RADP-based hypersonic interception missile adaptive mid-guidance method and system

Abstract

The invention provides an RADP-based hypersonic-velocity interception missile adaptive mid-guidance method and system, and belongs to the technical field of hypersonic-velocity interception missiles. Firstly, calculation and judgment are carried out according to obtained motion information of an interception bomb body and a target, a lateral guidance instruction is solved through an RADP method after triggering conditions are met, and the problems of high strike precision and large calculation amount caused by uncertain external interference are solved through the method in combination with a robust self-adaptive dynamic programming method. According to simulation verification, the scheme can obviously improve the resolving efficiency, and the method has a good guidance effect.

Description

Technical Field

[0001] The invention belongs to the technical field of missile interception, and in particular relates to a RADP-based adaptive mid-course guidance method and system for a hypersonic interceptor missile. Background Art

[0002] Hypersonic interceptor missiles have the characteristics of high speed, long range and strong maneuverability. They have great civilian value and have become the focus of research.

[0003] The flight state and airflow disturbance of a hypersonic interceptor during hypersonic flight will cause its aerodynamic parameters to change, resulting in parameter uncertainty; secondly, during large maneuvering flight, some unmodeled dynamics may be stimulated, and the errors of the rudder and sensor themselves will also lead to unmodeled uncertainty; finally, the faster the interceptor is, the faster it responds to disturbances and is more susceptible to atmospheric interference. All of the above factors affect flight performance, and the design of its robust closed-loop control system is challenging. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a RADP-based adaptive mid-course guidance method and system for a hypersonic interceptor missile, so as to solve the problem in the prior art that the uncertainty of parameters and unmodeled uncertainty in the flight process of a hypersonic interceptor missile will affect the flight performance.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP, comprising the following steps:

[0007] S1, through the event trigger device to determine whether the interceptor missile's trajectory deviation is greater than the set threshold, if so, execute S2, otherwise continue to monitor the actual trajectory deviation;

[0008] S2, obtaining the expected trajectory deviation angle according to the positions of the interceptor missile and the target;

[0009] S3, constructing an infinite time domain value function of the trajectory deviation angle tracking error, wherein the infinite time domain value function is composed of the sideslip angle guidance law, the upper limit of the disturbance and the tracking error of the interceptor missile system; constructing an optimal value function of the infinite time domain value function, expressing the optimal value function through the Hamilton-Jacobi equation, and the optimization goal of the optimal value function is that the tracking error converges to zero;

[0010] S4, updating the weight of the evaluation network structure by the gradient descent method, solving the Hamilton-Jacobi equation by the evaluation network structure, obtaining the optimal index function, and then obtaining the sideslip angle controller, wherein the sideslip angle controller outputs a control instruction so that the actual trajectory deviation angle is close to the expected trajectory deviation angle.

[0011] A further improvement of the present invention is:

[0012] Preferably, in S2, the calculation formula of the expected trajectory deviation angle is:

[0013]

[0014] Among them, V G is the desired velocity converted to the velocity in the original emission system.

[0015] Preferably, in S3, the infinite time domain value function is:

[0016]

[0017] Among them, e σ (t) is the tracking error of the trajectory angle at time t, α is a positive number, Δ m (τ) is a known bounded function, e σ is the tracking error of the trajectory angle, τ is the time constant, β B is the component of the roll angle to the lateral plane, and both R and Q are positive definite matrices.

[0018] Preferably, in S3, the formula of the sideslip angle guidance law is:

[0019]

[0020] Where m is mass, V is velocity, P is thrust, J * (e σ ) is the optimal value function of the trajectory deviation angle.

[0021] Preferably, in S3, the tracking error of the interceptor missile system is:

[0022]

[0023] Among them, m represents mass, V represents velocity, P represents thrust, β B is the sideslip angle, Z B The component of the lateral force on the lateral plane, Δ is the uncertain disturbance, is the derivative of the ballistic angle.

[0024] Preferably, in S3, the optimal value function is:

[0025]

[0026] Among them, β B is the sideslip angle.

[0027] Preferably, in S4, the update law for evaluating the network structure weight is:

[0028]

[0029] Among them, λ is the learning rate of the neural network, Ψ J (e σ ) is the activation function, e J is the Hamiltonian residual.

[0030] Preferably, the Hamilton-Jacobi equation is solved by evaluating the network structure as:

[0031]

[0032] The optimal indicator function is:

[0033]

[0034] Among them, Θ J is the ideal weight.

[0035] Preferably, the controller of the sideslip angle is verified by stability analysis.

[0036] A RADP-based adaptive mid-range guidance system for hypersonic interceptor missiles, comprising:

[0037] The trigger unit determines whether the estimated deviation of the actual ballistic deviation angle of the interceptor missile is greater than the set threshold through the event trigger device. If so, the expected unit is executed, otherwise the actual ballistic deviation angle is continuously monitored;

[0038] The expectation unit obtains the expected trajectory deviation angle according to the positions of the interceptor missile and the target;

[0039] A function construction unit is used to construct an infinite time domain value function of the trajectory deviation angle tracking error, wherein the infinite time domain value function is composed of a sideslip angle guidance law, a disturbance upper limit value, and a tracking error of the interceptor missile system; an optimal value function of the infinite time domain value function is constructed, and the optimal value function is expressed by the Hamilton-Jacobi equation, wherein the optimization goal of the optimal value function is that the tracking error converges to zero;

[0040] The solving unit updates the weight of the evaluation network structure through the gradient descent method, solves the Hamilton-Jacobi equation through the evaluation network structure, obtains the optimal index function, and then obtains the sideslip angle controller, and the sideslip angle controller outputs a control instruction so that the actual trajectory deviation angle is close to the expected trajectory deviation angle.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] The invention discloses a method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP. The method first establishes a mathematical model of the interceptor missile under a ground ballistic system. Calculation and judgment are performed based on the acquired motion information of the missile body and the target. Secondly, the missile calculates the expected direction according to its own and the target position and speed information under the premise of satisfying the constraints of the kinematic model and the no-fly zone constraints. The event trigger device is designed by the deviation value from the current direction. After the trigger condition is met, the lateral guidance instruction is solved by the RADP method. Then, the control instruction of the interceptor missile is solved and evaluated by the robust adaptive dynamic programming method (RADP). The evaluation network adopts a neural network and is updated by the gradient descent method. Then, the guidance method is verified by stability analysis. Finally, the method is verified by MATLAB simulation software to judge the effectiveness of the method. According to the simulation verification, it is shown that the scheme can significantly improve the solution efficiency and has a good guidance effect. The invention obtains the adaptive mid-course guidance method of a hypersonic interceptor missile by establishing a model for simulation verification. The method combines the robust adaptive dynamic programming method to solve the problems of strike accuracy and high calculation caused by uncertain external interference, and improves the accuracy and efficiency of guidance. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of the present invention;

[0044] Figure 2 It is a schematic diagram of the system structure of the present invention;

[0045] Figure 3 This is a diagram showing the trajectory deviation tracking effect of the simulation of the present invention;

[0046] Figure 4 A diagram of the system tracking error simulated by the present invention;

[0047] Figure 5 It is a timing diagram of the simulation event triggering of the present invention;

[0048] Figure 6 This is an interception simulation diagram including terminal guidance of the present invention. DETAILED DESCRIPTION

[0049] In the following, the terms "first", "second", "third", and "fourth" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, a feature defined as "first", "second", "third", and "fourth" may explicitly or implicitly include one or more of the features.

[0050] The co-shooting method provided in the embodiment of the present application can be applied to terminal devices such as mobile phones, tablet computers, wearable devices, vehicle-mounted devices, augmented reality (AR) / virtual reality (VR) devices, laptop computers, ultra-mobile personal computers (UMPC), netbooks, personal digital assistants (PDA), etc. The embodiment of the present application does not impose any restrictions on the specific type of the terminal device.

[0051] It should be noted that the terms "first", "second", etc. in the specification and drawings of the present invention are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0052] The first aspect of the present invention discloses a method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP, the method comprising the following steps:

[0053] Step 1: Establish a mathematical model for interceptor missiles

[0054] The derivation results of the missile's center of mass motion equations in the horizontal plane can be expressed as follows:

[0055]

[0056] The above equations have 6 equations, including unknowns V(t), σ(t), x(t), m(t), β(t), P(t), etc. Due to the "instantaneous balance" assumption, the equations are closed. The main requirement for controlling and solving the model is that the thrust P can be controlled by controlling the inlet flow and the sideslip angle β to obtain the expected trajectory. m represents mass, V represents speed, P is thrust, X, Y B ,Z B They represent drag, lift, and lateral force components, σ ​​is the trajectory deflection angle, g is the gravitational acceleration, α B ,β B is the lateral attack angle component and the roll angle component, Δ is the uncertain interference, m c is the fuel mass consumption rate.

[0057] Step 2: Design of event trigger device

[0058] As an optimization control method, the event-triggered control method aims to reduce the communication bandwidth and save the computational load of the processor. Compared with the traditional periodic adoption method, event-triggered control provides a mechanism to determine the sampling period without affecting the expected performance of the system. In the event-triggered control scheme, a monotonically increasing time series must first be defined as the sampling moment or trigger moment, and the controller only updates the control strategy at the trigger moment. For the controlled system, its state variable x(t) will be passed to the event generator. The function of the event generator is to determine whether the set event trigger condition is met. If it is met, the current moment is the trigger moment, and the current state is the trigger state. The state variables include velocity V, trajectory deflection angle sigma, and the coordinate values ​​of xz in the ground coordinate system.

[0059] The system state sampling and its corresponding error system state sampling are recorded as:

[0060]

[0061] The correction event triggering device can be expressed based on the correction time. The time of the k+1th trajectory correction can be expressed as:

[0062]

[0063] in, is the estimated deviation of the trajectory angle, and ε is the threshold of the shooting direction deviation. If the value is higher than this, trajectory correction must be performed.

[0064] The correction method decision device mainly relies on the line of sight angle deviation judgment. It is expressed as follows using the interval inequality:

[0065]

[0066] in is the estimated terminal deviation, f NN is the deviation calculation function, M is the motion information of the interceptor missile, and T is the motion information of the target.

[0067] Step 3: Calculation of desired trajectory angle

[0068] In the same coordinate system, the coordinates of the interceptor missile's current position A are defined as The longitude and latitude coordinates of the target point T are Then the angle from the current point A to the target point T is AdIt can be calculated that the firing angle Ad is the angle between the line from the current point A to the target point T and the true north direction. It is easy to know that when the interceptor missile's velocity direction coincides with the current firing direction, the interceptor missile can accurately strike the target. Therefore, this velocity direction is defined as the expected velocity direction, and a North Celestial East coordinate system is established at point A to represent this velocity, as shown below:

[0069]

[0070] The conversion process of converting this velocity to the original emission system is as follows:

[0071]

[0072] In the formula, R ET-NRE The transformation matrix from the North Celestial East coordinate system to the geocentric rectangular coordinate system is expressed as follows:

[0073]

[0074] R G-ET The transformation matrix from the geocentric rectangular coordinate system to the launch coordinate system is expressed as follows:

[0075]

[0076] Therefore, the desired trajectory angle σ d It can be expressed as follows:

[0077]

[0078] Combining formula (1), it can be seen that by controlling the trajectory angle, the speed direction of the interceptor missile can be controlled.

[0079] Step 4: Design of control instructions based on RADP

[0080] (1) Basic framework of RADP

[0081] To describe the basic framework of RADP, consider the following continuous-time interceptor missile control system dynamics:

[0082] x(t)=F(x(t),u(t)),x(t0)=x0 (10)

[0083] in, is the control input. Continuously differentiable, and let (0, 0) be an equilibrium point of the system. The above system can be further considered as an affine nonlinear system of the following form:

[0084]

[0085] in, is the internal dynamic matrix, is the input gain matrix, Δ(x) is the unknown disturbance, assuming that Δ(x) is about the known function Δ M (x), that is, ||Δ(x)||≤Δ M (x).

[0086] The performance indicator function is designed as follows:

[0087]

[0088] where U(x(t),u(t)) is a positive definite utility function and α is a positive number.

[0089] Definition: If u(x) is continuous on the set Ω, u(0) = 0, u(t) can keep the system stable and are all finite, then the control u(t) is called admissible control.

[0090] In order to solve the optimal control strategy of the system and find the optimal performance index function, that is, the control strategy that can make the performance index reach the optimal level, according to dynamic programming theory, the performance index satisfies the following Hamilton-Jacobby-Bellman equation (HJB).

[0091]

[0092] For convenience use Indicates that the following Hamiltonian function is defined:

[0093]

[0094] Represents the partial derivative of the performance indicator function with respect to x.

[0095] The optimal performance index function can be expressed as:

[0096]

[0097] Since J is continuously differentiable, the minimum value of the above formula exists, so we have:

[0098]

[0099] Thus, the optimal control strategy of the system is obtained. To facilitate the subsequent discussion, the quadratic utility function is selected for analysis, that is, U(x(t),u(t))=Q(x)+u T Ru, where Q(x) is a semi-positive definite penalty function in quadratic form and R is a positive definite symmetric matrix.

[0100] At this time, the optimal control strategy can be expressed as follows:

[0101]

[0102] At this time, the optimal control rate expression only The general strategy iteration steps are as follows. However, the J expression is often difficult to calculate and may have the problem of dimensionality explosion. Therefore, it can be approximated by an evaluation network.

[0103] According to the generalized approximation properties of neural networks, the specific form of the evaluation network is often designed as follows:

[0104]

[0105] in, is the ideal weight vector, is the number of hidden layer nodes, represents the activation function, ε(x) is the evaluation network approximation error, and we can get:

[0106]

[0107] in, is the gradient of the activation function, is the approximation error gradient. Based on the above formula, the Lyapunov equation can be transformed into the following form:

[0108]

[0109] Given the ideal weight vector W c Unknown, so the current weight output of the evaluation network is introduced As an estimate of the ideal weight, the optimal performance indicator estimate is:

[0110]

[0111] In formula (21), select an appropriate σ(x), when x≠0, and Always holds true. Then the gradient function of (21) is:

[0112]

[0113] The following approximate Hamiltonian function is derived:

[0114]

[0115] The standard fastest gradient descent method is used to train the evaluation network and minimize the objective function That is, the weight update law of the evaluation network is:

[0116]

[0117] Among them, l cTo evaluate the learning rate of the network. At this time, the Hamiltonian function is:

[0118]

[0119] in, is the residual error due to the neural network approximation.

[0120] definition And there exists a positive integer ∑ m , such that ||∑||≤∑ m The weight estimation error of the evaluation network is defined as have Then the state equation of the weight estimation error system is:

[0121]

[0122] To ensure Can converge to W c , the continuous incentive conditions must be met, that is, to ensure

[0123]

[0124] The estimated control strategy is:

[0125]

[0126] By inputting the estimated control rate into the dynamic system to obtain the corresponding system output, and comparing it with the theoretical output under the optimal control mode, the degree of deviation between the estimated weight and the ideal weight can be determined. The commonly used comparison index is the value of the Hamiltom function. When the system control signal is the optimal control rate, At this time, the equation is the famous HJB equation. Under the estimated control rate, Through Perform iterative updates to make e c When it approaches 0, the optimal control rate design can be completed.

[0127] (2) Design and solve the control law for the interceptor trajectory deviation angle based on RADP;

[0128] Combining formula (1), it can be seen that by controlling the sideslip angle β, the trajectory angle σ can be controlled, and then the three velocity directions of the interceptor missile can be controlled. The main task of the lateral guidance is to eliminate the trajectory angle error by designing the command of the sideslip angle β and guide the interceptor missile to the target point. The design process of the sideslip angle guidance law is as follows:

[0129] Define the tracking error of the trajectory deviation as e σ =σ-σ d . Then, the expected rate of change of tracking error is:

[0130]

[0131] The infinite time domain value function of definition (28) is:

[0132]

[0133] Combining formula (1), we can see that the goal is to find the optimal sideslip angle guidance law β B (t), so that the tracking error (28) of the interceptor system converges to the neighborhood of zero, and the performance index reaches a minimum, that is,

[0134]

[0135] Therefore, the following Hamiltonian function is defined:

[0136]

[0137] Therefore, according to the first-order necessary condition, the optimal tracking error control law, that is, the optimal sideslip angle guidance law, can be expressed as:

[0138]

[0139] At this time, substituting equation (32) into equation (31), the coupled Hamilton-Jacobi equation is expressed as:

[0140]

[0141] Furthermore, the event trigger error is defined as:

[0142]

[0143] To avoid confusion, the optimal tracking error control input under event triggering is recorded as v, which can be expressed as:

[0144]

[0145] Since the most valuable function in equation (33) is unknown, it is difficult to obtain an analytical solution to the equation; therefore, using the RADP theory and the universal approximation properties of neural networks, an evaluation network structure is designed to approximately solve the coupled equation (33). First, the optimal value function can be expressed as:

[0146]

[0147] in, is the ideal weight, is the activation function, N J is the number of neurons in the hidden layer, ε J (e σ ) is the reconstruction error. Formula (36) is about eσ The gradient of can be expressed as

[0148]

[0149] Since Θ J is unknown, so the approximate output of the evaluation network can be expressed as:

[0150]

[0151] Similarly, there are

[0152]

[0153] Since the input of the system is event-triggered, that is, the control law is updated non-periodically, the event-triggered approximate optimal error control law can be rewritten as:

[0154]

[0155] According to formula (39), the estimation of formula (40) can be expressed as

[0156]

[0157] At this time, the approximate Hamiltonian function can be expressed as

[0158]

[0159] Use the gradient descent method to update the evaluation network structure. Define the following function about the Hamiltonian residual:

[0160]

[0161] Then the update law for evaluating the network structure weight can be expressed as

[0162]

[0163] Among them, λ is the learning rate of the neural network.

[0164] Step 5: Stability Analysis

[0165] Before giving a proof of stability, the following assumptions need to be made:

[0166] Assumption 1: The control law satisfies the Lipschitz condition, that is,

[0167]

[0168] Among them, C e is a constant greater than zero.

[0169] Assumption 2: and The norm of is bounded, that is,

[0170]

[0171] Among them, Θ JM ,Ψ JM ,ε JM are all constants greater than zero.

[0172] Proof: The Lyapunov candidate function is selected as follows

[0173] V(t)=V1(t)+V2(t) (46)

[0174] in, Case 1: When the event is not triggered, that is, t∈[t k ,t k+1 ], k∈N, we can find the time derivative of (46),

[0175]

[0176] From (33), we can see that:

[0177]

[0178] Substituting (48) into (47) yields:

[0179]

[0180] remember (49) further becomes

[0181]

[0182] According to (32), (35), (41) and Assumption 1, we have:

[0183]

[0184] Substituting (51) into (50) yields:

[0185]

[0186]

[0187] Therefore, if the trigger condition is met and z falls in the tight set

[0188]

[0189] at this time, Established.

[0190] Case 2: When the event is triggered, t = t k+1, k∈N. The differential form of the Lyapunov candidate function (46) can be expressed as:

[0191]

[0192] According to situation 1, there are Notice Knowable It can be further seen that ΔV1(t)≤0. Note that Furthermore, we can know

[0193]

[0194] Among them, κ is a Therefore, in the case of event triggering, ΔV(t)≤0 holds.

[0195] Simulation Example

[0196] Based on the above method, the hypersonic interception guidance is simulated, and the simulation results are as follows: Figure 3-Figure 6 As shown in the figure, it can be seen that the RADP method can track the change of the expected trajectory angle faster and better. The control law is updated only when the event trigger condition is violated, which greatly saves the calculation and communication volume. The system tracking error does not exceed 0.2 after 100s and finally converges to 0, and the tracking control effect is good.

[0197] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A RADP-based adaptive mid-course guidance method for a hypersonic interceptor missile, characterized in that: The following steps are involved: S1, through the event trigger device to determine whether the interceptor missile's trajectory deviation is greater than the set threshold, if so, execute S2, otherwise continue to monitor the actual trajectory deviation; S2, obtaining the expected trajectory deviation angle according to the positions of the interceptor missile and the target; S3, constructing an infinite time domain cost function of the trajectory deviation angle tracking error, wherein the infinite time domain cost function is composed of the sideslip angle guidance law, the upper limit of the disturbance and the tracking error of the interceptor missile system; Constructing an optimal value function of the infinite time domain value function, expressing the optimal value function by the Hamilton-Jacobi equation, wherein the optimization goal of the optimal value function is that the tracking error converges to zero; S4, updating the weight of the evaluation network structure by the gradient descent method, solving the Hamilton-Jacobi equation by the evaluation network structure, obtaining the optimal index function, and then obtaining the sideslip angle controller, wherein the sideslip angle controller outputs a control instruction so that the actual trajectory deviation angle is close to the expected trajectory deviation angle.

2. The method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP according to claim 1, characterized in that: In S2, the calculation formula of the expected trajectory deviation angle is: Among them, V G is the desired velocity converted to the velocity in the original emission system.

3. The method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP according to claim 1, characterized in that: In S3, the infinite time domain value function of the trajectory deviation angle is: Among them, e σ (t) is the tracking error of the trajectory angle at time t, α is a positive number, Δ m (τ) is a known bounded function, e σ is the tracking error of the trajectory angle, τ is the time constant, β B is the component of the roll angle to the lateral plane, and both R and Q are positive definite matrices.

4. The method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP according to claim 1, characterized in that: In S3, the formula of the sideslip angle guidance law is: Where m is mass, V is velocity, P is thrust, J * (e σ ) is the optimal value function of the trajectory deviation angle.

5. The method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP according to claim 1, characterized in that: In S3, the tracking error of the interceptor missile system is: Among them, m represents mass, V represents velocity, P represents thrust, β B is the sideslip angle, Z B The component of the lateral force on the lateral plane, Δ is the uncertain disturbance, is the derivative of the ballistic angle.

6. The method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP according to claim 1, characterized in that: In S3, the optimal value function is: Among them, β B is the sideslip angle.

7. The method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP according to claim 1, characterized in that: In S4, the update law of the evaluation network structure weight is: Among them, λ is the learning rate of the neural network, Ψ J (e σ ) is the activation function, e J is the Hamiltonian residual.

8. The method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP according to claim 7 is characterized in that: The Hamilton-Jacobi equation is solved by evaluating the network structure: The optimal indicator function is: Among them, Θ J is the ideal weight.

9. The method for adaptive mid-course guidance of a hypersonic interceptor missile based on RADP according to claim 1, characterized in that: The controller of the sideslip angle is verified by stability analysis.

10. A RADP-based adaptive mid-range guidance system for hypersonic interceptor missiles, characterized in that: include: A trigger unit, used to determine whether the estimated deviation of the actual ballistic deviation angle of the interceptor missile is greater than a set threshold through an event trigger device, and if so, execute the expected unit, otherwise continue to monitor the actual ballistic deviation angle; An expectation unit is used to obtain an expected trajectory deviation angle according to the positions of the interceptor missile and the target; A function construction unit, used to construct an infinite time domain cost function of a trajectory deviation angle tracking error, wherein the infinite time domain cost function is composed of a sideslip angle guidance law, a disturbance upper limit value, and a tracking error of an interceptor missile system; Constructing an optimal value function of the infinite time domain value function, expressing the optimal value function by the Hamilton-Jacobi equation, wherein the optimization goal of the optimal value function is that the tracking error converges to zero; The solving unit is used to update the weight of the evaluation network structure through the gradient descent method, solve the Hamilton-Jacobi equation through the evaluation network structure, obtain the optimal index function, and then obtain the sideslip angle controller, and the sideslip angle controller outputs a control instruction so that the actual trajectory deviation angle is close to the expected trajectory deviation angle.

Images