CN116222310B - Two-pair synchronous region coverage interception method based on RBF_G in three-dimensional space - Google Patents
Two-pair synchronous region coverage interception method based on RBF_G in three-dimensional space Download PDFInfo
- Publication number
- CN116222310B CN116222310B CN202310393166.3A CN202310393166A CN116222310B CN 116222310 B CN116222310 B CN 116222310B CN 202310393166 A CN202310393166 A CN 202310393166A CN 116222310 B CN116222310 B CN 116222310B
- Authority
- CN
- China
- Prior art keywords
- missile
- time
- target
- interception
- rbf
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000001360 synchronised effect Effects 0.000 title claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 53
- 238000004364 calculation method Methods 0.000 claims abstract description 13
- 230000001133 acceleration Effects 0.000 claims description 40
- 238000013528 artificial neural network Methods 0.000 claims description 32
- 239000013598 vector Substances 0.000 claims description 25
- 239000011159 matrix material Substances 0.000 claims description 20
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 12
- 230000002068 genetic effect Effects 0.000 claims description 10
- 238000004088 simulation Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000035772 mutation Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 14
- 230000007123 defense Effects 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F41—WEAPONS
- F41H—ARMOUR; ARMOURED TURRETS; ARMOURED OR ARMED VEHICLES; MEANS OF ATTACK OR DEFENCE, e.g. CAMOUFLAGE, IN GENERAL
- F41H11/00—Defence installations; Defence devices
- F41H11/02—Anti-aircraft or anti-guided missile or anti-torpedo defence installations or systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Aviation & Aerospace Engineering (AREA)
- Aiming, Guidance, Guns With A Light Source, Armor, Camouflage, And Targets (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A two-to-one synchronous region coverage interception method based on RBF_G in a three-dimensional space belongs to the field of multi-to-one missile synchronous interception. The invention solves the problems that different maneuvering levels and types of the target are not considered and the randomness of normal overload of the target is not considered in the existing synchronous interception method. According to the method, firstly, a calculation method of target interception time by a missile in a three-dimensional space is provided, secondly, a training data set is generated for training and generating an RBF_G network, and then a proportional guidance strategy of a variable proportional coefficient is provided based on a proportional guidance rate, so that an interceptor is allowed to intercept a maneuvering target at expected interception time, even if the target adopts maneuvering of different levels and types, normal overload of the target is a random fixed value, and the missile can realize two-to-one synchronous region coverage interception through the expected interception time and a current time error. The method can be applied to the synchronous interception of the two-to-one guided missiles.
Description
Technical Field
The invention belongs to the field of synchronous interception of many-to-one missiles, and particularly relates to a two-to-synchronous area coverage interception method based on RBF_G in a three-dimensional space.
Background
For the synchronous time interception problem, many scholars have conducted intensive research before, a new guidance rate is designed to give a calculation method (S.R.Kumar and D.Mukherjee,"Terminal time-constrained nonlinear interception strategies against maneuvering targets,"Journal of Guidance,Control,and Dynamics,vol.44,no.1,pp.200–209,2021.[Online].Available:https://doi.org/10.2514/1.G005455). of interception time, a method based on forward time prediction is proposed to solve the interception problem of the marine-air anti-ship missile, a time estimation scheme based on the guidance strategy proposed by the scheme can effectively realize attack on an accurate target (M-J.Tahk,S.-W.Shim,S.-M.Hong,H.-L.Choi and C.-H.Lee,"Impact Time Control Based on Time-to-Go Prediction for Sea-Skimming Antiship Missiles,"in IEEE Transactions on Aerospace and Electronic Systems,vol.54,no.4,pp.2043-2052,Aug.2018). in a specified time and consider the proportional guidance rate of resistance, a calculation problem (B.Zhang,D.Zhou and C.Shao,"Closed-Form Time-to-Go Estimation for Proportional Navigation Guidance Considering Drag,"in IEEE Transactions on Aerospace and Electronic Systems,vol.58,no.5,pp.4705-4717,Oct.2022,doi:10.1109/TAES.2022.3164863.). of the residual time when the missile flies is proposed to solve the problem (Rusnak I.Optimal guidance laws with uncertain time-of-flight[J].IEEE Transactions on Aerospace and Electronic Systems,2000,36(2):721-725.). of uncertain flying time, a recursion-based time estimation method is proposed to solve the estimation problem of the residual flying time of the missile, and a recursion time calculation method for updating the flying time in a non-iterative manner is proposed. The recursive method includes an error compensation feature that explicitly calculates the dead time error (Min-JeaTahk,Chang-Kyung Ryoo and Hangju Cho,"Recursive time-to-go estimation for homing guidance missiles,"in IEEE Transactions on Aerospace and Electronic Systems,vol.38,no.1,pp.13-24,Jan.2002,doi:10.1109/7.993225.). resulting from a non-zero initial heading error, in document (C.Y.Wang,X.J.Ding,J.N.Wang,and J.Y.Shan,"A robust threedimensional cooperative guidance law against maneuvering target,"Journal of the Franklin Institute,vol.357,no.10,pp.5735-5752,Jul.2020.), using the relative distance divided by the relative velocity, and in document (J.Zhao,R.Zhou,and Z.N.Dong,"Three-dimensional cooperative guidance laws against stationary and maneuvering targets,"Chinese Journal of Aeronautics,vol.28,no.4,pp.1104-1120,Apr.2015.), using the relative distance, the velocities of the target and missile, and the heading angle of the missile. In many cases, neither of the above estimation methods can accurately estimate the advance time, so that the missile cannot attack the target at the same time under CPN's law.
Many scholars have made many researches on many-to-one interception strategies in recent years, an optimal strategy is proposed for solving non-collision game in three-dimensional space, in the scheme, a 2-to-1 interception problem (Garcia E,Casbeer D W,Pachter M.Optimal Strategies for a Class of Multi-Player Reach-Avoid Differential Games in 3D Space[J].IEEE Robotics and Automation Letters,2020,PP(99):1-1.). based on an optimal state feedback strategy is proposed for solving the three-dimensional space, a two-to-one pursuit strategy is proposed for differential game (Garcia E,Fuchs Z E,Milutinovic D,et al.A Geometric Approach for the Cooperative Two-Pursuer One-Evader Differential Game[J].IfacPapersonline,2017,50(1):15209-15214.). of two pursuers and one evacuator, in the scheme, the position relation between the two pursuers and the one evacuator is mainly analyzed, and according to the solution of the HJI equation in the differential game, a new non-motorized plane defense rule of an optimal motion equation (M.Pachter,A.Von Moll,E.García,D.Casbeer,and D.Milutinovi′c,"Twoon-one pursuit,"Journal of Guidance,Control,and Dynamics,vol.42,pp.1–7,02 2019.). of the tracker and the gambler is finally proposed for solving the cooperative interception problem between the pursuers, and a cooperative navigation strategy based on the coverage range control flying path angle is proposed for solving the problem of coverage interception based on the idea of dynamic game, which is mainly based on the problem (Sinha,Nandan,Kumar,et al.A New Guidance Law for the Defense Missile of NonmaneuverableAircraft[J].IEEE transactions on control systems technology:A publication of the IEEE Control Systems Society,2015,23(6):2424-2431.), of the optimal trajectory in real time (Bolun Zhang,Di Zhou,Junlong Li,and Yuhan Yao Coverage-Based Cooperative Guidance Strategy by Controlling Flight Path Angle,Journal of Guidance,Control,and Dynamics 2022 45:5,972-981).
In terms of stability analysis, a lyapunov candidate function proposed to analyze the stability (Ren X,Lewis F L,Zhang J.Neural network compensation control for mechanical systems with disturbances[J].Automatica,2009,45(5):1221-1226.). of a neural network in a control system, a lyapunov function proposed to analyze the range of values of the weighted norms of the neural network in the presence of disturbances, is limited by the parameters of the designed system. A new RBF network is proposed, wherein the center of the RBF network is selected based on a three-dimensional model of an initial missile by a i 0 norm selection method (Wang H,Shi Z,Wong H T,et al.An l0-Norm-Based Centers Selection for Failure Tolerant RBF Networks[J].IEEE Access,2019,7:151902-151914.)., and a Lyapunov-like method is proposed for analyzing the performance of pure PNG manufacturing conductivity in a three-dimensional space in interception.
In summary, in view of the existing work discussed at present, the factor that different maneuvering levels and types exist in the target are not considered in the existing synchronous interception method, and the normal overload of the target is set to a fixed value, that is, the randomness of the normal overload of the target is not considered, so that the application effect of the existing synchronous interception method in practice needs to be further improved.
Disclosure of Invention
The invention aims to solve the problems that different maneuvering levels and types of targets are not considered and the randomness of normal overload of the targets is not considered in the existing synchronous interception method, and provides a two-pair synchronous area coverage interception method based on RBF_G in a three-dimensional space.
The technical scheme adopted by the invention for solving the technical problems is as follows:
The method for intercepting the coverage of two pairs of synchronous areas based on RBF_G in the three-dimensional space comprises the following steps:
Firstly, establishing a missile interception model in a three-dimensional space, namely establishing a relative kinematics equation of two missiles and a target in the three-dimensional space;
step two, designing a calculation method of target interception time of the missile in the three-dimensional space;
step three, constructing a training set of the RBF_G neural network based on the relative kinematics equation of the step one, and training the RBF_G neural network by utilizing the constructed training set;
and step four, obtaining time deviation according to the synchronous interception time output by the trained RBF_G neural network and the interception time calculated in the step two, then giving a guidance law of synchronous interception through the controller, substituting the time deviation into the guidance law to obtain the normal acceleration of the missile, and giving a final guidance instruction according to the normal acceleration.
The beneficial effects of the invention are as follows:
According to the method, firstly, a calculation method of target interception time by a missile in a three-dimensional space is provided, secondly, a training data set is generated for training and generating an RBF_G network, and then a proportional guidance strategy of a variable proportional coefficient is provided based on a proportional guidance rate, so that an interceptor is allowed to intercept a maneuvering target at expected interception time, even if the target adopts maneuvering of different levels and types, normal overload of the target is a random fixed value, and the missile can realize two-to-one synchronous region coverage interception through the expected interception time and a current time error.
Drawings
FIG. 1 is a block diagram of the overall design of the method of the present invention;
Wherein I represents an input layer, R_B represents a radial basis function layer, L represents a linear layer, and O represents an output layer; regularizing and anti-regularizing ballistic data to prevent excessive training errors caused by too large data phase differences;
FIG. 2 is a schematic view of missile interception in three dimensions;
In the figure, (X I,YI,ZI) represents an inertial reference coordinate system, (X M,YM,ZM) represents an projectile coordinate system, a M represents missile acceleration, a T represents target acceleration, gamma T represents an included angle between normal acceleration and y axis in a speed coordinate system of a target, gamma M represents an included angle between normal acceleration and y axis in a speed coordinate system of the missile, theta L represents elevation angle between a P-E vision line system and the inertial reference coordinate system, Representing the inclination angle between the P-E vision line system and the inertial reference coordinate system;
The inertial reference coordinate system takes the mass center of the earth as an origin, takes the straight spring point of the equatorial plane as an x-axis, takes the rotation axis of the earth as a y-axis, and takes the equatorial plane vertical to the x-axis as a z-axis;
The missile body coordinate system takes the mass center of the missile as an origin, takes the symmetry axis of the missile body shell, points to the head of the missile as an x axis, takes the longitudinal symmetry plane of the missile as a y axis and takes the longitudinal symmetry plane of the missile as a z axis;
The speed coordinate system of the target takes the mass center of the target as an origin, takes the speed direction as an x-axis, and a y-axis is positioned on the symmetry plane of the missile and is perpendicular to the x-axis, and a z-axis is determined according to a right-hand rule;
The speed coordinate system of the missile takes the mass center of the missile as an origin, takes the speed direction as an x-axis, and a y-axis is positioned on the symmetrical plane of the missile and is perpendicular to the x-axis, and a z-axis is determined according to a right-hand rule;
the P-E vision line takes the mass center of the missile as an origin, takes the connecting line of the missile and the target as an x-axis, takes the vertical x-axis as a y-axis according to the right-hand rule, and takes the vertical x-axis as a z-axis according to the right-hand rule;
FIG. 3 is a flow chart for generating training data;
FIG. 4 is a two-to-one coverage intercept schematic;
FIG. 5 is a model diagram of an RBF_G neural network;
FIG. 6 is a control block diagram of the guidance system;
FIG. 7 is a screenshot of partial data of an input dataset;
FIG. 8 is a second screenshot of partial data of an input dataset;
FIG. 9 is a screenshot of partial data of an output dataset;
FIG. 10 is a graph of training error versus three algorithms;
FIG. 11 is a graph of prediction accuracy versus three algorithms;
FIG. 12 is a schematic diagram of synchronized area coverage interception;
FIG. 13 is a time error diagram of sync area interception;
FIG. 14 is a schematic diagram of unsynchronized area coverage interception;
Fig. 15 is a time error diagram of asynchronous area interception.
Detailed Description
Detailed description of the inventionin the first embodiment, this embodiment will be described with reference to fig. 1 and 5. The method for intercepting two pairs of synchronous region coverage based on RBF_G in the three-dimensional space specifically comprises the following steps:
Firstly, establishing a missile interception model in a three-dimensional space, namely establishing a relative kinematics equation of two missiles and a target in the three-dimensional space;
step two, designing a calculation method of target interception time of the missile in the three-dimensional space;
step three, constructing a training set of the RBF_G neural network based on the relative kinematics equation of the step one, and training the RBF_G neural network by utilizing the constructed training set;
And step four, as shown in fig. 6, obtaining time deviation according to the synchronous interception time output by the trained RBF_G neural network and the interception time calculated in the step two, then giving a guidance law of synchronous interception through the controller, substituting the time deviation into the guidance law to obtain normal acceleration of the missile, and giving a final guidance instruction according to the normal acceleration.
By generating a large amount of missile simulated flight data, the RBF_G neural network is utilized to predict the estimated time of missile interception targets, energy consumption and maximum normal overload. Finally, a set of guidance system is designed to realize the area coverage interception of the possible acceleration of the target projectile. In practical application, the synchronous interception time of the missile to the target is predicted through the RBF_G neural network, and then the time deviation is generated by making a difference with the calculated interception time. Guidance instructions may be generated based on the time offset, which may be regenerated with each iteration update. Finally, the coverage interception of the missile to the target area is realized, and the experimental result shows that the time deviation of each missile reaching the preset position is small enough, so that the coverage interception of the missile synchronous area can be realized.
The second embodiment will be described with reference to fig. 2. The first difference between the present embodiment and the specific embodiment is that the relative kinematic equation of the two missiles and the target in the three-dimensional space is:
assuming the missile and target are point masses, the autopilot and seeker dynamics of the missile are sufficiently fast to be ignored. Further assume that the velocity of the missile and target is constant and that the angle of attack is sufficiently small to be negligible.
Wherein,Is the first derivative of r i, r i is the distance between the ith missile and the target, v T is the velocity vector of the target, θ T is the elevation angle between the velocity coordinate system of the target and the P-E line of sight,Is the tilt angle between the speed coordinate system of the target and the P-E line of sight, i=1, 2, v Mi is the speed vector of the ith missile, θ Mi is the elevation angle between the speed coordinate system of the ith missile and the P-E line of sight,Is the inclination angle between the speed coordinate system of the ith missile and the P-E sight system, q yMi is the sight elevation angle of the ith missile, q zMi is the sight inclination angle of the ith missile,Is the first derivative of q yMi,Is the first derivative of q zMi;
Is the first derivative of θ Mi, A zMi is the z-axis acceleration component of the ith missile in the own speed coordinate system, A yMi is the y-axis acceleration component of the ith missile in the own speed coordinate system,/> Is the first derivative of θ T, A zT is the z-axis acceleration component of the target in its own velocity coordinate system,IsA yT is the y-axis acceleration component of the target in its own velocity coordinate system.
Other steps and parameters are the same as in the first embodiment.
And a third specific embodiment: the difference between this embodiment and the first or second embodiment is that the specific process of the second step is:
Calculating attack time of the missile along a straight line:
wherein t goL represents attack time of the missile along a straight line, Δy is the relative distance between the missile and the target in the y-axis direction of the inertial reference frame, v M represents the speed of the missile, a ty is the projection of the target acceleration a t in the y-axis direction of the target speed frame, n= 2*v T ×Δx, and Δx is the relative distance between the missile and the target in the x-axis direction of the inertial reference frame;
In the time t goL, the target moves to a virtual position TF V under a speed coordinate system, and the virtual position TF under an inertial reference system and the terminal speed V t of the target are calculated through a conversion matrix;
converting the speed of the missile into a virtual P-E vision system to obtain an included angle sigma MF between the speed of the missile and the P-E vision system, wherein the time for the missile to reach a virtual point is as follows:
Wherein, the absolute value is represented, alpha= - (N-2) (N-1), N is a proportionality coefficient, the value is between 2 and 6, Yta is the lift-drag coefficient of missile,R represents the distance between the missile and the target;
Where Δt go is the extension of time due to the proportional-derivative arc, V mt2＝vM*eΞ, e is the base of the natural logarithm, xi= yta x N x sign (σ MF)*(0-σMF)/N-1, sign (·) is the sign function, V t is the terminal velocity of the target, Is the terminal angle at the interception moment, t d＝tgoF-tgoL;
The interception time t goS is:
tgoS＝tgoL+Δtgo
Other steps and parameters are the same as in the first or second embodiment.
For both missile 1 and missile 2, the method of the present embodiment is used to calculate the target interception time, for example, when missile 1 is calculated, the corresponding parameters are substituted into the data of missile 1.
The specific embodiment IV is as follows: this embodiment will be described with reference to fig. 3. The difference between this embodiment and one to three embodiments is that the training set for constructing the rbf_g neural network includes the following specific procedures:
step 1, setting a maximum cycle number Num, a stop trajectory simulation flag bit stop_flag, an initial range of missile operation parameters and an initial range of target operation parameters;
Step 2, initializing an initial position of a target and a missile, an elevation angle of a sight angle, a deflection angle of the sight angle, an initial speed, an initial acceleration, an initial distance and estimated interception time; during initialization, taking a value according to the initial range given in the step 1;
Step 3, calculating the maximum interception time max (t go) and the minimum interception time min (t go) of the missile by adopting the method of the step two for the initialization parameters of the step 2;
Step 4, assigning the minimum interception time min (t go) to the interception time t go;
Step 5, judging whether the interception time t go is smaller than the maximum interception time max (t go), if t go is smaller than the maximum interception time max (t go), executing step 6, otherwise returning to step 2;
Step 6, generating ballistic data according to the initialization parameters in the step 2 and the kinematics equation established in the step one, judging whether the generated ballistic data meets d 1 <1 and d 2 <1, if yes, executing the step 7, otherwise executing the step 19;
Wherein d 1 represents the distance between the target and the 1 st missile after the interception is completed, and d 2 represents the distance between the target and the 2 nd missile after the interception is completed;
Step 7, judging whether the ballistic data meets the following conditions: if the maximum normal acceleration a max is less than 100, executing the step 8, and if the maximum normal acceleration a max is not satisfied, executing the step 19;
Step 8, judging whether the ballistic data meets the following conditions: r1 is less than 100 and R2 is less than 100, if yes, executing step 9, and if not, executing step 19;
wherein R1 represents the initial distance between the 1 st missile and the target, and R2 represents the initial distance between the 2 nd missile and the target;
Step 9, judging whether the ballistic data meets the following conditions: the time_error is less than 0.5, the time_error represents the difference value of the two missile interception times, if the difference value is satisfied, the step 10 is executed, and if the difference value is not satisfied, the step 11 is executed;
Step 10, calculating an overall numerical value: total = weight × energy + (1-weight) a max, weight representing the weight, energy representing the sum of the energy consumed by the missile and the maximum normal overload; total has no actual physical meaning, is an intermediate value;
If Total is less than Total min, then let Total min =total, save and output ballistic data (i.e. write=1), and then execute step 19; wherein Total min represents the minimum before the current iteration;
If Total is more than or equal to Total min, directly executing the step 19;
Step 11, let time_up=max (t go),Time_low＝min(tgo), allocate the interception Time Tgo1 of the 1 st missile as time_low, and allocate the interception Time Tgo2 of the 2 nd missile as time_up;
step 12, generating ballistic data according to the relative kinematic equation of the missile and the target in the step one, and judging whether Tgo < max (t go);
If Tgo is less than max (t go), then step 13 is performed;
If Tgo is not less than max (t go), then step 20 is performed;
Step 13, assigning the interception Time Tgo of the 2 nd missile as time_up;
step 14, judging whether the interception Time Tgo of the 2 nd missile is greater than time_low, if Tgo is greater than time_low, executing step 15, otherwise executing step 17;
Step 15, judging whether the generated ballistic data meets d 1 <1 and d 2 <1, if yes, executing step 16, otherwise executing step 18;
Step 16, judging whether the ballistic data meets the following conditions: the maximum normal acceleration a max is less than 100, R1 is less than 100 and R2 is less than 100;
If the maximum normal acceleration a max is less than 100, R1 is less than 100 and R2 is less than 100, calculating the overall value: total=weight =energy+ (1-weight) a max, weight represents weight, energy represents energy; if Total is less than Total min, letting Total min =total, storing and outputting ballistic data, and then executing step 18, wherein Total min represents the minimum value before the current iteration; if Total is greater than or equal to Total min, directly executing the step 18;
If the maximum normal acceleration a max is less than 100, R1 is less than 100 and R2 is less than 100, executing the step 18;
step 17, tgo 1= Tgo1+dt, and returning to step 12;
step 18, let Tgo 2= Tgo2-dt, return to step 14;
Step 19, re-giving the interception time t go＝tgo +dt, returning to the execution step5, and executing the step 20 while returning to the step 5;
Step 20, judging whether the stored ballistic data are stored in a training set;
If the stored trajectory data is stored in the training set, judging whether the current cycle number turn is smaller than the maximum cycle number Num, if turn is smaller than Num, enabling the cycle number turn=turn+1, and returning to the step 2; if turn=num, then step 21 is performed;
if the stored ballistic data is not stored in the training set, directly returning to the step 2;
And step 21, ending.
The invention performs random on all data within a certain reasonable range, and specifically comprises an initial position, a speed, a sight angle and acceleration. That is, each time the ballistic data is not controllable, which results in a more even distribution of the ballistic in space, and a greater range of consideration. The coverage interception schematic diagram of the two-to-one missile is shown in fig. 4, and two virtual movement points of the target are intercepted by two missiles simultaneously, wherein at the initial moment, other initial parameters except acceleration are consistent. Thus, two missiles can intercept two virtual targets, and synchronous interception of the targets can be realized. If two missiles can intercept two virtual targets with constant acceleration values at the same time, we can default that when the targets fly at one acceleration, the missiles can intercept the targets. Thus, coverage interception of the target can be achieved. The initial aim of the method is achieved, and the problem of interception of the missile is solved.
The parameters initialized in step 2 of this embodiment, i.e., each column of the input dataset are shown in table 1, all missile parameters are capitalized with M1 and M2. All virtual target points are also capitalized with T1 and T2.
TABLE 1 input data
The first virtual target point and the second virtual target point are identical except acceleration at the initial moment, so that the same data in the first virtual target point and the second virtual target point can be directly assigned to the second virtual target point during initialization assignment.
The output data of the training set is shown in table 2:
TABLE 2 output data
For the output data set, when the synchronization flag bit inter =1, interception of the first and second targets can be realized on behalf of the first and second missiles, and the time difference is smaller than the expected time difference. t M1,tM2 is the predicted time of flight. When the synchronization flag bit inter =0, interception of the first and second targets can be realized on behalf of the first and second missiles, but the time difference is larger than the expected time difference. Then t M1,tM2 is the set time of flight for missile one and missile two, respectively. The purpose of this is to minimize this time difference even though the two missiles cannot intercept the virtual target point synchronously. The remaining output parameters are consistent. The simulation environment adopted by the invention is shown in table 3:
TABLE 3 simulation Environment
| Project | Data |
| CPU | E5-2696V4 |
| Memory | 128g |
| Disk capacity | 1T |
| System and method for controlling a system | Win10 professional edition |
In this embodiment, ballistic data is screened according to performance indexes, and the satisfied data is retained:
1. Our aim is to let the missile intercept the target head-on, but if the target cannot be intercepted head-on, i.e. the target is caught head-to-tail, then this is also the intercept trajectory, but not what we expect, so the corresponding data is not preserved.
Here, since "Nan" may occur when the line of sight angle is excessively large, the calculation formula of the line of sight angle is as follows:
as the distance between the missile and the target gets closer and closer, The calculated error is very large, and once the head tracking cannot be realized, the head tracking mode is converted into the tail tracking mode, and the view angle is infinite. Thus, when the proportional conductivity is used, the normal overload tends to be infinite and does not meet the condition of constraint on the normal overload, so that data of Nan is deleted in the ballistic iteration process.
2. If the initial positions of the two missiles are too far apart, synchronous interception cannot be realized. In addition, we hope that both missiles are intercepted under the condition of head-tracking, the tail-tracking time is too long, and the data are not in the range of consideration.
3. Only if the time difference between the two missile interception virtual points is smaller than the given time difference, synchronous interception is considered to be realized, otherwise, the corresponding data is not reserved. For data intercepted asynchronously we will recalculate to give the appropriate time of flight, respectively.
Aiming at the problem of interception of guided segments of missile terminals, we do parameter constraint of table 4:
TABLE 4 setting ranges of simulation parameters
Where N t is a constant that is used to determine the proportionality coefficient between the maximum allowable time error and the desired difference. Next, we give a definition about the time parameter, as shown in table 5.
TABLE 5 definition of time parameters for missile one and missile two
| Definition of time | Missile 1 | Missile 2 |
| Optimum time | tgo | tgo |
| Calculated missile flight time | tgos1 | tgos2 |
| Actual missile flight time | tm1 | tm2 |
Wherein the optimal time is a human given time parameter, and the reference range is a constant value added or subtracted to the calculated time parameters of the two missiles. The calculated missile flight time is given by an algorithm calculated through three-dimensional space interception time. The actual missile flight time is iterated step by step through the program. The maximum allowable time error range is the difference between the actual missile flight times.
By means of tables 4 and 5, the parameter settings and the desired parameter indicators during the simulation are given, so that the data meeting the requirements can be missed, and the data not meeting the requirements can be deleted, for example, when the missile can be intercepted synchronously, but the required normal overload is too large. Such data cannot be practically applied. Also, although interception can be achieved, the difference in interception time exceeds the maximum allowable time error. The final desire is to be able to acquire enough data for training so that the accuracy of the RBF neural network model is higher. When we have a set of initial data we get a suitable interception time, which can greatly reduce the time spent in practical applications.
Other steps and parameters are the same as in one to three embodiments.
Fifth embodiment: the difference between the present embodiment and one to four embodiments is that the training set constructed is used to train the rbf_g neural network, and the specific process is as follows:
initializing parameters of a radial base layer, parameters of a genetic algorithm, linear layer parameters and deflection of the radial base layer, and initializing training times l=1;
step (2), inputting input data into an input layer of the RBF_G neural network;
Step (3), inputting the output of the input layer into a radial base layer, and generating a radial base layer matrix through radial base function calculation;
Step (4), the radial base matrix generated in the step (3) is transmitted into an initial population of a genetic algorithm, and whether the number of the initial population is larger than 2 is judged;
If the initial population number is more than 2, executing the step (5);
Otherwise, let l=l+1, return to step (2);
step (5), calculating the second norms of the training error vectors of each individual in the initial population, taking the individual with the smallest second norms of the training error vectors in the initial population as the parent class of the genetic algorithm, and taking the two Fan Shuci small individuals of the training error vectors in the initial population as the parent class of the genetic algorithm;
Each individual in the population comprises a radial basis matrix And a training error vector e i;
Step (6), selecting each row of the parent class and the parent class to randomly generate different constants N ', and performing cross operation according to the constants N';
step (7), randomly generating a natural number M for the first line of the processing result of the step (6), and then carrying out mutation operation on the M-th element of the first line;
And (3) processing each row of the processing result in the step (6) in turn until each row is processed, and generating a new radial base matrix of the subclass;
step (8), calculating the output of the RBF_G neural network through a linear layer by using the new sub-class radial base matrix generated in the step (7);
Step (9), calculating errors according to the output of the step (8), and adding the new radial base matrix of the subclasses into the initial population of the genetic algorithm;
Step (10), reversely transferring and updating the parameters of the RBF_G neural network according to the error calculated in the step (9);
And (11) judging whether the set maximum iteration number or the error is smaller than a set threshold value, and ending the training process if the set maximum iteration number or the error is smaller than the set threshold value.
Otherwise, let l=l+1, return to step (2).
Other steps and parameters are the same as in one to four embodiments.
Specific embodiment six: the difference between this embodiment and one to fifth embodiments is that in the step (9), the error is calculated according to the output of the step (8), and the specific process is as follows:
The input data vector X of the rbf_g neural network is expressed as:
X＝[x1,x2,x3,...,xn]T
Wherein X 1,x2,x3,...,xn is the 1 st, 2 nd, 3 rd, … th, n-th input data in X;
The output of the rbf_g neural network is:
Wherein y (x i) is the actual output value corresponding to the ith input data, q is the number of hidden layer neurons, ω j is the weight of the jth hidden layer neuron, b j is the bias of the jth hidden layer neuron, For the radial basis function of the j-th hidden layer neuron,Dist (x i-cj) denotes the Euclidean distance between x i and c j, c j is the center of the radial basis function, σ j is the variance;
The error is:
Wherein E i represents a deviation corresponding to the ith input data, y' (x i) represents an actual value of an output corresponding to the ith input data, and m represents the number of neurons of the output.
Other steps and parameters are the same as in one of the first to fifth embodiments.
Seventh embodiment: this embodiment differs from one of the first to sixth embodiments in that the specific process of step (10) is:
Wherein E (n, 1) represents the learning rate of the radial base layer deflection vector b with the number of rows of E being n and the number of columns being 1, e= [ E 1,E2,E3,...,En]T,ηb ] b= [ b 1,b2,b3,...,bq]T, Δb being the deviation of the radial base layer deflection vector;
after the delta b is calculated, summing the element in the delta b and the corresponding element in the deflection vector b before updating to obtain deflection of the updated hidden layer neuron;
wherein, Is a radial basis matrix with the number of rows being q and the number of columns being n,E epsilon (n, 1), eta ω is the learning rate of the weight vector omega, and delta omega is the deviation of the weight vector;
after the delta omega is calculated, summing the elements in the delta omega and the corresponding elements in the weight vector omega before updating to obtain the weight of the updated hidden layer neuron;
wherein, For the learning rate of c j, Δc j is the deviation of c j,Is a radial base matrix with 1 row and n columns;
After Δc j is calculated, Δc j is summed with c j before updating to obtain updated c j;
wherein, For the learning rate of σ j, Δσ j is the deviation of σ j, which represents the 2 norm.
After Δσ j is calculated, Δσ j is summed with σ j before updating to obtain updated σ j.
Other steps and parameters are the same as in one of the first to sixth embodiments.
Eighth embodiment: this embodiment differs from one of the first to seventh embodiments in that the proportionality coefficient of the guidance law is:
Nc＝N*(1-Et*terror*R(t)/R(t0))
wherein, N is a proportionality coefficient adopted when calculating the interception time, t error is the difference between the set interception time and the calculated interception time, R (t) is the distance between the missile and the target at the time t, R (t 0) is the distance between the missile and the target at the initial time, and E t is a constant.
Other steps and parameters are the same as those of one of the first to seventh embodiments.
Detailed description nine: this embodiment differs from one to eight embodiments in that the normal acceleration a M of the missile is:
aM＝Nc*Ω×vM
Wherein Ω represents the P-E line-of-sight angular rate, For the angular velocity in the y-direction of the P-E line of sight,V M represents the missile velocity, which is the angular velocity in the z-direction.
Other steps and parameters are the same as in one to eight of the embodiments.
Detailed description ten: this embodiment differs from one of the first to ninth embodiments in that the guidance system is stable when the guidance law satisfies the condition ① or ②;
Condition ①:
Condition ②:
where e v is the error vector, For the estimated error of the linear weights, · F represents the F-norm, the constant σ > 0, λ v is the minimum eigenvalue of K v, K v is the error control gain,M * is the binary norm of the ideal radial base matrix.
Other steps and parameters are the same as in one of the first to ninth embodiments.
Guidance system stability demonstration:
suppose 1: ideal omega * is bounded and omega * F≤M* is a tight set of omega 1, then one can give WhereinIs the estimation error of the linear weight.
Suppose 2: reference errorWhere the constant lambda c＞0,M＞0,Kv is the given error control gain.
Suppose 3: an adaptive law of parameter ω based on σ correction is given.
Where Γ > 0 is the gain matrix and σ > 0 is the scalar parameter.
Proof, please refer to the literature for a detailed proof procedure assuming 1-3 (Ren X,Lewis F L,Zhang J.Neural network compensation control for mechanical systems with disturbances[J].Automatica,2009,45(5):1221-1226.).
Theorem one: if the guidance system is stable, any one of the following equations needs to be satisfied.
Or
Wherein lambda v is the minimum eigenvalue of K v and epsilon is a small constant andAnd (5) correlation.
And (3) proving: we assume that the lyapunov candidate function
The derivative of V with respect to time can be obtained:
Where lambda v > 0 is the minimum eigenvalue of K v. Then the first time period of the first time period,
Thus, we can obtain
Wherein the method comprises the steps ofIs a constant. If V < 0 is required, we only need to satisfyThen/>, can be obtainedThen
Or So we can rewrite V as
Y is the minimum eigenvalue of M. Kappa is the eigenvalue of Γ minimum.
Because of e v andAre all bounded by upper bounds. Then there is also an upper bound for V. Then for the whole control system, it is only necessary to meet/>, if stable convergence of the system is desired
Or
And (5) finishing the verification.
Simulation results
Through the training set data generating part, 3000 groups of data are finally selected, finally, the generated partial data are shown in fig. 7, 8 and 9, the input data meaning of each column is shown in table 6, the output data meaning of each column is shown in table 4, wherein the training set input data total 20 columns comprise initial information of missiles and targets, the training set output data total 11 columns comprise final missile interception time, energy consumption and maximum normal overload. Wherein the partial data of the input data set is shown and the partial data of the output data set is shown.
TABLE 6 definition of parameters
The training errors of the three algorithms RBF, BP neural network and RBF_G are compared respectively, and as shown in the following figure 10, the training errors of the three algorithms can be obtained, the errors of the three algorithms are large at the beginning of training, but after about 5000 times of training, the training errors of RBF and BP NN are stabilized between 1% and 1.5%, and the training error of RBF_G is less than 0.1%. After 10000 times of data training, the training error of the final RBF is 1.15%, the training error of BP is 1.24%, and the training error of RBF_G is 0.00042%.
Next, we compare the prediction accuracy rates of the three algorithms RBF, BP neural network and rbf_g, and as shown in fig. 11, we can obtain the prediction of the three algorithms, the prediction accuracy rate of BP is 72%, the prediction accuracy rate of RBF is 75%, the prediction accuracy rate of rbf_g is 86%, and the prediction error is mainly caused by the inaccurate energy prediction in the flight phase. But this does not affect our predictions of time of flight, which we mainly use to achieve synchronized coverage interception of missiles to targets in a guidance system.
Finally, a schematic diagram of the synchronous area coverage interception can be realized through the guidance system, as shown in fig. 12, and a two-to-two interception schematic diagram is shown in fig. 12. By means of a small area schematic diagram, we can get the solution proposed by the present invention to achieve area coverage interception, and by means of a time error schematic diagram 13 we can see that the final time error is stable, the estimated flight time is 19.69s. The time of flight of M1 is 20.14s, the time of flight of M2 is 20.63s, and the final time of flight error is 0.49s and less than 0.5s.
For the missile incapable of realizing synchronous interception, the interception time is set respectively, so that the flight time is close enough to achieve the aim of minimum interception time error, as shown in fig. 14, the scheme provided by the invention can realize area coverage interception through a small area schematic diagram, and the final time error is stable and the set flight time is 64.3s and 63.4s respectively as shown in a time error schematic diagram 15. The final M1 time of flight is 64.31s, the final M2 time of flight is 63.64s, and the final time of flight error is 0.67 to greater than 0.5s. Wherein the maximum normal overload of M1 is 31.4647 and-15.2397. The maximum normal overload for M2 is 95.1541 and-80.6831.
The above examples of the present invention are only for describing the calculation model and calculation flow of the present invention in detail, and are not limiting of the embodiments of the present invention. Other variations and modifications of the above description will be apparent to those of ordinary skill in the art, and it is not intended to be exhaustive of all embodiments, all of which are within the scope of the invention.
Claims (7)
1. The method for intercepting the coverage of two pairs of synchronous areas based on RBF_G in the three-dimensional space is characterized by comprising the following steps:
Firstly, establishing a missile interception model in a three-dimensional space, namely establishing a relative kinematics equation of two missiles and a target in the three-dimensional space;
The relative kinematic equation of the two missiles and the target in the three-dimensional space is as follows:
wherein, Is the first derivative of r i, r i is the distance between the ith missile and the target, v T is the velocity vector of the target, θ T is the elevation angle between the velocity coordinate system of the target and the P-E line of sight,Is the tilt angle between the speed coordinate system of the target and the P-E line of sight, i=1, 2, v Mi is the speed vector of the ith missile, θ Mi is the elevation angle between the speed coordinate system of the ith missile and the P-E line of sight,Is the inclination angle between the speed coordinate system of the ith missile and the P-E sight system, q yMi is the sight elevation angle of the ith missile, q zMi is the sight inclination angle of the ith missile,Is the first derivative of q yMi,Is the first derivative of q zMi;
Is the first derivative of θ Mi, A zMi is the z-axis acceleration component of the ith missile in the own speed coordinate system, A yMi is the y-axis acceleration component of the ith missile in the own speed coordinate system,/> Is the first derivative of θ T, A zT is the z-axis acceleration component of the target in its own velocity coordinate system,IsA yT is the y-axis acceleration component of the target in its own velocity coordinate system;
step two, designing a calculation method of target interception time of the missile in the three-dimensional space;
The specific process of the second step is as follows:
Calculating attack time of the missile along a straight line:
wherein t goL represents attack time of the missile along a straight line, Δy is the relative distance between the missile and the target in the y-axis direction of the inertial reference frame, v M represents the speed of the missile, a ty is the projection of the target acceleration a t in the y-axis direction of the target speed frame, n= 2*v T ×Δx, and Δx is the relative distance between the missile and the target in the x-axis direction of the inertial reference frame;
In the time t goL, the target moves to a virtual position TF V under a speed coordinate system, and the virtual position TF under an inertial reference system and the terminal speed V t of the target are calculated through a conversion matrix;
converting the speed of the missile into a virtual P-E vision system to obtain an included angle sigma MF between the speed of the missile and the P-E vision system, wherein the time for the missile to reach a virtual point is as follows:
Wherein, |·| represents the absolute value, α= - (N-2)/(N-1), N is the scaling factor, Yta is the lift-drag coefficient of missile,R represents the distance between the missile and the target;
Where Δt go is the extension of time due to the proportional-derivative arc, V mt2＝vM*eΞ, e is the base of the natural logarithm, xi= yta x N x sign (σ MF)*(0-σMF)/N-1, sign (·) is the sign function, V t is the terminal velocity of the target, Is the terminal angle at the interception moment, t d＝tgoF-tgoL;
The interception time t goS is:
tgoS＝tgoL+Δtgo
step three, constructing a training set of the RBF_G neural network based on the relative kinematics equation of the step one, and training the RBF_G neural network by utilizing the constructed training set;
The specific process of constructing the training set of the RBF_G neural network is as follows:
step 1, setting a maximum cycle number Num, a stop trajectory simulation flag bit stop_flag, an initial range of missile operation parameters and an initial range of target operation parameters;
step 2, initializing an initial position of a target and a missile, an elevation angle of a sight angle, a deflection angle of the sight angle, an initial speed, an initial acceleration, an initial distance and estimated interception time;
Step 3, calculating the maximum interception time max (t go) and the minimum interception time min (t go) of the missile by adopting the method of the step two for the initialization parameters of the step 2;
Step 4, assigning the minimum interception time min (t go) to the interception time t go;
Step 5, judging whether the interception time t go is smaller than the maximum interception time max (t go), if t go is smaller than the maximum interception time max (t go), executing step 6, otherwise returning to step 2;
Step 6, generating ballistic data according to the initialization parameters in the step 2 and the kinematics equation established in the step one, judging whether the generated ballistic data meets d 1 <1 and d 2 <1, if yes, executing the step 7, otherwise executing the step 19;
Wherein d 1 represents the distance between the target and the 1 st missile after the interception is completed, and d 2 represents the distance between the target and the 2 nd missile after the interception is completed;
Step 7, judging whether the ballistic data meets the following conditions: if the maximum normal acceleration a max is less than 100, executing the step 8, and if the maximum normal acceleration a max is not satisfied, executing the step 19;
Step 8, judging whether the ballistic data meets the following conditions: r1 is less than 100 and R2 is less than 100, if yes, executing step 9, and if not, executing step 19;
wherein R1 represents the initial distance between the 1 st missile and the target, and R2 represents the initial distance between the 2 nd missile and the target;
Step 9, judging whether the ballistic data meets the following conditions: the time_error is less than 0.5, the time_error represents the difference value of the two missile interception times, if the difference value is satisfied, the step 10 is executed, and if the difference value is not satisfied, the step 11 is executed;
Step 10, calculating an overall numerical value: total = weight × energy + (1-weight) a max, weight representing the weight, energy representing the sum of the energy consumed by the missile and the maximum normal overload;
If Total is less than Total min, letting Total min =total, storing and outputting ballistic data, and executing step 19; wherein Total min represents the minimum before the current iteration;
If Total is more than or equal to Total min, directly executing the step 19;
Step 11, let time_up=max (t go),Time_low＝min(tgo), allocate the interception Time Tgo1 of the 1 st missile as time_low, and allocate the interception Time Tgo2 of the 2 nd missile as time_up;
step 12, generating ballistic data according to the relative kinematic equation of the missile and the target in the step one, and judging whether Tgo < max (t go);
If Tgo is less than max (t go), then step 13 is performed;
If Tgo is not less than max (t go), then step 20 is performed;
Step 13, assigning the interception Time Tgo of the 2 nd missile as time_up;
step 14, judging whether the interception Time Tgo of the 2 nd missile is greater than time_low, if Tgo is greater than time_low, executing step 15, otherwise executing step 17;
Step 15, judging whether the generated ballistic data meets d 1 <1 and d 2 <1, if yes, executing step 16, otherwise executing step 18;
Step 16, judging whether the ballistic data meets the following conditions: the maximum normal acceleration a max is less than 100, R1 is less than 100 and R2 is less than 100;
If the maximum normal acceleration a max is less than 100, R1 is less than 100 and R2 is less than 100, calculating the overall value: total=weight =energy+ (1-weight) a max, weight represents weight, energy represents energy; if Total is less than Total min, letting Total min =total, storing and outputting ballistic data, and then executing step 18, wherein Total min represents the minimum value before the current iteration; if Total is greater than or equal to Total min, directly executing the step 18;
If the maximum normal acceleration a max is less than 100, R1 is less than 100 and R2 is less than 100, executing the step 18;
step 17, tgo 1= Tgo1+dt, and returning to step 12;
step 18, let Tgo 2= Tgo2-dt, return to step 14;
Step 19, re-giving the interception time t go＝tgo +dt, returning to the execution step5, and executing the step 20 while returning to the step 5;
Step 20, judging whether the stored ballistic data are stored in a training set;
If the stored trajectory data is stored in the training set, judging whether the current cycle number turn is smaller than the maximum cycle number Num, if turn is smaller than Num, enabling the cycle number turn=turn+1, and returning to the step 2; if turn=num, then step 21 is performed;
if the stored ballistic data is not stored in the training set, directly returning to the step 2;
Step 21, ending;
and step four, obtaining time deviation according to the synchronous interception time output by the trained RBF_G neural network and the interception time calculated in the step two, then giving a guidance law of synchronous interception through the controller, substituting the time deviation into the guidance law to obtain the normal acceleration of the missile, and giving a final guidance instruction according to the normal acceleration.
2. The method for intercepting two pairs of synchronous region coverage based on rbf_g in three-dimensional space according to claim 1, wherein the training of rbf_g neural network by using the constructed training set comprises the following specific steps:
initializing parameters of a radial base layer, parameters of a genetic algorithm, linear layer parameters and deflection of the radial base layer, and initializing training times l=1;
step (2), inputting input data into an input layer of the RBF_G neural network;
Step (3), inputting the output of the input layer into a radial base layer, and generating a radial base layer matrix through radial base function calculation;
Step (4), the radial base matrix generated in the step (3) is transmitted into an initial population of a genetic algorithm, and whether the number of the initial population is larger than 2 is judged;
If the initial population number is more than 2, executing the step (5);
Otherwise, let l=l+1, return to step (2);
step (5), calculating the second norms of the training error vectors of each individual in the initial population, taking the individual with the smallest second norms of the training error vectors in the initial population as the parent class of the genetic algorithm, and taking the two Fan Shuci small individuals of the training error vectors in the initial population as the parent class of the genetic algorithm;
Step (6), selecting each row of the parent class and the parent class to randomly generate different constants N ', and performing cross operation according to the constants N';
step (7), randomly generating a natural number M for the first line of the processing result of the step (6), and then carrying out mutation operation on the M-th element of the first line;
And (3) processing each row of the processing result in the step (6) in turn until each row is processed, and generating a new radial base matrix of the subclass;
step (8), calculating the output of the RBF_G neural network through a linear layer by using the new sub-class radial base matrix generated in the step (7);
Step (9), calculating errors according to the output of the step (8), and adding the new radial base matrix of the subclasses into the initial population of the genetic algorithm;
Step (10), reversely transferring and updating the parameters of the RBF_G neural network according to the error calculated in the step (9);
step (11), judging whether the set maximum iteration number or the error is smaller than a set threshold value, and ending the training process if the set maximum iteration number or the error is smaller than the set threshold value;
Otherwise, let l=l+1, return to step (2).
3. The method for intercepting two pairs of synchronous region coverage based on rbf_g in three-dimensional space according to claim 2, wherein in the step (9), the error is calculated according to the output of the step (8), which comprises the following specific steps:
The input data vector X of the rbf_g neural network is expressed as:
X＝[x1,x2,x3,...,xn]T
Wherein X 1,x2,x3,...,xn is the 1 st, 2 nd, 3 rd, … th, n-th input data in X;
The output of the rbf_g neural network is:
Wherein y (x i) is the actual output value corresponding to the ith input data, q is the number of hidden layer neurons, ω j is the weight of the jth hidden layer neuron, b j is the bias of the jth hidden layer neuron, For the radial basis function of the j-th hidden layer neuron,Dist (x i-cj) denotes the Euclidean distance between x i and c j, c j is the center of the radial basis function, σ j is the variance;
The error is:
Wherein E i represents a deviation corresponding to the ith input data, y' (x i) represents an actual value of an output corresponding to the ith input data, and m represents the number of neurons of the output.
4. The method for intercepting two pairs of synchronous area coverage based on rbf_g in three-dimensional space according to claim 3, wherein the specific process of the step (10) is as follows:
wherein e= [ E 1,E2,E3,...,En]T,ηb ] is the learning rate of the radial base layer deflection vector b, b= [ b 1,b2,b3,...,bq]T, Δb is the deviation of the radial base layer deflection vector;
wherein, Is a radial basis matrix with the number of rows being q and the number of columns being n, eta ω is the learning rate of the weight vector omega, and delta omega is the deviation of the weight vector;
wherein, For the learning rate of c j, Δc j is the deviation of c j,Is a radial base matrix with 1 row and n columns;
wherein, For a learning rate of σ j, Δσ j is the deviation of σ j, representing 2 norms.
5. The method for intercepting a two-pair synchronous area coverage based on rbf_g in a three-dimensional space according to claim 4, wherein a scaling factor of said guidance law is:
Nc＝N*(1-Et*terror*R(t)/R(t0))
wherein, N is a proportionality coefficient adopted when calculating the interception time, t error is the difference between the set interception time and the calculated interception time, R (t) is the distance between the missile and the target at the time t, R (t 0) is the distance between the missile and the target at the initial time, and E t is a constant.
6. The method for intercepting a two-pair synchronous area coverage based on rbf_g in three-dimensional space according to claim 5, wherein a normal acceleration a M of said missile is:
aM＝Nc*Ω×vM
wherein Ω represents the P-E line of sight angular velocity and V M represents the missile velocity.
7. The method for intercepting a two-pair synchronous area coverage based on rbf_g in three-dimensional space according to claim 6, wherein a guidance system is stable when said guidance law satisfies a condition ① or ②;
Condition ①:
Condition ②:
where e v is the error vector, For the estimated error of linear weights, |·| F represents the F-norm, the constant σ > 0, λ v is the minimum eigenvalue of K v, K v is the error control gain,M * is the binary norm of the ideal radial base matrix.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310393166.3A CN116222310B (en) | 2023-04-13 | 2023-04-13 | Two-pair synchronous region coverage interception method based on RBF_G in three-dimensional space |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310393166.3A CN116222310B (en) | 2023-04-13 | 2023-04-13 | Two-pair synchronous region coverage interception method based on RBF_G in three-dimensional space |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116222310A CN116222310A (en) | 2023-06-06 |
| CN116222310B true CN116222310B (en) | 2024-04-26 |
Family
ID=86578979
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310393166.3A Active CN116222310B (en) | 2023-04-13 | 2023-04-13 | Two-pair synchronous region coverage interception method based on RBF_G in three-dimensional space |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116222310B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8358238B1 (en) * | 2009-11-04 | 2013-01-22 | Lockheed Martin Corporation | Maneuvering missile engagement |
| CN109284530A (en) * | 2018-08-02 | 2019-01-29 | 西北工业大学 | Space non-cooperative target appearance rail integration method for parameter estimation based on deep learning |
| CN111832121A (en) * | 2020-07-17 | 2020-10-27 | 中国人民解放军火箭军工程大学 | Multi-aircraft cooperative detection and guidance integrated method and system |
| CN112256055A (en) * | 2020-10-12 | 2021-01-22 | 清华大学 | Three-body confrontation defense prediction guidance method adopting fixed point optimization |
| CN112902767A (en) * | 2021-01-28 | 2021-06-04 | 西安交通大学 | Multi-missile time collaborative missile guidance method and system |
| CN114415723A (en) * | 2022-01-11 | 2022-04-29 | 北京科技大学 | Multi-aircraft cooperative capture space division method and device and electronic equipment |
| KR102396924B1 (en) * | 2021-10-19 | 2022-05-12 | 한화시스템 주식회사 | Intercepting method, filtering method and intercepting apparatus |
| CN115329594A (en) * | 2022-08-31 | 2022-11-11 | 哈尔滨工业大学 | Large-scale missile cluster attack and defense confrontation simulation acceleration method and system |
| CN115755969A (en) * | 2022-11-22 | 2023-03-07 | 中国人民解放军海军工程大学 | Aircraft guidance method and device based on zero-control interception flow pattern |
| CN115857548A (en) * | 2022-11-29 | 2023-03-28 | 南京理工大学 | Terminal guidance law design method based on deep reinforcement learning |
-
2023
- 2023-04-13 CN CN202310393166.3A patent/CN116222310B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8358238B1 (en) * | 2009-11-04 | 2013-01-22 | Lockheed Martin Corporation | Maneuvering missile engagement |
| CN109284530A (en) * | 2018-08-02 | 2019-01-29 | 西北工业大学 | Space non-cooperative target appearance rail integration method for parameter estimation based on deep learning |
| CN111832121A (en) * | 2020-07-17 | 2020-10-27 | 中国人民解放军火箭军工程大学 | Multi-aircraft cooperative detection and guidance integrated method and system |
| CN112256055A (en) * | 2020-10-12 | 2021-01-22 | 清华大学 | Three-body confrontation defense prediction guidance method adopting fixed point optimization |
| CN112902767A (en) * | 2021-01-28 | 2021-06-04 | 西安交通大学 | Multi-missile time collaborative missile guidance method and system |
| KR102396924B1 (en) * | 2021-10-19 | 2022-05-12 | 한화시스템 주식회사 | Intercepting method, filtering method and intercepting apparatus |
| CN114415723A (en) * | 2022-01-11 | 2022-04-29 | 北京科技大学 | Multi-aircraft cooperative capture space division method and device and electronic equipment |
| CN115329594A (en) * | 2022-08-31 | 2022-11-11 | 哈尔滨工业大学 | Large-scale missile cluster attack and defense confrontation simulation acceleration method and system |
| CN115755969A (en) * | 2022-11-22 | 2023-03-07 | 中国人民解放军海军工程大学 | Aircraft guidance method and device based on zero-control interception flow pattern |
| CN115857548A (en) * | 2022-11-29 | 2023-03-28 | 南京理工大学 | Terminal guidance law design method based on deep reinforcement learning |
Non-Patent Citations (1)
| Title |
|---|
| 子拦截弹拦截无人机集群防碰撞制导律;罗瑞宁;黄树彩;赵岩;张振;张飞;航空兵器;20230228;第1卷(第1期);51-58 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116222310A (en) | 2023-06-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jiang et al. | Cooperative guidance with multiple constraints using convex optimization | |
| Ming et al. | A novel non-singular terminal sliding mode control-based integrated missile guidance and control with impact angle constraint | |
| Lin et al. | Missile guidance law design using adaptive cerebellar model articulation controller | |
| CN112462792B (en) | Actor-Critic algorithm-based underwater robot motion control method | |
| CN114003050B (en) | Active defense guidance method of three-body countermeasure strategy based on differential game | |
| CN111898201B (en) | High-precision autonomous attack guiding method for fighter in air combat simulation environment | |
| CN114020021B (en) | Design method and system for multi-missile distributed cooperative guidance law | |
| CN115329594B (en) | Large-scale missile cluster attack and defense confrontation simulation acceleration method and system | |
| Lukacs et al. | Trajectory-shape-varying missile guidance for interception of ballistic missiles during the boost phase | |
| CN115047769A (en) | Unmanned combat platform obstacle avoidance-arrival control method based on constraint following | |
| CN116222310B (en) | Two-pair synchronous region coverage interception method based on RBF_G in three-dimensional space | |
| Zhuang et al. | Optimization of high-speed fixed-wing UAV penetration strategy based on deep reinforcement learning | |
| Van et al. | Synthesis of Suboptimal Guidance Law for Anti-Tank Guided Missile with Terminal Impact Angle Constraint Based on the SDRE Technique | |
| Zhang et al. | Enhanced fruit fly optimization algorithm based backstepping-HOSMC for integrated guidance and control of hypersonic gliding vehicle | |
| CN116992952A (en) | Pre-training method, training method and system for collaborative guidance law model | |
| CN114020018B (en) | Determination method and device of missile control strategy, storage medium and electronic equipment | |
| CN112257259B (en) | Method and system for estimating whole-course trajectory of ballistic missile based on improved autonomous multiple models | |
| CN114815878A (en) | Hypersonic aircraft cooperative guidance method based on real-time optimization and deep learning | |
| Zhang et al. | The time-to-go consensus of multi-missiles with communication delay | |
| Chen et al. | A Cooperative Guidance Law for Multiple Missiles based on Reinforcement Learning | |
| CN113361196A (en) | Missile killing probability evaluation method, system, equipment and readable medium | |
| Fan et al. | An Optimization Method of Attitude Control Parameters Based on Genetic Algorithm for the Boost-Glide Rocket | |
| Pirzadeh et al. | Design of generalized predictive control for the stabilizing loop from a two-axis gimbal seeker, considering cross-coupling in between two channels | |
| CN117540639A (en) | Finite time convergence missile terminal guidance law design method based on deep neural network | |
| Hughes | Multi-objective evolutionary guidance for swarms |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |