CN117556718B - Thermal power distribution method based on neural network and multi-strategy combined gray wolf optimization algorithm - Google Patents

Thermal power distribution method based on neural network and multi-strategy combined gray wolf optimization algorithm Download PDF

Info

Publication number
CN117556718B
CN117556718B CN202410046968.1A CN202410046968A CN117556718B CN 117556718 B CN117556718 B CN 117556718B CN 202410046968 A CN202410046968 A CN 202410046968A CN 117556718 B CN117556718 B CN 117556718B
Authority
CN
China
Prior art keywords
wolf
strategy
target
information
algorithm
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
Application number
CN202410046968.1A
Other languages
Chinese (zh)
Other versions
CN117556718A (en
Inventor
闫天
蒋子健
黄汉桥
程昊宇
张蓬
李桐
刘灿
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN202410046968.1A priority Critical patent/CN117556718B/en
Publication of CN117556718A publication Critical patent/CN117556718A/en
Application granted granted Critical
Publication of CN117556718B publication Critical patent/CN117556718B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • G06N3/0499—Feedforward networks
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Geometry (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The embodiment of the disclosure relates to a fire distribution method based on a neural network and a multi-strategy combined gray wolf optimization algorithm. According to the embodiment of the disclosure, a multi-strategy combined wolf optimizing algorithm is provided according to a four-dimensional wolf information strategy, a collaborative parameter self-adaptive optimizing strategy, a weighted combination position updating strategy, a nonlinear factor convergence strategy and a mutation factor strategy, a fire power distribution scheme is evaluated by combining a neural network striking capability evaluation model, and the problem of multi-missile air-to-air striking multi-target fire power distribution is solved.

Description

Thermal power distribution method based on neural network and multi-strategy combined gray wolf optimization algorithm
Technical Field
The embodiment of the disclosure relates to the technical field of fire distribution, in particular to a fire distribution method based on a neural network and a multi-strategy combined gray wolf optimization algorithm.
Background
With the continuous updating of the informatization armed device, the corresponding striking capability is also improved, and the optimization of fire distribution directly influences the course and win-lose of war. Under modern battle scenes, the battle missions need to be tightly surrounded, and the battle situation and primary and secondary targets are combined to reasonably distribute firepower so as to act on the adversary and joints with comprehensive efficiency, and the battle missions can be striven for being achieved to the greatest extent with minimum cost. In the process of multi-missile cooperative hit combat, the problem of multi-target firepower distribution is a key problem needing to be studied in an important way.
In solving the fire distribution problem using intelligent algorithms, commonly used algorithms include particle swarm algorithm (Particle Swarm Optimization, PSO), ant colony algorithm (ant colony optimization, ACO), simulated annealing algorithm (Simulated annealing, SA), gray wolf optimization algorithm (Grey Wolf Optimizer, GWO), and the like. And (3) applying the algorithm to a model of the specific problem, and converting the fire distribution problem into an optimization problem to solve. However, the above solution method is limited to generating an optimal solution by an algorithm, and the feasibility of the allocation scheme is not considered, and the subsequent verification of the allocation scheme is lacking, which is unacceptable for a high-dynamic, irreversible missile. In addition, the traditional heuristic algorithm has the defects of easy sinking into local optimum, low convergence speed, incapability of global search and the like.
Disclosure of Invention
In order to avoid the defects of the prior art, the application provides a fire distribution method based on a neural network and a multi-strategy combined gray wolf optimization algorithm, which is used for solving the problems that the traditional heuristic algorithm in the prior art is easy to fall into local optimum, has low convergence speed, cannot judge the feasibility of a distribution scheme and the like.
According to an embodiment of the present disclosure, there is provided a fire distribution method based on a neural network and a multi-strategy combined wolf optimization algorithm, the method including:
acquiring missile group information and ground target information, dividing the ground target into a wayside target and a destination target according to the geographic position, and calculating threat probability of the ground target and damage probability of the missile to the ground target according to the ground target information; wherein the endpoint targets include a primary target and a secondary target;
setting basic parameters of a Multi-strategy combined gray wolf optimization (Multi-strategy combination grey Wolf optimization, MSC-GWO) algorithm, and defining time slices; dividing the striking time into time slices according to the along-road target and the end point target, wherein one time slice corresponds to primary fire distribution;
determining a task target and the type of the task target based on the ground target information, determining a striking time slice according to the task target and the type of the task target, and generating an initial allocation scheme; wherein the task targets include a along-the-way task target, a secondary task target, and a primary task target;
combining the striking time slices, and optimizing a fire power distribution scheme based on an MSC-GWO algorithm;
constructing a neural network striking capability evaluation model, judging whether an optimized initial allocation scheme is feasible or not by using the neural network striking capability evaluation model, if so, determining that the optimized initial allocation scheme is an optimal firepower allocation scheme of a task target, updating bullet guide group information at the end of a striking time slice, and stepping the striking time slice; if not, regenerating an initial allocation scheme;
judging whether the task target is a main task target, if so, ending the task; if not, the task target and the type of the task target are redetermined.
Further, the MSC-GWO algorithm includes:
initializing a wolf population and expanding population information according to a four-dimensional wolf information strategy, and initializing basic parametersInitializing->Wolf and jersey>Wolf and->Wolf position->；
Processing the individual wolves outside the boundary, and calculating the fitness value of each individual wolfAnd calculating the average fitness value of the wolf population +.>；
Comparing fitness values of individual wolves, and updatingWolf and jersey>Wolf and->Wolf position->；
Generating control parameters according to a nonlinear factor convergence strategyCalculating to obtain a cooperative parameter ++according to a cooperative parameter self-adaptive optimizing strategy>；
Updating the position of the current wolf individual according to the weighted combination position updating strategy, and calculating the fitness value of the current wolf individualAnd average fitness value of the wolf population +.>；
Introducing a first mutation operator and a second mutation operator according to a mutation factor strategy to improve global searching capability and avoid falling into local optimum;
outputting the optimal solutionAnd an optimal fitness value->；
Will best solveInputting the optimal solution into a neural network striking capability evaluation model>Whether or not it is feasible;
if feasible, ending the algorithm; otherwise, the calculation is performed again.
Further, the four-dimensional wolf information strategy includes:
at population initialization, the positions of the wolf individuals are represented in a four-dimensional coordinate system (namely a three-dimensional space coordinate system and a one-dimensional time coordinate system), namely:
(1)
in the method, in the process of the invention,is the firstpInformation contained in individual wolves, item 1pFour-dimensional wolf information of individual wolves>Expressed in the form of four-dimensional vectors recorded as +.>，i、jAndkthree directions of the individual wolf in the space coordinate system are respectively +.>For the direction of the wolf individuals in the time dimension,I、JandKrespectively corresponding coordinates of the gray wolf individuals on three coordinate axes in a space coordinate system, and (I)>A time-consuming threshold in the time dimension for the wolf individual; in the space coordinate system, the variable interval of the function is set to +.>，A L AndB L are all constant and randomly generate a modulus +.>Pitch angle->And yaw angle->The method comprises the steps of carrying out a first treatment on the surface of the Setting a time-consuming threshold in the time dimension>，/>The method comprises the steps of carrying out a first treatment on the surface of the The relation among the module value, the pitch angle, the yaw angle and the time consumption threshold value and four-dimensional gray wolf information is as follows:
(2)
(3)
in three dimensions, a total ofInformation of individual wolves, specific number of wolves +.>Time-dependent ++f for initializing the wolf information>The method comprises the steps of carrying out a first treatment on the surface of the The wolf information in three directions is updated according to formula (4), wherein +.>And the updating modes in three directions are the same:
(4)
wherein,
(5)
in the method, in the process of the invention,for the relative distance between the individual gray wolves and the optimal solution,>for the relative distance between the individual gray wolves and the suboptimal solution,>for the relative distance between the individual gray wolves and the suboptimal solution, ++>Is the optimal position of the individual gray wolves, +.>Is the suboptimal position of the individual gray wolf,>is the next time of the individual wolfThe optimal position of the device is provided,X H1 for the calculated first position of the wolf,X H2 for the resolved second position of the wolf,X H3 for the third position of the calculated wolf, < ->And->Are all [0,1 ]]Random numbers in between;
when solving the fitness function value, changing the coordinate value into a real number:
(6)
(7)
in the method, in the process of the invention,representing the nth dimension module->、/>And->For three gene information at n +.>Is a real variable.
Further, the collaborative parameter adaptive optimization strategy includes:
prior to each iteration update, a collaborative parameter is calculatedWhen the method is used, a cooperative parameter self-adaptive optimizing strategy is adopted:
(8)
wherein,and->Are all [0,1 ]]Random number between->For the current iteration number>The maximum iteration number; defining the +.>The updated particle swarm position information and speed information of the wheel are as follows:
(9)
(10)
wherein,for the speed definition domain constraint of the particle swarm algorithm, the update formula of the particle position and the speed is as follows:
(11)
(12)
wherein,is->Particles (1)>；/>Is the>Wei (dimension)>The method comprises the steps of carrying out a first treatment on the surface of the The PSO algorithm searches the particles in parallel and records the optimal position of each particle in the iterative process +.>Optimal position of particles within a population；/>And->For acceleration constant or learning factor, take +.>；/>And->Is->The interval oral administration is from uniformly distributed random numbers.
Further, the nonlinear factor convergence strategy includes:
in the MSC-GWO algorithm, nonlinear control parameters are introduced:
(13)
in the method, in the process of the invention,for controlling the initial value of the parameter, is a constant, < ->Is the base of natural logarithm, +.>For the current iteration number>Is the maximum number of iterations.
Further, the weighted combination location update strategy includes:
based on the formula (4), an adaptive position adjustment strategy is further adopted to adjust the fitness value of the current gray wolf individualAverage fitness value with the wolf population +.>Compare if->Is superior to->The position of the wolf is updated continuously by using a classical GWO algorithm; if->Second to->The following variant gray wolf positions were used:
(14)
in the method, in the process of the invention,is the sum of the maximum fitness value and the minimum fitness value, i.e. +.>，/>Is thatX H1 、X H2 And (3) withX H3 Sum (S)/(S)>For optimal fitness value, +.>For suboptimal fitness value, +.>Is a suboptimal fitness value.
Further, the mutant factor strategy includes:
introducing a first mutation operator and a second mutation operator into an MSC-GWO algorithm;
first mutation operator: searching mutation vectors in the first mutation operator in a search space around the optimal solution by a vector band randomly selected in the population;
(15)
in the method, in the process of the invention,for the first mutation operator generated based on formula (10), the formula (10) is given by->For the optimal gray wolf position->For optimal solution weights, take a value between 0 and 1,/for>For a randomly selected first mutation position, +.>For a randomly selected second mutation position, +.>For a randomly selected third mutation position, +.>Is a first coefficient>Is a second coefficient>Is a third coefficient, and->，/>And->The values of (2) are all between 0 and 1;
second mutation operator: searching in different directions by using different mutation vectors by the second mutation operator to obtain a global optimal solution;
(16)
in the method, in the process of the invention,for the second mutation operator generated based on formula (11), the ++>Taking an integer greater than 4 as a constant value,for a randomly selected fourth mutation position, +.>For randomly selected->Mutation position, ->，/>Is the fourth coefficient, and->Has a value between 0 and 1;
the first mutation operator and the second mutation operator respectively represent mutation of the position near the optimal solution and mutation of the global random position, so that weak global searching capability of a classical GWO algorithm can be effectively improved.
Further, the step of constructing a neural network striking capability evaluation model includes:
establishing a missile three-degree-of-freedom model according to the dynamics and kinematic characteristics of the missile;
based on a missile three-degree-of-freedom model, acquiring missile ranges under different initial situations by utilizing Montrealo simulation, and sampling the initial situations and the corresponding missile ranges as a training data set; the initial situation comprises the altitude, the speed and the roll angle of the missile;
inputting a training data set obtained by Monte Carlo simulation into a neural network for training to obtain a neural network striking capability evaluation model;
the neural network algorithm comprises forward information transmission and reverse error training, and neuron nodes are arranged in the forward information transmissionAnd node->The weight between them is->Node->Is +.>The node output value is +.>The output value calculation formula is as follows:
(17)
(18)
wherein,the method specifically selects a Sigmoid function for activating the function;
in the error reverse training, the neural network is trained based on the Levenberg-Marquardt algorithm, and the Levenberg-Marquardt algorithm update formula is as follows:
(19)
wherein,for the Hessian matrix,>is a unitary matrix->Optimizing step size for iteration ++>For an increased correction amount->A decreasing gradient that is an objective function;
and inputting the current missile situation to the neural network striking capability evaluation model to obtain the corresponding current missile range, so that whether the generated initial allocation scheme is feasible or not is evaluated online.
The technical scheme provided by the embodiment of the disclosure can comprise the following beneficial effects:
according to the fire distribution method based on the neural network and the multi-strategy combined wolf optimization algorithm, the multi-strategy combined wolf optimization algorithm is provided according to the four-dimensional wolf information strategy, the collaborative parameter self-adaptive optimizing strategy, the weighted combination position updating strategy, the nonlinear factor convergence strategy and the mutation factor strategy, the fire distribution scheme is evaluated by combining a neural network striking capability evaluation model, and the problem of multi-missile air-to-face striking multi-target fire distribution is solved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure. It will be apparent to those of ordinary skill in the art that the drawings in the following description are merely examples of the disclosure and that other drawings may be derived from them without undue effort.
FIG. 1 shows a step diagram of a fire distribution method based on a neural network and a multi-strategy combined gray wolf optimization algorithm in an exemplary embodiment of the present disclosure;
FIG. 2 illustrates a conceptual diagram of a multi-missile air-to-ground strike model in an exemplary embodiment of the present disclosure;
FIG. 3 illustrates a simulation of a multi-missile air-to-ground strike model in an exemplary embodiment of the present disclosure;
FIG. 4 shows a flow chart of a fire distribution method based on a neural network and a multi-strategy combined gray wolf optimization algorithm in an exemplary embodiment of the present disclosure;
FIG. 5 illustrates a neural network regression performance metric in an exemplary embodiment of the present disclosure;
FIG. 6 illustrates optimal fitness values and average fitness values of an iterative process of the MSC-GWO algorithm in an exemplary embodiment of the present disclosure;
FIG. 7 illustrates a comparison of multi-missile cooperative batting fire power distribution optimization results based on different algorithms in an exemplary embodiment of the present disclosure.
Detailed Description
Example embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments may be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Furthermore, the drawings are merely schematic illustrations of embodiments of the disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities.
In the present exemplary embodiment, a fire distribution method based on a neural network and a multi-strategy combined wolf optimization algorithm is provided first. Referring to fig. 1, the fire distribution method based on the neural network and the multi-strategy combined gray wolf optimization algorithm may include: step S101 to step S106.
Step S101: acquiring missile group information and ground target information, dividing the ground target into a wayside target and a destination target according to the geographic position, and calculating threat probability of the ground target and damage probability of the missile to the ground target according to the ground target information; wherein the endpoint targets include a primary target and a secondary target;
step S102: setting basic parameters of MSC-GWO algorithm, and defining time slices; dividing the striking time into time slices according to the along-road target and the end point target, wherein one time slice corresponds to primary fire distribution;
step S103: determining a task target and the type of the task target based on the ground target information, determining a striking time slice according to the task target and the type of the task target, and generating an initial allocation scheme; wherein the task targets include along-the-way task targets, secondary task targets, or primary task targets;
step S104: combining the striking time slices, and optimizing a fire power distribution scheme based on an MSC-GWO algorithm;
step S105: constructing a neural network striking capability evaluation model, judging whether an optimized initial allocation scheme is feasible or not by using the neural network striking capability evaluation model, if so, determining that the optimized initial allocation scheme is an optimal firepower allocation scheme of a task target, updating bullet guide group information at the end of a striking time slice, and stepping the striking time slice; if not, regenerating an initial allocation scheme;
step S106: judging whether the task target is a main task target, if so, ending the task; if not, the task target and the type of the task target are redetermined.
According to the fire distribution method based on the neural network and the multi-strategy combined wolf optimization algorithm, the multi-strategy combined wolf optimization algorithm is provided according to the four-dimensional wolf information strategy, the collaborative parameter self-adaptive optimizing strategy, the weighted combined position updating strategy, the nonlinear factor convergence strategy and the mutation factor strategy, a fire distribution scheme is evaluated by combining a neural network striking capability evaluation model, and the problem of multi-missile air-to-face striking multi-target fire distribution is solved.
Next, the respective steps of the above-described fire distribution method based on the neural network and multi-strategy combined wolf optimization algorithm in the present exemplary embodiment will be described in more detail with reference to fig. 1 to 7.
In step S101, initial information such as the position, number, speed, heading angle, etc. of the bullet guide group is acquired before the allocation is started, and the position, number, mobility, and type of the ground target are detected. The targets to be hit are divided into a along-way target and a main task target according to the position information, and then the destination target is divided into a main target and a secondary target. And calculating threat probability of the target and damage probability of the missile to the target according to the types.
In step S102, basic parameters of MSC-GWO are set, and initial values of parameters are controlledSynergistic parameter->Inertial weight of search equation +.>Maximum number of iterations->The number of individual gray wolves in the search space is +.>Mutation operator->；
The striking time is divided into n+1 time slices according to n along-way targets and 1 end point target, and 1 time slice corresponds to 1 fire distribution.
In step S103, the task targets are one or more targets in the ground target information, the type and the number of the task targets are determined by referring to the ground target information, the striking time slices are determined according to the number of the task targets, and an initial allocation scheme is generated according to the type of the task targets; the task targets of each time include one of a along-road task target, a secondary task target, and a primary task target; in the initial allocation scheme, the target of the along-way task with threat is destroyed first, and then the target of the secondary task and the target of the main task are destroyed.
In step S104, the MSC-GWO algorithm includes:
initializing a wolf population and expanding population information according to a four-dimensional wolf information strategy, and initializing basic parametersInitializing->Wolf and jersey>Wolf and jersey>Wolf position->；
Processing the individual wolves outside the boundary, and calculating the fitness value of each individual wolfAnd calculating the average fitness value of the wolf population +.>；
Comparing fitness values of individual wolves, and updatingWolf and jersey>Wolf and jersey>Wolf position->；
Generating control parameters according to a nonlinear factor convergence strategyCalculating to obtain a cooperative parameter ++according to a cooperative parameter self-adaptive optimizing strategy>；
Updating the position of the current wolf individual according to the weighted combination position updating strategy, and calculating the fitness value of the current wolf individualAnd average fitness value of the wolf population +.>；
Introducing a first mutation operator and a second mutation operator according to a mutation factor strategy to improve global searching capability and avoid falling into local optimum;
outputting the optimal solutionAnd an optimal fitness value->；
Will best solveInputting the optimal solution into a neural network striking capability evaluation model>Whether or not it is feasible;
if feasible, ending the algorithm; otherwise, the calculation is performed again.
The four-dimensional gray wolf information strategy comprises:
at population initialization, the individual positions of the gray wolves are represented in a four-dimensional coordinate system, namely:
(1)
in the method, in the process of the invention,is the firstpInformation contained in individual wolves, item 1pFour-dimensional wolf information of individual wolves>Expressed in the form of four-dimensional vectors recorded as +.>，i、jAndkthree directions of the individual wolf in the space coordinate system are respectively +.>For the direction of the wolf individuals in the time dimension,I、JandKrespectively corresponding coordinates of the gray wolf individuals on three coordinate axes in a space coordinate system, and (I)>A time-consuming threshold in the time dimension for the wolf individual; in the space coordinate system, the variable interval of the function is set to +.>，A L AndB L are all constant and randomly generate a modulus +.>Pitch angle->And yaw angle->The method comprises the steps of carrying out a first treatment on the surface of the Setting a time-consuming threshold in the time dimension>，/>The method comprises the steps of carrying out a first treatment on the surface of the The relation among the module value, the pitch angle, the yaw angle and the time consumption threshold value and four-dimensional gray wolf information is as follows:
(2)
(3)
in three dimensions, a total ofInformation of individual wolves, specific number of wolves +.>Time-dependent ++f for initializing the wolf information>The method comprises the steps of carrying out a first treatment on the surface of the The wolf information in three directions is updated according to formula (4), wherein +.>And the updating modes in three directions are the same:
(4)
wherein,
(5)
in the method, in the process of the invention,for the relative distance between the individual gray wolves and the optimal solution,>for the relative distance between the individual gray wolves and the suboptimal solution,>for the relative distance between the individual gray wolves and the suboptimal solution, ++>Is the most of the individual wolvesPosition of good, good>Is the suboptimal position of the individual gray wolf,>for sub-optimal position of the gray wolf individuals,X H1 for the calculated first position of the wolf,X H2 for the resolved second position of the wolf,X H3 for the third position of the calculated wolf, < ->And->Are all [0,1 ]]Random numbers in between;
when solving the fitness function value, changing the coordinate value into a real number:
(6)
(7)
in the method, in the process of the invention,representing the nth dimension module->、/>And->For three gene information at n +.>Is a real variable.
The collaborative parameter self-adaptive optimizing strategy comprises the following steps:
at each timeBefore iterative updating of the round, the collaborative parameters are calculatedWhen the method is used, a cooperative parameter self-adaptive optimizing strategy is adopted:
(8)
wherein,and->Are all [0,1 ]]Random number between->For the current iteration number>The maximum iteration number; definition of the +.sup.th in Particle Swarm (PSO) algorithm optimization parameters>The updated particle swarm position information and speed information of the wheel are as follows:
(9)
(10)
wherein,for the speed definition domain constraint of the particle swarm algorithm, the update formula of the particle position and the speed is as follows:
(11)
(12)
wherein,is->Particles (1)>；/>Is the>Wei (dimension)>The method comprises the steps of carrying out a first treatment on the surface of the The PSO algorithm searches the particles in parallel and records the optimal position of each particle in the iterative process +.>Optimal position of particles within a population；/>Is inertial weight, ++>And->For acceleration constant or learning factor, take +.>；/>And->Is thatThe interval oral administration is from uniformly distributed random numbers.
The nonlinear factor convergence strategy includes:
in the algorithm of the MSC-GWO,is the value of (2) and the convergence factora(i.e. control parameters) are related,athe value of the convergence factor in the basic MSC-GWO algorithm is linearly reduced from 2 to 0, the process ignores the nonlinearity of the algorithm in the convergence process, and nonlinear control parameters are introduced for improving the performance of the algorithm:
(13)
in the method, in the process of the invention,for controlling the initial value of the parameter, is a constant, < ->Is the base of natural logarithm, +.>For the current iteration number>Is the maximum number of iterations. This non-linear decrementing fashion searches for the early stagesaThe value attenuation degree is gradually reduced, global searching can be better carried out, the attenuation degree in the later period of searching is accelerated, local searching can be better carried out, and the balance capability of the algorithm global searching and the local searching is improved.
The weighted combination location update strategy includes:
in order to further improve the convergence rate of the algorithm, on the basis of the formula (4), an adaptive position adjustment strategy is further adopted to adjust the adaptation value of the current wolf individualAverage fitness value with the wolf population +.>Compare if->Is superior to->The position of the wolf is updated continuously by using a classical GWO algorithm; if->Second to->The following variant gray wolf positions were used:
(14)
in the method, in the process of the invention,is the sum of the maximum fitness value and the minimum fitness value, i.e. +.>，/>Is thatX H1 、X H2 And (3) withX H3 Sum (S)/(S)>For optimal fitness value, +.>For suboptimal fitness value, +.>Is a suboptimal fitness value.
The mutant factor strategy includes:
introducing a first mutation operator and a second mutation operator into an MSC-GWO algorithm;
first mutation operator: searching mutation vectors in the first mutation operator in a search space around the optimal solution by a vector band randomly selected in the population;
(15)
in the method, in the process of the invention,for the first mutation operator generated based on formula (10), the formula (10) is given by->For the optimal gray wolf position->For optimal solution weights, take a value between 0 and 1,/for>For a randomly selected first mutation position, +.>For a randomly selected second mutation position, +.>For a randomly selected third mutation position, +.>Is a first coefficient>Is a second coefficient>Is a third coefficient, and->，/>And->The values of (2) are all between 0 and 1;
second mutation operator: searching in different directions by using different mutation vectors by the second mutation operator to obtain a global optimal solution;
(16)
in the method, in the process of the invention,for the second mutation operator generated based on formula (11), the ++>Taking an integer greater than 4 as a constant value,for a randomly selected fourth mutation position, +.>For randomly selected->Mutation position, ->，/>Is the fourth coefficient, and->Has a value between 0 and 1;
the first mutation operator and the second mutation operator respectively represent mutation of the position near the optimal solution and mutation of the global random position, so that weak global searching capability of a classical GWO algorithm can be effectively improved.
In step S105, the step of constructing a neural network striking power evaluation model includes:
establishing a missile three-degree-of-freedom model according to the dynamics and kinematic characteristics of the missile;
based on a missile three-degree-of-freedom model, acquiring missile ranges under different initial situations by utilizing Montrealo simulation, and sampling the initial situations and the corresponding missile ranges as a training data set; the initial situation comprises the altitude, the speed and the roll angle of the missile;
inputting a training data set obtained by Monte Carlo simulation into a neural network for training to obtain a neural network striking capability evaluation model;
the neural network algorithm comprises forward information transmission and reverse error training, and neuron nodes are arranged in the forward information transmissionAnd node->The weight between them is->Node->Is +.>The node output value is +.>The output value calculation formula is as follows:
(17)
(18)
wherein,to activate the function, a Sigmoid function is selected here.
In the error reverse training, a neural network is trained based on a Levenberg-Marquardt (L-M) algorithm, and the L-M algorithm updates the formula as follows:
(19)
wherein,for the Hessian matrix,>is a unitary matrix->Optimizing step size for iteration ++>For an increased correction amount->A decreasing gradient that is an objective function;
and inputting the current missile situation to the neural network striking capability evaluation model to obtain the corresponding current missile range, so that whether the generated initial allocation scheme is feasible or not is evaluated online.
Judging whether the initial allocation scheme is feasible or not according to the neural network striking capability evaluation model, if so, obtaining an optimal firepower allocation scheme of a task target and updating missile group information at the end of a time slice; if it is not feasible to go to step S104;
in step S106, it is determined whether the task target is a main task target, if so, a final result is output (i.e., the initial distribution scheme is an optimal fire distribution scheme); if not, go to step S103.
In a specific embodiment, to verify the effectiveness of the algorithm, a simulation was performed for the fire distribution of 30 missiles striking 1+7+1 enemy targets. Fig. 2 is a conceptual diagram of a multi-missile air-to-ground strike model, wherein 1 target with threat (i.e., a target along the way) exists on the strike path of a missile group, the destination target comprises 7 secondary targets and 1 primary target, and different targets have different detection distances and strike ranges. Fig. 3 shows a simulation diagram of a multi-missile air-to-ground strike model, the relevant information of a missile (i.e., hypersonic missile) is shown in table 1, the threat targets on the path are an outer layer of a us shield ship 8, the secondary targets of the task end points are an aeus shield ship 2, an aeus shield ship 3, an aeus shield ship 4, an aeus shield ship 5, an aeus shield ship 6 and an aeus shield ship 7, the main targets of the end points are aircraft carriers, and the relevant information is shown in table 2.
TABLE 1 missile group configuration information table
Table 1 shows missile group information such as number, altitude, speed and course angle of missiles, and in order to better simulate real environment, only central coordinates of the missile groups are set, and specific initial positions of each missile are subjected to random processing.
Table 2: target setting information table
Table 2 shows the number, type, number, coordinates, and value amounts of task targets. Wherein the radar has a maximum rangePerformance parameters of an air defense missile: horizontal distance of killing area->Performance parameters of a B-type air defense missile: horizontal distance of killing area->。
Table 3: neural network training parameters
Table 3 shows neural network parameters used to train the missile hit ability assessment model. The initial position and the speed of the sample are used as input values of training samples, and the predicted range is used as the label quantity for training. To prevent overfitting, the present application divides the data into three parts, training, validation, test. The whole network consists of two layers, wherein the first layer is a hidden layer, the second layer is an output layer, and the first layer consists of 10 neurons.
As shown in fig. 4, a flow chart of a fire distribution method based on a neural network and a multi-strategy combined gray wolf optimization algorithm comprises the following steps:
initializing information of the missile group and the ground target. Specific information is shown in tables 1 and 2.
Initializing parameters. The number of the gray wolves of the MSC-GWO algorithm is set to 50, the maximum iteration number is set to 80, the initial value of the control parameter is set to 2, the inertial weight of the search equation is set to 0.8, the first coefficient, the second coefficient and the third coefficientSet as 0.6,0.4,0.7 respectively;
determining a striking time slice, and dividing the striking time into 2 time slices;
determining the type of a ground target currently ready for striking;
generating an initial allocation scheme based on the coding according to the target type;
optimizing a fire distribution scheme based on an MSC-GWO algorithm in a time slice;
judging whether the allocation scheme is feasible or not according to the neural network striking capability evaluation model, if so, obtaining an optimal firepower allocation scheme of a task target and updating missile group information at the end of a time slice; if not, generating an initial allocation scheme based on the coding again according to the target type;
step time slices;
judging whether the task target is a main task target or not, and if so, outputting a final result; if not, the type of ground target currently being prepared for striking is re-determined.
The effect of the present application can be further demonstrated by the following simulation experiments.
Matlab 2021a is adopted for simulation, the running environment is an Inter (R) Core (TM) i5-10300H processor, and the operating system is Windows10.
Simulation results:
the time consumption for solving the multi-missile air-to-face hit fire distribution problem based on the neural network and the MSC-GWO algorithm is 2.55721s, and the optimal function value is 1.4128. The specific allocations are shown in table 4.
Table 4: fire distribution scheme
Fig. 5 shows regression of the fitting value of the trained neural network striking power evaluation model against the true value, and the regression is represented by a coefficient R, if R is closer to 1, the model fitting power is stronger, and the trained neural network striking power evaluation model has better performance. Fig. 5 verifies that the neural network striking capability evaluation model obtained based on neural network training has stronger fitting capability, and can meet the requirement of online evaluation of the range of the missile.
As can be seen from fig. 6, the MSC-GWO algorithm gradually decreases and stabilizes the value of the optimum function as the number of iterations increases, and the average fitness function decreases while oscillating, although there is a slight fluctuation, the overall trend is no longer changing, indicating that an optimum fire distribution scheme is obtained by the MSC-GWO algorithm.
From fig. 7, it can be seen that the MSC-GWO algorithm proposed in the present application is compared with the common intelligent optimization algorithms (GWO, GA, PSO and ACA). When solving the same fire distribution problem, the adaptability value of the final convergence of the PSO algorithm and the ACA algorithm is too large, a good optimization effect is not obtained, the PSO algorithm converges too early compared with other algorithms, and the iteration is carried out until the 30 th time of the PSO algorithm falls into local optimum. Compared with the three algorithms, the MSC-GWO algorithm and the GWO algorithm have better optimization performance, wherein the MSC-GWO algorithm has stronger global searching capability and higher calculation accuracy based on various proposed strategy combinations, and the GWO algorithm has a slightly higher optimization speed because of the simplicity of the algorithm.
According to the four-dimensional gray wolf information strategy, the collaborative parameter self-adaptive optimizing strategy, the weighted combination position updating strategy, the nonlinear factor convergence strategy and the mutation factor strategy, a multi-strategy combination gray wolf optimizing algorithm is provided, a striking capability evaluation scheme based on a neural network is combined, and the problem of multi-missile air-to-air striking multi-target firepower distribution is solved. The striking capability evaluation is introduced when the fitness function is designed; the simulation experiment result verifies the effectiveness and reliability of dynamic solution of the fire distribution problem based on neural network evaluation and MSC-GWO algorithm, and compared with the fire distribution simulation experiment based on IPSO, ACO, SA and GWO algorithms, the simulation experiment result verifies the solution precision and optimization speed of MSC-GWO algorithm. The method is superior to the existing distribution method in the aspects of multi-bullet cooperation, air-to-face cooperation, striking and multi-target firepower distribution, and the MSC-GWO algorithm in the method is superior to the IPSO, ACO, SA, GWO general algorithm, so that the method has important application value.
Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. In the description of the embodiments of the present disclosure, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present disclosure. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, one skilled in the art can combine and combine the different embodiments or examples described in this specification.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

Claims (2)

1. A fire distribution method based on a neural network and a multi-strategy combined wolf optimization algorithm is characterized by comprising the following steps:
acquiring missile group information and ground target information, dividing the ground target into a wayside target and a destination target according to the geographic position, and calculating threat probability of the ground target and damage probability of the missile to the ground target according to the ground target information; wherein the endpoint targets include a primary target and a secondary target;
setting basic parameters of MSC-GWO algorithm, and defining time slices; dividing the striking time into time slices according to the along-road target and the end point target, wherein one time slice corresponds to primary fire distribution;
determining a task target and the type of the task target based on the ground target information, determining a striking time slice according to the task target and the type of the task target, and generating an initial allocation scheme; wherein the task targets include a along-the-way task target, a secondary task target, and a primary task target;
combining the striking time slices, and optimizing a fire power distribution scheme based on an MSC-GWO algorithm;
constructing a neural network striking capability evaluation model, judging whether an optimized initial allocation scheme is feasible or not by using the neural network striking capability evaluation model, if so, determining that the optimized initial allocation scheme is an optimal firepower allocation scheme of a task target, updating bullet guide group information at the end of a striking time slice, and stepping the striking time slice; if not, regenerating an initial allocation scheme;
judging whether the task target is a main task target, if so, ending the task; if not, re-determining the task target and the type of the task target;
the MSC-GWO algorithm includes:
initializing a wolf population and expanding population information according to a four-dimensional wolf information strategy, and initializing basic parametersInitializing->Wolf and jersey>Wolf and->Wolf position->；
Processing the individual wolves outside the boundary, and calculating the fitness value of each individual wolfAnd calculating the average fitness value of the wolf population +.>；
Comparing fitness values of individual wolves, and updatingWolf and jersey>Wolf and->Wolf position->；
Generating control parameters according to a nonlinear factor convergence strategyCalculating to obtain a cooperative parameter ++according to a cooperative parameter self-adaptive optimizing strategy>；
Updating the position of the current wolf individual according to the weighted combination position updating strategy, and calculating the fitness value of the current wolf individualAnd average fitness value of the wolf population +.>；
Introducing a first mutation operator and a second mutation operator according to a mutation factor strategy to improve global searching capability and avoid falling into local optimum;
outputting the optimal solutionAnd an optimal fitness value->；
Will best solveInputting the optimal solution into a neural network striking capability evaluation model>Whether or not to be able toA row;
if feasible, ending the algorithm; otherwise, re-calculating;
the four-dimensional gray wolf information strategy comprises:
at population initialization, the positions of the wolf individuals are represented in a four-dimensional coordinate system, namely:
(1)
in the method, in the process of the invention,is the firstpInformation contained in individual wolves, item 1pFour-dimensional wolf information of individual wolves>Expressed in the form of four-dimensional vectors recorded as +.>，i、jAndkthree directions of the individual wolf in the space coordinate system are respectively +.>For the direction of the wolf individuals in the time dimension,I、JandKrespectively corresponding coordinates of the gray wolf individuals on three coordinate axes in a space coordinate system, and (I)>A time-consuming threshold in the time dimension for the wolf individual; in the space coordinate system, the variable interval of the function is set to +.>，A L AndB L are all constant and randomly generate a modulus +.>Pitch angle->And yaw angle->The method comprises the steps of carrying out a first treatment on the surface of the Setting a time-consuming threshold in the time dimension>，/>The method comprises the steps of carrying out a first treatment on the surface of the The relation among the module value, the pitch angle, the yaw angle and the time consumption threshold value and four-dimensional gray wolf information is as follows:
(2)
(3)
in three dimensions, a total ofInformation of individual wolves, specific number of wolves +.>Time-dependent ++f for initializing the wolf information>The method comprises the steps of carrying out a first treatment on the surface of the The wolf information in three directions is updated according to formula (4), wherein +.>And the updating modes in three directions are the same:
(4)
wherein,
(5)
in the method, in the process of the invention,is the position of the gray wolf at time t +.>Is the position of the gray wolf at time t+1,>for the relative distance between the individual gray wolves and the optimal solution,>for the relative distance between the individual gray wolves and the suboptimal solution,>for the relative distance between the individual gray wolves and the suboptimal solution, ++>Is the optimal position of the individual gray wolves, +.>For the sub-optimal position of the gray wolf individual,for sub-optimal position of the gray wolf individuals,X H1 to solve forThe first position of the gray wolf is provided with a first position,X H2 for the resolved second position of the wolf,X H3 for the third position of the calculated wolf, < ->And->Are all [0,1 ]]Random numbers in between;
when solving the fitness function value, changing the coordinate value into a real number:
(6)
(7)
in the method, in the process of the invention,representing the nth dimension module->、/>And->For three gene information at n +.>Is a real variable;
the collaborative parameter self-adaptive optimizing strategy comprises the following steps:
prior to each iteration update, a collaborative parameter is calculatedWhen the method is used, a cooperative parameter self-adaptive optimizing strategy is adopted:
(8)
wherein,and->Are all [0,1 ]]Random number between->For the current iteration number>The maximum iteration number; defining the +.>The updated particle swarm position information and speed information of the wheel are as follows:
(9)
(10)
wherein,for the speed definition domain constraint of the particle swarm algorithm, the update formula of the particle position and the speed is as follows:
(11)
(12)
wherein,is->Particles (1)>；/>Is the>Wei (dimension)>The method comprises the steps of carrying out a first treatment on the surface of the The PSO algorithm searches the particles in parallel and records the optimal position of each particle in the iterative process +.>Optimal position of particles within a population；/>Is inertial weight, ++>And->For acceleration constant or learning factor, take +.>；/>And->Is thatThe interval oral administration is from the random number of the uniform distribution;
the nonlinear factor convergence strategy includes:
in the MSC-GWO algorithm, nonlinear control parameters are introduced:
(13)
in the method, in the process of the invention,for controlling the initial value of the parameter, is a constant, < ->Is the base of natural logarithm, +.>For the current number of iterations,the maximum iteration number;
the weighted combination location update strategy includes:
based on the formula (4), an adaptive position adjustment strategy is further adopted to adjust the fitness value of the current gray wolf individualAverage fitness value with the wolf population +.>Compare if->Is superior to->The position of the wolf is updated continuously by using a classical GWO algorithm; if->Second to->The following variant gray wolf positions were used:
(14)
in the method, in the process of the invention,is the sum of the maximum fitness value and the minimum fitness value, i.e. +.>，/>Is thatX H1 、X H2 And (3) withX H3 Sum (S)/(S)>For optimal fitness value, +.>For suboptimal fitness value, +.>Is a sub-optimal fitness value;
the step of constructing a neural network striking capability evaluation model comprises the following steps:
establishing a missile three-degree-of-freedom model according to the dynamics and kinematic characteristics of the missile;
based on a missile three-degree-of-freedom model, acquiring missile ranges under different initial situations by utilizing Montrealo simulation, and sampling the initial situations and the corresponding missile ranges as a training data set; the initial situation comprises the altitude, the speed and the roll angle of the missile;
inputting a training data set obtained by Monte Carlo simulation into a neural network for training to obtain a neural network striking capability evaluation model;
the neural network algorithm comprises forward information transmission and reverse error training, and neuron nodes are arranged in the forward information transmissionAnd node->The weight between them is->Node->Is +.>The node output value is +.>The output value calculation formula is as follows:
(17)
(18)
wherein,the method specifically selects a Sigmoid function for activating the function;
in the error reverse training, the neural network is trained based on the Levenberg-Marquardt algorithm, and the Levenberg-Marquardt algorithm update formula is as follows:
(19)
wherein,for the Hessian matrix,>is a unitary matrix->Optimizing step size for iteration ++>In order to increase the amount of correction,a decreasing gradient that is an objective function;
and inputting the current missile situation to the neural network striking capability evaluation model to obtain the corresponding current missile range, so that whether the generated initial allocation scheme is feasible or not is evaluated online.
2. The fire distribution method based on the neural network and the multi-strategy combined gray wolf optimization algorithm according to claim 1, wherein the mutation factor strategy comprises:
introducing a first mutation operator and a second mutation operator into an MSC-GWO algorithm;
first mutation operator: searching mutation vectors in the first mutation operator in a search space around the optimal solution by a vector band randomly selected in the population;
(15)
in the method, in the process of the invention,for the first mutation operator generated based on formula (10), the formula (10) is given by->For the optimal gray wolf position->For optimal solution weights, take a value between 0 and 1,/for>For a randomly selected first mutation position, +.>For a randomly selected second mutation position, +.>For a randomly selected third mutation position, +.>Is a first coefficient>Is a second coefficient>Is of a third coefficient, an，/>And->The values of (2) are all between 0 and 1;
second mutation operator: searching in different directions by using different mutation vectors by the second mutation operator to obtain a global optimal solution;
(16)
in the method, in the process of the invention,for the second mutation operator generated based on formula (11), the ++>Taking the integer greater than 4 as a constant value, < ->For a randomly selected fourth mutation position, +.>For randomly selected->Mutation position, ->，/>Is the fourth coefficient, and->And has a value between 0 and 1.
CN202410046968.1A 2024-01-12 2024-01-12 Thermal power distribution method based on neural network and multi-strategy combined gray wolf optimization algorithm Active CN117556718B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410046968.1A CN117556718B (en) 2024-01-12 2024-01-12 Thermal power distribution method based on neural network and multi-strategy combined gray wolf optimization algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410046968.1A CN117556718B (en) 2024-01-12 2024-01-12 Thermal power distribution method based on neural network and multi-strategy combined gray wolf optimization algorithm

Publications (2)

Publication Number Publication Date
CN117556718A CN117556718A (en) 2024-02-13
CN117556718B true CN117556718B (en) 2024-03-29

Family

ID=89823719

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410046968.1A Active CN117556718B (en) 2024-01-12 2024-01-12 Thermal power distribution method based on neural network and multi-strategy combined gray wolf optimization algorithm

Country Status (1)

Country Link
CN (1) CN117556718B (en)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103425840A (en) * 2013-08-14 2013-12-04 西北工业大学 Cooperative air combat firepower distribution method based on improved multi-target leapfrog algorithm
CN106779210A (en) * 2016-12-13 2017-05-31 中国人民解放军国防科学技术大学 Algorithm of Firepower Allocation based on ant group algorithm
CN107067121A (en) * 2017-06-13 2017-08-18 广东工业大学 A kind of improvement grey wolf optimized algorithm based on multiple target
CN107832885A (en) * 2017-11-02 2018-03-23 南京航空航天大学 A kind of fleet Algorithm of Firepower Allocation based on adaptive-migration strategy BBO algorithms
CN111443728A (en) * 2020-03-25 2020-07-24 北京航空航天大学 Chaos wolf optimization-based unmanned aerial vehicle formation control method
EP3979124A1 (en) * 2020-10-01 2022-04-06 HENSOLDT Sensors GmbH An active protection system and method of operating active protection systems
CN116088572A (en) * 2023-01-31 2023-05-09 西北工业大学 Unmanned plane collaborative attack multi-target task allocation method based on self-adaptive gray wolf optimization algorithm
CN116976525A (en) * 2023-08-18 2023-10-31 嵩山实验室 Resource allocation method based on improved artificial bee colony

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103425840A (en) * 2013-08-14 2013-12-04 西北工业大学 Cooperative air combat firepower distribution method based on improved multi-target leapfrog algorithm
CN106779210A (en) * 2016-12-13 2017-05-31 中国人民解放军国防科学技术大学 Algorithm of Firepower Allocation based on ant group algorithm
CN107067121A (en) * 2017-06-13 2017-08-18 广东工业大学 A kind of improvement grey wolf optimized algorithm based on multiple target
CN107832885A (en) * 2017-11-02 2018-03-23 南京航空航天大学 A kind of fleet Algorithm of Firepower Allocation based on adaptive-migration strategy BBO algorithms
CN111443728A (en) * 2020-03-25 2020-07-24 北京航空航天大学 Chaos wolf optimization-based unmanned aerial vehicle formation control method
EP3979124A1 (en) * 2020-10-01 2022-04-06 HENSOLDT Sensors GmbH An active protection system and method of operating active protection systems
CN116088572A (en) * 2023-01-31 2023-05-09 西北工业大学 Unmanned plane collaborative attack multi-target task allocation method based on self-adaptive gray wolf optimization algorithm
CN116976525A (en) * 2023-08-18 2023-10-31 嵩山实验室 Resource allocation method based on improved artificial bee colony

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Intelligent Guidance and Control Methods for Missile Swarm;Hanqiao Huang 等;《Computational Intelligence and Neuroscience》;20220125;全文 *
智能空战体系下无人协同作战发展现状及关键技术;黄汉桥 等;《导航与控制》;20190228;全文 *
武器目标分配问题研究进展：模型、算法与应用;李梦杰 等;《系统工程与电子技术》;20230430;全文 *

Also Published As

Publication number Publication date
CN117556718A (en) 2024-02-13

Similar Documents

Publication Publication Date Title
CN106779210B (en) Algorithm of Firepower Allocation based on ant group algorithm
CN107832885B (en) Ship formation fire power distribution method based on self-adaptive migration strategy BBO algorithm
CN108549402A (en) Unmanned aerial vehicle group method for allocating tasks based on quantum crow group hunting mechanism
Ab. Rashid Tiki-taka algorithm: a novel metaheuristic inspired by football playing style
CN114460959A (en) Unmanned aerial vehicle group cooperative autonomous decision-making method and device based on multi-body game
CN115222271A (en) Weapon target distribution method based on neural network
CN113177583B (en) Aerial target clustering method
CN117556718B (en) Thermal power distribution method based on neural network and multi-strategy combined gray wolf optimization algorithm
CN116090356B (en) Heterogeneous warhead multi-objective task planning method based on task reliability constraint
CN116088572A (en) Unmanned plane collaborative attack multi-target task allocation method based on self-adaptive gray wolf optimization algorithm
CN110782062A (en) Many-to-many packet interception target distribution method and system for air defense system
CN116050515B (en) XGBoost-based parallel deduction multi-branch situation prediction method
CN115328189B (en) Multi-unmanned plane cooperative game decision-making method and system
CN111797966A (en) Multi-machine cooperative global target distribution method based on improved sheep swarm algorithm
CN115859778A (en) Air combat maneuver decision method based on DCL-GWOO algorithm
CN112612300B (en) Multi-platform intelligent decision-making multi-target game method and device
CN114814741A (en) DQN radar interference decision method and device based on priority important sampling fusion
CN111382896B (en) WTA target optimization method of self-adaptive chaotic parallel clone selection algorithm
CN114548674A (en) Multi-agent confrontation scene-oriented threat situation assessment method, device and equipment
CN113553777A (en) Anti-unmanned aerial vehicle swarm air defense deployment method, device, equipment and medium
RU2684963C1 (en) Method for automatic group target allocation of fighters based on eventual elimination of participants
Tianhan et al. UAV cooperative attack and route planning based on DPSO algorithm
Li et al. Double Deep Q-learning for Anti-saturation Attack Problem of Warship Group
CN113359817A (en) Method for intercepting targets by multi-missile formation under driving restraint guidance system
CN112612206B (en) Multi-agent collaborative decision-making method and system for uncertain events

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