Intelligent adaptive control method for gliding aircraft based on dynamic model updating driving

Abstract

The application discloses a method for intelligent adaptive control of a gliding aircraft based on dynamic model updating driving, and belongs to the technical field of aircraft control.The method is as follows: constructing an offline gliding aircraft agent model;designing a gliding aircraft controller;calculating an actual control instruction;and dynamically updating the gliding aircraft agent model and adjusting the adaptive control.The application adopts a neural network to construct a gliding aircraft agent model, avoids modeling errors, and provides a good foundation for online dynamic updating of the aircraft model;an online dynamic updating of the aircraft agent model is carried out by using a sliding mechanism and an incremental learning method based on a memory perception synapse, the adaptability of the aircraft control system to the environment and tasks is enhanced, and the precision of the aircraft agent model is improved;and the actual control instruction is calculated and designed based on the aircraft model agent network and nonlinear model interpolation, the adaptability of the aircraft to interference and uncertainty is improved, and online high-precision control of the aircraft is realized.

Description

TECHNICAL FIELD

[0001] The application relates to a gliding aircraft intelligent adaptive control method based on dynamic model updating driving and belongs to the technical field of aircraft control. BACKGROUND

[0002] The existing aircraft control system design method mainly depends on a fixed model, however, in actual application, the dynamic characteristics of an aircraft and a model thereof are affected by environmental disturbances, task changes and other factors, so that a traditional control method is poor in adaptability and low in control precision. In recent years, an intelligent control method can obtain an approximately optimal control strategy through training, but the intelligent control method is designed and trained offline according to a specific model, is poor in generalization ability and is difficult to be applied to a complex and changeable flight environment.

[0003] Therefore, an intelligent adaptive control method capable of dynamically updating an aircraft model online is urgently needed to improve the robustness and adaptability of an aircraft control system. SUMMARY

[0004] To solve the problems in the background art, the application provides a gliding aircraft intelligent adaptive control method based on dynamic model updating driving.

[0005] To achieve the above object, the application adopts the following technical scheme: a gliding aircraft intelligent adaptive control method based on dynamic model updating driving, the method comprising the following steps:

[0006] S1: offline gliding aircraft agent model construction;

[0007] The S1 comprises the following steps:

[0008] S101: collecting gliding aircraft historical flight data and respectively constructing three-channel aerodynamic moment model identification sample libraries of pitch, yaw and roll;

[0009] S102: establishing a fully connected neural network, training a three-channel aerodynamic moment model agent network of the gliding aircraft by using the sample libraries and forming a gliding aircraft model basic agent network:

[0010] The network input is six-dimensional flight state, and the output is one-dimensional aerodynamic moment coefficient;

[0011] Wherein:

[0012] The pitch channel agent network model is as follows:

[0013] C mz = net z (Ma,α,β,δ x ,δ y ,δ z ) (1)

[0014] The yaw channel proxy network model is as follows:

[0015] C my = net y (Ma, a, b, d x , d y , d z ) (2)

[0016] The roll channel proxy network model is as follows:

[0017] C mx = net x (Ma, a, b, d x , d y , d z ) (3)

[0018] In formulas (1)-(3):

[0019] Ma is the Mach number;

[0020] a is the attack angle;

[0021] b is the sideslip angle;

[0022] d x is the roll channel rudder deflection angle;

[0023] d y is the yaw channel rudder deflection angle;

[0024] d z is the pitch channel rudder deflection angle;

[0025] C mx is the roll channel one-dimensional aerodynamic moment coefficient;

[0026] C my is the yaw channel one-dimensional aerodynamic moment coefficient;

[0027] C mz is the pitch channel one-dimensional aerodynamic moment coefficient.

[0028] S2: glider controller design;

[0029] The S2 includes the following steps:

[0030] S201: construct a control-oriented glider model, and a glider dynamics model around the center of mass is:

[0031]

[0032] In formula (4):

[0033] Q = [a b s] Tis the attitude angle vector, where: a is the roll angle;

[0034] is the first derivative of the attitude angle vector;

[0035] Q is the attitude angular rate transformation matrix;

[0036] ω b = [ω x ω y ω z T is the attitude angular rate vector, where: ω x is the roll angular rate, ω y is the yaw angular rate, and ω z is the pitch angular rate;

[0037] is the first derivative of the attitude angular rate vector;

[0038] Δf F is the model uncertainty term;

[0039] I is the aircraft rotational inertia;

[0040] M is the control moment;

[0041] M d is the disturbance moment;

[0042] S202: define the system state variable as x = [a b s ω x ω y ω z T Rewrite equation (4) into the MIMO system form as follows:

[0043]

[0044] In equation (5):

[0045] is the first derivative of the system state variable;

[0046] f(x) is the system state nonlinear term;

[0047] g(x) is the system input nonlinear term;

[0048] d is the system uncertainty term;

[0049] S203: perform feedback linearization processing on equation (5) to obtain:

[0050]

[0051] In equation (6):

[0052] ​​ is the second derivative of the attitude angle vector;

[0053] F is the state matrix after feedback linearization;

[0054] E is the equivalent control matrix;

[0055] is the disturbance;

[0056] S204: design the state feedback control law:

[0057] M = -E -1 F + Π (7)

[0058] In formula (7):

[0059] Π = [π1 π2 π3] T is the auxiliary control variable, wherein: π1 is the auxiliary control variable of the pitch channel, π2 is the auxiliary control variable of the yaw channel, and π3 is the auxiliary control variable of the roll channel;

[0060] Then:

[0061]

[0062] Introducing the virtual control variable U = EΠ, we can obtain:

[0063]

[0064] S205: the system is decoupled into three second-order systems, and a non-singular fast terminal sliding mode control law is designed for each channel to obtain the virtual control variable U;

[0065] The S205 includes the following steps:

[0066] S20501: define the expected control output as Wherein: α d is the expected attack angle, is the first derivative of the expected attack angle;

[0067] Then the state error vector e α is:

[0068] e α = [e α1 , e α2 ] T = x d -x (10)

[0069] In formula (10):

[0070] e α1 is the attack angle error;

[0071] eα2 is the angle of attack variation rate error;

[0072] S20502: Select nonsingular fast terminal sliding mode surface:

[0073]

[0074] In formula (11):

[0075] k1>0, k2>0, p, q, m and n are constants and all take odd numbers, satisfying

[0076] S20503: Obtain the first derivative of the sliding mode surface:

[0077]

[0078] In formula (12):

[0079] is the second derivative of the expected angle of attack a d ;

[0080] u1 is the pitch channel control variable;

[0081] S20504: Divide the control variable into the equivalent control law u eq on the sliding mode surface and the switching control u sw , that is, the pitch channel control variable u1=u 1eq +u 1sw , let can be obtained:

[0082]

[0083] S20505: Adopt the idea of terminal attractor to design the switching control u 1sw :

[0084]

[0085] In formula (14):

[0086] k3>0, k4>0, k and l are constants and all take odd numbers, satisfying

[0087] S20506: Substitute formula (14) into formula (12) and solve formula (13) to obtain:

[0088]

[0089] Then the pitch channel control variable is:

[0090]

[0091] S20507: similarly, we have

[0092] The yaw channel control variable is

[0093]

[0094] The roll channel control variable is

[0095]

[0096] S20508: we have virtual control variable U = [u1 u2 u3] T .

[0097] S206: we have expected control moment M by formula (7)

[0098] M = -E -1 F + E -1 U (19).

[0099] S3: actual control command calculation;

[0100] The S3 includes the following steps:

[0101] S301: according to the current state of the glider and the expected control moment M of the pitch channel z , we have expected aerodynamic moment coefficient of the pitch channel

[0102]

[0103] In formula (20):

[0104] q is dynamic pressure;

[0105] S is reference area;

[0106] L is reference length;

[0107] S302: calculate the actual control command of the pitch channel;

[0108] The S302 includes the following steps:

[0109] S30201: according to the glider's ability, divide the pitch rudder channel deflection angle into feature point intervals, input the current Mach number, angle of attack, sideslip angle and pitch rudder deflection angle sequence into the aerodynamic network proxy model, i.e. formula (1), to get the corresponding expected aerodynamic moment coefficient sequence of the pitch channel

[0110] S30202: find the maximum and minimum values of the sequence , and judge whether the expected aerodynamic moment coefficient is greater than the sequence the maximum value in the sequence or less than the minimum value in the sequence ,

[0111] When the above conditions are met, the rudder deflection angle corresponding to the maximum value or the minimum value in the sequence is taken as the actual control instruction

[0112] Conversely, starting from the starting value of the sequence , the expected aerodynamic moment coefficient is sequentially determined whether it is located between the two adjacent values in the sequence , and when the condition is met, the expected aerodynamic moment coefficient corresponding to the rudder deflection angle δ z is obtained by interpolation. When there are multiple rudder deflection angles that meet the conditions, the rudder deflection angle with the smallest absolute value of deviation from the rudder deflection angle at the previous moment is taken as the actual control instruction

[0113] S303: The actual control instruction of the yaw channel and the actual control instruction of the roll channel are obtained in the same way.

[0114] S4: Dynamic updating of the glider agent model and adaptive control adjustment.

[0115] The S4 includes the following steps:

[0116] S401: Using flight mechanics principles, the measured data during the flight of the glider is converted into the input and output of the agent network model, the data is accumulated, and a sliding mechanism is used to form a model dynamic updating sample set;

[0117] S402: Using the model dynamic updating sample set and the MAS incremental learning method, the glider agent network model is continuously updated and adjusted online, so that the agent network model is updated online and dynamically;

[0118] The S402 includes the following steps:

[0119] S40201: Assuming that the parameters of the original agent network model are θ old , the importance Ω of the network parameters is calculated using the gradient of the output to the input;

[0120] S40202: Calculate the network loss function Y:

[0121] Y=Y new +λΩ(θ-θ old ) 2 (21)

[0122] In equation (21):

[0123] Y newa loss function calculated by using a model to dynamically update a sample set;

[0124] lambda is a parameter for measuring importance;

[0125] theta is a proxy network model parameter of current generation optimization;

[0126] S40203: the optimization of formula (21) is realized by using the gradient descent method, that is, the optimization of the proxy network model parameter theta is realized, thereby realizing the dynamic update of the proxy network model.

[0127] S403: the aerodynamic network proxy model used in S3 is replaced by the dynamically updated model in S402 to obtain more accurate aircraft control instructions delta * .

[0128] Compared with the prior art, the beneficial effects of the present application are:

[0129] The present application adopts a neural network to construct a gliding aircraft proxy model, avoids modeling errors, and provides a good foundation for online dynamic update of the aircraft model; the online dynamic update of the aircraft proxy model is realized by using a sliding mechanism and an incremental learning method based on memory perception synapses, the online update of the aircraft model and the adaptive adjustment of the control system are realized, the adaptability of the aircraft control system to the environment and the task is enhanced, and the precision of the aircraft proxy model is effectively improved; the actual control instruction calculation and design based on the aircraft model proxy network and nonlinear model interpolation effectively improve the adaptability of the aircraft to interference and uncertainty, and realize online high-precision control of the aircraft. BRIEF DESCRIPTION OF DRAWINGS

[0130] Figure 1 is a flowchart of the present application. DETAILED DESCRIPTION

[0131] The technical solutions in the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0132] A gliding aircraft intelligent adaptive control method based on dynamic model update driving, the method comprising the following steps:

[0133] S1: offline gliding aircraft proxy model construction;

[0134] The S1 comprises collecting gliding aircraft simulation and flight test data to form a sample set, constructing an aircraft agent network model, and training the sample set to form an aircraft model basic agent network, and specifically comprises the following steps:

[0135] S101: Collecting gliding aircraft historical flight data, and constructing a three-channel aerodynamic moment model identification sample library for pitch, yaw and roll, respectively;

[0136] S102: Establishing a fully connected neural network, and using the sample library to train the three-channel aerodynamic moment model agent network of the gliding aircraft by using the Adam algorithm to form a gliding aircraft model basic agent network:

[0137] The network input is six-dimensional flight state, including: Mach number, attack angle, side slip angle and pitch, yaw and roll three-channel rudder deflection angle; and the output is one-dimensional aerodynamic moment coefficient;

[0138] Wherein:

[0139] The pitch channel agent network model is as follows:

[0140] C mz =net z (Ma,α,β,δ x ,δ y ,δ z ) (1)

[0141] The yaw channel agent network model is as follows:

[0142] C my =net y (Ma,α,β,δ x ,δ y ,δ z ) (2)

[0143] The roll channel agent network model is as follows:

[0144] C mx =net x (Ma,α,β,δ x ,δ y ,δ z ) (3)

[0145] In formula (1)-(3):

[0146] Ma is the Mach number;

[0147] α is the attack angle;

[0148] β is the side slip angle;

[0149] δ x is the roll channel rudder deflection angle;

[0150] δ y is the rudder deflection angle of the yaw channel;

[0151] δ z is the elevator deflection angle of the pitch channel;

[0152] C mx is the one-dimensional aerodynamic moment coefficient of the roll channel;

[0153] C my is the one-dimensional aerodynamic moment coefficient of the yaw channel;

[0154] C mz is the one-dimensional aerodynamic moment coefficient of the pitch channel.

[0155] S2: glider controller design;

[0156] The S2 comprises constructing a control-oriented model based on glider characteristics and designing a non-singular fast terminal sliding mode control law to obtain a required desired control moment, and specifically comprises the following steps:

[0157] S201: constructing a control-oriented glider model, and the glider dynamics model around the center of mass is:

[0158]

[0159] In formula (4):

[0160] Θ=[αβσ] T is an attitude angle vector, wherein: σ is the roll angle;

[0161] is the first-order derivative of the attitude angle vector;

[0162] Q is an attitude angle speed conversion matrix;

[0163] ω b =[ω x ω y ω z ] T is an attitude angle speed vector, wherein: ω x is the roll angle speed, ω y is the yaw angle speed, and ω z is the pitch angle speed;

[0164] is the first-order derivative of the attitude angle speed vector;

[0165] Δf F is a model uncertainty term;

[0166] I is the moment of inertia of the aircraft;

[0167] M is the control moment;

[0168] M d is the disturbance moment;

[0169] S202: define the system state variable as x = [a b s w x w y w z ] T Rewrite equation (4) into the form of MIMO system as follows:

[0170]

[0171] In equation (5):

[0172] is the first order derivative of the system state variable;

[0173] f(x) is the nonlinear term of the system state;

[0174] g(x) is the nonlinear term of the system input;

[0175] d is the system uncertainty term;

[0176] S203: perform feedback linearization on equation (5) to obtain:

[0177]

[0178] In equation (6):

[0179] is the second order derivative of the attitude angle vector;

[0180] F is the state matrix after feedback linearization;

[0181] E is the equivalent control matrix;

[0182] is the disturbance quantity;

[0183] S204: design the state feedback control law:

[0184] M = -E -1 F + Π (7)

[0185] In equation (7):

[0186] Π = [π1 π2 π3] T is the auxiliary control variable, where: π1 is the auxiliary control variable in the pitch channel, π2 is the auxiliary control variable in the yaw channel, and π3 is the auxiliary control variable in the roll channel;

[0187] Then, we can obtain:

[0188]

[0189] Introducing virtual control variable U = EΠ, we have:

[0190]

[0191] S205: The system is decoupled into three second-order systems, and the non-singular fast terminal sliding mode control law is designed for each channel to obtain the virtual control variable U;

[0192] The S205 includes the following steps:

[0193] S20501: Define the expected control output as Wherein: α d is the expected attack angle, is the first derivative of the expected attack angle;

[0194] The state error vector e α is:

[0195] e α = [e α1 , e α2 ] T = x d -x (10)

[0196] In equation (10):

[0197] e α1 is the attack angle error;

[0198] e α2 is the attack angle rate error;

[0199] S20502: Select the non-singular fast terminal sliding surface:

[0200]

[0201] In equation (11):

[0202] k1>0, k2>0, p, q, m and n are constants and are all odd numbers, satisfying

[0203] S20503: Obtain the first derivative of the sliding surface:

[0204]

[0205] In equation (12):

[0206] is the second derivative of the expected attack angle α d ;

[0207] u1 is the pitch channel control variable;

[0208] S20504: Divide the control quantity into equivalent control law on the sliding surface u eq and switching control u sw , i.e. the pitch channel control quantity u1 = u 1eq + u 1sw , let The following can be obtained:

[0209]

[0210] S20505: The idea of terminal attractor is adopted to design switching control u 1sw :

[0211]

[0212] In formula (14):

[0213] k3> 0, k4> 0, k and l are all constants and are all odd numbers, and satisfy

[0214] S20506: Substitute formula (14) into formula (12) and combine formula (13) to obtain:

[0215]

[0216] Then the pitch channel control quantity is:

[0217]

[0218] S20507: Similarly, the following can be obtained:

[0219] The yaw channel control quantity is:

[0220]

[0221] The roll channel control quantity is:

[0222]

[0223] S20508: The virtual control quantity U = [u1 u2 u3] T is obtained.

[0224] S206: The expected control moment is obtained by using formula (7):

[0225] M = -E -1 F + E -1 U (19).

[0226] S3: Actual control command acquisition;

[0227] The S3 comprises calculating corresponding aerodynamic moment coefficients according to the expected control moment obtained in the S2, and then calculating actual control instructions (rudder deflection angle) based on a glider model agent network and a nonlinear model interpolation method; specifically comprising the following steps:

[0228] S301: calculating expected aerodynamic moment coefficients of the pitch channel according to current state of the glider and expected control moment M z of the pitch channel can be obtained

[0229]

[0230] In formula (20):

[0231] q is dynamic pressure;

[0232] S is reference area;

[0233] L is reference length;

[0234] S302: calculating actual control instructions of the pitch channel;

[0235] The S302 comprises the following steps:

[0236] S30201: dividing the pitch rudder deflection angle into characteristic point intervals (such as [-20, -10, -5, 0, 5, 10, 20]°) according to the glider capability, inputting current Mach number, attack angle, sideslip angle and pitch rudder deflection angle sequence into the aerodynamic network agent model, i.e. formula (1), to obtain corresponding expected aerodynamic moment coefficient sequence of the pitch channel

[0237] S30202: considering that the aerodynamic model is highly nonlinear, multiple rudder deflection angles can achieve the same aerodynamic moment coefficient, in order to avoid sudden changes in the calculated rudder deflection angle and affect the control performance, a nonlinear interpolation method is used to solve the rudder deflection angle.

[0238] finding the maximum value and the minimum value of the sequence judging whether the expected aerodynamic moment coefficient is greater than the maximum value in the sequence or less than the minimum value in the sequence

[0239] When the above conditions are met, the rudder deflection angle corresponding to the maximum value or the minimum value in the sequence is taken as the actual control instruction

[0240] Otherwise, starting from the starting value of the sequence , the expected aerodynamic moment coefficient is sequentially judged whether it is located in the sequence ​When the condition is met between two adjacent values, the desired aerodynamic moment coefficient is obtained by interpolation The corresponding rudder deflection angle δ z When there are multiple rudder deflection angles that meet the condition, the rudder deflection angle with the smallest absolute value of deviation from the rudder deflection angle at the previous time is taken as the actual control command

[0241] S303: Similarly, the actual control command of the yaw channel and the actual control command of the roll channel are obtained.

[0242] S4: Dynamic updating of glider agent model and adaptive control adjustment.

[0243] The S4 includes collecting online flight data to form a model dynamic updating sample set, using a sliding mechanism and a memory-based perceptual synapse (MAS) incremental learning method to update the glider agent network model parameters, so that the agent model is dynamically updated online, and an accurate agent model is provided for S3 to obtain more accurate aircraft control commands. Specifically, the following steps are included:

[0244] S401: Using flight mechanics principles to convert the measured data in the glider flight process into agent network model inputs and outputs, accumulating the data, and using a sliding mechanism to form a model dynamic updating sample set;

[0245] S402: Using the model dynamic updating sample set and the MAS incremental learning method to continuously update and adjust the glider agent network model online, so that the agent network model is dynamically updated online, thereby more accurately representing the aircraft model. The MAS incremental learning method has low computational complexity and can meet the online adjustment time requirements;

[0246] The S402 includes the following steps:

[0247] S40201: Assume that the parameters of the original agent network model are θ old , and use the gradient (i.e. sensitivity) of the output to the input to calculate the importance of the network parameters Ω;

[0248] S40202: Calculate the network loss function Y:

[0249] Y=Y new +λΩ(θ-θ old ) 2 (21)

[0250] In equation (21):

[0251] Y new is the loss function calculated using the model dynamic updating sample set;

[0252] λ is a parameter that measures importance;

[0253] θ is the agent network model parameter of the current generation optimization;

[0254] S40203: the optimization of formula (21) is realized by using the gradient descent method, that is, the optimization of the agent network model parameter θ is realized, thereby realizing the dynamic update of the agent network model.

[0255] S403: the aerodynamic network agent model used in S3 is replaced by the dynamically updated model in S402, thereby providing an accurate agent model for S3 to obtain more accurate aircraft control instructions δ * .

[0256] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be regarded as exemplary and non-limiting, the scope of the present application is defined by the appended claims rather than the above description, and all changes falling within the meaning and scope of the equivalent conditions of the claims are intended to be included in the present application. Any reference signs in the claims should not be regarded as limiting the claims involved.

[0257] In addition, it should be understood that although the present specification is described in terms of embodiments, not every embodiment contains only one independent technical solution, and the description of the specification is only for the sake of clarity, and those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be combined appropriately to form other embodiments that those skilled in the art can understand.

Claims

1. A method for intelligent adaptive control of a gliding aircraft based on dynamic model updating driving, characterized in that: The method comprises the following steps: S1: constructing an offline glider agent model; S2: designing a glider controller; S3: obtaining actual control instructions; The S3 comprises the following steps: S301: Obtain the expected aerodynamic moment coefficient of the pitch channel according to the current state of the gliding aircraft and the expected control moment of the pitch channel , the expected aerodynamic moment coefficient of the pitch channel can be obtained : (20) In formula (20): Dynamic pressure; for reference area; Reference length; S302: obtaining actual control instructions of the pitch channel; The S302 comprises the following steps: S30201: According to the glider ability, the pitch control surface deflection angle is divided into characteristic point intervals, and the current Mach number , attack angle , sideslip angle and pitch control surface deflection angle sequence input aerodynamic agent network model, get the corresponding expected aerodynamic moment coefficient sequence of the pitch channel , the pitch channel agent network model is as follows: (1) In formula (1): roll channel rudder angle; for yaw channel rudder angle; Cp is the one-dimensional aerodynamic force coefficient for the roll channel; S30202: Finding a Sequence The maximum and minimum values ​​are used to determine the desired aerodynamic moment coefficient. Is it greater than the sequence? The maximum value in the sequence is less than the maximum value in the sequence. The minimum value in, When the above conditions are satisfied, the maximum or minimum value in the sequence corresponding to the rudder deflection angle is the actual control command ; Conversely, starting from the initial value of the sequence , the expected aerodynamic moment coefficient is sequentially determined whether it is located between two adjacent values in the sequence , and when the condition is met, the expected aerodynamic moment coefficient corresponding to the rudder deflection angle is obtained by interpolation; when there are multiple rudder deflection angles satisfying the condition, the rudder deflection angle with the minimum absolute value of deviation from the rudder deflection angle at the previous moment is taken as the actual control instruction . S303: obtaining actual control instructions of the yaw channel and actual control instructions of the roll channel in the same way; S4: dynamic updating of the glider agent model and adaptive control adjustment; The S4 comprises the following steps: S401: converting measured data in the flight process of the glider into inputs and outputs of the agent network model by using flight mechanics principles, accumulating the data, and forming a model dynamic updating sample set by using a sliding mechanism; S402: continuously updating and adjusting the glider agent network model online by using the model dynamic updating sample set and the MAS incremental learning method, so that the agent network model is updated online and dynamically; The S402 comprises the following steps: S40201: assuming that the parameters of the original agent network model are , the importance of the network parameters is calculated using the gradient of the output with respect to the input ; S40202: compute network loss function : (21) In formula (21): a loss function calculated using model dynamic updates to the sample set; a parameter to measure importance; Agent network model parameters optimized for the current generation; S40203: optimizing formula (21) by using a gradient descent method, i.e. updating the parameters of the agent network model , thereby realizing dynamic updating of the agent network model; S403: replace the pneumatic network agent model used in S3 with the dynamically updated model in S402 to obtain more accurate aircraft control instructions .

2. The intelligent adaptive control method for a glider based on dynamic model updating driving according to claim 1, characterized in that: The S1 comprises the following steps: S101: collecting historical flight data of the glider, and respectively constructing identification sample libraries of pitch, yaw and roll three-channel aerodynamic moment models of the glider; S102: establishing a fully connected neural network, training the three-channel aerodynamic moment model agent network of the glider by using the sample libraries, and forming a glider model basic agent network: The network input is six-dimensional flight state, and the output is one-dimensional aerodynamic moment coefficient; Wherein: The yaw channel agent network model is as follows: (2) The roll channel agent network model is as follows: (3) In formulas (2)-(3): Cp is the one-dimensional aerodynamic moment coefficient for the yaw channel; Cp is the one-dimensional aerodynamic moment coefficient for the pitch channel.

3. The intelligent adaptive control method for a glider based on dynamic model updating driving according to claim 2, characterized in that: The S2 comprises the following steps: S201: constructing a control-oriented glider model, and a glider dynamics model around the center of mass is as follows: (4) In formula (4): is the vector of attitude angles, where: is the angle of roll; is the first derivative of the attitude angle vector; is the attitude angular velocity conversion matrix; is the vector of attitude angular rates, where: is the roll angular rate, is the yaw angular rate, is the pitch angular rate; is the first derivative of the attitude angular velocity vector; is the model uncertainty term; moment of inertia for the aircraft; To control the torque; is the disturbance torque; S202: define system state variables as Rewrite equation (4) into MIMO system form as follows: (5) In formula (5): is the first derivative of the system state variable; is the system state nonlinear term; inputting a non-linear term for the system; is a system uncertainty term; S203: performing feedback linearization processing on formula (5), and the following can be obtained: (6) In formula (6): is the second derivative of the attitude angle vector; is the feedback linearized state matrix; is the equivalent control matrix; is the perturbation quantity; S204: designing a state feedback control law: (7) In formula (7): is an auxiliary control variable for the control variable, is an auxiliary control variable for the pitch channel, is an auxiliary control variable for the yaw channel, is an auxiliary control variable for the roll channel; Then the following can be obtained: (8) Introducing a virtual control variable It follows that (9) S205: The system is decoupled into three second-order systems, and a non-singular fast terminal sliding mode control law is designed for each channel to obtain a virtual control variable ; S20501 : define the desired control output as wherein: is the desired angle of attack, is the first derivative of the desired angle of attack; the state error vector is given by: (10) In formula (10): for the angle of attack error; for the angle of attack rate error; S20502: selecting a non-singular fast terminal sliding mode surface: (11) In formula (11): , , , , and are constants and are all odd numbers, satisfying , ; S20503: obtaining the first derivative of the sliding mode surface: (12) In formula (12): second derivative of the desired angle of attack of the second derivative; pitch channel control quantity; S20504: Dividing the control quantity into equivalent control law on the sliding surface and switching control , i.e. the pitch channel control quantity , let , we get: (13) S20505: Design the switching control by using the idea of terminal attractor : (14) In formula (14): , , and are constants and are all odd numbers, satisfying ; S20506: substituting formula (14) into formula (12) and connecting formula (13) to obtain: (15) Then the pitch channel control quantity is as follows: (16) S20507: in the same way, the following can be obtained: The yaw channel control quantity is as follows: (17) The roll channel control quantity is as follows: (18) S20508: obtain virtual control quantity ; S206: obtaining the expected control moment by using formula (7): (19)。

Images