Hypersonic aircraft cluster collaborative guidance method and system based on neural network

Abstract

The invention discloses a hypersonic flight vehicle cluster collaborative guidance method and system based on a neural network, and relates to the technical field of aviation, and the method comprises the steps: obtaining the state information of each flight vehicle; based on a kinematics principle, constructing a motion model of the attack projectile and the interception projectile; constructing an interception point prediction neural network according to the distance between the interception bomb and the attack bomb at the moment when the interception bomb engine is shut down, the prepositive angle of the attack bomb, the prepositive angle of the interception bomb, the speed of the attack bomb and the speed of the interception bomb; according to the cooperative defense penetration guidance strategy and the consistency guidance law of the attack missile cluster, generating a cooperative guidance instruction for the attack missile cluster based on the coordinates of the interception points predicted by the interception point prediction neural network; the consistency guidance law of the attack projectile cluster comprises an initial maneuvering section, a consistency control maneuvering section and a maneuvering escape section. The method can improve the penetration success rate of the attack projectile cluster on the premise that the position information of an enemy does not need to be obtained in the whole process.

Description

Technical Field

[0001] This application relates to the field of aviation technology, and in particular to a method and system for cooperative guidance of hypersonic vehicle swarms based on neural networks. Background Technology

[0002] With the rapid development of military technology, modern warfare is increasingly characterized by unmanned and intelligent features. This paradigm shift has made cooperative strategies increasingly important in the field of guidance and control. Among these, cooperative swarm guidance and penetration has become a key research area, demonstrating stronger adaptability, survivability, and combat effectiveness in adversarial environments. Therefore, swarm penetration strategies and their guidance are becoming a research focus for addressing the challenges of next-generation warfare scenarios.

[0003] Unlike the classic "one-on-one" combat scenarios in most current research, swarm penetration focuses on "many-on-many" adversarial scenarios. However, existing research mostly concentrates on small-scale swarm adversarial scenarios, usually limited to "two-on-two" engagement modes, and its scalability is limited in large-scale swarm operations. Furthermore, these methods heavily rely on accurate real-time detection of the interceptor's information, a prerequisite often difficult to achieve in modern warfare, thus posing significant challenges and limitations to their application in actual combat. Simultaneously, the widely adopted differential game theory and optimization algorithms place high demands on high-performance computing hardware, which is typically difficult to implement in missile-borne computing systems. Compared to specific and simple scenarios like "one-on-one" and "two-on-two," research results focusing on cooperative penetration strategies for hypersonic vehicles in swarm modes are currently scarce. To date, a cooperative penetration strategy and guidance law design algorithm for hypersonic vehicle swarm scenarios is lacking. Summary of the Invention

[0004] The purpose of this application is to provide a collaborative guidance method and system for hypersonic vehicle swarms based on neural networks, which can improve the penetration success rate of attack missile swarms.

[0005] To achieve the above objectives, this application provides the following solution:

[0006] Firstly, this application provides a cooperative guidance method for hypersonic vehicle swarms based on neural networks, including:

[0007] Acquire the status information of each aircraft; the aircraft include offensive missiles and interceptor missiles; the status information of the offensive missiles is real-time status information; the status information of the interceptor missiles includes status information before the interceptor missile engine shuts down and during the terminal phase of interception.

[0008] Based on kinematic principles, a motion model is constructed for the offensive and interceptor missiles; this motion model describes the relative motion relationship between the offensive and interceptor missiles.

[0009] An interception point prediction neural network is constructed based on the distance between the interceptor and the attacking missile, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity at the moment the interceptor's engine is shut down. The output of the interception point prediction neural network is the coordinates of the interception point. The distance between the interceptor and the attacking missile, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity at the moment the interceptor's engine is shut down are determined based on real-time status information. The distance between the interceptor and the attacking missile, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity at the moment the interceptor's engine is shut down are determined based on the status information of each aircraft.

[0010] Based on the coordinated penetration guidance strategy and consistency guidance law of the offensive missile cluster, coordinated guidance commands are generated for the offensive missile cluster based on the coordinates of the interception point predicted by the interception point prediction neural network. The coordinated penetration guidance strategy refers to selecting one offensive missile in the offensive missile cluster as a sacrificial missile to attract the attack target of the interceptor missiles, while the remaining offensive missiles change course to break through the interceptor missiles' defense system. The consistency guidance law includes an initial maneuver phase, a consistency control maneuver phase, and a maneuver escape phase. The initial maneuver phase is used to control the offensive missiles to concentrate towards the central axis of the lateral maneuver corridor. The consistency control maneuver phase is used to calculate the remaining seeker activation time of the offensive missiles based on the interception point predicted by the neural network, and to make the remaining seeker activation time of each offensive missile tend to be consistent. The maneuver escape phase is used to select a sacrificial missile among the offensive missiles to activate the decoy and control the remaining offensive missiles to escape with maximum overload.

[0011] Optionally, the formula expression for the motion model is:

[0012]

[0013] In the formula, q i and r i Indicates the i-th interceptor missile M i and the i-th attack missile A i The line-of-sight angle between the projectile and the target, and the relative distance between the projectile and the target, θ Mi ,η Mi They are interceptor missiles M i The velocity direction angle and lead angle, θ Ti ,η Ti These are offensive missiles A. i velocity direction angle and lead angle, a Ai It is the i-th attack missile A i normal acceleration, V Mi and V Ai These are the i-th interceptor missiles M i With the i-th attack missile A ispeed, D Mi and D Ai These are the i-th interceptor missiles M i With the i-th attack missile A i The drag acceleration, r i ,q i ,θ Ai V Mi V Ai The first derivative with respect to time t.

[0014] Optionally, based on the distance between the interceptor and the attacking missile at the moment the interceptor's engine shuts down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity, an interception point prediction neural network is constructed, specifically including:

[0015] Set the input variables for the interception point prediction neural network. Where r i s , y Ai ,y Mi , respectively represent the distance between interceptor i and attacking missile i at the moment the interceptor missile engine is shut down, the leading angle of attacking missile i, the leading angle of interceptor i, the y-coordinate of attacking missile i, and the y-coordinate of interceptor i; These are the velocities of the attack missile i and the interceptor missile i, respectively.

[0016] The target constraint for the interception point prediction neural network is the remaining seeker startup time. Consistency control; among which, t represents the activation time of the guidance system of interceptor missile i, where t is the current time.

[0017] Set the activation function of the interception point prediction neural network as follows:

[0018] Set the output layer function of the intercept point prediction neural network as: purelin(x) = x.

[0019] The Levenberg-Marquardt algorithm is used as the backpropagation algorithm for the interception point prediction neural network to construct the interception point prediction neural network.

[0020] Optionally, after constructing the interception point prediction neural network based on the distance between the interceptor and the attacking missile at the moment the interceptor's engine shuts down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity, the following further steps are included:

[0021] Based on the sample dataset, the interception point prediction neural network is trained to obtain the trained interception point prediction neural network; the sample dataset includes real-time status information of interceptor missiles and attack missiles in multiple historical interception scenarios and the corresponding interception point coordinates.

[0022] Optionally, the initial maneuver phase is used to control the offensive missile to concentrate towards the central axis of the lateral maneuver corridor, specifically including:

[0023] According to the formula Control the offensive missiles to concentrate along the central axis of the lateral maneuver corridor; among them, K py and K dy If y is a constant, then y = y c Represents the central axis of the lateral mobility corridor, sgn(·) is the sign function, sgn(u Ai ) = 1 indicates that the overload direction is θ Ai The direction of increase, u Ai Indicates overload.

[0024] When all the aforementioned offensive missiles are concentrated to satisfy max(y) Ai )-min(y Ai )<y s1 At that time, the guidance law is switched to the consistent maneuver phase; where y s1 It is a constant.

[0025] Optionally, the consistency control maneuver segment is used to calculate the remaining seeker activation time of the attack missile based on the interception point predicted by the neural network, and to make the remaining seeker activation time of each attack missile more consistent, specifically including:

[0026] According to the formula Calculate the remaining seeker activation time of the offensive missile. Where, x pi ,y pi These are the predicted interception point coordinates for the interceptor missile.

[0027] Optionally, in calculating the remaining seeker activation time of the offensive missile. Following that, it also includes:

[0028] Based on the remaining seeker activation time of each of the aforementioned offensive missiles And the maximum value of the remaining seeker opening time among the remaining seeker opening times. Sure and The difference

[0029] When the difference is greater than a set threshold, according to the formula Calculate the guidance status flag variables; where tgo c >0, a>0 are constants, yc1 and y c2 It is a constant, and satisfies 0 < y c1 <y c2 .

[0030] Based on the guidance status flag variable, according to formula u Ai =sgn(u Ai )·|u Ai |Calculate offensive missile A i Overload; among them, K θ ,K t ,K dt All are constants.

[0031] Optionally, the maneuvering escape phase is used to select a sacrificial missile to activate the decoy among the attack missiles and control the remaining attack missiles to escape with maximum overload, specifically including:

[0032] The field of view angle of the interceptor missile's seeker is ±σ.

[0033] For the offensive missile k, if at t l At any given moment, if the attacking missile k is within the field of view of the seeker of the interceptor missile i, then it is determined that the interceptor missile i can be lured by the attacking missile k, and the set of lurable interceptor missiles is obtained as follows: for:

[0034] Based on the set of luring interceptor missiles, and according to the principle of sacrificial missile selection, the sacrificial missiles in the set of luring interceptor missiles are determined; where x Ak (t l ), y Ak (t l ) for t l The coordinates of the attack missile k at any given moment, x Mi (t l ), y Mi (t l ) for t l The coordinates of intercepting bullet i are constantly monitored.

[0035] According to the formula Calculate the overload of the remaining offensive missiles; among them, This represents the maximum overload value of the offensive missile.

[0036] Optionally, the selection principles for sacrificial rounds include:

[0037] For each attacking missile in the set of lurable interceptor missiles, determine whether the attacking missile is within the field of view of the interceptor missile seeker, and determine the number of interceptor missiles within the field of view of the interceptor missile seeker.

[0038] The offensive missile with the largest number of interceptor missiles within the field of view of the interceptor missile's seeker is selected as the sacrificial missile.

[0039] If multiple attack missiles meet the conditions for a sacrificial missile, then among the attack missiles that meet the conditions, the one that is furthest from the interceptor missile is selected as the sacrificial missile.

[0040] Secondly, this application provides a neural network-based hypersonic vehicle swarm cooperative guidance system, comprising:

[0041] The information acquisition module is used to acquire the status information of each aircraft, including offensive missiles and interceptor missiles; the status information of the offensive missiles is real-time status information; the status information of the interceptor missiles includes status information before the interceptor missile engine shuts down and during the terminal phase of interception.

[0042] The motion model construction module is used to construct motion models of the offensive and interceptor missiles based on kinematic principles; the motion models are used to describe the relative motion relationship between the offensive and interceptor missiles.

[0043] The neural network construction module is used to construct an interception point prediction neural network based on the distance between the interceptor and the attacking missile at the moment the interceptor's engine is shut down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity. The output of the interception point prediction neural network is the coordinates of the interception point. The distance between the interceptor and the attacking missile at the moment the interceptor's engine is shut down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity are determined based on real-time status information. The distance between the interceptor and the attacking missile at the moment the interceptor's engine is shut down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity are determined based on the status information of each aircraft.

[0044] The cooperative guidance module is used to generate cooperative guidance commands for the attack missile cluster based on the cooperative penetration guidance strategy and consistency guidance law of the attack missile cluster, and on the coordinates of the interception point predicted by the interception point prediction neural network. The cooperative penetration guidance strategy refers to selecting one attack missile in the attack missile cluster as a sacrificial missile to attract the attack target of the interceptor missiles, while the remaining attack missiles change course to break through the interceptor missiles' defense system. The consistency guidance law includes an initial maneuver phase, a consistency control maneuver phase, and a maneuver escape phase. The initial maneuver phase is used to control the attack missiles to concentrate towards the central axis of the lateral maneuver corridor. The consistency control maneuver phase is used to calculate the remaining seeker activation time of the attack missiles based on the interception point predicted by the neural network, and to make the remaining seeker activation time of each attack missile tend to be consistent. The maneuver escape phase is used to select a sacrificial missile among the attack missiles to activate the decoy and control the remaining attack missiles to escape with maximum overload.

[0045] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0046] This application provides a neural network-based cooperative guidance method and system for hypersonic vehicle swarms. First, it acquires the state information of each vehicle and constructs accurate motion models for both attack and interceptor missiles. Second, it introduces an interception point prediction neural network. This network can predict the coordinates of the interception point based on multiple key parameters at the moment the interceptor missile's engine shuts down, helping the attack missiles avoid potential interception threats in advance. Furthermore, it designs a cooperative penetration guidance strategy and a consistent guidance law for the attack missile swarm, including an initial maneuver phase, a consistent control maneuver phase, and a maneuver escape phase. In the initial maneuver phase, the attack missiles are controlled to converge towards the central axis of the lateral maneuver corridor, which helps enhance the coordination and stealth of the attack missile swarm. In the consistent control maneuver phase, based on the interception point predicted by the neural network, the remaining seeker activation time of the attack missiles can be accurately calculated, ensuring that the attack missiles can launch their attack at the optimal time. In the maneuver escape phase, by selecting a sacrificial missile to activate the decoy and controlling the remaining attack missiles to escape with maximum overload, the survivability and penetration success rate of the attack missile swarm are further improved. This application effectively improves the penetration success rate of offensive missile clusters by constructing an accurate motion model, introducing an interception point prediction neural network, and designing a cooperative penetration guidance strategy. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart illustrating a neural network-based cooperative guidance method for hypersonic vehicle swarms, as provided in one embodiment of this application.

[0049] Figure 2 This is a schematic diagram illustrating the guidance relationship between an interceptor missile and an offensive missile, provided in one embodiment of this application.

[0050] Figure 3 This is a schematic diagram of a cluster-based collaborative penetration strategy provided in an embodiment of this application.

[0051] Figure 4 A lateral mobility corridor diagram provided for an embodiment of this application.

[0052] Figure 5 This is a diagram of a neural network hierarchy provided in an embodiment of this application.

[0053] Figure 6A cluster trajectory diagram of offensive and interceptor missiles provided in one embodiment of this application.

[0054] Figure 7 This is a magnified view of a portion of the vicinity of the convergence point of the trajectories of offensive and interceptor missiles, provided in an embodiment of this application.

[0055] Figure 8 This is a schematic diagram of the functional modules of a neural network-based hypersonic vehicle swarm cooperative guidance system provided in an embodiment of this application. Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0058] Example 1

[0059] like Figure 1 As shown, this embodiment provides a cooperative guidance method for hypersonic vehicle swarms based on neural networks, including:

[0060] Step 101: Obtain the status information of each aircraft; the aircraft include offensive missiles and interceptor missiles; the status information of the offensive missiles is real-time status information; the status information of the interceptor missiles includes the status information before the interceptor missile engine shuts down and the status information at the end of the interception phase.

[0061] Step 102: Based on kinematic principles, construct motion models for the offensive and interceptor missiles; these models describe the relative motion between the offensive and interceptor missiles.

[0062] Step 103: Construct an interception point prediction neural network based on the distance between the interceptor and the attacking missile at the moment the interceptor's engine is shut down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity; the output of the interception point prediction neural network is the coordinates of the interception point; the distance between the interceptor and the attacking missile at the moment the interceptor's engine is shut down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity are determined based on real-time status information; the distance between the interceptor and the attacking missile at the moment the interceptor's engine is shut down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity are determined based on the status information of each aircraft.

[0063] Step 104: Based on the coordinated penetration guidance strategy and consistency guidance law of the offensive missile cluster, generate a coordinated guidance command for the offensive missile cluster based on the coordinates of the interception point predicted by the interception point prediction neural network. The coordinated penetration guidance strategy refers to selecting one offensive missile in the offensive missile cluster as a sacrificial missile to attract the attack target of the interceptor missiles, while the remaining offensive missiles change course to break through the interceptor missiles' defense system. The consistency guidance law includes an initial maneuver phase, a consistency control maneuver phase, and a maneuver escape phase. The initial maneuver phase is used to control the offensive missiles to concentrate towards the central axis of the lateral maneuver corridor. The consistency control maneuver phase is used to calculate the remaining seeker activation time of the offensive missiles based on the interception point predicted by the neural network, and to make the remaining seeker activation time of each offensive missile tend to be consistent. The maneuver escape phase is used to select a sacrificial missile among the offensive missiles to activate the decoy and control the remaining offensive missiles to escape with maximum overload.

[0064] Specifically, this embodiment proposes a cluster-based coordinated cover penetration strategy. The main objective of this strategy is to select one offensive missile as a sacrificial missile before the offensive missile cluster intersects with the enemy interceptor missile cluster. This sacrificial missile uses certain methods (such as enhancing the reflection signal of the enemy's detection radar) to lure all enemy interceptor missiles, while ensuring that all other offensive missiles still have sufficient escape distance from the interceptor missiles. This allows the remaining offensive missiles to successfully penetrate the enemy's defenses after being lured into a trap. Furthermore, this strategy uses a neural network to predict the interception point during the mid-course guidance phase. This prediction does not require specific information about the enemy interceptor missiles' position and velocity during the mid-course guidance phase, thus improving the feasibility of this strategy in practical applications.

[0065] In some embodiments, when performing steps 101-104, the specific steps may be as follows:

[0066] First, the motion model of the aircraft is modeled.

[0067] (a1) Assume that both our side and the enemy side have high-speed aircraft, and the number of them is equal, denoted as N. m Offensive missiles (our side) and interceptor missiles (enemy side) are respectively launched using clusters. and This embodiment discusses the penetration phase of an offensive missile that occurs during the reentry gliding phase. During this phase, the aircraft is in a quasi-equilibrium gliding state, and its trajectory altitude changes little; therefore, changes in altitude are ignored. This embodiment studies the design of a cover-penetration guidance law in a horizontal two-dimensional plane. Figure 2 The diagram shown illustrates the guidance relationship between interceptor and attack missiles, where xOy is the inertial coordinate system.

[0068] Assuming the interceptor missile uses proportional guidance to intercept our corresponding attacking missile, the motion model (relative motion equations) between the interceptor missile and the corresponding attacking missile is as follows:

[0069]

[0070] Where, q i and r i Indicates the i-th interceptor missile M i and the i-th attack missile A i The line-of-sight angle between the projectile and the target, and the relative distance between the projectile and the target, θ Mi ,η Mi They are interceptor missiles M i The velocity direction angle and lead angle, θ Ti ,η Ti These are offensive missiles A. i The velocity direction angle and lead angle. Ai It is the i-th attack missile A i normal acceleration, V Mi and V Ai These are the i-th interceptor missiles M i With the i-th attack missile A i speed, D Mi and D Ai These are the i-th interceptor missiles M i With the i-th attack missile A i The drag acceleration. r i ,q i ,θ Ai V Mi V Ai The first derivative with respect to time t. This embodiment calculates the corresponding normal overload u by designing the guidance law for the attack missile. Ai This is to achieve a cluster-based, coordinated penetration strategy, thereby improving the overall penetration rate of the offensive missile cluster.

[0071] (a2) Based on the above set of equations, the motion model of the interceptor missile and the attack missile can be simulated and modeled by methods such as Euler method or Runge-Kutta method.

[0072] Then, an interception point prediction neural network is constructed and trained.

[0073] By constructing such Figure 5 The neural network shown is used to train the neural network and obtain the interception point prediction neural network.

[0074] Define remaining seeker power-on time for:

[0075]

[0076] in, t represents the activation time of the guidance system of interceptor missile i, where t is the current time.

[0077] One of the goals of the cluster-based collaborative penetration strategy proposed in this embodiment is to achieve... Consistent control is achieved to ensure that sacrificial missiles lure as many interceptor missiles as possible, thereby increasing the penetration success rate of the attack missile cluster.

[0078] (b1) Construct the training dataset:

[0079] The neural network used to predict the intercept point is a multiple-input single-output (MISO) network. The output variable is the predicted intercept point (PIP), while the input variables need to be carefully designed. Generally, the input variables should comprehensively cover all independent variables related to the output, while avoiding redundant or irrelevant variables. In the scenario discussed in this paper, information about the interceptor missile is unavailable for most of the process, therefore, real-time prediction of the intercept point and the remaining seeker activation time is not possible. Therefore, this embodiment only uses the information when the interceptor missile engine is shut down to train the neural network and obtain the predicted interception point, where the superscript s indicates the moment when the interceptor missile engine is shut down.

[0080] Variables related to predicting intercept points include: These are, respectively, the distance between interceptor missile i and attacking missile i at the moment the interceptor missile's engine shuts down, the lead angle of attacking missile i, the lead angle of interceptor missile i, the velocity of attacking missile i, and the velocity of interceptor missile i. Furthermore, since the swarm will converge towards the central maneuver corridor, the y-coordinates of attacking missile i and interceptor missile i are also relevant variables. Meanwhile, variable V Ai V Mi This can be simplified to their ratio: Therefore, the input variables of the intercept point prediction neural network include:

[0081]

[0082] The output variable of the interception point prediction neural network is the interception point (x) at which the interceptor missile successfully intercepts the corresponding attack missile. pi ,y pi ).

[0083] The training dataset was generated through numerical simulation. In the numerical simulation used to collect the dataset, the attack missile cluster only performed the initial maneuver towards the central maneuver corridor, without subsequent maneuvers to achieve... Consistent maneuverability. Based on the application scenario, multiple simulations are conducted with certain initial conditional random variables. Each simulation collects the 6-dimensional input variables corresponding to the engine shutdown of each interceptor missile. And the output variable interception point coordinates [xpi ,y pi ] T This forms the neural network training dataset.

[0084] (b2) Constructing a neural network:

[0085] Design a hierarchical structure for a neural network. The input to the neural network is 6-dimensional. The output is 2D [x pi ,y pi ] T Generally, one to two hidden layers and one output layer are sufficient. Specifically, a hidden layer containing 10 neurons and an output layer can be designed, as shown in the attached diagram. Figure 4 As shown.

[0086] Then, the activation function of the neural network is designed as follows:

[0087]

[0088] The output layer function of the neural network is designed as follows:

[0089] purelin(x) = x (5).

[0090] Finally, the backpropagation algorithm for the neural network is designed. Considering that the proposed neural network is a small-scale network and is trained offline, this embodiment selects the Levenberg-Marquardt algorithm as its backpropagation algorithm, which has the advantages of fast convergence, strong robustness and high degree of automation.

[0091] (b3) Training the neural network:

[0092] Based on the sample dataset, the interception point prediction neural network is trained to obtain the trained interception point prediction neural network; the sample dataset includes real-time status information of interceptor missiles and attack missiles in multiple historical interception scenarios and the corresponding interception point coordinates.

[0093] Specifically, after constructing the neural network, 80% of the dataset is randomly selected as the training set, and the neural network is trained using the Levenberg-Marquardt algorithm. After training, the remaining 20% ​​of the data is selected as the validation set to verify the training results of the neural network.

[0094] Next, a coordinated penetration guidance strategy for offensive missile clusters will be established.

[0095] (c1) Design a cluster-coordinated attack strategy:

[0096] This embodiment proposes a cluster-based coordinated penetration strategy. Specifically, it involves concentrating both sides within a certain lateral range, then selecting a sacrificial attack missile as a target, which uses certain methods (such as enhancing its radar reflection signal) to induce the interceptor missile to target that missile. Subsequently, the sacrificial missile and other attack missiles maneuver in different directions, allowing the protected attack missile to escape the seeker's field of view (e.g., ...) of the interceptor missile. Figure 2 (As shown in the light red fan-shaped area), thus achieving breakthrough.

[0097] To achieve the highest possible success rate of cluster penetration using the proposed cover-and-penetrate strategy, the following two points must be ensured:

[0098] 1) Sacrificial missiles can attract as many interceptor missiles as possible.

[0099] 2) Maximize the escape time of the covered attack missiles.

[0100] The first point is to ensure that the sacrificial missile is within the detection range of multiple interceptor missile seekers as much as possible; this is also the principle for selecting the sacrificial missile. If multiple attack missiles can attract the same number of interceptor missiles, then the attack missile at the rear is selected as the decoy. Therefore, it is necessary to concentrate the attack and interceptor missiles in the lateral direction to ensure a high degree of overlap in the field of view of the interceptor missile detectors.

[0101] (c2) Design a consistent guidance law for offensive missile clusters:

[0102] i) Initial maneuver phase:

[0103] First, the offensive missile cluster should be directed towards... Figure 3 The lateral maneuver corridor shown is used for concentrated maneuvering.

[0104] At this stage, u Ai The design is as follows:

[0105]

[0106] Among them, K py and K dy If y is a constant, then y = y c This represents the central axis of the transverse mobility corridor. sgn(·) is the sign function, sgn(u Ai ) = 1 indicates that the overload direction is θ Ai The direction of increase is the same as the direction of decrease.

[0107] When all attack missiles are concentrated to satisfy max(y) Ai )-min(y Ai )<y s1 At that time, the guidance law switches to the consistent maneuver phase, where y s1 It is a constant.

[0108] ii) Consistency control maneuver segment:

[0109] y Ai With y c The difference is denoted as Δy Ai =y Ai -y c The remaining seeker activation time can be estimated using the following formula.

[0110]

[0111] Where, x pi ,y pi For interceptor missiles and offensive missiles M i A i The predicted interception point coordinates are output by the neural network constructed and trained in step two.

[0112] remember i = 1, 2, ..., N m The maximum value in is Since hypersonic attack missiles cannot accelerate during guidance, they achieve this through maneuvering. The only way to achieve consistency is to make smaller near Will and The difference is denoted as Obviously, there are

[0113] Define and calculate the guidance status flag variable sta as follows:

[0114]

[0115] Among them, tgo c >0 is a constant, when That is to identify Consensus has been reached.

[0116] If sta ≠ 0, that is sta is defined as follows:

[0117]

[0118] Among them, tgo c >0, a>0 are constants, y c1 and y c2 It is also a constant, and satisfies 0 < y c1 <y c2 .

[0119] Finally, calculate the attack missile A. i Overload. Overload uAi The amplitude is calculated using the following formula:

[0120]

[0121] Among them, K θ ,K t ,K dt All are constants.

[0122] Calculate the control overload direction sgn(u) Ai ):

[0123]

[0124] Calculate offensive missile A i Overload: u Ai =sgn(u Ai )·|u Ai |

[0125] iii) Escape phase:

[0126] Let r be the terminal guidance range (i.e., the range at which the seeker is activated) of the enemy interceptor missile cluster. f Assume that at time t p All attack missiles entered the interception range of their corresponding interceptor missiles, that is:

[0127]

[0128] At that moment, the attack missile cluster enters the maneuver escape phase.

[0129] Select the sacrificial ammunition according to the following principles and activate it to decoy.

[0130] Assuming the seeker's field of view angle of the interceptor missile is ±σ (σ>0), for the attack missile k, assuming that at time t p If an interceptor missile i is within the field of view of its seeker, then that interceptor missile can be lured by the attacking missile k. Define the set of interceptor missiles that can be lured. for:

[0131]

[0132] Where, x Ak (t l ), y Ak (t l ) for t l The coordinates of the attack missile k at any given moment, x Mi (t l ), y Mi (t l ) for t l The coordinates of intercepting bullet i are constantly monitored.

[0133] Based on Sacrificial Bullet M k The selection principles are as follows:

[0134] For each attacking missile in the set of lurable interceptor missiles, determine whether the attacking missile is within the field of view of the interceptor missile seeker, and determine the number of interceptor missiles within the field of view of the interceptor missile seeker.

[0135] The offensive missile with the largest number of interceptor missiles within the field of view of the interceptor missile's seeker is selected as the sacrificial missile.

[0136] If multiple attack missiles meet the conditions for a sacrificial missile, then among the attack missiles that meet the conditions, the one that is furthest from the interceptor missile is selected as the sacrificial missile.

[0137] Specifically, i) within the field of view of the seeker of the most interceptor missiles, i.e., cluster maximum:

[0138]

[0139] ii) If multiple attack missiles satisfy condition i), then select the last attack missile in the cluster from all attack missiles that satisfy condition i), i.e., x. Ai The smallest one, designated as the sacrificial bullet, is denoted as A. d .

[0140] After selecting the sacrificial missile and activating its decoy function, calculate the overload of the attack missile cluster using the following formula:

[0141]

[0142] in, This represents the maximum overload value of the offensive missile.

[0143] Based on the above calculations, the overload u of the offensive missile can be obtained. Ai It also controls the attack missile cluster and completes the cluster's coordinated penetration.

[0144] In some embodiments, numerical simulations of a "10 vs. 10" scenario are performed using this strategy algorithm, and magnified views of the trajectories of the attack and interceptor missile clusters and the vicinity of their convergence point are shown below. Figure 6 and Figure 7 As shown, Figure 6 and Figure 7 The lines at the end of the interceptor missile trajectories indicate the field of view of each interceptor missile. It can be seen that all attack missiles are within the field of view of the interceptors; * indicates that the selected sacrificial missile is attack missile number 2, meaning that attack missile number 2 is used as a decoy, while the remaining attack missiles maneuver in the opposite direction. When the sacrificial missile is intercepted by an interceptor missile, nine attack missiles are now outside the field of view of all the interceptor missiles' seekers, thus the swarm's penetration success rate is 90%.

[0145] Example 2

[0146] like Figure 8 As shown, this embodiment provides a hypersonic vehicle swarm cooperative guidance system based on neural networks, including:

[0147] The information acquisition module 801 is used to acquire the status information of each aircraft; the aircraft include offensive missiles and interceptor missiles; the status information of the offensive missiles is real-time status information; the status information of the interceptor missiles includes the status information before the interceptor missile engine shuts down and the status information at the end of the interception phase.

[0148] The motion model construction module 802 is used to construct motion models of the offensive and interceptor missiles based on kinematic principles; the motion models are used to describe the relative motion relationship between the offensive and interceptor missiles.

[0149] The neural network construction module 803 is used to construct an interception point prediction neural network based on the distance between the interceptor and the attacking missile, the lead angle of the attacking missile, the lead angle of the interceptor, the velocity of the attacking missile, and the velocity of the interceptor at the moment the interceptor's engine is shut down. The output of the interception point prediction neural network is the coordinates of the interception point. The distance between the interceptor and the attacking missile, the lead angle of the attacking missile, the lead angle of the interceptor, the velocity of the attacking missile, and the velocity of the interceptor at the moment the interceptor's engine is shut down are determined based on real-time status information. The distance between the interceptor and the attacking missile, the lead angle of the attacking missile, the lead angle of the interceptor, the velocity of the attacking missile, and the velocity of the interceptor at the moment the interceptor's engine is shut down are determined based on the status information of each aircraft.

[0150] The cooperative guidance module 804 is used to generate cooperative guidance commands for the attack missile cluster based on the cooperative penetration guidance strategy and consistency guidance law of the attack missile cluster, and on the coordinates of the interception point predicted by the interception point prediction neural network. The cooperative penetration guidance strategy refers to selecting one attack missile in the attack missile cluster as a sacrificial missile to attract the attack target of the interceptor missiles, while the remaining attack missiles change course to break through the interceptor missiles' defense system. The consistency guidance law includes an initial maneuver phase, a consistency control maneuver phase, and a maneuver escape phase. The initial maneuver phase is used to control the attack missiles to concentrate towards the central axis of the lateral maneuver corridor. The consistency control maneuver phase is used to calculate the remaining seeker activation time of the attack missiles based on the interception point predicted by the neural network, and to make the remaining seeker activation time of each attack missile tend to be consistent. The maneuver escape phase is used to select a sacrificial missile among the attack missiles to activate the decoy and control the remaining attack missiles to escape with maximum overload.

[0151] In summary, this application has the following technical effects:

[0152] This application addresses the challenges of small swarm size, high reliance on enemy information, and limited practical application in hypersonic vehicle swarm cooperative penetration techniques. It proposes a cooperative swarm penetration strategy combining interception point prediction and consistency control. First, this application implements interception point and remaining time prediction based on a neural network. The neural network only requires parameters of the interceptor missile's engine shutdown moment as input, and its output is used to predict the mid-course interception point and remaining flight time. Second, this application proposes a guidance law based on consistency control theory, achieving consistency control over the remaining time. Furthermore, by designing a strategy of using sacrificial missiles to lure interceptor missiles, the success rate of attack missile swarm penetration is significantly improved.

[0153] The method provided in this application only needs to detect the information of the interceptor missile cluster before the interceptor missile engine is shut down and after the distance between the two clusters is reduced to the point where the seeker of the interceptor missile is turned on. In the mid-course guidance when the distance between the two sides is far and it is difficult to detect enemy information, there is no need to use any real-time information of the enemy interceptor missile cluster, including position, speed, etc., which greatly improves the feasibility of this application in practical application scenarios.

[0154] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0155] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A cooperative guidance method for hypersonic vehicle swarms based on neural networks, characterized in that, include: Acquire the status information of each aircraft; the aircraft include offensive missiles and interceptor missiles; the status information of the offensive missiles is real-time status information. The status information of the interceptor missile includes the status information before the interceptor missile engine shuts down and during the terminal interception phase; Based on kinematic principles, a motion model is constructed for the offensive and interceptor missiles; this motion model describes the relative motion relationship between the offensive and interceptor missiles. An interception point prediction neural network is constructed based on the distance between the interceptor and the attacking missile, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity at the moment the interceptor's engine is shut down. The output of the interception point prediction neural network is the coordinates of the interception point. The distance between the interceptor and the attacking missile, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity at the moment the interceptor's engine is shut down are determined based on real-time status information. The distance between the interceptor and the attacking missile, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity at the moment the interceptor's engine is shut down are determined based on the status information of each aircraft. Based on the coordinated penetration guidance strategy and consistency guidance law of the offensive missile cluster, coordinated guidance commands are generated for the offensive missile cluster based on the coordinates of the interception point predicted by the interception point prediction neural network. The coordinated penetration guidance strategy refers to selecting one offensive missile in the offensive missile cluster as a sacrificial missile to attract the attack target of the interceptor missiles, while the remaining offensive missiles change course to break through the interceptor missiles' defense system. The consistency guidance law includes an initial maneuver phase, a consistency control maneuver phase, and a maneuver escape phase. The initial maneuver phase is used to control the offensive missiles to concentrate towards the central axis of the lateral maneuver corridor. The consistency control maneuver phase is used to calculate the remaining seeker activation time of the offensive missiles based on the interception point predicted by the neural network, and to make the remaining seeker activation time of each offensive missile tend to be consistent. The maneuver escape phase is used to select a sacrificial missile among the offensive missiles to activate the decoy and control the remaining offensive missiles to escape with maximum overload.

2. The method for cooperative guidance of hypersonic vehicle swarms based on neural networks according to claim 1, characterized in that, The formula for the motion model is as follows: In the formula, q i and r i Indicates the i-th interceptor missile M i and the i-th attack missile A i The line-of-sight angle between the projectile and the target, and the relative distance between the projectile and the target, θ Mi ,η Mi They are interceptor missiles M i The velocity direction angle and the lead angle, θ Ti ,η Ti These are offensive missiles A. i velocity direction angle and lead angle, a Ai It is the i-th attack missile A i normal acceleration, V Mi and V Ai These are the i-th interceptor missiles M i With the i-th attack missile A i speed, D Mi and D Ai These are the i-th interceptor missiles M i With the i-th attack missile A i The drag acceleration, r i ,q i ,θ Ai V Mi V Ai The first derivative with respect to time t.

3. The method for cooperative guidance of hypersonic vehicle swarms based on neural networks according to claim 2, characterized in that, Based on the distance between the interceptor and the attacking missile at the moment the interceptor's engine shuts down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity, an interception point prediction neural network is constructed, specifically including: Set the input variables for the interception point prediction neural network. in y Ai ,y Mi , respectively represent the distance between interceptor i and attacking missile i at the moment the interceptor missile engine is shut down, the leading angle of attacking missile i, the leading angle of interceptor i, the y-coordinate of attacking missile i, and the y-coordinate of interceptor i; These are the velocities of the attack missile i and the interceptor missile i, respectively. The target constraint for the interception point prediction neural network is the remaining seeker startup time. Consistency control; among which, t represents the activation time of the guidance system of interceptor missile i, where t is the current time. Set the activation function of the interception point prediction neural network as follows: Set the output layer function of the interception point prediction neural network as: purelin(x) = x; The Levenberg-Marquardt algorithm is used as the backpropagation algorithm for the interception point prediction neural network to construct the interception point prediction neural network.

4. The method for cooperative guidance of hypersonic vehicle swarms based on neural networks according to claim 3, characterized in that, After constructing an interception point prediction neural network based on the distance between the interceptor and the attacking missile at the moment the interceptor's engine shuts down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity, the following steps are also included: Based on the sample dataset, the interception point prediction neural network is trained to obtain the trained interception point prediction neural network; the sample dataset includes real-time status information of interceptor missiles and attack missiles in multiple historical interception scenarios and the corresponding interception point coordinates.

5. The method for cooperative guidance of hypersonic vehicle swarms based on neural networks according to claim 4, characterized in that, The initial maneuver phase is used to control the offensive missiles to converge toward the central axis of the lateral maneuver corridor, specifically including: According to the formula Control the offensive missiles to concentrate along the central axis of the lateral maneuver corridor; among them, K py and K dy If y is a constant, then y = y c Represents the central axis of the lateral mobility corridor, sgn(·) is the sign function, sgn(u Ai ) = 1 indicates that the overload direction is θ Ai The direction of increase, u Ai Indicates overload; When all the aforementioned offensive missiles are concentrated to satisfy max(y) Ai )-min(y Ai )<y s1 At that time, the guidance law is switched to the consistent maneuver phase; where y s1 It is a constant.

6. The method for cooperative guidance of hypersonic vehicle swarms based on neural networks according to claim 5, characterized in that, The consistency control maneuver segment is used to calculate the remaining seeker activation time of the attack missiles based on the interception point predicted by the neural network, and to make the remaining seeker activation time of each attack missile more consistent, specifically including: According to the formula Calculate the remaining seeker activation time of the offensive missile. Where, x pi ,y pi These are the predicted interception point coordinates for the interceptor missile.

7. The method for cooperative guidance of hypersonic vehicle swarms based on neural networks according to claim 6, characterized in that, Calculating the remaining seeker activation time of the offensive missile Following that, it also includes: Based on the remaining seeker activation time of each of the aforementioned offensive missiles And the maximum value of the remaining seeker opening time among the remaining seeker opening times. Sure and The difference When the difference is greater than a set threshold, according to the formula Calculate the guidance status flag variables; where tgo c >0, a>0 are constants, y c1 and y c2 It is a constant, and satisfies 0 < y c1 <y c2 ; Based on the guidance status flag variable, according to formula u Ai =sgn(u Ai )·|u Ai |Calculate offensive missile A i Overload; among them, K θ ,K t ,K dt All are constants.

8. The method for cooperative guidance of hypersonic vehicle swarms based on neural networks according to claim 7, characterized in that, The maneuvering escape phase is used to select a sacrificial missile from among the attack missiles to activate the decoy, and to control the remaining attack missiles to escape with maximum overload, specifically including: The field of view angle of the interceptor missile's seeker is ±σ; For the offensive missile k, if at t l At any given moment, if the attacking missile k is within the field of view of the seeker of the interceptor missile i, then it is determined that the interceptor missile i can be lured by the attacking missile k, and the set of lurable interceptor missiles is obtained as follows: for: Based on the set of luring interceptor missiles, and according to the principle of sacrificial missile selection, the sacrificial missiles in the set of luring interceptor missiles are determined; where x Ak (t l ), y Ak (t l ) for t l The coordinates of the attack missile k at any given moment, x Mi (t l ), y Mi (t l ) for t l The coordinates of intercepting bullet i at all times; According to the formula Calculate the overload of the remaining offensive missiles; among them, This represents the maximum overload value of the offensive missile.

9. A method for cooperative guidance of hypersonic vehicle swarms based on neural networks according to claim 8, characterized in that, The selection principles for sacrificial bullets specifically include: For each attacking missile in the set of lurable interceptor missiles, determine whether the attacking missile is within the field of view of the interceptor missile seeker, and determine the number of interceptor missiles within the field of view of the interceptor missile seeker. The offensive missile with the largest number of interceptor missiles within the field of view of the interceptor missile's seeker is selected as the sacrificial missile; If multiple attack missiles meet the conditions for a sacrificial missile, then among the attack missiles that meet the conditions, the one that is furthest from the interceptor missile is selected as the sacrificial missile.

10. A hypersonic vehicle swarm cooperative guidance system based on neural networks, characterized in that, include: The information acquisition module is used to acquire the status information of each aircraft, including offensive missiles and interceptor missiles; the status information of the offensive missiles is real-time status information; the status information of the interceptor missiles includes status information before the interceptor missile's engine shuts down and during the terminal interception phase. The motion model construction module is used to construct motion models of the offensive and interceptor missiles based on kinematic principles; the motion models are used to describe the relative motion relationship between the offensive and interceptor missiles. A neural network construction module is used to construct an interception point prediction neural network based on the distance between the interceptor and the attacking missile at the moment the interceptor's engine is shut down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity. The output of the interception point prediction neural network is the coordinates of the interception point. The distance between the interceptor and the attacking missile at the moment the interceptor's engine is shut down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity are determined based on real-time status information. The distance between the interceptor and the attacking missile at the moment the interceptor's engine is shut down, the attacking missile's lead angle, the interceptor's lead angle, the attacking missile's velocity, and the interceptor's velocity are determined based on the status information of each aircraft. The cooperative guidance module is used to generate cooperative guidance commands for the attack missile cluster based on the cooperative penetration guidance strategy and consistency guidance law of the attack missile cluster, and on the coordinates of the interception point predicted by the interception point prediction neural network. The cooperative penetration guidance strategy refers to selecting one attack missile in the attack missile cluster as a sacrificial missile to attract the attack target of the interceptor missiles, while the remaining attack missiles change course to break through the interceptor missiles' defense system. The consistency guidance law includes an initial maneuver phase, a consistency control maneuver phase, and a maneuver escape phase. The initial maneuver phase is used to control the attack missiles to concentrate towards the central axis of the lateral maneuver corridor. The consistency control maneuver phase is used to calculate the remaining seeker activation time of the attack missiles based on the interception point predicted by the neural network, and to make the remaining seeker activation time of each attack missile tend to be consistent. The maneuver escape phase is used to select a sacrificial missile among the attack missiles to activate the decoy and control the remaining attack missiles to escape with maximum overload.

Images