Finite state feedback guidance control method based on autoregressive cerebellar neural network

Abstract

The embodiment of the invention provides a finite state feedback guidance control method based on an autoregressive cerebellar neural network, and belongs to the technical field of control. The finite state feedback guidance control method specifically comprises the steps that a dive section guidance control integrated model is established; estimating the line-of-sight angular rate and guidance loop composite disturbance, designing a self-adaptive time-varying sliding mode surface based on an estimated value, and designing a guidance instruction on the basis of the self-adaptive time-varying sliding mode surface; designing an adaptive law to compensate airflow angle loop composite disturbance, designing a virtual attitude angular rate instruction on the basis, and performing amplitude limiting on the virtual attitude angular rate instruction to obtain an attitude angular rate instruction; designing an autoregressive cerebellar neural network to estimate the composite disturbance of the angular rate loop, compensating the influence brought by input saturation through an anti-saturation auxiliary system, and designing a control law on the basis; and obtaining a finite state feedback guidance control integrated overall control scheme based on the autoregressive cerebellar neural network in the subduction section of the hypersonic aircraft. Through the scheme of the invention, the adaptability and robustness of guidance control are improved.

Description

Technical Field

[0001] The disclosed embodiments relate to the field of control technology, and in particular to a finite state feedback guidance control method based on an autoregressive cerebellar neural network. Background Art

[0002] At present, with the development of hypersonic related technologies, the application potential of hypersonic aircraft in different fields is becoming more and more obvious, and it has become a weapon to seize the strategic advantage in the air and space. Compared with traditional aircraft, the complex mission environment and special aerodynamic structure of hypersonic aircraft make its fast time-varying, strong coupling, nonlinear, and uncertain characteristics more prominent. On the other hand, the mission requirements of directional precision strike in the dive phase require hypersonic aircraft to meet many complex constraints. This puts higher requirements on the design of guidance and control systems.

[0003] At present, the guidance and control integration method has been applied to the dive mission scenario to a certain extent, but there are still some limitations. First, the long angle of attack saturation time may cause the aircraft to be unable to press down in time and miss the target under the limited airspace and maneuverability constraints; second, the time-varying constraints are rarely considered in the design of the attitude control loop control law, the initial state value range is conservative, and the robustness is insufficient.

[0004] It can be seen that there is an urgent need for a finite state feedback guidance control method based on autoregressive cerebellar neural network with strong adaptability and robustness. Summary of the invention

[0005] In view of this, an embodiment of the present disclosure provides a finite state feedback guidance control method based on an autoregressive cerebellar neural network, which at least partially solves the problem of poor adaptability and robustness in the prior art.

[0006] The present disclosure provides a finite state feedback guidance control method based on an autoregressive cerebellar neural network, comprising:

[0007] Step 1: Establish a dive phase guidance and control integrated model based on the dive phase mission requirements of hypersonic aircraft;

[0008] Step 2: Use a finite-time observer to estimate the sight line angular rate and the guidance loop composite disturbance, design an adaptive time-varying sliding surface based on the estimated value that takes into account the target position and terminal attack angle constraints, and design the guidance command based on this;

[0009] Step 3, design an adaptive law to compensate for the composite disturbance of the airflow angle loop, and on this basis design a virtual attitude angular rate command based on the asymmetric obstacle Lyapunov function, and limit it to obtain the attitude angular rate command;

[0010] Step 4: Design an autoregressive cerebellar neural network to estimate the composite disturbance of the angular rate loop, compensate for the impact of input saturation through the anti-saturation auxiliary system, and design the control law based on the asymmetric barrier Lyapunov function;

[0011] Step 5, according to the adaptive time-varying sliding surface, virtual guidance instructions, guidance instructions, virtual attitude angular rate instructions, attitude angular rate instructions and control law, an integrated overall control scheme of finite state feedback guidance and control based on autoregressive cerebellar neural network in the dive phase of hypersonic vehicle is obtained.

[0012] According to a specific implementation of the embodiment of the present disclosure, the expression of the integrated guidance and control model of the dive phase is:

[0013]

[0014] Among them, x 1 、x 2 , x 3 、x 4 is the system state variable, f 2 、f 3 、f 4 , g 2 , g 3 , g 4 is the known dynamics of the model, d 2 ,d 3 ,d 4 is the model disturbance, u is the rudder angle command, sat (·) is a saturation function.

[0015] According to a specific implementation of the embodiment of the present disclosure, step 2 specifically includes:

[0016] Design a finite-time observer, where the expression of the finite-time observer is

[0017]

[0018] Among them, v 0 、v 1 、v 2 is the auxiliary variable defined, z 0 、z 1 and z 2 They are x 2 ,d 2 and An estimated value of

[0019] The line-of-sight angular rate and the guidance loop composite disturbance in the integrated guidance and control model of the dive phase are estimated using a finite-time observer, and an adaptive time-varying sliding surface S is designed based on the estimated value.g =[S g1 ,S g2 ] T for:

[0020]

[0021] Among them, r is the relative distance between the projectile and the target, r 0 is the initial relative distance between the projectile and the target, x 1i is the line of sight angle error component, x 2i is the line of sight angular velocity component, 0<β g2i <π / 2、β g1i >0, r f is the parameter to be designed;

[0022] Design of virtual guidance instructions based on adaptive time-varying sliding surface As shown below:

[0023]

[0024] in, and is the virtual guidance command component, T go is the estimated value of the remaining waiting time, K g1 , ε g1 is the positive definite diagonal matrix to be designed; K gε =ε g2 exp(-ε g3 t)x 1 is the angle of attack saturation control term, v g1 , ε g2 , ε g3 is a positive constant, is a finite-time observer pair d 2 The observed value of The finite-time observer x 2 Observed value Then the nonlinear term The expression is

[0025]

[0026]

[0027] in, As auxiliary variables, let x 3c =[x 3c,1 ,x 3c,2 ] T , where x 3c,1 、x 3c,2 are the required angle of attack and the required roll angle, respectively, and are given by The inverse solution is:

[0028]

[0029] Among them, C Lc is the required lift coefficient, δ x , δ z are the roll and pitch rudder deflection angles, δ f is the folding angle, and are aerodynamic coefficients, A is the aerodynamic coefficient matrix, A 11 , A 12 , A 21 , A 22 is the corresponding weight;

[0030] make By finite time tracking differentiator we get:

[0031]

[0032] Among them, H 1 , β T1,i , β T2,i are constants to be designed. and Track the output of the differentiator for finite time;

[0033] After further limiting, the guidance command x is obtained 3d

[0034]

[0035] According to a specific implementation of the embodiment of the present disclosure, step 3 specifically includes:

[0036] Designing adaptive laws for

[0037]

[0038] Among them, ξ is an auxiliary variable, a 1 、a 2 is the normal number to be designed;

[0039] Define the asymmetric barrier Lyapunov function V g1 for

[0040]

[0041] in, is the adaptive law estimation error, ε Ai =q 1i e 3i / k bi +(1-q 1i )3i / k ai is an auxiliary variable, k bi , k ai are the upper and lower boundaries of the three-channel airflow angle tracking errors, e 3i is the three-channel airflow angle tracking error, and the auxiliary variable q 1 The expression of i is

[0042]

[0043] According to the asymmetric barrier Lyapunov function V g1 , design virtual attitude angular rate instruction x 4c =[x 4c,1 ,x 4c,2 ,x 4c,3 ] T as follows:

[0044]

[0045] Among them, x 4c,1 、x 4c,2 、x 4c,3 is the virtual attitude angular rate command component, K′ 2 is the positive definite diagonal matrix to be designed, s 3 , σ 3 is the tracking error and filtering error, e 3 is the airflow angle tracking error vector, auxiliary variable λ 3 =[λ 31 ,λ 32 ,λ 33 ] T The specific form is

[0046]

[0047] x 4c By finite-time tracking of the differentiator we obtain:

[0048]

[0049] Among them, H 2 , β T3,i , β T4,i are constants to be designed. Track the output of the differentiator for finite time;

[0050] After further limiting, the attitude angular rate command x is obtained 4d

[0051]

[0052] According to a specific implementation of the embodiment of the present disclosure, step 4 specifically includes:

[0053] Designing an autoregressive cerebellar neural network, wherein the autoregressive cerebellar neural network comprises an input layer, a recursive associative memory space, a receptive field space, a weight space and an output layer;

[0054] The recursive associative memory space includes n blocks, each block includes m neurons, and the input p of the jth neuron in the i-th block is ij And output h ij It can be expressed as:

[0055]

[0056] Among them, r ij is the recursive loop weight, h ij (tN) is the output of the previous N steps, v ij and z ij Respectively represent the mean and width of the Gaussian basis function;

[0057] Each component φ in the receptive field space i is the cumulative multiplication of the output in the i-th block in the previous layer;

[0058] The expression of the network weight matrix W corresponding to the weight space is

[0059]

[0060] x 4 =[ω x ,ω y ,ω z ] T As the input of the network, the output of the network is the composite disturbance d in the attitude angular rate loop 4 Estimated value of

[0061] Define the asymmetric barrier Lyapunov function V g2

[0062]

[0063] in, is the estimation error of the ideal value of the corresponding neural network parameter, ε ωi =q 2i e 4i / k di +(1-q 2i ) 4i / k ci is an auxiliary variable, e 4i is the three-channel attitude angular rate tracking error, k di , kci They are the upper and lower bounds of the three-channel attitude angular rate tracking error, and the auxiliary variable q 2i The expression is

[0064]

[0065] V g2 , design control law u

[0066]

[0067] Among them, K 3 ′, K h is the positive definite diagonal matrix to be designed, h s is the state of the anti-saturation auxiliary system, and the expression of the anti-saturation auxiliary system is

[0068]

[0069] Where Δ is the saturation error, σ 4 is the filtering error, K h , L h is the positive definite diagonal matrix to be designed, ε s is the positive constant to be designed, and the auxiliary variable λ 4 =[λ 41 ,λ 42 ,λ 43 ] T The specific form is

[0070]

[0071] According to a specific implementation of the embodiment of the present disclosure, the expression of the finite state feedback guidance control integrated overall control scheme is:

[0072]

[0073] The finite state feedback guidance and control scheme based on the autoregressive cerebellar neural network in the disclosed embodiment includes: step 1, establishing a dive section guidance and control integrated model according to the dive section mission requirements of the hypersonic vehicle; step 2, using a finite time observer to estimate the line of sight angular velocity and the guidance loop composite disturbance, designing an adaptive time-varying sliding surface that takes into account the target position and terminal attack angle constraints based on the estimated values, and designing guidance instructions on this basis; step 3, designing an adaptive law to compensate for the composite disturbance of the airflow angle loop, and on this basis designing a guidance instruction based on the asymmetric obstacle Lyapunov function. A virtual attitude angular rate instruction is generated, and the attitude angular rate instruction is obtained by limiting it; in step 4, an autoregressive cerebellar neural network is designed to estimate the composite disturbance of the angular rate loop, and the influence of input saturation is compensated by the anti-saturation auxiliary system. On this basis, a control law is designed based on the asymmetric obstacle Lyapunov function; in step 5, according to the adaptive time-varying sliding surface, virtual guidance instructions, guidance instructions, virtual attitude angular rate instructions, attitude angular rate instructions and control laws, an integrated control scheme of finite state feedback guidance control based on an autoregressive cerebellar neural network for the dive phase of a hypersonic vehicle is obtained.

[0074] The beneficial effects of the embodiments of the present disclosure are as follows: through the scheme of the present disclosure, a dive-stage finite state feedback guidance and control integrated scheme based on an autoregressive cerebellar neural network is designed for a hypersonic aircraft dive-stage strike mission in which the line-of-sight angular velocity is unmeasurable, taking into account the target position, strike angle, airflow angle, attitude angular velocity and input saturation constraints. The scheme can effectively improve the saturation of the aircraft's angle of attack, reduce the conservatism of the initial value range of the state, and has good robustness and adaptability, and is suitable for finite state feedback dive strike missions under multi-constraint conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0076] Figure 1 A schematic flow chart of a finite state feedback guidance control method based on an autoregressive cerebellar neural network provided in an embodiment of the present disclosure;

[0077] Figure 2 A structural block diagram corresponding to a finite state feedback guidance control method based on an autoregressive cerebellar neural network provided in an embodiment of the present disclosure;

[0078] Figure 3 A schematic diagram of the structure of an autoregressive cerebellar neural network provided in an embodiment of the present disclosure;

[0079] Figure 4 A schematic diagram of an aerodynamic shape of a hypersonic aircraft provided in an embodiment of the present disclosure;

[0080] Figure 5 A schematic diagram of a three-dimensional dive trajectory of an aircraft provided in an embodiment of the present disclosure;

[0081] Figure 6 A schematic diagram of an airflow angle variation curve provided in an embodiment of the present disclosure;

[0082] Figure 7 A schematic diagram of a posture angular rate change curve provided in an embodiment of the present disclosure;

[0083] Figure 8 A schematic diagram of a rudder angle variation curve provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0084] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0085] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0086] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein may be used to implement this device and / or practice this method.

[0087] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0088] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0089] The disclosed embodiment provides a finite state feedback guidance control method based on an autoregressive cerebellar neural network, which can be applied to the control process of a hypersonic aircraft.

[0090] See also Figure 1 , is a flow chart of a finite state feedback guidance control method based on an autoregressive cerebellar neural network provided by an embodiment of the present disclosure. Figure 1 and Figure 2 As shown, the method mainly comprises the following steps:

[0091] Step S1: According to the dive mission requirements of hypersonic vehicle, a dive guidance and control integrated model is established.

[0092]

[0093] Among them, x 1 =[θ L -θ Lf ,ψ L -ψ Lf ] T is the line of sight angle error; is the control input of the guidance loop, C L is the lift coefficient, σ is the roll angle; θ L , L are respectively the line of sight inclination angle and the line of sight deflection angle; θ Lf , Lf are the expected sight line inclination angle and sight line deflection angle respectively; and is the known dynamics of the model; is the total disturbance of the guidance loop; T SB is the transformation matrix of the track coordinate system and the guide line coordinate system; they are expressed as follows:

[0094]

[0095] Among them, S r , Q A represent the vehicle reference area and dynamic pressure respectively.

[0096] x 3 =[α,β,σ] T is the airflow angle, α and β are the angle of attack and sideslip respectively, d 3 is a composite disturbance, the model has known dynamics f 3 , g 3 The specific form is as follows:

[0097]

[0098] x 4 =[ω x ,ω y ,ω z ] T is the attitude angular velocity vector, u=[δ x ,δ y ,δ z ] is the rudder angle command, and the model nonlinear term f 4 The specific form is as follows:

[0099]

[0100] in, and The specific forms of are shown in formula (7) and formula (8):

[0101]

[0102] g 4 and d 4 The specific form is as follows:

[0103]

[0104] Where d is the external disturbance torque.

[0105] sat(·) is the saturation function, and its specific form is as follows:

[0106]

[0107] Step 2: A finite-time observer is used to estimate the line-of-sight angular rate and the composite disturbance of the guidance loop. Based on the estimated values, a new global adaptive time-varying sliding surface is designed which takes into account the target position and terminal attack angle constraints. Guidance instructions are designed on this basis.

[0108] The finite-time state observer is shown below:

[0109]

[0110] Among them, z 0 =[z 01,z 02 ] T ∈R 2 、z 1 =[z 11 ,z 12 ] T ∈R 2 and z 2 =[z 21 ,z 22 ] T ∈R 2 They are x 2 , d and The estimated value of v k =[v k1 ,v k2 ] T ∈R 2 , the components for k=0,1,2 are as follows:

[0111]

[0112] in, and k >0, k=1,2,3 are the parameters to be designed.

[0113] On this basis, an adaptive global time-varying sliding surface is designed to eliminate the approaching phase and make the guidance law globally robust. g =[S g1 ,S g2 ] T As shown in formula (14):

[0114]

[0115] Among them, r is the relative distance, 0<β g2i <π / 2、r f Let β be the parameter to be designed. g1i satisfy:

[0116]

[0117] Among them, x 1i0 、x 2i0 and are the line of sight angle error, line of sight angle rate and relative distance change rate at the initial moment respectively. Assuming that at the initial moment of the dive, there is x 1i0 ·x 2i0 <0, then β g1i >0, and there is S gi (r 0 )=0, thus eliminating the approaching stage and making the system state on the sliding surface at the initial moment.

[0118] When S gi =0, we have:

[0119]

[0120] Further:

[0121]

[0122] Solving equation (17) yields:

[0123]

[0124] From formula (18), we can see that when S gi = 0, x 1i It decreases monotonically with respect to r, and when r = r f Converges to 0 at . Therefore, the adaptive global time-varying sliding surface S g It enables the aircraft to meet the terminal position and attack angle constraints.

[0125] The virtual control law to be designed is defined as The required lift coefficient and the required bank angle C are Lc and σ c It can be obtained by inversely solving the following formula:

[0126]

[0127] Required angle of attack α c It can be solved by the following formula:

[0128]

[0129] Among them C Lc is the required lift coefficient, δ x , δ z are the roll and pitch rudder deflection angles, δ f is the folding angle, and Both are aerodynamic coefficients.

[0130] Let x 3c =[x 3c,1 ,x 3c,2 ] T =[α c ,σ c ] T , x 3c By finite time tracking differentiator we can obtain:

[0131]

[0132] Since this project adopts the BTT control scheme, the sideslip angle of the aircraft should always be 0. Limiting the angle of attack command and adding the sideslip angle command can obtain the airflow angle command x 3d As shown below:

[0133]

[0134] where x 3d,1 The amplitude is taken as T 31 , and solve for x 3d The corresponding lift coefficient C Ld ,definition The tracking error s 3 and filtering error σ 3 As shown below:

[0135]

[0136] For S g The derivative is:

[0137]

[0138] Let H g =[H g1 ,H g2 ] T , H gi The specific form is shown as follows:

[0139]

[0140] Define the asymmetric barrier Lyapunov function V g1 As shown below:

[0141]

[0142] For equation (26), design the virtual control law for:

[0143]

[0144] Among them, T go is the estimated value of the remaining waiting time; K g1 , ε g1 is the positive definite diagonal matrix to be designed; K gε =ε g2 exp(-ε g3 t)x 1 is the angle of attack saturation control term; v g1 , ε g2 , ε g3 is a positive constant; is a finite-time observer pair d 2The observed value of The finite-time observer x 2 Observed value but The specific form is as follows:

[0145]

[0146] Step 3: Design an adaptive law to compensate for the compound disturbance of the airflow angle loop, and on this basis design a virtual attitude angular rate command based on the asymmetric obstacle Lyapunov function.

[0147] Define the virtual control law to be designed as x 4c , x 4c By tracking the differentiator in finite time and limiting the instruction, we can obtain:

[0148]

[0149]

[0150] x 4d,i The amplitude is taken as T 4i . Define tracking error e 4 and filtering error σ 4 As shown below:

[0151] e 4 =x 4 -x 4d ,σ 4 =x 4d -x 4c (32)

[0152] Let e 3 =[e 31 ,e 32 ,e 33 ] T =x 3 -x 3d , for e 3 The derivative is:

[0153]

[0154] Define the asymmetric barrier Lyapunov function V g2 as follows:

[0155]

[0156] Among them, ε Ai =q 1i e 3i / k bi +(1-q 1i ) 3i / kai ;k bi , k ai are the upper and lower boundaries of the three-channel airflow angle tracking errors, respectively. The specific forms are as follows:

[0157] k a1 =x 3d,1 -α min ,k b1 =α max -x 3d,1 ,k a2 =k b2 =β b ,k a3 =k b3 =σ b (35)

[0158] α min , α max , β b and σ b are all positive numbers to be designed, q 1i The specific form is as follows:

[0159]

[0160] Design virtual attitude angular rate instruction x 4c for:

[0161]

[0162] Among them, K′ 2 =diag(K′ 21 ,K′ 22 ,K′ 23 ), the specific form is as follows:

[0163]

[0164] η 1i , K 2i is the positive constant to be designed. 3 =[λ 31 ,λ 32 ,λ 33 ] T , the specific form is as follows:

[0165]

[0166] is an adaptive law, and its specific form is as follows:

[0167]

[0168] Where ξ=[η 31|e 31 |,η 32 |e 32 |,η 33 |e 33 |] T , a 1 、a 2 is the normal number to be designed.

[0169] Step 4: Design an autoregressive cerebellar neural network to accurately estimate the compound disturbance of the angular rate loop, compensate for the impact of input saturation through the anti-saturation auxiliary system, and then design the control law, i.e., the rudder angle command, based on the asymmetric obstacle Lyapunov function.

[0170] The structure of the autoregressive cerebellar neural network is as follows Figure 3 As shown:

[0171] The detailed introduction of each layer of the network structure is as follows:

[0172] Input layer: where x = [x 1 ,x 2 ,…,x m ] is the input of the network.

[0173] Recursive associative memory space: This layer consists of n blocks, each block consists of m neurons, and the input p of the jth neuron in the i-th block is ij And output h ij It can be expressed as:

[0174]

[0175] Among them, r ij is the recursive loop weight, h ij (tN) is the output of the previous N steps, v ij and z ij They represent the mean and width of the Gaussian basis function respectively.

[0176] Receptive field space: Each component φ in this layer i is the cumulative multiplication of the output of the i-th block in the previous layer, that is, φ i The specific form is as follows:

[0177]

[0178] The column vector Φ is defined as follows:

[0179]

[0180] Define column vector Similarly, we can define column vectors and

[0181] Weight space: The network weight matrix W is defined as follows:

[0182]

[0183] Output layer: Neural network output y = [y 1 ,y 2 ,…,y m The specific form of ] is as follows:

[0184] y=W T Φ (45)

[0185] x 4 =[ω x ,ω y ,ω z ] T As the network input, the network output is the composite disturbance d in the attitude angular rate loop 4 Estimated value of According to the properties of the autoregressive cerebellar neural network, there exists W * and Φ * , so that the following holds:

[0186] d 4 =(W * ) T Φ * +Γ (46)

[0187] Where Γ is the approximation error of the neural network. Definition and The disturbance estimation error As shown below:

[0188]

[0189] in, and W * and Φ * The estimated value of Will Taylor expansion gives:

[0190]

[0191] in, The partial derivative matrix is ​​defined as follows:

[0192]

[0193] Definition Φ v , Φ z , Φ r and As shown below:

[0194]

[0195] τ Φ1 is the higher-order term in Taylor expansion, τ Φ =τ Φ1 +τ Φ2 , τ Φ2 The specific form is as follows:

[0196]

[0197] From formula (48) and The definition of can be obtained:

[0198]

[0199] Define saturation error Δ=sat(u)-u, then for e 4 The derivative is:

[0200]

[0201] The asymmetric barrier Lyapunov function is defined as follows:

[0202]

[0203] Among them, ε ωi =q 2i e 4i / k di +(1-q 2i ) 4i / k ci ;k di , k ci They are the upper and lower boundaries of the three-channel attitude angular rate tracking errors, and the specific forms are as follows:

[0204] k di =ω j,max -ω jd ,k ci =ω jd -ω j,min ,i=1,2,3 j=x,y,z (55)

[0205] ω j,min ,ω j,max is a positive constant to be defined, q 2i The specific form is as follows:

[0206]

[0207] The designed control law u is as follows:

[0208]

[0209] K h is the positive definite diagonal matrix to be designed, K′ 3 =diag(K′ 31 ,K′ 32 ,K′ 33 ), the specific form is as follows:

[0210]

[0211] η 2i , K 3i is the positive constant to be designed. s The status of the anti-saturation auxiliary system:

[0212]

[0213] Among them, K h , L h is the positive definite diagonal matrix to be designed, ε s Is a normal number.

[0214] Step 5: Combining the adaptive time-varying sliding surface, virtual guidance command, guidance command, virtual attitude angular rate command, attitude angular rate command and control law, the integrated overall scheme of finite state feedback guidance and control based on autoregressive cerebellar neural network in the dive phase of hypersonic vehicle is obtained as follows:

[0215]

[0216] In order to verify the effectiveness and robustness of the proposed guidance and control integration method, the present invention designs a representative simulation example and performs numerical simulation. Figure 4 shown.

[0217] Simulation example:

[0218] The parameters of the guidance and control integration method disclosed in the present invention are set as follows:

[0219] The finite-time observer parameters are set as: ξ 0 =3,ξ 1 =1.5,ξ 2 =1.1

[0220] Finite time tracking differentiator parameter setting: H 1 =H 2 =100;β T1 =β T2 =β T3 =β T4 =[2,2,2] T

[0221] The guidance loop parameters are set as: f =30; K 1 =diag(2,1.8);v 1 =1;ε 1 =diag(1×10 -4 ,1×10 -4 );β 11 =0.78+0.0028×(|x 11 (0)|-50); β 12 =0.625+0.003×(|x 12 (0)-30|); ε 2 =1×10 -4 ; ε 3 =0.2.

[0222] The attitude control loop parameters are set as: K 2 =diag(30,30,30);b 31 =b 32 =3°; b 33 =60°; T 31 =17°; K 3 =diag(30,20,30);b 41 =25°; b 42 =40°; b 43 =35 ° ; T 41 =30°; T 42 =T 43 =25°; K h =diag(0.25,0.25,0.25); k h =diag(2,2,2);γ h =0.5; ε = 1 × 10 -5 ;h 01 =h 02 =h 03 =1×10 -4 .

[0223] The external disturbance torque is set as follows:

[0224]

[0225] The initial state of the aircraft, target position and expected impact angle, as well as the airflow angle and attitude angular rate constraints are set as follows:

[0226] Table 1

[0227]

[0228]

[0229] Table 2

[0230]

[0231] The single simulation results are shown in Table 2 and Figures 5 to 8 As shown in Table 2 f 、V f , θ, ψ represent the strike time, terminal falling velocity, terminal velocity inclination angle and velocity deviation angle, respectively. Figure 5 The three-dimensional dive trajectory of the aircraft is shown. It can be seen from the figure that the method disclosed in the present invention can enable the hypersonic aircraft to accurately hit the target. Figure 6 and Figure 7 are the change curves of airflow angle and attitude angle rate respectively. The dotted line in the figure is the state boundary, and the solid line is the state curve. Figure 6 It can be seen that the method disclosed in the present invention has a shorter angle of attack saturation time, a higher trajectory control margin, and both the angle of attack and sideslip angle constraints can be satisfied. Figure 7 It can be seen that the three-axis attitude angular rate of the method disclosed in the present invention is always within the state boundary, which shows that the method disclosed in the present invention can meet the attitude angular rate constraint in the presence of external interference and input saturation. Figure 8 It is a rudder angle curve. It can be seen from the figure that the rudder angle of the method disclosed in the present invention changes more smoothly and the saturation time is shorter.

[0232] To further verify the task adaptability of the method disclosed in the present invention, simulation tests were carried out under different strike angle constraints. The simulation results are shown in Table 3. f =180° and ψ f =240°, ψ(0)=130° and σ(0)=50°, and the other parameters are consistent with the single simulation task. As shown in Table 3, the method disclosed in the present invention has good task adaptability.

[0233] Table 3

[0234]

[0235]

[0236] It should be understood that various parts of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.

[0237] The above is only a specific implementation of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present disclosure should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A finite state feedback guidance control method based on autoregressive cerebellar neural network, characterized in that: include: Step 1: Establish a dive phase guidance and control integrated model based on the dive phase mission requirements of hypersonic aircraft; Step 2: Use a finite-time observer to estimate the sight line angular rate and the guidance loop composite disturbance, design an adaptive time-varying sliding surface based on the estimated value that takes into account the target position and terminal attack angle constraints, and design the guidance command based on this; Step 3, design an adaptive law to compensate for the composite disturbance of the airflow angle loop, and on this basis design a virtual attitude angular rate command based on the asymmetric obstacle Lyapunov function, and limit it to obtain the attitude angular rate command; Step 4: Design an autoregressive cerebellar neural network to estimate the composite disturbance of the angular rate loop, compensate for the impact of input saturation through the anti-saturation auxiliary system, and design the control law based on the asymmetric barrier Lyapunov function; Step 5, according to the adaptive time-varying sliding surface, virtual guidance instructions, guidance instructions, virtual attitude angular rate instructions, attitude angular rate instructions and control law, an integrated overall control scheme of finite state feedback guidance and control based on autoregressive cerebellar neural network in the dive phase of hypersonic vehicle is obtained.

2. The method according to claim 1, characterized in that , the expression of the integrated guidance and control model of the dive phase is: Among them, x1, x2, x3 and x4 are system state variables, f2, f3, f4, g2, g3, g4 are known model dynamics, d2, d3, d4 are model disturbances, and u is the rudder angle command. sat (·) is a saturation function.

3. The method according to claim 2, characterized in that , the step 2 specifically includes: Design a finite-time observer, where the expression of the finite-time observer is Among them, v0, v1, v2 are defined auxiliary variables, z0, z1 and z2 are x2, d2 and An estimated value of The line-of-sight angular rate and the guidance loop composite disturbance in the integrated guidance and control model of the dive phase are estimated using a finite-time observer, and an adaptive time-varying sliding surface S is designed based on the estimated value. g =[S g1 ,S g2 ] T for: Among them, r is the relative distance between the projectile and the target, r0 is the initial relative distance between the projectile and the target, and x 1i is the line of sight angle error component, x 2i is the line of sight angular velocity component, 0<β g2i <π / 2, β g1i >0, r f is the parameter to be designed; Design of virtual guidance instructions based on adaptive time-varying sliding surface As shown below: in, and is the virtual guidance command component, T go is the estimated value of the remaining waiting time, K g1 , ε g1 is the positive definite diagonal matrix to be designed; K gε =ε g2 exp(-ε g3 t)x1 is the angle of attack saturation control term, v g1 , ε g2 , ε g3 is a normal number, is the observation value of d2 by the finite-time observer, let The observation value of x2 by the finite-time observer Then the nonlinear term The expression is in, As auxiliary variables, let x 3c =[x 3c,1 ,x 3c,2 ] T , where x 3c,1 、x 3c,2 are the required angle of attack and the required roll angle, respectively, and are given by The inverse solution is obtained, the specific form is: Among them, C Lc is the required lift coefficient, δ x , δ z are the roll and pitch rudder deflection angles, δ f is the folding angle, and are aerodynamic coefficients, A is the aerodynamic coefficient matrix, A 11 , A 12 , A 21 , A 22 is the corresponding weight; make By finite time tracking differentiator we get: Among them, H1, β T1,i , β T2,i are constants to be designed, θ 1,i and θ 2i Track the output of the differentiator for finite time; After further limiting, the guidance command x is obtained 3d x 3d =[sat(θ 1,1 ),0,θ 1,2 ] T 。 4. The method according to claim 3, characterized in that , the step 3 specifically includes: Designing adaptive laws for Among them, ξ is an auxiliary variable, a1 and a2 are normal constants to be designed; Define the asymmetric barrier Lyapunov function V g1 for in, is the adaptive law estimation error, ε Ai =q 1i e 3i / k bi +(1-q 1i ) 3i / k ai is an auxiliary variable, k bi , k ai are the upper and lower boundaries of the three-channel airflow angle tracking errors, e 3i is the three-channel airflow angle tracking error, and the auxiliary variable q 1i The expression is According to the asymmetric barrier Lyapunov function V g1 , design virtual attitude angular rate instruction x 4c =[x 4c,1 ,x 4c,2 ,x 4c,3 ] T as follows: Among them, x 4c,1 、x 4c,2 、x 4c,3 is the virtual attitude angular rate command component, K′2 is the positive definite diagonal matrix to be designed, s 3 , σ 3 is the tracking error and filtering error, e3 is the airflow angle tracking error vector, and the auxiliary variable λ3=[λ 31 ,λ 32 ,λ 33 ] T The specific form is x 4c By finite-time tracking of the differentiator we obtain: Among them, H2, β T3,i , β T4,i are constants to be designed, θ 3i ,θ 4i Track the output of the differentiator for finite time; After further limiting, the attitude angular rate command x is obtained 4d x 4d =[sat(θ 3,1 ),sat(θ 3,2 ),sat(θ 3,3 )] T 。 5. The method according to claim 4, characterized in that , the step 4 specifically includes: Designing an autoregressive cerebellar neural network, wherein the autoregressive cerebellar neural network comprises an input layer, a recursive associative memory space, a receptive field space, a weight space and an output layer; The recursive associative memory space includes n blocks, each block includes m neurons, and the input p of the jth neuron in the i-th block is ij And output h ij It can be expressed as: Among them, r ij is the recursive loop weight, h ij (tN) is the output of the previous N steps, v ij and z ij Respectively represent the mean and width of the Gaussian basis function; Each component φ in the receptive field space i is the cumulative multiplication of the output in the i-th block in the previous layer; The expression of the network weight matrix W corresponding to the weight space is With x4=[ω x ,ω y ,ω z ] T As the input of the network, the output of the network is the estimated value of the composite disturbance d4 in the attitude angular rate loop Define the asymmetric barrier Lyapunov function V g2 in, is the estimation error of the ideal value of the corresponding neural network parameter, ε ωi =q 2i e 4i / k di +(1-q 2i ) 4i / k ci is an auxiliary variable, e 4i is the three-channel attitude angular rate tracking error, k di , k ci They are the upper and lower bounds of the three-channel attitude angular rate tracking error, and the auxiliary variable q 2i The expression is V g2 , design control law u Among them, K′3, K h is the positive definite diagonal matrix to be designed, h s is the state of the anti-saturation auxiliary system, and the expression of the anti-saturation auxiliary system is Among them, Δ is the saturation error, σ4 is the filtering error, K h , L h is the positive definite diagonal matrix to be designed, ε s is a positive constant to be designed, and the auxiliary variable λ4=[λ 41 ,λ 42 ,λ 43 ] T The specific form is 6. The method according to claim 5, characterized in that , the expression of the integrated overall control scheme of finite state feedback guidance control is:

Images