function of the RBF neural network is expressed as: Among them, h(x) is the approximation error, μ(x) is the neural network approximation error, and W* is the ideal weight; Among them, c j is the center value of the Gaussian function, b j is the width.

8. An adaptive control system for an unmanned aerial vehicle system, characterized in that: include: The acquisition module is used to obtain the position and attitude information of the UAV, and divide the UAV system into a position subsystem and an attitude subsystem according to the position and attitude information of the UAV; The position subsystem controller construction module is used to construct an asymmetric time-varying barrier Lyapunov function based on the system error and system constraints of the position subsystem; select a tubular preset performance boundary based on the asymmetric time-varying barrier Lyapunov function, and impose quantitative constraints on the overshoot of the system output by limiting the difference between the actual trajectory and the expected trajectory of the UAV within the tubular preset performance boundary; at the same time, the RBF neural network is used to approximate the disturbance to each subsystem of the UAV, estimate the external environmental interference of the UAV, generate the adaptive law of the neural network, and then construct the position subsystem controller; The attitude subsystem controller construction module is used to construct the attitude subsystem controller using the backstepping sliding mode control method according to the position channel control law generated by the asymmetric time-varying obstacle Lyapunov function; The control module is used to adaptively control the UAV system through the position subsystem controller and the attitude subsystem controller.

9. An adaptive control computer device for an unmanned aerial vehicle system, characterized in that: include: A memory, a processor, and a computer program stored in the memory, wherein the processor implements the steps of the adaptive control method for the unmanned aerial vehicle system according to any one of claims 1 to 7 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which includes program instructions. When the program instructions are executed by a processor, they are used to execute the steps of the adaptive control method for the drone system according to any one of claims 1 to 7.

Images