Method for designing hypersonic vehicle maneuvering penetration strategy based on timing game

Abstract

This invention discloses a method for designing a hypersonic vehicle maneuvering penetration strategy based on timing game theory. First, based on an analysis of the advantages and disadvantages of hypersonic vehicles in close-range combat, a game-theoretic penetration strategy based on a penetration window is proposed. Then, a mathematical model of offensive and defensive confrontation under an approximate reverse-orbit interception situation is established in the lateral plane. Next, based on the game-theoretic penetration strategy using the penetration window, Monte Carlo simulations are performed on different combat situations, enemy and friendly maneuvering strategies, and maneuvering capabilities in engineering applications to generate a game-theoretic confrontation database. Finally, the database is learned offline through a neural network, and a penetration window is calculated online to guide the hypersonic vehicle in completing penetration in different scenarios. During the online calculation process, for uncertain enemy and situational information, the input item can be taken to extreme values ​​or multiple values ​​in conjunction with the database. Finally, a common interval is found among the generated multiple windows, which is the common penetration window.

Description

Technical Field

[0001] This invention belongs to the field of aircraft technology, specifically relating to a design method for a hypersonic aircraft maneuvering penetration strategy. Background Technology

[0002] In the process of close-range penetration of hypersonic vehicles, the constraints of their own maneuverability, as well as the uncertainties of the maneuverability and guidance law of interceptor missiles, pose great difficulties for effective penetration in close-range and short-term situations. Therefore, how to utilize the attacker's maneuver initiative and high-speed characteristics, and combine them with intelligent learning methods, to propose an intelligent maneuvering penetration strategy for hypersonic vehicles, so as to achieve successful penetration of hypersonic vehicles under the above-mentioned constraints and uncertainties, is the key scientific problem to be solved.

[0003] Based on the current state of research both domestically and internationally, the focus of research on hypersonic vehicle penetration strategies has been primarily on programmed maneuvering penetration strategies. However, for game-theoretic maneuvering penetration, research has mainly focused on penetration guidance laws based on unilateral optimality theory and differential game theory. The hypersonic vehicle maneuvering penetration strategies derived using these two methods introduce the single / double-sided extremum problem of mathematical functionals into the offensive and defensive confrontation model between hypersonic vehicles and anti-missile interceptors. The penetrating vehicle and the interceptor missile are treated as the two sides in a game, incorporating terminal constraints such as reentry point position and velocity, maneuvering overload, and control variable constraints. Performance indicators include the terminal miss distance. Hamiltonian functions are constructed using the motion models of the penetrating vehicle and the interceptor missile, and the optimal penetration and interception strategies are solved using the necessary conditions for extrema. Penetration guidance laws based on unilateral optimality theory require prior assumptions about the enemy interceptor's guidance law. In real offensive and defensive scenarios, obtaining the enemy's guidance information is difficult, posing challenges to engineering implementation. Penetration guidance laws based on differential game theory assume both sides employ optimal maneuvering strategies simultaneously, resulting in conservative laws that fail to fully utilize the penetration capabilities of hypersonic vehicles. While the combination of intelligent methods and penetration strategies has made some progress in scenarios involving drones and cruise missiles, existing intelligent maneuvering penetration strategies for these scenarios are not applicable due to the highly dynamic and rapidly changing nature of hypersonic vehicle offensive and defensive scenarios.

[0004] Considering the engineering implementation of penetration strategies, real-world offensive and defensive scenarios often suffer from problems such as information acquisition delays and poor information accuracy. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention provides a method for designing maneuvering penetration strategies for hypersonic vehicles based on timing game theory. First, based on an analysis of the advantages and disadvantages of hypersonic vehicles in close-range combat, a game-theoretic penetration strategy based on a penetration window is proposed. Then, a mathematical model of offensive and defensive confrontation under an approximate reverse-orbit interception situation is established in the lateral plane. Next, based on the game-theoretic penetration strategy using the penetration window, Monte Carlo simulations are performed on different combat situations, enemy and friendly maneuvering strategies, and maneuvering capabilities in engineering applications to generate a game-theoretic confrontation database. Finally, the database is learned offline through a neural network, and a penetration window is calculated online to guide the hypersonic vehicle in completing penetration in different scenarios. During the online calculation process, for uncertain enemy and situational information, the input item can be taken as an extreme value or multiple values ​​in conjunction with the database. Finally, a common interval is found among the generated multiple windows, which is the common penetration window.

[0006] The technical solution adopted by this invention to solve its technical problem includes the following steps:

[0007] Step 1: Based on the analysis of the advantages and disadvantages of hypersonic vehicles in close-range combat, the operational concept of penetration window is proposed.

[0008] The penetration window refers to the relative distance range within which a hypersonic vehicle can penetrate an interceptor missile through full-G maneuvers; the close range refers to the detection range of the hypersonic vehicle.

[0009] Step 2: Establish a mathematical model of attack and defense confrontation under an approximate reverse-track interception situation in the lateral plane;

[0010] Step 3: Use Monte Carlo simulation to traverse different combat situations, enemy and friendly maneuver strategies, and maneuver capabilities in engineering applications to generate a game-theoretic database;

[0011] Step 4: Perform offline learning on the game adversarial database generated in Step 3 based on the BP neural network.

[0012] Furthermore, step 2 specifically includes:

[0013] The approximate reverse trajectory interception situation refers to a specific angular range. When the initial velocity deviation between the hypersonic vehicle and the interceptor missile is greater than or equal to this angular range, the hypersonic vehicle does not need to perform maneuvering penetration and can complete the penetration using its own speed. When the initial velocity deviation between the two sides is less than this angular range, maneuvering penetration is required.

[0014] The mathematical description of the interceptor missile in the terminal guidance phase is as follows:

[0015]

[0016] In the formula, V m(t) represents the velocity vector of the interceptor missile; γ m Indicates the trajectory deflection angle of the interceptor missile; n zm This represents the actual flight overload of the interceptor missile; g represents the acceleration due to gravity. This represents the velocity component of the interceptor missile on the x-axis; This represents the velocity component of the interceptor missile along the z-axis; The guidance laws for interceptor missiles used in engineering practice include proportional guidance law (PN), modified proportional guidance law (APN), and adaptive sliding mode guidance law (ASMG). These three interceptor missile guidance laws are expressed in the lateral plane as follows:

[0017] Proportional guidance law:

[0018]

[0019] Modified proportional guidance law:

[0020]

[0021] Adaptive sliding mode guidance law:

[0022]

[0023] In the formula: V is the line-of-sight angular velocity from the perspective of the interceptor missile; c The approach speed of the attacking and defending sides; N, ε, and δ are all guidance law parameters; a h This represents the actual acceleration of the hypersonic vehicle. This correction term ensures that the interceptor missile takes overload compensation measures when the target performs a constant-value maneuver; u m This indicates the command acceleration used when intercepting missiles with different guidance laws; This indicates the azimuth angular velocity from the perspective of the interceptor missile;

[0024] Meanwhile, the interceptor missile employs STT control throughout the entire interception process, and its constraint is limited to the maximum available overload constraint n. zmmax , is represented as:

[0025] |n zm |≤n zmmax

[0026] The specific description of hypersonic vehicles is as follows:

[0027]

[0028] Among them, V h (t) represents the velocity vector of the hypersonic vehicle, γ h τ represents the ballistic deflection angle of a hypersonic vehicle. h n represents the first-order element time constant of a hypersonic vehicle.zhc Indicates the overload command for a hypersonic vehicle, n zh This indicates the actual flight overload of a hypersonic vehicle. This represents the velocity component of a hypersonic vehicle on the x-axis. This represents the velocity component of a hypersonic vehicle on the z-axis.

[0029] The control constraints on hypersonic vehicles are:

[0030] n zh ≤n zhmax

[0031] Based on existing simulation results, the angle range for the approximate reverse-orbit interception situation is set to ±3.2°, that is:

[0032] |γ h0 -γ m0 +π|<3.2°

[0033] In the formula, n zhmax Indicates the maximum available overload of a hypersonic vehicle; γ h0 ,γ m0 These represent the initial trajectory deflection angles of the hypersonic vehicle and the interceptor missile, respectively.

[0034] Preferably, step 3 is as follows:

[0035] Step 3-1: Simulation with reference parameters as initial conditions and without parameter bias: Take a fixed value at the initial penetration moment, obtain the state parameter changes during the aircraft game process, and obtain the initial distance between the two sides, the interceptor missile overload, and the launch angle of both sides as reference quantities for the initial bias amount;

[0036] Step 3-2: Using the conclusions and angle constraints in Step 3-1 as initial conditions, conduct simulations under multiple parameters of random deflection: By performing multiple maneuvering penetration simulations under different initial conditions, construct an offensive and defensive adversarial database, traverse the initial situation information and the maneuvering strategies that the interceptor may take, and provide training and verification data for subsequent offline learning of the database through neural networks.

[0037] Preferably, the BP neural network described in step 4 has two hidden layers and selects the tansig activation function. The specific parameter settings of the neural network are shown in the table below:

[0038] Table 1 Neural Network Parameter Settings

[0039]

[0040] The beneficial effects of this invention are as follows:

[0041] This invention proposes a design method for hypersonic vehicle penetration strategies based on active game theory maneuvers in real-world offensive and defensive scenarios, under conditions of information acquisition delay and inaccuracy. It offers two main advantages. First, compared to optimal control and differential games, the self-maneuvering strategy primarily relies on its own short-range early warning and maneuverability capabilities, without depending on assumptions about the interceptor missile's guidance law, resulting in lower uncertainty and easier engineering implementation. Second, the designed penetration strategy itself has simple maneuvering command forms, requires no complex calculations, and faces no real-time limitations, making it easy to implement in engineering. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating the calculation of the penetration window for the hypersonic vehicle of this invention.

[0043] Figure 2 This is a schematic diagram of the penetration window concept of the hypersonic vehicle of the present invention, wherein (a) the hypersonic vehicle maneuvers within the "penetration window", (b) the hypersonic vehicle maneuvers prematurely outside the "penetration window", and (c) the hypersonic vehicle maneuvers too late outside the "penetration window".

[0044] Figure 3 This invention presents a schematic diagram of an approximate reverse-orbit interception situation.

[0045] Figure 4 This invention provides a schematic diagram of offensive and defensive confrontation under an approximate reverse-track interception situation in the lateral plane.

[0046] Figure 5 The following is a Monte Carlo simulation diagram of the "penetration window" in the case of no deflection in an embodiment of the present invention. (a) The interception missile guidance law is a proportional guidance law, (b) The interception missile guidance law is a modified proportional guidance law, and (c) The interception missile guidance law is an adaptive sliding mode guidance law.

[0047] Figure 6 The diagram shows a Monte Carlo simulation of the "penetration window" under random deflection conditions in an embodiment of the present invention. (a) The interceptor missile guidance law is a proportional guidance law, (b) The interceptor missile guidance law is a modified proportional guidance law, and (c) The interceptor missile guidance law is an adaptive sliding mode guidance law.

[0048] Figure 7 The following are examples of the fitting results of the "penetration window" neural network in the embodiments of the present invention when the minimum error of the training target is 0.001: (a) the interception missile guidance law is a proportional guidance law, (b) the interception missile guidance law is a modified proportional guidance law, and (c) the interception missile guidance law is an adaptive sliding mode guidance law. Detailed Implementation

[0049] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0050] To address the issue of poor robustness of existing penetration strategies designed based on classical and intelligent methods in real-world offensive and defensive scenarios, this paper combines the advantages of neural networks, such as independence from explicit mathematical models and the ability to fit complex nonlinear mapping relationships. By fitting the full overload maneuver range of a hypersonic vehicle using neural networks, not only can successful penetration against enemy interceptor missiles be achieved, but the robustness of the penetration strategy to information delays and inaccuracies can also be enhanced.

[0051] The technical solution of this invention is as follows:

[0052] Step 1: Based on the analysis of the advantages and disadvantages of hypersonic vehicles in close-range combat, the concept of "penetration window" is proposed.

[0053] like Figure 2 As shown, the "penetration window" refers to the relative distance range within which a hypersonic vehicle can achieve penetration against an interceptor missile through full-G maneuvers. When the interceptor missile has already formed a near-reverse-track interception posture against the maneuverable vehicle, the speed advantage of the maneuverable vehicle cannot be fully utilized during the penetration process. The main advantages that can be utilized are the maneuver initiative of the attacker and the smaller miss distance required for successful penetration, while the disadvantage is that its overload capacity is far weaker than that of the interceptor missile. Therefore, if the maneuver is performed too early, the exposure time of the overload disadvantage will be longer, leading to interception; conversely, if the maneuver is performed too late, the required miss distance cannot be achieved laterally. Penetration maneuvers can be performed only within the "penetration window," fully utilizing the attacker's maneuver initiative and combining the rapid approach speeds of both sides to create a miss distance of more than 1 meter required for penetration in a short time, while avoiding the overload disadvantage as much as possible, and achieving effective penetration against the interceptor missile within a small area.

[0054] Step 2: Establish a mathematical model of attack and defense confrontation under an approximate reverse-track interception situation in the lateral plane.

[0055] Hypersonic aircraft maneuvering can be categorized into longitudinal and lateral maneuvers based on the direction of the maneuver. These two types of maneuvers can occur individually or simultaneously. This invention prioritizes lateral maneuvers for penetration because they can be performed at a constant altitude and speed, avoiding the impact on the controller caused by changes in speed and altitude during the penetration process.

[0056] like Figure 3 As shown, the near-reverse trajectory interception situation refers to a specific angular range. When the initial velocity deviation between the hypersonic vehicle and the interceptor missile is greater than this angular range, the hypersonic vehicle does not need to perform maneuvering penetration and can complete the penetration using its own speed. When the initial velocity deviation between the two sides is less than this angular range, maneuvering penetration is required.

[0057] Step 3: Based on the concept of "penetration window", Monte Carlo simulation is used to traverse as many different combat situations, enemy and friendly maneuver strategies, and maneuver capabilities as possible in engineering applications to generate a game-theoretic database.

[0058] Step 4: Perform offline learning on the database based on a BP neural network.

[0059] In the new offensive and defensive confrontation scenario, situational parameters, aircraft capability parameters, and enemy and friendly maneuver strategies are used as inputs. Unknown parameters can be selected from multiple sets of values ​​within a boundary range, generating multiple windows. Offline training yields reliable results, while online output provides the penetration maneuver window (a common area of ​​multiple windows). Maneuverable, trajectory-changing aircraft can utilize custom maneuver strategies within this window to achieve efficient, small-range penetration against interceptor missiles. The calculation process for the "penetration window" is attached. Figure 1 As shown.

[0060] Example:

[0061] Step 1: Establish a mathematical model of attack and defense confrontation under an approximate reverse trajectory interception situation in the lateral plane.

[0062] When a single interceptor missile is present, a schematic diagram of the offensive and defensive confrontation in the lateral plane and the definitions of relevant angles are attached. Figure 4 As shown. Where H represents a hypersonic vehicle and M represents an interceptor missile. Figure 4 Without loss of generality, the initial line of sight at HM is taken as the X-axis, and the Z-axis is perpendicular to the X-axis in the constant-height plane. The relevant symbols in the figure are defined as shown in Table 1.

[0063] Table 1 Comparison of Parameter Names of Aircraft on Both Offensive and Defensive Sides

[0064]

[0065] In the table: i = h, m. h represents the hypersonic vehicle, and m represents the interceptor missile. Line-of-sight azimuth λ hm Defined as the direction from the interceptor missile to the hypersonic vehicle. Figure 4 λ hm <0.

[0066] Combined with appendix Figure 4 The two-dimensional planar offensive and defensive confrontation shown can be used to construct the following mathematical description of the interceptor missile in the terminal guidance phase:

[0067]

[0068] In the formula, Typical interceptor missile guidance laws used in engineering practice include Proportional Navigation (PN), Augmented Proportional Navigation (APN), and Adaptive Sliding Mode Guidance (ASMG). These three interceptor missile guidance laws in the lateral plane can be expressed as follows:

[0069] Proportional guidance law:

[0070]

[0071] Modified proportional guidance law:

[0072]

[0073] Adaptive sliding mode guidance law:

[0074]

[0075] In the formula: V is the line-of-sight angular velocity from the perspective of the interceptor missile; c The approach speeds of the attacking and defending sides are denoted as N, ε, and δ, which are guidance law parameters.

[0076] Meanwhile, the interceptor missile employs STT control throughout the entire interception process, and its constraint is limited to the maximum available...

[0077] Overload constraints are represented as:

[0078] |n zm |≤n zmmax

[0079] The specific description of hypersonic vehicles is as follows:

[0080]

[0081] The control constraints on hypersonic vehicles are:

[0082] n zh ≤n zhmax

[0083] Based on existing simulation results, the angle range for the approximate reverse-orbit interception situation is set to ±3.2°, that is:

[0084] |γ h0 -γ m0 +π|<3.2°

[0085] In the formula, γ h0 ,γ m0These represent the initial trajectory deflection angles of the hypersonic vehicle and the interceptor missile, respectively.

[0086] Step 2, based on the concept of "penetration window," utilizes Monte Carlo simulation to traverse as many different combat situations, enemy and friendly maneuver strategies, and maneuver capabilities as possible in engineering applications to generate a game-theoretic database. This includes the following sub-steps.

[0087] Step 2.1, Simulation with reference parameters as initial conditions and without parameter bias: Take a fixed value at the initial penetration moment and obtain the state parameter changes during the aircraft game process to obtain reference quantities such as the initial distance between the two sides, the interceptor missile overload, and the launch angle of both sides.

[0088] Table 2. Simulation parameter settings for maneuver penetration methods based on the concept of "penetration window".

[0089]

[0090]

[0091] The offensive and defensive confrontation between the hypersonic vehicle and the interceptor missile is shown in the appendix when the interceptor missile employs proportional guidance, modified proportional guidance, and adaptive sliding mode guidance respectively. Figure 5 As shown.

[0092] Step 2.2, using the conclusions and angle constraints in Step 2.1 as initial conditions, conduct simulations under multiple parameters of random deflection: By performing multiple maneuvering penetration simulations under different initial conditions, construct an offensive and defensive adversarial database, traverse the initial situation information and the maneuvering strategies that the interceptor may take, and provide training and verification data for subsequent offline learning of the database through neural networks.

[0093] Table 3 Relevant Pull-off Parameter Settings

[0094]

[0095] Monte Carlo simulation results are attached. Figure 6 As shown.

[0096] Step 3: Using the 500 sets of "penetration window" data obtained from the random parameter adjustment in Step 2.2 as the training set, a backpropagation (BP) neural network is used for fitting. The network has two hidden layers and the tansig activation function is selected. After multiple training iterations and parameter adjustments, the network with the best results is retained for subsequent testing.

[0097] Table 4 Neural Network Parameter Settings

[0098]

[0099] The fitting results of the "penetration window" neural network are attached. Figure 7As shown.

[0100] Ultimately, the trained neural network can be used to predict the "penetration window" of a hypersonic vehicle under given initial conditions. Assume the enemy interceptor missile's initial state is as follows: initial relative distance R0 = 11 km, trajectory deflection angle γ... m0 =181°, maximum usable overload u mmax =7g;

[0101] 1) When the enemy interceptor missile adopts the proportional guidance law, the actual upper limit of the hypersonic vehicle's "penetration window" is 10km and the lower limit is 1.3km. Using the finally trained BP neural network to predict the window, the upper limit of the "penetration window" is predicted to be 9.99km and the lower limit to be 1.33km, with errors of 0.01 and 0.03 respectively.

[0102] 2) When the enemy interceptor missile adopts the modified proportional guidance law, the actual upper limit of the hypersonic vehicle's "penetration window" is 5.3km and the lower limit is 1.4km. Using the finally trained BP neural network for window prediction, the predicted upper limit of the "penetration window" is 5.215km and the lower limit of the "penetration window" is 1.422km, with errors of 0.085 and 0.022 respectively.

[0103] 3) When the enemy interceptor missile adopts the adaptive sliding mode guidance law, the upper limit of the actual "penetration window" of the hypersonic vehicle is 8.1km and the lower limit is 1.4km. The window is predicted by using the finally trained BP neural network, and the upper limit of the "penetration window" is predicted to be 8.175km and the lower limit is predicted to be 1.475km, with errors of 0.075 and 0.075 respectively.

[0104] Combining the three "penetration windows" calculated using different guidance laws against enemy interceptor missiles, the common "penetration window" is finally obtained as △R = [1.475, 5.215]. Hypersonic vehicles can successfully penetrate enemy interceptor missiles by performing full-G maneuvers within this range.

[0105] Conclusion: Based on the analysis of the advantages and disadvantages of hypersonic vehicles in close-range combat, this invention proposes the operational concept of a "penetration window." It constructs an offensive and defensive adversarial database, traversing initial situational information and possible maneuver strategies of the interceptor. Through offline learning of the database using a BP neural network, it calculates and generates "penetration windows" online to guide hypersonic vehicles in penetrating different scenarios. During online calculation, for uncertain enemy and situational information, the input item can be taken as an extreme value or multiple values ​​based on the database. Finally, a common interval is found among the generated windows, which is the common "penetration window." The design concept of the common "penetration window" avoids subjective estimation of uncertain information, making the designed penetration strategy universal and robust.

Claims

1. A method for designing a hypersonic vehicle maneuvering penetration strategy based on timing game theory, characterized in that, The steps include the following: Step 1: Based on the analysis of the advantages and disadvantages of hypersonic vehicles in close-range combat, the operational concept of penetration window is proposed; The penetration window refers to the relative distance range within which a hypersonic vehicle can penetrate an interceptor missile through full-G maneuvers; the close range refers to the detection range of the hypersonic vehicle. Step 2: Establish an offensive and defensive mathematical model in the lateral plane under an approximate reverse-track interception situation; specifically: The approximate reverse trajectory interception situation refers to the situation where, when the initial velocity deviation between the hypersonic vehicle and the interceptor missile is greater than or equal to a specific angular range, the hypersonic vehicle does not need to perform maneuvering penetration and can complete the penetration using its own speed; when the initial velocity deviation between the two sides is less than this angular range, maneuvering penetration is required. The mathematical description of the interceptor missile in the terminal guidance phase is as follows: In the formula, This represents the velocity vector of the interceptor missile; Indicates the trajectory deviation angle of the interceptor missile; This indicates the actual flight overload of the interceptor missile; Represents gravitational acceleration; Indicates that the interceptor missile is in Velocity components on the axis; Indicates that the interceptor missile is in Velocity components on the axis; The guidance laws for interceptor missiles used in engineering practice include proportional guidance law (PN), modified proportional guidance law (APN), and adaptive sliding mode guidance law (ASMG). These three interceptor missile guidance laws are expressed in the lateral plane as follows: Proportional guidance law: Modified proportional guidance law: Adaptive sliding mode guidance law: In the formula: The line-of-sight angular velocity is the velocity from the perspective of the interceptor missile. The approach speed of the attacking and defending sides; and All are guidance law parameters; This represents the actual acceleration of a hypersonic vehicle during flight; This indicates the command acceleration used when intercepting missiles with different guidance laws; This indicates the azimuth angular velocity from the perspective of the interceptor missile; Meanwhile, the interceptor missile employs STT control throughout the entire interception process, and its only constraint is the maximum available overload constraint. , represented as: The specific description of hypersonic vehicles is as follows: in, This represents the velocity vector of a hypersonic vehicle. Indicates the ballistic deflection angle of a hypersonic vehicle. This represents the first-order element time constant of a hypersonic vehicle. This indicates an overload command for a hypersonic aircraft. This indicates the actual flight overload of a hypersonic vehicle. Indicates that hypersonic vehicles are in Velocity components on the axis, Indicates that hypersonic vehicles are in Velocity components on the axis; The control constraints on hypersonic vehicles are: Based on existing simulation results, the angle range for the approximate reverse-orbit interception situation is set as follows: ,Right now: In the formula, This indicates the maximum available overload of a hypersonic vehicle. These are the initial trajectory deflections of the hypersonic vehicle and the interceptor missile, respectively. Step 3: Use Monte Carlo simulation to traverse different combat situations, enemy and friendly maneuver strategies, and maneuver capabilities in engineering applications to generate a game-theoretic database; Step 4: Perform offline learning on the game adversarial database generated in Step 3 based on the BP neural network.

2. The method for designing a hypersonic vehicle maneuvering penetration strategy based on timing game theory as described in claim 1, characterized in that, Step 3 is as follows: Step 3-1: Simulation with reference parameters as initial conditions and without parameter bias: Take a fixed value at the initial penetration moment, obtain the state parameter changes during the aircraft game process, and obtain the initial distance between the two sides, the interceptor missile overload, and the launch angle of both sides as reference quantities for the initial bias amount; Step 3-2: Using the initial deflection reference value and angle limit in Step 3-1 as initial conditions, conduct simulations under multiple parameters of random deflection: By performing multiple maneuvering penetration simulations under different initial conditions, construct an offensive and defensive confrontation database, traverse the initial situation information and the maneuvering strategies that the interceptor may take, and provide training and verification data for subsequent offline learning of the database through neural networks.

3. The method for designing a hypersonic vehicle maneuvering penetration strategy based on timing game theory according to claim 2, characterized in that, The BP neural network described in step 4 has two hidden layers and selects the tansig activation function; its parameters are set as follows: Training sessions: 1000; Learning rate: 0.01; Minimum training error: 0.001; Minimum performance gradient: 1*10 -6 ; Damping factor: 1*10 10 ; Maximum number of failed confirmations: 50.

Images