A UAV sideslip turning control method based on neural network intelligent control

Abstract

The invention provides a UAV sideslip turning control method based on neural network intelligent control, which solves the first-order derivative and the second-order derivative signal of the UAV lateral expected trajectory signal through a combined differentiator, and then solves the lateral velocity and lateral acceleration error signal through the yaw angular rate, the yaw angle and the flight speed signal, and then forms the yaw angle expected signal through error synthesis, and obtains the yaw angle error signal by comparing with the yaw angle, and then generates the yaw angular rate error signal and the yaw angular acceleration error signal through the combined differentiator and the angular velocity signal, and finally trains a radial basis function neural network with the above position error, velocity error, acceleration error, angle error, angular velocity error and angular acceleration error, and uses the neural network to approximate the uncertainty error of the system, and superimposes the error and the error integral signal to form the final UAV yaw channel control total signal, so as to realize the intelligent control of the UAV sideslip turning.

Description

Technical Field

[0001] The present invention relates to the field of aircraft control, and in particular to a lateral snaking maneuver turning control method for a high-speed aircraft based on neural network intelligent control. Background Art

[0002] The flight of low-speed UAV is difficult to obtain an accurate model due to the uncertainty of aerodynamic parameters, airflow and modeling errors of low-speed wind tunnels. There must be model errors or unmodeled dynamics. Therefore, although the traditional PID control can achieve system stability, it is very difficult to further improve the rapidity and anti-interference ability after reaching a certain level, and there are great bottleneck restrictions. At the same time, angular acceleration plays a very important role in the stable control of UAVs, but the current angular acceleration measurement instrument has low accuracy, making it difficult to achieve satisfactory control effects. Based on the above background reasons, the present invention considers using a type of combined differentiator to realize the solution of angle error to angular acceleration error, and proposes a method that combines traditional error feedback with neural network intelligent control, and uses the big data processing ability of neural networks in intelligent control to eliminate the interference caused by system uncertainty, so that the entire control method has good anti-uncertainty interference ability. At the same time, the method of combined differentiator is used to realize the integration of six types of error signals from position translation to center of mass rotation, providing a large amount of error data for the neural network, and also making the invention method produce better results in experiments, so that the method not only has theoretical value, but also has high engineering practical value.

[0003] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the invention

[0004] The purpose of the present invention is to provide a UAV sideslip turning control method based on neural network intelligent control, thereby overcoming the problem of weak anti-uncertainty interference ability of sideslip turning caused by limitations and defects of related technologies.

[0005] According to one aspect of the present invention, a method for controlling a sideslip turn of an unmanned aerial vehicle based on neural network intelligent control is provided, comprising the following steps:

[0006] Step S10, install an attitude gyroscope to measure the yaw angle of the UAV, install a rate gyroscope to measure the yaw angle rate of the UAV, install an inertial navigation device to measure the lateral displacement signal and flight speed of the UAV, and then set the lateral expected trajectory signal according to the task; compare the lateral displacement signal with the lateral expected trajectory signal to obtain a lateral position error signal; then design first-order and second-order transfer functions to form a combined differentiator, input the lateral displacement signal and the lateral expected trajectory into the combined differentiator respectively, and obtain the first-order derivative signal of the lateral displacement and the second-order derivative signal of the lateral expected trajectory.

[0007] In step S20, the first-order derivative signal of the lateral displacement and the first-order derivative signal of the lateral desired trajectory are passed through a combined differentiator again to obtain a second-order derivative signal of the lateral displacement and a second-order derivative signal of the lateral desired trajectory, and then the flight speed signal, the yaw angle signal, the first-order derivative signal of the lateral displacement and the first-order derivative signal of the lateral desired trajectory are used to calculate the lateral velocity error signal; the flight speed signal, the yaw angle rate signal, the second-order derivative signal of the lateral displacement and the second-order derivative signal of the lateral desired trajectory are used to calculate the lateral acceleration error signal.

[0008] Step S30, performing nonlinear transformation and integration on the lateral position error signal to obtain a nonlinear integral signal of the lateral position error; then integrating the lateral position error signal, the lateral velocity error signal, the lateral acceleration error signal and the nonlinear transformation signal of the lateral position error to form an expected yaw angle signal; then comparing with the yaw angle signal to obtain a yaw angle error signal; then designing first-order and second-order transfer functions to form a combined differentiator, inputting the expected yaw angle signal into the combined differentiator to obtain an expected first-order derivative signal of the yaw angle; and then inputting the expected first-order derivative signal of the yaw angle into the combined differentiator to obtain an expected second-order derivative signal of the yaw angle.

[0009] Step S40, input the yaw angle rate into the combined differentiator to obtain the first-order derivative signal of the yaw angle rate; then compare the yaw angle rate signal with the expected first-order derivative signal of the yaw angle to obtain the yaw angle rate error signal; compare the first-order derivative signal of the yaw angle rate with the expected second-order derivative signal of the yaw angle to obtain the yaw angle acceleration error signal.

[0010] Step S50, performing nonlinear transformation on the yaw angle error signal to obtain a yaw angle error nonlinear transformation signal and then integrating it to obtain a yaw angle error nonlinear integration signal; then establishing a small six-input radial basis function neural network, using the lateral position error signal, the lateral velocity error signal, the lateral acceleration error signal, the yaw angle error signal, the yaw angle rate error signal, and the yaw angle acceleration error signal as network inputs respectively, and designing a neural network comprehensive error signal according to the yaw angle expected signal, the yaw angle error signal, the yaw angle rate error signal, the yaw angle acceleration error signal, the yaw angle error nonlinear transformation signal and the yaw angle error nonlinear integration signal; then designing a weight adaptive adjustment rule of the neural network according to the neural network comprehensive error signal, and obtaining the weight of the neural network after integration, and then solving the six sub-outputs of the neural network respectively, and finally obtaining the total output of the neural network by integration, and finally superimposing the yaw angle error signal, the yaw angle rate error signal, the yaw angle acceleration error signal, the yaw angle error nonlinear transformation signal and the yaw angle error nonlinear integration signal to form the final UAV yaw channel control total signal.

[0011] In an exemplary embodiment of the present invention, an attitude gyroscope is installed to measure the yaw angle of the UAV, a rate gyroscope is installed to measure the yaw angle rate of the UAV, an inertial navigation device is installed to measure the lateral displacement signal and the flight speed of the UAV, and then a lateral desired trajectory signal is set according to the task; the lateral displacement signal is compared with the lateral desired trajectory signal to obtain a lateral position error signal; then a first-order and second-order transfer function is designed to form a combined differentiator, and the lateral displacement signal and the lateral desired trajectory are respectively input into the combined differentiator to obtain a first-order derivative signal of the lateral displacement and a second-order derivative signal of the lateral desired trajectory, which includes:

[0012] e1=zz d

[0013]

[0014]

[0015] Where ψ is the yaw angle of the UAV measured by the attitude gyroscope, ω y is the yaw rate of the UAV measured by the rate gyroscope, z is the lateral displacement signal of the UAV measured by the inertial navigation device, v is the flight speed of the UAV measured by the inertial navigation device, z d The lateral desired trajectory signal is set for the task; e1 is the lateral position error signal; T1 and T2 are the time constants of the combined differentiator formed by the first-order and second-order transfer functions, which are constant parameters, and s is the differential operator of the transfer function; z1 is the first-order derivative signal of the lateral displacement; z3 is the second-order derivative signal of the lateral desired trajectory; z2 is the second-order derivative signal of the lateral displacement.

[0016] In an exemplary embodiment of the present invention, the first-order derivative signal of the lateral displacement and the first-order derivative signal of the lateral desired trajectory are again passed through a combined differentiator to obtain a second-order derivative signal of the lateral displacement and a second-order derivative signal of the lateral desired trajectory, and then the flight speed signal and the yaw angle signal, the first-order derivative signal of the lateral displacement and the first-order derivative signal of the lateral desired trajectory are used to resolve the lateral velocity error signal; the flight speed signal and the yaw angle rate signal, the second-order derivative signal of the lateral displacement and the second-order derivative signal of the lateral desired trajectory are used to resolve the lateral acceleration error signal, including:

[0017]

[0018] e2=vsin(ψ)+z1-2z3;

[0019] e3=vω y +z2-2z4;

[0020] Wherein z4 is the second-order derivative signal of the lateral desired trajectory; e2 is the lateral velocity error signal; and e3 is the lateral acceleration error signal.

[0021] In an exemplary embodiment of the present invention, the lateral position error signal is nonlinearly transformed and then integrated to obtain a lateral position error nonlinear integral signal; then the lateral position error signal, the lateral velocity error signal, the lateral acceleration error signal and the lateral position error nonlinear transformation signal are integrated to form a yaw angle desired signal; then the yaw angle error signal is compared with the yaw angle signal to obtain a yaw angle error signal; then a first-order and a second-order transfer function are designed to form a combined differentiator, the yaw angle desired signal is input into the combined differentiator, and a yaw angle desired first-order derivative signal is obtained; then the yaw angle desired first-order derivative signal is input into the combined differentiator to obtain a yaw angle desired second-order derivative signal, which includes:

[0022]

[0023] s1=∫e a dt;

[0024] ψ 1 =c1e1+c2e2+c3e3+c4s1+c5e a ;

[0025] e4=ψ-ψ 1 ;

[0026]

[0027] where e a is the nonlinear transformation signal of the lateral position error, s1 is the nonlinear integral signal of the lateral position error; c6 and ε1 are constant parameters of the nonlinear transformation; ψ 1is the expected signal of yaw angle, c1, c2, c3, c4, c5 are constant parameters; e4 is the yaw angle error signal; T1 and T2 are the time constants of the combined differentiator formed by the first-order and second-order transfer functions, which are constant parameters, and s is the differential operator of the transfer function; ψ2 is the expected first-order derivative signal of yaw angle; ψ3 is the expected second-order derivative signal of yaw angle;

[0028] In an exemplary embodiment of the present invention, the yaw angular rate is input into a combined differentiator to obtain a first-order derivative signal of the yaw angular rate; then the yaw angular rate signal is compared with an expected first-order derivative signal of the yaw angle to obtain a yaw angular rate error signal; and the yaw angular rate first-order derivative signal is compared with an expected second-order derivative signal of the yaw angle to obtain a yaw angle acceleration error signal, including:

[0029]

[0030] e5=ω y -ψ2;

[0031] e6 = ψ4 - ψ3;

[0032] Wherein, ψ4 is the first-order derivative signal of the yaw angular rate; e5 is the yaw angular rate error signal; and e6 is the yaw angular acceleration error signal.

[0033] In an exemplary embodiment of the present invention, a nonlinear transformation is performed on the yaw angle error signal to obtain a nonlinear transformation signal of the yaw angle error, and then the signal is integrated to obtain a nonlinear integral signal of the yaw angle error; then a small six-input radial basis function neural network is established, and the lateral position error signal, the lateral velocity error signal, the lateral acceleration error signal, the yaw angle error signal, the yaw angle rate error signal, and the yaw angle acceleration error signal are used as network inputs respectively, and a neural network comprehensive error signal is designed according to the yaw angle expected signal, the yaw angle error signal, the yaw angle rate error signal, the yaw angle acceleration error signal, the yaw angle error nonlinear transformation signal and the yaw angle error nonlinear integral signal; then, a weight adaptive adjustment rule of the neural network is designed according to the neural network comprehensive error signal, and the weight of the neural network is obtained after integration, and then the six sub-outputs of the neural network are solved respectively, and finally the total output of the neural network is obtained by integration, and finally the yaw angle error signal, the yaw angle rate error signal, the yaw angle acceleration error signal, the yaw angle error nonlinear transformation signal and the yaw angle error nonlinear integral signal are superimposed to form the final UAV yaw channel control total signal including:

[0034]

[0035] s2=∫(c 12 e4+e f )dt;

[0036]

[0037]

[0038]

[0039] u w =u w1 +u w2 +u w3 +u w4 +u w5 +u w6 ;

[0040] u=c 11 e4+c7e5+c8e6+c9s2+c 10 e f +u w ;

[0041] where e f is the nonlinear transformation signal of the yaw angle error; s2 is the nonlinear integration signal of the yaw angle error; c7, c8, c9, c 10 、c 11 、c 12 is a constant parameter; s w is the neural network integrated error signal; All of them are the adaptive adjustment rules of the weights of the neural network, and the weights of the neural network are obtained after integration; w 11i 、w 12i 、w 21i 、w 22i 、w 31i 、w 32i 、w 41i 、w 42i 、w 51i 、w 52i 、w 61i 、w 62i is the weight of the neural network; k 11 , k 12 , k 21 , k 22 , k 31 , k 32 , k 41 , k 42 , k 51 , k 52 , k 61 , k 62 is a constant parameter used to adjust the convergence speed of the neural network; u w1 is the first output of the neural network, u w2 is the second output of the neural network, u w3is the third output of the neural network, u w4 is the fourth output of the neural network, u w5 is the fifth output of the neural network, u w6 is the sixth output of the neural network; u w is the total output of the neural network; ε 1i , σ 1i is a constant parameter used to adjust the center point position of the first input; ε 2i , σ 2i is a constant parameter used to adjust the center point position of the second input; ε 3i , σ 3i is a constant parameter used to adjust the center point position of the third input; ε4i and σ4i are constant parameters used to adjust the center point position of the fourth input; ε 5i , σ 5i is a constant parameter used to adjust the center point position of the fifth input; ε 6i , σ 6i is a constant parameter used to adjust the center point position of the sixth input; u is the final total signal of the UAV yaw channel control, which is transmitted to the UAV yaw channel yaw rudder system to realize the task of intelligently controlling the UAV sideslip turning by the neural network.

[0042] Beneficial Effects

[0043] The invention provides a UAV side-slip turning control method based on neural network intelligent control, and its main innovations are as follows: using a combined differentiator method, the six errors of position error, velocity error, acceleration error at the translational level in the UAV side-slip turning, and yaw angle error, yaw angular rate error, and yaw angular acceleration error in the rotation of the UAV are uniformly controlled, providing all-round error feedback information for high-performance side-slip turning. Secondly, a type of radial basis neural network is provided, so that the neural network can use the above-mentioned massive error data to perform a large amount of data processing, thereby using the intelligent data processing capability of the neural network to eliminate the adverse effects of uncertain interference in lateral position control, so that the entire scheme has good stability and robustness.

[0044] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present invention, and together with the specification are used to explain the principles of the present invention. Obviously, the accompanying drawings described below are only some embodiments of the present invention, and for those of ordinary skill in the art, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0046] Figure 1 This is a flow chart of a UAV side-slip turning control method based on neural network intelligent control provided by the present invention;

[0047] Figure 2 is a yaw rate signal curve of the UAV according to the method provided in the embodiment of the present invention (unit: degree per second);

[0048] Figure 3 is a yaw angle signal curve of the UAV according to the method provided in the embodiment of the present invention (unit: degree);

[0049] Figure 4 is a signal curve of the lateral displacement signal of the UAV according to the method provided in the embodiment of the present invention (unit: meter);

[0050] Figure 5 is a signal curve of the lateral velocity error signal of the method provided in the embodiment of the present invention (unit: meters per second);

[0051] Figure 6 is a lateral acceleration error signal curve of the method provided in the embodiment of the present invention (unit: meter per second squared);

[0052] Figure 7 is the total output signal curve of the neural network of the method provided in the embodiment of the present invention (unit: degree);

[0053] Figure 8 It is the total signal curve of the UAV yaw channel control of the method provided in the embodiment of the present invention (unit: degree). DETAILED DESCRIPTION

[0054] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present invention will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present invention. However, those skilled in the art will appreciate that the technical solutions of the present invention may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid blurring out various aspects of the present invention.

[0055] The present invention provides a method for solving the first-order derivative and the second-order derivative signal of the lateral desired trajectory signal of the unmanned aerial vehicle through a combined differentiator, and then solving the lateral velocity and the lateral acceleration error signal through the yaw angle rate, the yaw angle and the flight speed signal, and then forming the yaw angle desired signal through error synthesis, and obtaining the yaw angle error signal by comparing with the yaw angle, and then generating the yaw angle rate error signal and the yaw angle acceleration error signal through the combined differentiator and the angular velocity signal, and finally training a class of radial basis function neural network with the above position error, velocity error, acceleration error, angle error, angular velocity error and angular acceleration rate error, using the neural network to approximate the uncertainty error of the system, and superimposing the yaw angle error signal, the yaw angle rate error signal, the yaw angle acceleration error signal, the yaw angle error nonlinear transformation signal and the yaw angle error nonlinear integral signal to form the final unmanned aerial vehicle yaw channel control total signal, so as to realize the intelligent control method of the unmanned aerial vehicle sideslip turning.

[0056] Next, a method for controlling a UAV side slip and turn based on neural network intelligent control according to the present invention will be further explained and illustrated in conjunction with the accompanying drawings. Figure 1 As shown, the UAV side-slip turning control method based on neural network intelligent control may include the following steps:

[0057] Step S10, install an attitude gyroscope to measure the yaw angle of the UAV, install a rate gyroscope to measure the yaw angle rate of the UAV, install an inertial navigation device to measure the lateral displacement signal and flight speed of the UAV, and then set the lateral expected trajectory signal according to the task; compare the lateral displacement signal with the lateral expected trajectory signal to obtain a lateral position error signal; then design first-order and second-order transfer functions to form a combined differentiator, input the lateral displacement signal and the lateral expected trajectory signal into the combined differentiator respectively, and obtain a first-order derivative signal of the lateral displacement and a first-order derivative signal of the lateral expected trajectory.

[0058] Specifically, first install the attitude gyroscope to measure the yaw angle of the drone, denoted as ψ. Install the rate gyroscope to measure the yaw rate of the drone, denoted as ω. y .

[0059] Secondly, an inertial navigation device is installed to measure the lateral displacement signal of the UAV, which is recorded as z. The flight speed of the UAV measured by the inertial navigation device is recorded as v.

[0060] Then, the lateral desired trajectory signal is set according to the task; the lateral displacement signal is compared with the lateral desired trajectory signal to obtain the lateral position error signal as follows:

[0061] e1=zz d ;

[0062] where z d is the lateral desired trajectory signal set according to the task; e1 is the lateral position error signal.

[0063] Next, the first-order and second-order transfer functions are designed to form a combined differentiator, and the lateral displacement signal is input to obtain the first-order derivative signal of the lateral displacement as follows:

[0064]

[0065] Where T1 and T2 are the time constants of the combined differentiator formed by the first-order and second-order transfer functions, which are constant parameters, s is the differential operator of the transfer function; z1 is the first-order derivative signal of the lateral displacement.

[0066] Then, the first-order and second-order transfer functions are designed to form a combined differentiator, and the lateral desired trajectory signal is input to obtain the first-order derivative signal of the lateral desired trajectory as follows:

[0067]

[0068] Where z3 is the first-order derivative signal of the lateral desired trajectory.

[0069] In step S20, the first-order derivative signal of the lateral displacement and the first-order derivative signal of the lateral desired trajectory are passed through a combined differentiator again to obtain a second-order derivative signal of the lateral displacement and a second-order derivative signal of the lateral desired trajectory, and then the flight speed signal, the yaw angle signal, the first-order derivative signal of the lateral displacement and the first-order derivative signal of the lateral desired trajectory are used to calculate the lateral velocity error signal; the flight speed signal, the yaw angle rate signal, the second-order derivative signal of the lateral displacement and the second-order derivative signal of the lateral desired trajectory are used to calculate the lateral acceleration error signal.

[0070] Specifically, the first-order derivative signal of the lateral displacement is firstly passed through a combined differentiator to obtain the second-order derivative signal of the lateral displacement as follows:

[0071]

[0072] Where z2 is the second-order derivative signal of the lateral displacement.

[0073] Secondly, the first-order derivative signal of the lateral desired trajectory is passed through the combined differentiator to obtain the second-order derivative signal of the lateral desired trajectory as follows:

[0074]

[0075] Where z4 is the second-order derivative signal of the lateral desired trajectory;

[0076] Then, the lateral velocity error signal is calculated by using the flight speed signal, the yaw angle signal, the first-order derivative signal of the lateral displacement, and the first-order derivative signal of the lateral desired trajectory as follows:

[0077] e2=vsin(ψ)+z1-2z3;

[0078] Where e2 is the lateral velocity error signal.

[0079] Finally, the lateral acceleration error signal is calculated using the flight speed signal, the yaw rate signal, the second-order derivative signal of the lateral displacement, and the second-order derivative signal of the lateral desired trajectory as follows:

[0080] e3=vω y +z2-2z4;

[0081] Where e3 is the lateral acceleration error signal.

[0082] Step S30, performing nonlinear transformation and integration on the lateral position error signal to obtain a nonlinear integral signal of the lateral position error; then integrating the lateral position error signal, the lateral velocity error signal, the lateral acceleration error signal and the nonlinear transformation signal of the lateral position error to form an expected yaw angle signal; then comparing with the yaw angle signal to obtain a yaw angle error signal; then designing first-order and second-order transfer functions to form a combined differentiator, inputting the expected yaw angle signal into the combined differentiator to obtain an expected first-order derivative signal of the yaw angle; and then inputting the expected first-order derivative signal of the yaw angle into the combined differentiator to obtain an expected second-order derivative signal of the yaw angle.

[0083] Specifically, firstly, the lateral position error signal is nonlinearly transformed and then integrated to obtain a lateral position error nonlinear integral signal as follows:

[0084]

[0085] s1=∫e a dt;

[0086] Where c6 and ε1 are constant parameters of nonlinear transformation; e a is the nonlinear transformation signal of the lateral position error, and s1 is the nonlinear integration signal of the lateral position error.

[0087] Then, the lateral position error signal, lateral velocity error signal, lateral acceleration error signal and lateral position error nonlinear transformation signal are integrated to form the desired yaw angle signal as follows:

[0088] ψ 1 =c1e1+c2e2+c3e3+c4s1+c5e a ;

[0089] where ψ 1 is the expected signal of yaw angle, c1, c2, c3, c4, c5 are constant parameters.

[0090] Secondly, the yaw angle expected signal is compared with the yaw angle signal to obtain the yaw angle error signal as follows:

[0091] e4=ψ-ψ 1 ;

[0092] Where e4 is the yaw angle error signal.

[0093] Next, the first-order and second-order transfer functions are designed to form a combined differentiator. The desired yaw angle signal is input into the combined differentiator to obtain the desired first-order derivative signal of the yaw angle as follows:

[0094]

[0095] Where T1 and T2 are the time constants of the combined differentiator formed by the first-order and second-order transfer functions, which are constant parameters, s is the differential operator of the transfer function; ψ2 is the expected first-order derivative signal of the yaw angle.

[0096] Finally, the desired first-order derivative signal of the yaw angle is input into the combined differentiator to obtain the desired second-order derivative signal of the yaw angle as follows:

[0097]

[0098] Where ψ3 is the expected second-order derivative signal of the yaw angle.

[0099] Step S40, inputting the yaw angular rate into a combined differentiator to obtain a first-order derivative signal of the yaw angular rate; then comparing the yaw angular rate signal with the expected first-order derivative signal of the yaw angle to obtain a yaw angular rate error signal; comparing the first-order derivative signal of the yaw angular rate with the expected second-order derivative signal of the yaw angle to obtain a yaw angle acceleration error signal includes:

[0100] Specifically, firstly, the yaw angular rate is input into the combined differentiator to obtain the first-order derivative signal of the yaw angular rate as follows:

[0101]

[0102] Where ψ4 is the first-order derivative signal of the yaw angular rate.

[0103] Secondly, the yaw rate signal is compared with the expected first-order derivative signal of the yaw angle to obtain the yaw rate error signal as follows:

[0104] e5=ω y -ψ2;

[0105] Where e5 is the yaw rate error signal.

[0106] Finally, the yaw angle acceleration error signal is obtained by comparing the first-order derivative signal of the yaw angle rate with the expected second-order derivative signal of the yaw angle as follows:

[0107] e6 = ψ4 - ψ3;

[0108] Where e6 is the yaw angle acceleration error signal.

[0109] Step S50, performing nonlinear transformation on the yaw angle error signal to obtain a yaw angle error nonlinear transformation signal and then integrating it to obtain a yaw angle error nonlinear integration signal; then establishing a small six-input radial basis function neural network, using the lateral position error signal, the lateral velocity error signal, the lateral acceleration error signal, the yaw angle error signal, the yaw angle rate error signal, and the yaw angle acceleration error signal as network inputs respectively, and designing a neural network comprehensive error signal according to the yaw angle expected signal, the yaw angle error signal, the yaw angle rate error signal, the yaw angle acceleration error signal, the yaw angle error nonlinear transformation signal and the yaw angle error nonlinear integration signal; then designing a weight adaptive adjustment rule of the neural network according to the neural network comprehensive error signal, and obtaining the weight of the neural network after integration, and then solving the six sub-outputs of the neural network respectively, and finally obtaining the total output of the neural network by integration, and finally superimposing the yaw angle error signal, the yaw angle rate error signal, the yaw angle acceleration error, the yaw angle error nonlinear transformation signal and the yaw angle error nonlinear integration signal to form the final UAV yaw channel control total signal.

[0110] Specifically, it is divided into the following twenty-one steps. The first step is to perform nonlinear transformation on the yaw angle error signal to obtain a yaw angle error nonlinear transformation signal and then integrate it to obtain a yaw angle error nonlinear integration signal

[0111]

[0112] s2=∫(c 12 e4+e f )dt;

[0113] where e f is the nonlinear transformation signal of the yaw angle error; s2 is the nonlinear integration signal of the yaw angle error, where c12 Is a constant parameter.

[0114] In the second step, the neural network comprehensive error signal is calculated according to the yaw angle expected signal, yaw angle error signal, yaw angle rate error signal, yaw angle acceleration error signal, yaw angle error nonlinear transformation signal and yaw angle error nonlinear integral signal as follows:

[0115]

[0116] where s w is the neural network comprehensive error signal; c7, c8, c9, c 10 、c 11 Is a constant parameter.

[0117] The third step is to solve the weight adaptive adjustment rule of the first sub-network of the neural network according to the comprehensive error signal of the neural network as follows:

[0118]

[0119] in is the adaptive adjustment rule of the weights of the first sub-network of the neural network, k 11 , k 12 is a constant parameter used to adjust the convergence speed of the neural network; ε 1i , σ 1i It is a constant parameter used to adjust the center point position of the first input.

[0120] The fourth step is to integrate the adaptive adjustment rule of the weights of the first sub-network of the neural network, and obtain the weights of the first sub-network of the neural network as follows:

[0121]

[0122] where w 11i 、w 12i is the weight of the first sub-network of the neural network.

[0123] The fifth step is to solve the output of the first sub-network of the neural network according to the weights of the first sub-network mentioned above as follows:

[0124]

[0125] Among them, u w1 It is the output of the first sub-network of the neural network.

[0126] In the sixth step, according to the neural network comprehensive error signal, the weight adaptive adjustment rule of the second sub-network of the neural network is solved as follows:

[0127]

[0128] in is the adaptive adjustment rule of the weights of the second sub-network of the neural network, k 21 , k 22 is a constant parameter used to adjust the convergence speed of the neural network; ε 2i , σ 2i It is a constant parameter used to adjust the center point position of the second input.

[0129] In the seventh step, the weight adaptive adjustment rule of the second sub-network of the neural network is integrated to obtain the weight of the second sub-network of the neural network as follows:

[0130]

[0131] where w 21i 、w 22i is the weight of the second sub-network of the neural network.

[0132] In the eighth step, the output of the second sub-network of the neural network is solved according to the weights of the second sub-network as follows:

[0133]

[0134] Among them, u w2 This is the output of the second subnetwork of the neural network.

[0135] In the ninth step, according to the neural network comprehensive error signal, the weight adaptive adjustment rule of the third sub-network of the neural network is solved as follows:

[0136]

[0137] in is the adaptive adjustment rule of the weights of the third sub-network of the neural network, k 21 , k 22 is a constant parameter used to adjust the convergence speed of the neural network; ε 3i , σ 3i It is a constant parameter used to adjust the center point position of the third input.

[0138] The tenth step is to integrate the adaptive adjustment rule of the weights of the third sub-network of the neural network, and obtain the weights of the third sub-network of the neural network as follows:

[0139]

[0140]

[0141] where w 31i 、w 32iis the weight of the third sub-network of the neural network.

[0142] In the eleventh step, according to the weights of the third sub-network mentioned above, the output of the third sub-network of the neural network is solved as follows:

[0143]

[0144] Among them, u w3 It is the output of the third sub-network of the neural network.

[0145] In the twelfth step, according to the comprehensive error signal of the neural network, the weight adaptive adjustment rule of the fourth sub-network of the neural network is solved as follows:

[0146]

[0147] in is the adaptive adjustment rule of the weights of the fourth sub-network of the neural network, k 41 , k 42 is a constant parameter used to adjust the convergence speed of the neural network; ε 4i , σ 4i It is a constant parameter used to adjust the center point position of the fourth input.

[0148] In the thirteenth step, the weight adaptive adjustment rule of the fourth sub-network of the neural network is integrated to obtain the weight of the fourth sub-network of the neural network as follows:

[0149]

[0150] where w 41i 、w 42i is the weight of the fourth sub-network of the neural network.

[0151] Step 14: Based on the weights of the fourth sub-network, the output of the fourth sub-network of the neural network is solved as follows:

[0152]

[0153] Among them, u w4 This is the fourth subnetwork output of the neural network.

[0154] In step 15, according to the neural network comprehensive error signal, the weight adaptive adjustment rule of the fifth sub-network of the neural network is solved as follows:

[0155]

[0156] in is the adaptive adjustment rule of the weights of the fifth sub-network of the neural network, k 51 , k 52is a constant parameter used to adjust the convergence speed of the neural network; ε 5i , σ 5i It is a constant parameter used to adjust the center point position of the fifth input.

[0157] In the sixteenth step, the weight adaptive adjustment rule of the fifth sub-network of the neural network is integrated to obtain the weight of the fifth sub-network of the neural network as follows:

[0158]

[0159] where w 51i 、w 52i is the weight of the fifth sub-network of the neural network.

[0160] Step 17: Based on the weights of the fifth sub-network, the output of the fourth sub-network of the neural network is as follows:

[0161]

[0162] Among them, u w5 This is the fifth subnetwork output of the neural network.

[0163] In step 18, according to the neural network comprehensive error signal, the weight adaptive adjustment rule of the sixth sub-network of the neural network is solved as follows:

[0164]

[0165] in is the adaptive adjustment rule of the weights of the sixth sub-network of the neural network, k 61 , k 62 is a constant parameter used to adjust the convergence speed of the neural network; ε 6i , σ 6i It is a constant parameter used to adjust the center point position of the sixth input.

[0166] In the nineteenth step, the weight adaptive adjustment rule of the sixth sub-network of the neural network is integrated to obtain the weight of the sixth sub-network of the neural network as follows:

[0167]

[0168] where w 61i 、w 62i is the weight of the sixth sub-network of the neural network.

[0169] Step 20: Based on the weights of the sixth sub-network, the output of the fourth sub-network of the neural network is as follows:

[0170]

[0171] Among them, u w6 This is the sixth subnetwork output of the neural network.

[0172] In the 21st step, the six sub-network outputs of the neural network are integrated to obtain the total output of the neural network, and the yaw angle error signal, yaw angle rate error signal, yaw angle acceleration error, yaw angle error nonlinear transformation signal and yaw angle error nonlinear integral signal are superimposed to form the final UAV yaw channel control total signal as follows:

[0173] u w =u w1 +u w2 +u w3 +u w4 +u w5 +u w6 ;

[0174] u=c 11 e4+c7e5+c8e6+c9s2+c 10 e f +u w ;

[0175] where u w is the total output of the neural network; u is the final total signal of the UAV yaw channel control, which is transmitted to the UAV yaw channel yaw rudder system to realize the task of the neural network intelligently controlling the UAV sideslip turning.

[0176] Case implementation and computer simulation results analysis

[0177] In step S10, the lateral desired trajectory signal z is set according to the task d =50sin(0.2t)+30cos(0.3t), select parameters T1=0.003, T2=0.2.

[0178] In step S30, parameters c1=0.02, c2=0.1, c3=0.13, c4=0.2, c5=2, c6=0.005, ε1=2.5 are selected.

[0179] In step S50, select parameter c 12 =0.3, c7=0.3, c8=0.1, c9=0.2, c 10 =0.53, c 11 =0.2, k 11 =0.02, k 12 =0.03, k 21 =0.008, k 22 =0.015, k 31 =0.018, k 32 =0.022, k41 =0.03, k 42 =0.024, k 51 =0.08, k 52 =0.005, k 61 =0.008, k 62 =0.005.

[0180] Finally, the UAV yaw rate signal is obtained as follows: Figure 2 As shown; the UAV yaw angle signal is obtained as Figure 3 As shown; the lateral displacement signal of the UAV is obtained as Figure 4 As shown; the lateral speed error signal is obtained as Figure 5 As shown; the lateral acceleration error signal is obtained as Figure 6 As shown; the total output signal of the neural network is Figure 7 As shown; the total signal of the UAV yaw channel control is obtained as Figure 8 As shown in the image, it can be seen that the UAV tracks the expected sinusoidal position signal with a yaw angle of about 2 degrees, and the output of the neural network accounts for about 50% of the total control signal after stabilization. At the same time, the speed error and acceleration error both make sinusoidal fluctuations around 0 as the mean value; from the perspective of the entire experimental effect, the method provided by the present invention is feasible and effective.

Claims

1. A UAV side-slip turning control method based on neural network intelligent control, characterized in that: An attitude gyroscope is installed to measure the yaw angle of the UAV, a rate gyroscope is installed to measure the yaw rate of the UAV, and an inertial navigation device is installed to measure the lateral displacement signal and the flight speed of the UAV, and then the lateral expected trajectory signal is set according to the task; the lateral displacement signal is compared with the lateral expected trajectory signal to obtain a lateral position error signal; then the first-order and second-order transfer functions are designed to form a combined differentiator, and the lateral displacement signal and the lateral expected trajectory signal are respectively input into the combined differentiator to obtain the first-order derivative signal of the lateral displacement and the first-order derivative signal of the lateral expected trajectory; then the first-order derivative signal of the lateral displacement and the first-order derivative signal of the lateral expected trajectory are passed through the combined differentiator again to obtain the second-order derivative signal of the lateral displacement and the second-order derivative signal of the lateral expected trajectory, and then the flight speed signal and the yaw angle signal, the first-order derivative signal of the lateral displacement and the first-order derivative signal of the lateral expected trajectory are used to solve the lateral velocity error signal; the flight speed signal and the yaw rate signal, the second-order derivative signal of the lateral displacement and the second-order derivative signal of the lateral expected trajectory are used to solve the lateral acceleration error signal, including: e1=zz d ; e2=vsin(ψ)+z1-2z3; <h2 style=";text-align:left;direction:ltr">e3 = vω<h2 style=";text-align:left;direction:ltr"> y <h2 style=";text-align:left;direction:ltr"> +z2-2z4; Where ψ is the yaw angle of the UAV measured by the attitude gyroscope, ω y is the yaw rate of the UAV measured by the rate gyroscope, z is the lateral displacement signal of the UAV measured by the inertial navigation device, v is the flight speed of the UAV measured by the inertial navigation device, z d The lateral desired trajectory signal is set for the task; e1 is the lateral position error signal; T1 and T2 are the time constants of the combined differentiator formed by the first-order and second-order transfer functions, which are constant parameters, and s is the differential operator of the transfer function; z1 is the first-order derivative signal of the lateral displacement; z3 is the first-order derivative signal of the lateral desired trajectory; z2 is the second-order derivative signal of the lateral displacement; z4 is the second-order derivative signal of the lateral desired trajectory; e2 is the lateral velocity error signal; e3 is the lateral acceleration error signal.

2. According to claim 1, a UAV side-slip turning control method based on neural network intelligent control is characterized in that: According to the lateral position error signal, a nonlinear transformation is performed and then integrated to obtain a lateral position error nonlinear integral signal; then the lateral position error signal, the lateral velocity error signal, the lateral acceleration error signal and the lateral position error nonlinear transformation signal are integrated to form a yaw angle expected signal; then the yaw angle error signal is compared with the yaw angle signal to obtain a yaw angle error signal; then a first-order and second-order transfer function is designed to form a combined differentiator, and the yaw angle expected signal is input into the combined differentiator to obtain a yaw angle expected first-order derivative signal; then the yaw angle expected first-order derivative signal is input into the combined differentiator to obtain a yaw angle expected second-order derivative signal; then the yaw angle rate is input into the combined differentiator to obtain a yaw angle rate first-order derivative signal; then the yaw angle rate signal is compared with the yaw angle expected first-order derivative signal to obtain a yaw angle rate error signal; the yaw angle rate first-order derivative signal is compared with the yaw angle expected second-order derivative signal to obtain a yaw angle acceleration error signal, which includes: s1=∫e a dt; ψ 1 =c1e1+c2e2+c3e3+c4s1+c5e a ; e4=ψ-ψ 1 ; e5=ω y -ψ2; e6 = ψ4 - ψ3; where e a is the nonlinear transformation signal of the lateral position error, s1 is the nonlinear integral signal of the lateral position error; c6 and ε1 are constant parameters of the nonlinear transformation; ψ 1 is the expected yaw angle signal, c1, c2, c3, c4, c5 are constant parameters; e4 is the yaw angle error signal; T1 and T2 are time constants of the combined differentiator formed by the first-order and second-order transfer functions, which are constant parameters, and s is the differential operator of the transfer function; ψ2 is the expected first-order derivative signal of the yaw angle; ψ3 is the expected second-order derivative signal of the yaw angle; ψ4 is the first-order derivative signal of the yaw angle rate; e5 is the yaw angle rate error signal; e6 is the yaw angle acceleration error signal.

3. The UAV side-slip turning control method based on neural network intelligent control according to claim 2 is characterized in that: According to the yaw angle error signal, a nonlinear transformation is performed to obtain a yaw angle error nonlinear transformation signal, and then the signal is integrated to obtain a yaw angle error nonlinear integration signal; then a small six-input radial basis neural network is established, and the lateral position error signal, the lateral velocity error signal, the lateral acceleration error signal, the yaw angle error signal, the yaw angle rate error signal, and the yaw angle acceleration error signal are used as network inputs respectively, and a neural network comprehensive error signal is designed according to the yaw angle expected signal, the yaw angle error signal, the yaw angle rate error signal, the yaw angle acceleration error signal, the yaw angle error nonlinear transformation signal and the yaw angle error nonlinear integration signal; then, according to the neural network comprehensive error signal, a weight adaptive adjustment rule of the neural network subnetwork is designed, and the weight of the neural network subnetwork is obtained after integration, and then the six subnetwork outputs of the neural network are solved respectively, and finally the total output of the neural network is obtained by synthesis, and finally the yaw angle error signal, the yaw angle rate error signal, the yaw angle acceleration error signal, the yaw angle error nonlinear transformation signal and the yaw angle error nonlinear integration signal are superimposed to form the final UAV yaw channel control total signal including: s2=∫(c 12 e4+e f )dt; in w =in w1 +in w2 +in w3 +in w4 +in w5 +in w6 ; u=c 11 e4+c7e5+c8e6+c9s2+c 10 e f +u w ; where e f is the nonlinear transformation signal of the yaw angle error; s2 is the nonlinear integration signal of the yaw angle error; c7, c8, c9, c 10 、c 11 、c 12 is a constant parameter; s w is the neural network integrated error signal; are all adaptive adjustment rules of the weights of the neural network sub-network, and the weights of the neural network sub-network are obtained after integration; w 11i 、w 12i 、w 21i 、w 22i 、w 31i 、w 32i 、w 41i 、w 42i 、w 51i 、w 52i 、w 61i 、w 62i is the weight of the neural network; k 11 , k 12 , k 21 , k 22 , k 31 , k 32 , k 41 , k 42 , k 51 , k 52 , k 61 , k 62 is a constant parameter used to adjust the convergence speed of the neural network; u w1 is the first sub-network output of the neural network, u w2 is the second sub-network output of the neural network, u w3 is the third sub-network output of the neural network, u w4 is the fourth sub-network output of the neural network, u w5 is the fifth subnetwork output of the neural network, u w6 is the sixth sub-network output of the neural network; u w is the total output of the neural network; ε 1i , σ 1i is a constant parameter used to adjust the center point position of the first input; ε 2i , σ 2i is a constant parameter used to adjust the center point position of the second input; ε 3i , σ 3i is a constant parameter used to adjust the center point position of the third input; ε 4i , σ 4i is a constant parameter used to adjust the center point position of the fourth input; ε 5i , σ 5i is a constant parameter used to adjust the center point position of the fifth input; ε 6i , σ 6i is a constant parameter used to adjust the center point position of the sixth input; u is the final total signal of the UAV yaw channel control, which is transmitted to the UAV yaw channel yaw rudder system to realize the task of intelligently controlling the UAV sideslip turning by the neural network.

Images