UAV flight control method based on adaptive neural network

Abstract

The invention discloses a UAV flight control method based on an adaptive neural network, including: using a static stable flapping model to replace the flapping dynamics model of the main rotor of an unmanned helicopter; designing the longitudinal-lateral sub-control system model of the unmanned helicopter according to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage, and designing the roll, pitch and yaw servo control inputs of the unmanned helicopter based on the backstepping control method according to the rigid body kinematics and dynamics equations of the fuselage and the longitudinal-lateral sub-control system model; determining the vertical position kinematics and dynamics equations of the unmanned helicopter, and designing the collective pitch servo control inputs of the unmanned helicopter based on the vertical position kinematics and dynamics equations based on the backstepping control method; wherein, for unmodeled dynamics and external disturbances, an adaptive radial basis function neural network is used for estimation compensation. The invention can improve the flight control accuracy and robustness of the unmanned helicopter and reduce the calculation amount of the controller.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned helicopter control, and in particular to a flight control method of an unmanned aerial vehicle based on an adaptive neural network. Background Art

[0002] Among the robust autonomous flight control methods in unmanned helicopters, the traditional flight control methods mainly include classical feedback control (such as PID control method) and linear control method. Among them, the advantages of the PID control method are simple structure and no need for accurate system model, which can avoid the complex modeling process of unmanned helicopters. However, the PID control method cannot give full play to the maneuverability of unmanned helicopters and is not robust. The parameter adjustment depends entirely on experience and the process is cumbersome. The development of linear control methods is relatively mature, but the linear control method relies on a linear model based on the linearization of a single characteristic point, ignoring many nonlinear characteristics of the system. When the state of the unmanned helicopter deviates from the equilibrium point, the corresponding control performance may decrease, and even lead to instability of the closed-loop system, and full envelope flight cannot be guaranteed.

[0003] In order to solve the problem of poor control performance of traditional classical feedback control and linear control methods, nonlinear control methods are currently used to design autonomous flight control systems for unmanned helicopters. However, the nonlinear control methods currently used directly select non-physical inputs such as main rotor thrust, tail rotor thrust, longitudinal and lateral torque as four control inputs, ignoring the flapping dynamics equations of the unmanned helicopter rotor system. Although it can significantly reduce the difficulty of controller design, these four control inputs are seriously coupled with the input signals (physical inputs) of the main rotor collective pitch servo, tail rotor collective pitch servo, and longitudinal and lateral periodic pitch servo that actually control the unmanned helicopter, making it difficult to directly apply them to engineering implementation. Summary of the invention

[0004] In order to solve some or all of the technical problems existing in the above-mentioned prior art, the present invention provides a UAV flight control method based on an adaptive neural network.

[0005] The technical solution of the present invention is as follows:

[0006] A UAV flight control method based on an adaptive neural network is provided, and the method is used for flight control of an unmanned helicopter, comprising:

[0007] The static stable flapping model is used to replace the flapping dynamics model of the main rotor of the unmanned helicopter;

[0008] According to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage, the longitudinal-lateral sub-control system model of the unmanned helicopter is designed and extracted. Based on the backstepping control method, the roll servo control input, pitch servo control input and yaw servo control input of the unmanned helicopter are designed according to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage and the longitudinal-lateral sub-control system model of the unmanned helicopter. Among them, for unmodeled dynamics and external disturbances, an adaptive radial basis function neural network is used for estimation and compensation;

[0009] The vertical position kinematics and dynamics equations of the unmanned helicopter are determined, and based on the backstepping control method, the collective pitch servo control input of the unmanned helicopter is designed according to the vertical position kinematics and dynamics equations of the unmanned helicopter. Among them, for unmodeled dynamics and external disturbances, an adaptive radial basis function neural network is used for estimation and compensation.

[0010] In some possible implementations, the flapping dynamics model of the main rotor of the unmanned helicopter is:

[0011]

[0012] Where a and b represent the longitudinal and lateral flapping angles of the main rotor blade tip plane of the unmanned helicopter, respectively. is the derivative of a, is the derivative of b. When the azimuth angle Ψ of the unmanned helicopter m = 0°, the main rotor flaps downward, a is positive, and when the azimuth angle Ψ m =270°, the main rotor flaps downward, b is positive, γ m is the total Locke number of the two blades of the unmanned helicopter, ρ is the air density, c m is the main rotor blade chord length, C lα,mr is the lift coefficient of the main rotor blade, R m is the radius of the main rotor blade, I β is the main rotor blade moment of inertia, Ω m is the main rotor speed, k β is the main rotor blade stiffness coefficient, p and q represent the rolling angular velocity and pitch angular velocity of the unmanned helicopter, θ lat and θ lon They represent the lateral cyclic pitch variation and longitudinal cyclic pitch variation of the main rotor respectively.

[0013] In some possible implementations, the static stable waving model is:

[0014]

[0015] in, δ latis the input signal of the lateral cyclic variable pitch servo of the unmanned helicopter, δ lon is the input signal of the longitudinal cyclic variable pitch servo of the unmanned helicopter, θ lat0 and θ lon0 They represent the initial main rotor lateral cyclic pitch deviation and longitudinal cyclic pitch deviation respectively.

[0016] In some possible implementations, the adaptive radial basis function neural network includes:

[0017] An input layer, a hidden layer and an output layer, the input layer is connected to the hidden layer, the hidden layer is connected to the output layer, the input layer includes a plurality of input nodes, the hidden layer includes a plurality of hidden nodes, and the output layer includes an output node;

[0018] The output of the adaptive radial basis function neural network is designed to be:

[0019]

[0020] in, is the estimated value of Γ, W represents the weight matrix between the hidden layer and the output layer, H represents the matrix composed of all radial basis functions of the neural network, ||·||2 represents the 2-norm, Represents the square of the 2-norm.

[0021] In some possible implementations, when performing flight control of the unmanned helicopter, the parameters The adaptive law is designed as:

[0022]

[0023] in, for The derivative of are the estimated disturbance and uncertainty values ​​in horizontal position control, for The derivative of is the estimated disturbance and uncertainty value in attitude position control, for The derivative of is the estimated disturbance and uncertainty value in altitude position control, ζ i , ε i and i is a design parameter greater than zero, i=1,2,3, V 2e is the horizontal velocity error vector, V 2e =V2-α P , V2=[v x ,v y ] T , v xand v y are the speeds of the center of gravity of the unmanned helicopter in the x-axis and y-axis directions of the inertial coordinate system, α P is the virtual control item of horizontal position, ω e is the angular velocity error of the unmanned helicopter body, ω e =ω-ω d ,ω andω d are the actual and expected angular velocity vectors of the unmanned helicopter body, ω=[p,q,r] T , r is the yaw angular velocity of the unmanned helicopter, v ze is the vertical velocity error, v ze =v z -α z , v z is the velocity of the center of gravity of the unmanned helicopter in the z-axis direction of the inertial coordinate system, α z are virtual control items, H1(V2), H2(ω) and H3(v z ) are Gaussian functions of the corresponding variables, θ is the pitch angle of the unmanned helicopter, and φ is the roll angle of the unmanned helicopter.

[0024] In some possible implementations, the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage are:

[0025]

[0026] According to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage, the longitudinal-lateral sub-control system model of the unmanned helicopter is designed and extracted as follows:

[0027]

[0028] in, is the derivative of P, P = [x t ,y t ,z t ] T is the position vector of the center of gravity of the unmanned helicopter in the inertial coordinate system, x t ,y t and z t are the positions of the center of gravity in the x-axis, y-axis and z-axis directions of the inertial coordinate system, V = [v x ,v y ,v z ] T is the velocity vector of the center of gravity of the unmanned helicopter in the inertial coordinate system, v x 、v y and v z are the velocities of the center of gravity in the x-axis, y-axis and z-axis directions of the inertial coordinate system, m is the mass of the unmanned helicopter system, is the derivative of V, Θ=[φ,θ,ψ] T is the attitude angle vector of the unmanned helicopter, φ, θ and ψ are the roll angle, pitch angle and yaw angle of the unmanned helicopter respectively, is the derivative of Θ, R(Θ) is the transformation matrix from the body coordinate system to the inertial coordinate system, F is all external forces except the gravity of the unmanned helicopter, g is the gravitational acceleration, E z is a unit vector, E z =[0,0,1] T , T(Θ) is a 3×3 matrix, J is the moment of inertia matrix, J=diag[I x ,I y ,I z ], I x ,I y and I z are the moments of inertia of the unmanned helicopter body around the x-axis, y-axis and z-axis of the body coordinate system, is the derivative of ω, M is all the torques on the unmanned helicopter, is the derivative of P2, P2=[x t ,y t ] T , V2=[v x ,v y ] T , is the derivative of V2, R2=[R 13 ,R 23 ] T , R ij is the element in the i-th row and j-th column of the rotation matrix R(θ), d z is the vertical disturbance force of the unmanned helicopter, T mr is the main rotor lift, d x and d y are the longitudinal and lateral interference forces of the unmanned helicopter, is the derivative of R2.

[0029] In some possible implementations, when designing the yaw servo control input of the unmanned helicopter, the following feedback expression is used for feedback control;

[0030] T tr,ctr =k ped,ctr δ ped -k r,ctr r

[0031] Among them, T tr,ctr is the feedback design force, δ ped is the input signal for the yaw servo control of the unmanned helicopter, k ped,ctr and kr,ctr is the design parameter.

[0032] In some possible implementations, the roll, pitch, and yaw servo control inputs are designed to be:

[0033]

[0034] Among them, δ aer are the roll, pitch and yaw servo control inputs, A, B, M1 and M2 are the model parameter matrices, K ω is the parameter to be designed, Sk(·) is a skew-symmetric matrix, R e is the roll and pitch attitude error, for The estimated value calculated by the command filter, ω d The derivative of e is the yaw angle tracking error, ∈2 is the design parameter, T hov is the lift generated by the main rotor of the unmanned helicopter when it is hovering, H mr H is the vertical distance from the center of gravity of the unmanned helicopter to the main rotor hub. tr D is the vertical distance from the center of gravity of the unmanned helicopter to the tail rotor hub. tr It is the longitudinal horizontal distance from the center of gravity of the unmanned helicopter to the tail rotor hub.

[0035] In some possible implementations, the vertical position kinematics and dynamics equations of the unmanned helicopter are:

[0036]

[0037] According to the vertical position kinematics and dynamics equations of the unmanned helicopter, the collective pitch servo control input of the unmanned helicopter is designed as follows:

[0038]

[0039] Among them, z t is the position of the center of gravity of the unmanned helicopter in the z-axis direction of the inertial coordinate system, For z t The derivative of v z The derivative of col is the collective pitch servo control input, is α z The derivative of z is a highly virtual control item. k z is a design parameter greater than zero, z e is the vertical height error, ze =z t -z r , z r is the desired height, For z r The derivative of k vz is a design parameter greater than zero, k int1 and k int2 are system model parameters, b m is the number of main rotor blades, k col is the linkage gain from the collective pitch of the main rotor blades of the unmanned helicopter to the collective pitch servo, v im,hov is the downwash speed of the main rotor of the unmanned helicopter in the hovering state, and w is the vertical speed of the unmanned helicopter in the body coordinate system.

[0040] In some possible implementations, when designing the collective servo control input of the unmanned helicopter, the following feedback expression is used for feedback control;

[0041] T int =k int1 w+k int1 k int2 k col δ col -k int1 v im,hov

[0042] Among them, T int It is the feedback force of the main rotor of the unmanned helicopter.

[0043] The main advantages of the technical solution of the present invention are as follows:

[0044] The UAV flight control method based on adaptive neural network of the present invention simplifies the analysis of the flapping dynamics model of the main rotor, then designs the longitudinal-lateral position control law and the internal three-axis attitude control law of the unmanned helicopter, as well as the altitude control law, and uses an improved adaptive radial basis function neural network to estimate and compensate for unmodeled dynamics and external interference, thereby improving the flight control accuracy and robustness of the UAV and reducing the calculation amount of the controller. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 This is a flow chart of a UAV flight control method based on an adaptive neural network according to an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of the structure of a traditional radial basis function neural network is given as an example;

[0048] Figure 3 Schematic diagram of the structure of an adaptive radial basis function neural network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0050] The technical solution provided by the embodiments of the present invention is described in detail below with reference to the accompanying drawings.

[0051] See also Figure 1 An embodiment of the present invention provides a UAV flight control method based on an adaptive neural network, which is used for flight control of an unmanned helicopter, and includes the following steps:

[0052] S1, using a static stable flapping model to replace the flapping dynamics model of the main rotor of the unmanned helicopter;

[0053] S2, according to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage, the longitudinal-lateral sub-control system model of the unmanned helicopter is designed and extracted. Based on the backstepping control method, the roll servo control input, pitch servo control input and yaw servo control input of the unmanned helicopter are designed according to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage and the longitudinal-lateral sub-control system model of the unmanned helicopter. Among them, for unmodeled dynamics and external disturbances, an adaptive radial basis function neural network is used for estimation and compensation;

[0054] S3, determine the vertical position kinematics and dynamics equations of the unmanned helicopter, and design the collective pitch servo control input of the unmanned helicopter based on the backstepping control method according to the vertical position kinematics and dynamics equations of the unmanned helicopter. Among them, for unmodeled dynamics and external disturbances, an adaptive radial basis function neural network is used for estimation and compensation.

[0055] The following is a detailed description of the various steps and principles of the UAV flight control method based on an adaptive neural network provided in one embodiment of the present invention.

[0056] In one embodiment of the present invention, the reference coordinate system of the state quantities such as force, torque, acceleration, and speed of the unmanned helicopter body adopts the unmanned helicopter body coordinate system. The unmanned helicopter body coordinate system takes the position of the center of gravity of the unmanned helicopter as the origin, the x-axis points to the nose direction along the fuselage, the y-axis points to the right side of the nose direction, and is perpendicular to the longitudinal section of the unmanned helicopter. The z-axis is on the longitudinal section of the unmanned helicopter, perpendicular to the plane where the x-axis and the y-axis are located, and points to the bottom of the body. Among them, the x-axis is the roll axis, the x-axis direction is the longitudinal axis direction, the y-axis is the pitch axis, the y-axis direction is the transverse axis direction, and the z-axis is the yaw axis, and the z-axis direction is the vertical axis direction.

[0057] Step S1, using a static stable flapping model to replace the flapping dynamics model of the main rotor of the unmanned helicopter.

[0058] Specifically, the flapping dynamics model of the main rotor of an unmanned helicopter can be expressed as:

[0059]

[0060] Where a and b represent the longitudinal and lateral flapping angles of the main rotor blade tip plane of the unmanned helicopter, respectively. is the derivative of a, is the derivative of b. When the azimuth angle Ψ of the unmanned helicopter m = 0°, the main rotor flaps downward, a is positive, and when the azimuth angle Ψ m =270°, the main rotor flaps downward, b is positive, γ m is the total Locke number of the two blades of the unmanned helicopter, ρ is the air density, c m is the main rotor blade chord length, C lα,mr is the lift coefficient of the main rotor blade, R m is the radius of the main rotor blade, I β is the main rotor blade moment of inertia, Ω m is the main rotor speed, k β is the main rotor blade stiffness coefficient, p and q represent the rolling angular velocity and pitch angular velocity of the unmanned helicopter, θ lat and θ lon They represent the lateral cyclic pitch variation and longitudinal cyclic pitch variation of the main rotor respectively.

[0061] For the above-mentioned flapping dynamics model of the main rotor of the unmanned helicopter, the static stable flapping model used to equivalently replace the flapping dynamics model of the main rotor can be designed as follows:

[0062]

[0063] in, δ latis the input signal of the lateral cyclic variable pitch servo of the unmanned helicopter, δ lon is the input signal of the longitudinal cyclic variable pitch servo of the unmanned helicopter, θ lat0 and θ lon0 They represent the initial main rotor lateral cyclic pitch deviation and longitudinal cyclic pitch deviation respectively.

[0064] The flapping dynamics model of the main rotor of an unmanned helicopter is an important motion characteristic of the unmanned helicopter. It has a strong coupling effect with the rigid body dynamics model of the fuselage and is related to the dynamic response of the entire system. It must be considered when designing the controller of the unmanned helicopter. However, the flapping angle of the main rotor of the unmanned helicopter during flight cannot be effectively measured by any sensor. If the flapping motion is completely ignored, the performance of the control system will be affected. In one embodiment of the present invention, by using a compromise static stable flapping model to approximate the equivalent complete main rotor flapping dynamics model, the main rotor flapping dynamics model can be simplified and analyzed, thereby improving the control accuracy and robustness of the unmanned helicopter.

[0065] Step S2, according to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage, design and extract the longitudinal-lateral sub-control system model of the unmanned helicopter, based on the backstepping control method, according to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage and the longitudinal-lateral sub-control system model of the unmanned helicopter, design the roll servo control input, pitch servo control input and yaw servo control input of the unmanned helicopter, wherein for unmodeled dynamics and external disturbances, an adaptive radial basis function neural network is used for estimation and compensation.

[0066] In one embodiment of the present invention, in step S2, an adaptive radial basis function neural network is used to perform estimation compensation for unmodeled dynamics and external disturbances.

[0067] Figure 2 A traditional radial basis function neural network is shown. The traditional radial basis function (RBF) neural network includes a three-layer network structure of an input layer (Input Layer), a hidden layer (Hidden Layer) and an output layer (Output Layer). The input layer has multiple input nodes, the hidden layer has multiple hidden nodes, and the output layer has multiple output nodes. The output of the j-th output node of the output layer can be expressed as:

[0068]

[0069] Where x represents the input of the radial basis function neural network, H i (·) represents the radial basis function corresponding to the i-th hidden node in the hidden layer, W ijrepresents the weight between the i-th hidden node in the hidden layer and the j-th output node in the output layer, K represents the number of hidden nodes in the hidden layer, M represents the number of output nodes in the output layer, and Figure 2 In, x N Represents the input of the Nth input node of the input layer.

[0070] When using traditional radial basis function neural networks to estimate and compensate for unmodeled dynamics and external disturbances, it is necessary to design an adaptive law for each weight when training the neural network online. However, as the number of neural network nodes increases, the required adaptive law will increase exponentially, resulting in a sharp increase in the amount of calculation, which is not conducive to engineering implementation.

[0071] To this end, in one embodiment of the present invention, an adaptive radial basis function neural network for estimating and compensating unmodeled dynamics and external disturbances is improved.

[0072] Specifically, see Figure 3 In one embodiment of the present invention, the adaptive radial basis function neural network includes:

[0073] An input layer, a hidden layer and an output layer, the input layer is connected to the hidden layer, the hidden layer is connected to the output layer, the input layer includes a plurality of input nodes, the hidden layer includes a plurality of hidden nodes, and the output layer includes an output node;

[0074] The output of the adaptive radial basis function neural network is designed to be:

[0075]

[0076] in, is the estimated value of Γ, W represents the weight matrix between the hidden layer and the output layer, H represents the matrix composed of all radial basis functions of the neural network, ||·||2 represents the 2-norm, Represents the square of the 2-norm.

[0077] For the improved adaptive radial basis function neural network mentioned above, when training the neural network online, only the parameters Design adaptive laws.

[0078] In one embodiment of the present invention, the parameters are designed according to Lyapunov stability theory. The adaptive law is used to ensure the stability of the closed-loop control system.

[0079] Specifically, when performing flight control of an unmanned helicopter, the parameters The adaptive law is designed as:

[0080]

[0081] in, for The derivative of are the estimated disturbance and uncertainty values ​​in horizontal position control, for The derivative of is the estimated disturbance and uncertainty value in attitude position control, for The derivative of is the estimated disturbance and uncertainty value in altitude position control, ζ i , ε i and i is a design parameter greater than zero, i=1,2,3, V 2e is the horizontal velocity error vector, V 2e =V2-α P , V2=[v x ,v y ] T , v x and v y are the speeds of the center of gravity of the unmanned helicopter in the x-axis and y-axis directions of the inertial coordinate system, α P is the virtual control item of horizontal position, ω e is the angular velocity error of the unmanned helicopter body, ω e =ω-ω d ,ω andω d are the actual and expected angular velocity vectors of the unmanned helicopter body, ω=[p,q,r] T , r is the yaw angular velocity of the unmanned helicopter, v ze is the vertical velocity error, v ze =v z -α z , v z is the velocity of the center of gravity of the unmanned helicopter in the z-axis direction of the inertial coordinate system, α z are virtual control items, H1(V2), H2(ω) and H3(v z ) are Gaussian functions of the corresponding variables, θ is the pitch angle of the unmanned helicopter, and φ is the roll angle of the unmanned helicopter.

[0082] Among them, the inertial coordinate system is a north-east coordinate system, which takes a set ground point as the origin, with the x-axis pointing to the North Pole of the earth, the y-axis pointing to the east of the earth, and the z-axis perpendicular to the earth's surface and pointing downward.

[0083] In one embodiment of the present invention, a ground point is set to select a take-off point for the unmanned helicopter.

[0084] In one embodiment of the present invention, by using the improved adaptive radial basis function neural network to estimate and compensate for unmodeled dynamics and external disturbances, the computational burden of the unmanned helicopter controller during operation can be significantly reduced.

[0085] Furthermore, in one embodiment of the present invention, the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage can be expressed as:

[0086]

[0087] in, is the derivative of P, P = [x t ,y t ,z t ] T is the position vector of the center of gravity of the unmanned helicopter in the inertial coordinate system, x t ,y t and z t are the positions of the center of gravity in the x-axis, y-axis and z-axis directions of the inertial coordinate system, V = [v x ,v y ,v z ] T is the velocity vector of the center of gravity of the unmanned helicopter in the inertial coordinate system, v x 、v y and v z are the velocities of the center of gravity in the x-axis, y-axis and z-axis directions of the inertial coordinate system, m is the mass of the unmanned helicopter system, is the derivative of V, Θ=[φ,θ,ψ] T is the attitude angle vector of the unmanned helicopter, φ, θ and ψ are the roll angle, pitch angle and yaw angle of the unmanned helicopter respectively, is the derivative of Θ, R(Θ) is the transformation matrix from the body coordinate system to the inertial coordinate system, F is all external forces except the gravity of the unmanned helicopter, g is the gravitational acceleration, E z is a unit vector, E z =[0,0,1] T , T(Θ) is a 3×3 matrix, J is the moment of inertia matrix, J=diag[I x ,I y ,I z ], I x ,I y and I z are the moments of inertia of the unmanned helicopter body around the x-axis, y-axis and z-axis of the body coordinate system, is the derivative of ω, and M is all the torques acting on the unmanned helicopter.

[0088] Among them, the transformation matrix R(θ) is defined as:

[0089]

[0090] The matrix T(Θ) is defined as:

[0091]

[0092] Among them, c(·) represents cos(·), s(·) represents sin(·), and t(·) represents tan(·).

[0093] According to the above unmanned helicopter fuselage rigid body kinematics and dynamics equations, the longitudinal-lateral sub-control system model of the unmanned helicopter is designed and extracted as follows:

[0094]

[0095] in, is the derivative of P2, P2=[x t ,y t ] T , V2=[v x ,v y ] T , is the derivative of V2, R2=[R 13 ,R 23 ] T , R ij is the element in the i-th row and j-th column of the rotation matrix R(θ), d z is the vertical disturbance force of the unmanned helicopter, T mr is the main rotor lift, d F =[d x ,d y ] T , d x and d y are the longitudinal and lateral interference forces of the unmanned helicopter, is the derivative of R2.

[0096] Based on the backstepping control method, the rotation matrix is ​​used to represent the attitude angle. The roll servo control input, pitch servo control input and yaw servo control input of the unmanned helicopter are derived from the above-mentioned rigid body kinematics and dynamics equations of the unmanned helicopter fuselage and the longitudinal-lateral sub-control system model design of the unmanned helicopter.

[0097] In one embodiment of the present invention, in step S2, when designing the yaw steering gear control input of the unmanned helicopter, the feedback expression shown in the following formula is used for feedback control;

[0098] T tr,ctr =k ped,ctr δ ped -k r,ctr r

[0099] Among them, T tr,ctr is the feedback design force, δ ped is the input signal for the yaw servo control of the unmanned helicopter, k ped,ctr and k r,ctr is the design parameter.

[0100] Feedback control can convert non-physical input into physical input, that is, convert the tail rotor thrust into yaw servo control input.

[0101] In one embodiment of the present invention, the actual roll, pitch and yaw servo control inputs of the final design are:

[0102]

[0103] Among them, δ aer are the roll, pitch and yaw servo control inputs, A, B, M1 and M2 are the model parameter matrices, K ω is the parameter to be designed, Sk(·) is a skew-symmetric matrix, R e is the roll and pitch attitude error, for The estimated value calculated by the command filter, ω d The derivative of e is the yaw angle tracking error, ∈2 is the design parameter, T hov is the lift generated by the main rotor of the unmanned helicopter when it is hovering, H mr H is the vertical distance from the center of gravity of the unmanned helicopter to the main rotor hub. tr D is the vertical distance from the center of gravity of the unmanned helicopter to the tail rotor hub. tr It is the longitudinal horizontal distance from the center of gravity of the unmanned helicopter to the tail rotor hub.

[0104] Step S3, determining the vertical position kinematics and dynamics equations of the unmanned helicopter, and designing the collective pitch servo control input of the unmanned helicopter based on the backstepping control method according to the vertical position kinematics and dynamics equations of the unmanned helicopter, wherein an adaptive radial basis function neural network is used for estimation and compensation of unmodeled dynamics and external disturbances.

[0105] In one embodiment of the present invention, in step S3, the improved adaptive radial basis function neural network is used to estimate and compensate for unmodeled dynamics and external disturbances.

[0106] Since the coupling effect between the altitude control channel of the unmanned helicopter and the three channels of roll, pitch and yaw is small, in one embodiment of the present invention, a controller is designed for the altitude control channel alone.

[0107] In one embodiment of the present invention, for an unmanned helicopter, the vertical position kinematics and dynamics equations can be expressed as:

[0108]

[0109] Among them, z t is the position of the center of gravity of the unmanned helicopter in the z-axis direction of the inertial coordinate system, that is, the actual height of the unmanned helicopter, For z t The derivative of v z The derivative of .

[0110] Based on the backstepping control method, the collective pitch servo control input of the unmanned helicopter is derived from the above vertical position kinematics and dynamics equations:

[0111]

[0112] Among them, δ col is the collective pitch servo control input, is α z The derivative of z is a highly virtual control item. k z is a design parameter greater than zero, z e is the vertical height error, z e =z t -z r , z r is the desired height, For z r The derivative of k vz is a design parameter greater than zero, k int1 and k int2 are system model parameters, b m is the number of main rotor blades, k col is the linkage gain from the collective pitch of the main rotor blades of the unmanned helicopter to the collective pitch servo, v im,hov is the downwash speed of the main rotor of the unmanned helicopter in the hovering state, and w is the vertical speed of the unmanned helicopter in the body coordinate system.

[0113] In one embodiment of the present invention, in step S3, when designing and deriving the collective pitch servo control input of the unmanned helicopter, the feedback expression shown in the following formula is used for feedback control;

[0114] T int =k int1 w+k int1 k int2 k col δ col -kint1 v im,hov

[0115] Among them, T int It is the feedback force of the main rotor of the unmanned helicopter.

[0116] Feedback control can convert non-physical input into physical input, that is, convert the main rotor lift into the collective pitch servo control input of the altitude control channel.

[0117] An adaptive neural network-based UAV flight control method provided in one embodiment of the present invention simplifies the analysis of the flapping dynamics model of the main rotor, and then designs the longitudinal-lateral position control law and the internal three-axis attitude control law of the unmanned helicopter, as well as the altitude control law. An improved adaptive radial basis function neural network is used to estimate and compensate for unmodeled dynamics and external disturbances, and a corresponding adaptive law is designed according to Lyapunov stability theory. This can improve the flight control accuracy and robustness of the UAV and reduce the calculation amount of the controller.

[0118] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In addition, "front", "back", "left", "right", "upper" and "lower" in this article are all referenced to the placement state shown in the accompanying drawings.

[0119] Finally, it should be noted that 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 make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A UAV flight control method based on adaptive neural network, characterized in that: The method is used for flight control of an unmanned helicopter, comprising: The static stable flapping model is used to replace the flapping dynamics model of the main rotor of the unmanned helicopter; According to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage, the longitudinal-lateral sub-control system model of the unmanned helicopter is designed and extracted. Based on the backstepping control method, the roll servo control input, pitch servo control input and yaw servo control input of the unmanned helicopter are designed according to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage and the longitudinal-lateral sub-control system model of the unmanned helicopter. Among them, for unmodeled dynamics and external disturbances, an adaptive radial basis function neural network is used for estimation and compensation; Determine the vertical position kinematics and dynamics equations of the unmanned helicopter, and design the collective pitch servo control input of the unmanned helicopter based on the backstepping control method according to the vertical position kinematics and dynamics equations of the unmanned helicopter. For unmodeled dynamics and external disturbances, an adaptive radial basis function neural network is used for estimation and compensation. The designed roll, pitch and yaw servo control inputs are: Among them, δ aer are the roll, pitch and yaw servo control inputs, k ped,ctr and k r,ctr are design parameters, A, B, M1 and M2 are model parameter matrices, J is the moment of inertia matrix, K ω is the parameter to be designed, ω e is the angular velocity error of the unmanned helicopter body, ω is the actual angular velocity vector of the unmanned helicopter body, Sk(·) is a skew-symmetric matrix, R e is the roll and pitch attitude error, R ij is the element in the i-th row and j-th column of the rotation matrix R(θ), R(θ) is the transformation matrix from the body coordinate system to the inertial coordinate system, for The estimated value calculated by the command filter, ω d The derivative of d is the desired angular velocity vector of the unmanned helicopter, θ is the pitch angle of the unmanned helicopter, φ is the roll angle of the unmanned helicopter, and ψ e is the yaw angle tracking error, E z is a unit vector, E z =[0,0,1] T ,∈2 is the design parameter, is the estimated disturbance and uncertainty value in attitude position control, H2(ω) is the Gaussian function of ω, θ lat0 and θ lon0 They represent the initial main rotor lateral cyclic pitch deviation and longitudinal cyclic pitch deviation respectively, k β is the main rotor blade stiffness coefficient, T hov is the lift generated by the main rotor of the unmanned helicopter when it is hovering, H mr H is the vertical distance from the center of gravity of the unmanned helicopter to the main rotor hub. tr D is the vertical distance from the center of gravity of the unmanned helicopter to the tail rotor hub. tr It is the longitudinal horizontal distance from the center of gravity of the unmanned helicopter to the tail rotor hub; The vertical position kinematics and dynamics equations of the unmanned helicopter are: According to the vertical position kinematics and dynamics equations of the unmanned helicopter, the collective pitch servo control input of the unmanned helicopter is designed as follows: Among them, z t is the position of the center of gravity of the unmanned helicopter in the z-axis direction of the inertial coordinate system, For z t The derivative of v z The derivative of z is the velocity of the center of gravity of the unmanned helicopter in the z-axis direction of the inertial coordinate system, m is the mass of the unmanned helicopter system, T mr is the main rotor lift, g is the acceleration due to gravity, δ col is the collective pitch servo control input, is α z The derivative of z is a highly virtual control item. k z is a design parameter greater than zero, z e is the vertical height error, z e =z t -z r , z r is the desired height, For z r The derivative of ze is the vertical velocity error, v ze =v z -α z , v z is the velocity of the center of gravity of the unmanned helicopter in the z-axis direction of the inertial coordinate system, α z is a highly virtual control term, k vz is a design parameter greater than zero, k int1 and k int2 are system model parameters, ρ is the air density, Ω m is the main rotor speed, R m is the radius of the main rotor blade, C lα,mr is the lift coefficient of the main rotor blade, b m is the number of main rotor blades, c m is the main rotor blade chord length, k col is the linkage gain from the collective pitch of the main rotor blades of the unmanned helicopter to the collective pitch servo, v im,hov is the downwash speed of the main rotor of the unmanned helicopter in the hovering state, w is the vertical speed of the unmanned helicopter in the body coordinate system, ε3 is a design parameter greater than zero, for The derivative of is the estimated disturbance and uncertainty value in altitude position control, H3(v z ) is v z Gaussian function.

2. The UAV flight control method based on adaptive neural network according to claim 1 is characterized in that: The flapping dynamics model of the main rotor of the unmanned helicopter is: Where a and b represent the longitudinal and lateral flapping angles of the main rotor blade tip plane of the unmanned helicopter, respectively. is the derivative of a, is the derivative of b. When the azimuth angle Ψ of the unmanned helicopter m = 0°, the main rotor flaps downward, a is positive, and when the azimuth angle Ψ m =270°, the main rotor flaps downward, b is positive, γ m is the total Locke number of the two blades of the unmanned helicopter, ρ is the air density, c m is the main rotor blade chord length, C lα,mr is the lift coefficient of the main rotor blade, R m is the radius of the main rotor blade, I β is the main rotor blade moment of inertia, Ω m is the main rotor speed, k β is the main rotor blade stiffness coefficient, p and q represent the rolling angular velocity and pitch angular velocity of the unmanned helicopter, θ lat and θ lon They represent the lateral cyclic pitch variation and longitudinal cyclic pitch variation of the main rotor respectively.

3. The UAV flight control method based on adaptive neural network according to claim 2 is characterized in that: The static stable waving model is: in, δ lat is the input signal of the lateral cyclic variable pitch servo of the unmanned helicopter, δ lon is the input signal of the longitudinal cyclic variable pitch servo of the unmanned helicopter, θ lat0 and θ lon0 They represent the initial main rotor lateral cyclic pitch deviation and longitudinal cyclic pitch deviation respectively.

4. The UAV flight control method based on adaptive neural network according to claim 3 is characterized in that: The adaptive radial basis function neural network comprises: An input layer, a hidden layer and an output layer, the input layer is connected to the hidden layer, the hidden layer is connected to the output layer, the input layer includes a plurality of input nodes, the hidden layer includes a plurality of hidden nodes, and the output layer includes an output node; The output of the adaptive radial basis function neural network is designed to be: in, is the estimated value of Γ, W represents the weight matrix between the hidden layer and the output layer, H represents the matrix composed of all radial basis functions of the neural network, ||·||2 represents the 2-norm, Represents the square of the 2-norm.

5. The UAV flight control method based on adaptive neural network according to claim 4 is characterized in that: When performing flight control of the unmanned helicopter, the parameters The adaptive law is designed as: in, for The derivative of are the estimated disturbance and uncertainty values ​​in horizontal position control, for The derivative of is the estimated disturbance and uncertainty value in attitude position control, for The derivative of is the estimated disturbance and uncertainty value in altitude position control, ζ i , ε i and i is a design parameter greater than zero, i=1,2,3, V 2e is the horizontal velocity error vector, V 2e =V2-α P , V2=[v x ,v y ] T , v x and v y are the speeds of the center of gravity of the unmanned helicopter in the x-axis and y-axis directions of the inertial coordinate system, α P is the virtual control item of horizontal position, ω e is the angular velocity error of the unmanned helicopter body, ω e =ω-ω d ,ω andω d are the actual and expected angular velocity vectors of the unmanned helicopter body, ω=[p,q,r] T , r is the yaw angular velocity of the unmanned helicopter, v ze is the vertical velocity error, v ze =v z -α z , v z is the velocity of the center of gravity of the unmanned helicopter in the z-axis direction of the inertial coordinate system, α z are virtual control items, H1(V2), H2(ω) and H3(v z ) are Gaussian functions of the corresponding variables, θ is the pitch angle of the unmanned helicopter, and φ is the roll angle of the unmanned helicopter.

6. The UAV flight control method based on adaptive neural network according to claim 5 is characterized in that: The rigid body kinematics and dynamics equations of the unmanned helicopter fuselage are: According to the rigid body kinematics and dynamics equations of the unmanned helicopter fuselage, the longitudinal-lateral sub-control system model of the unmanned helicopter is designed and extracted as follows: in, is the derivative of P, P = [x t ,y t ,z t ] T is the position vector of the center of gravity of the unmanned helicopter in the inertial coordinate system, x t ,y t and z t are the positions of the center of gravity in the x-axis, y-axis and z-axis directions of the inertial coordinate system, V = [v x ,v y ,v z ] T is the velocity vector of the center of gravity of the unmanned helicopter in the inertial coordinate system, v x 、v y and v z are the velocities of the center of gravity in the x-axis, y-axis and z-axis directions of the inertial coordinate system, m is the mass of the unmanned helicopter system, is the derivative of V, Θ=[φ,θ,ψ] T is the attitude angle vector of the unmanned helicopter, φ, θ and ψ are the roll angle, pitch angle and yaw angle of the unmanned helicopter respectively, is the derivative of Θ, R(Θ) is the transformation matrix from the body coordinate system to the inertial coordinate system, F is all external forces except the gravity of the unmanned helicopter, g is the gravitational acceleration, E z is a unit vector, E z =[0,0,1] T , T(Θ) is a 3×3 matrix, J is the moment of inertia matrix, J=diag[I x ,I y ,I z ], I x ,I y and I z are the moments of inertia of the unmanned helicopter body around the x-axis, y-axis and z-axis of the body coordinate system, is the derivative of ω, M is all the torques on the unmanned helicopter, is the derivative of P2, P2=[x t ,y t ] T , V2=[v x ,v y ] T , is the derivative of V2, R2=[R 13 ,R 23 ] T , R ij is the element in the i-th row and j-th column of the rotation matrix R(θ), d z is the vertical disturbance force of the unmanned helicopter, T mr is the main rotor lift, d F =[d x ,d y ] T , d x and d y are the longitudinal and lateral interference forces of the unmanned helicopter, is the derivative of R2.

7. The UAV flight control method based on adaptive neural network according to claim 6 is characterized in that: When designing the yaw servo control input of the unmanned helicopter, the following feedback expression is used for feedback control; T tr,ctr =k ped,ctr δ ped -k r,ctr r Among them, T tr,ctr is the feedback design force, δ ped is the input signal for the yaw servo control of the unmanned helicopter, k ped,ctr and k r,ctr is the design parameter.

8. The UAV flight control method based on adaptive neural network according to claim 7 is characterized in that: When designing the collective pitch servo control input of the unmanned helicopter, the following feedback expression is used for feedback control; T int =k int1 w+k int1 k int2 k col δ col -k int1 v im,hov Among them, T int It is the feedback force of the main rotor of the unmanned helicopter.

Images