CN109242290B - Automatic generation method for unmanned aerial vehicle group action scheme - Google Patents

Automatic generation method for unmanned aerial vehicle group action scheme Download PDF

Info

Publication number
CN109242290B
CN109242290B CN201810988692.3A CN201810988692A CN109242290B CN 109242290 B CN109242290 B CN 109242290B CN 201810988692 A CN201810988692 A CN 201810988692A CN 109242290 B CN109242290 B CN 109242290B
Authority
CN
China
Prior art keywords
harmony
task
platform
priority
time
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
CN201810988692.3A
Other languages
Chinese (zh)
Other versions
CN109242290A (en
Inventor
周尧明
赵浩然
陈俊锋
郑江安
李�昊
姜晓爱
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.)
China's Aviation System Engineering Research Institute
Beihang University
Original Assignee
China's Aviation System Engineering Research Institute
Beihang 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 China's Aviation System Engineering Research Institute, Beihang University filed Critical China's Aviation System Engineering Research Institute
Priority to CN201810988692.3A priority Critical patent/CN109242290B/en
Publication of CN109242290A publication Critical patent/CN109242290A/en
Application granted granted Critical
Publication of CN109242290B publication Critical patent/CN109242290B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06316—Sequencing of tasks or work
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Strategic Management (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Development Economics (AREA)
  • Molecular Biology (AREA)
  • Educational Administration (AREA)
  • Genetics & Genomics (AREA)
  • Game Theory and Decision Science (AREA)
  • Physiology (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses an automatic generation method of an unmanned aerial vehicle group action scheme, which comprises the following steps: firstly, determining task core and scene information; a subtask decomposition module which decomposes the task core into a plurality of subtasks by using a genetic algorithm; thirdly, a subtask priority automatic sequencing module-harmony population generation; fourthly, the resource scheduling module performs harmony evaluation; fifthly, updating the harmony memory library by the subtask priority automatic sequencing module; sixthly, outputting an excellent scheme. The invention gets rid of human participation, realizes high automation, quick response rate and intelligent synergy; the generated scheme ensures the excellence and diversity and can be selected according to the excellence degree; the generated action scheme has higher execution efficiency and more definite purpose; the efficiency is higher during searching, and excellent solutions can be achieved without traversing all the arrays; the optimal scheduling and the evaluation of the optimal priority sequence and the sound individuals can be simultaneously carried out, the accuracy of time sequence and resource scheduling is ensured, and the calculation efficiency is increased; the modular algorithm is more compatible.

Description

Automatic generation method for unmanned aerial vehicle group action scheme
Technical Field
The invention relates to an automatic generation method of an unmanned aerial vehicle group action scheme, in particular to an automatic generation method of the unmanned aerial vehicle group action scheme based on a subtask decomposition genetic algorithm, a subtask priority ordering and sound searching algorithm and a resource scheduling multidimensional dynamic list planning algorithm, which can quickly, effectively and completely automatically generate the unmanned aerial vehicle group action scheme without human intervention and belongs to the fields of intelligent algorithm modules and unmanned aerial vehicle group task command scheduling.
Background
In the process of generating an intelligent and cooperative action scheme of a complex unmanned aerial vehicle cluster, various intelligent algorithms are introduced to automatically generate the scheme by using a computer in order to increase the generation efficiency and the excellence and diversity of the scheme. In recent years, with the development of unmanned aerial vehicle technology, unmanned aerial vehicles are widely applied in various fields. And single unmanned aerial vehicle limitation is stronger, and in large-scale task execution process, unmanned aerial vehicle often is in the form of unmanned aerial vehicle crowd formation row dress. The unmanned aerial vehicle group has the characteristics of quick response, high-efficiency collaboration, accurate task execution and the like, and the characteristic advantages of the unmanned aerial vehicle group cannot be clearly exerted if people are used as commanders for manual command. Therefore, the research of the automatic generation method of the unmanned aerial vehicle group action scheme is necessary. The excellent automatic generation method can enable the unmanned aerial vehicle cluster to execute tasks more quickly and efficiently, has strong openness and high response speed, and can correct the tasks and face more diversified and complex tasks faster than the human command speed. Therefore, the automatic generation method of the unmanned aerial vehicle group action scheme has good development prospect and practical application value.
At present, most of the generation of the unmanned aerial vehicle group action scheme is a sub-task which is divided manually, the priority is arranged manually according to experience, and the sub-task is used as input to generate the action scheme by using an algorithm. However, since the generation method involves human beings in the sub-task decomposition and priority generation stages, the generation method cannot be called an automatic generation method, has great limitations, and cannot be separated from human factors. At present, the academic world has no fully automatic action scheme generation method, and learners use a CS (cuckoo search) algorithm and an MPDLS (multi-priority list dynamic programming) algorithm to be combined, so that the automation degree is certain, but the automation precision required by the automatic generation of the action scheme of the unmanned aerial vehicle cluster is far not reached due to the fact that the learners also need to manually decompose subtasks and set the priority. Therefore, in order to meet the requirement of generating the unmanned aerial vehicle group action scheme, it is significant to provide a method for automatically generating the unmanned aerial vehicle group action scheme.
Disclosure of Invention
The invention provides an automatic generation method of an unmanned aerial vehicle group action scheme, which aims at improving the traditional generation method of the unmanned aerial vehicle group action scheme. The method is based on three modules, and each module is based on an algorithm, namely a subtask decomposition module, a subtask priority ordering module and a resource scheduling module, which respectively correspond to a subtask decomposition genetic algorithm, a priority sequence cuckoo search algorithm and a multidimensional list dynamic planning algorithm. The method can realize a given task core, namely a total task, such as object acquisition, object elimination, arrival at a designated position or rescue object, and automatically generate an action scheme for executing the task core, namely a time sequence for executing specific actions of each platform of the unmanned aerial vehicle group through modular operation. The method is mainly characterized in that a subtask decomposition module and a subtask priority automatic sequencing module are provided, no human participation is caused, and various excellent schemes can be generated efficiently and quickly.
The basic idea of the invention is as follows:
(1) subtask decomposition module
The input of the method is a task core, and the profit value of each task target is determined through the task core. First, modeling needs to be performed according to the task environment. The unmanned aerial vehicle of our party is modeled as a resource platform, namely a platform carrying resources with different capabilities, and one unmanned aerial vehicle cluster is a platform group. The function formed by weighting the moving distance required by the platform and the profit value is the comprehensive profit degree, the subtask decomposition module uses a genetic algorithm to perform clustering according to the comprehensive profit degree, and task targets with relatively short comprehensive profit degrees are clustered together to form the same subtask.
(2) Automatic subtask priority sorting module
The second module of the method is a subtask priority automatic sequencing module, a plurality of harmony populations are generated by using a harmony search algorithm, each harmony population represents a priority sequence, each generated priority sequence enters a resource scheduling module for optimal scheduling, then a harmony evaluation is returned, and then the optimal harmony individuals are obtained by iteration.
(3) Resource scheduling module
The third module of the method is a resource scheduling module, and the resource scheduling is carried out according to the priority of the priority sequence in the second module by using a multidimensional dynamic list planning algorithm, and the optimal scheduling scheme is evaluated at the moment and returned to the second module to be used as the harmony evaluation. The optimal scheduling scheme corresponding to the optimal sound individuals is the optimal action scheme, and the suboptimal optimal scheme can be reserved as required.
The invention discloses an automatic generation method of an unmanned aerial vehicle group action scheme, which comprises the following steps:
the method comprises the following steps: task core and context information are determined.
Step two: and the subtask decomposition module is used for decomposing the task core into a plurality of subtasks by using a genetic algorithm.
(1) Genetic algorithm chromosome representation
The chromosome representation of the genetic algorithm is in a binary coding form, and each chromosome is composed of a matrix F based on the scene information determined in the step one, wherein n is the number of subtasks, and m is the number of known targets. The value of each element in the F matrix is determined by the following equation, and the encoding condition of the F matrix is shown in table 1.
Figure GDA0002596710500000031
Figure GDA0002596710500000032
TABLE 1
After the target allocation scheme matrix F is determined, each enemy target under the battlefield environment can be respectively allocated to subtasks of different levels according to the F matrix result. Each chromosome of the genetic algorithm is an F matrix.
(2) Fitness function of subtask decomposition genetic algorithm
Firstly, when the existing perception target of an enemy is subjected to task planning, a comprehensive income index weight is obtained according to the position and income degree of the target:
weight＝location×α+gain×(1-α)
the weight of the coefficient alpha is given by the method user, and means the priority of the method user for the position index and the income index. And arranging the comprehensive income indexes matched with each target and the subtasks into a matrix W. And performing point multiplication on the matrix W and the matrix F to obtain a distributed comprehensive income index matrix A.
And calculating the variance of each row in the comprehensive income index matrix A, then adding the variances of the rows in the comprehensive income index matrix A, and if the sum of the variances is smaller, considering that the fitness of the individual is higher. Thus, the concrete form of the fitness function is:
Figure GDA0002596710500000041
in the formula, weight (f)ab) Elements of the row a and the column b in the comprehensive income indicator matrix A are shown; e (weight)x) The expectation of a row of data in the comprehensive income index matrix A;
(3) selection operation of subtask decomposition algorithm
Adopting a wheel disc method, namely adopting a fitness proportion selection mode to select individuals, wherein the selection probability of each individual is in direct proportion to the fitness of the individual; the population scale is set as M, and the fitness of the w (w ═ 1, 2.. multidot.M) th individual in the population is set as fwThe selection probability p of the w-th individualwAs shown in the following formula：
Figure GDA0002596710500000042
(4) Interleaving of subtask decomposition algorithms
Since the object operated by the subtask decomposition algorithm and the output result are target allocation schemes, that is, the number of "1" in each individual is equal to the allocated target number, if the structure of the original individual is changed in the crossing process to make the number of "1" in the individual become more or less, the problem that the target is idle or redundant target appears in the task boundary scheme given by the algorithm will be caused, which is obviously not the expected result.
In order to ensure that the number of '1' in each row of individuals is constant after the crossover operation, an adjustment measure is provided to improve the basic crossover operation of the genetic algorithm. The method comprises the following specific steps: assuming that A and B are two parents to be crossed and a last-in first-out stack is provided, As and Bs represent values on the s-th bits of chromosomes of the two parents, if the two As and Bs are different, the two bits are not exchanged for the moment, but are put into the stack for storage, then the subsequent bits of A and B are searched continuously, if the pair of different bits of Ai and Bi is found, and Ai and Bi are also different, the two groups of gene bits of As, Bs and Ai and Bi are exchanged at the same time, and therefore the number of '1' in the individuals can be ensured to be constant after the crossing operation of the two individuals A and B.
(5) Mutation operation of subtask decomposition algorithm
Similarly, it is not desirable that the mutation operation of the genetic algorithm destroys the stability of the individuals in the original scheme, and a special mutation method is required to ensure that the number of '1's in each individual is constant after the mutation operation. The method comprises the following specific steps: and (3) determining whether a certain bit s of the gene sequence of an individual in a row is mutated according to a given mutation probability, if so, selecting another random bit i in the same row, and if the two bits are also different, directly exchanging the bit s and the bit i, otherwise, not exchanging the bit s and the bit i, thereby achieving the effect of mutation.
(6) Genetic termination conditions
And stopping the genetic algorithm when the specified genetic algebra is reached.
Through the improved genetic algorithm, the task core can be decomposed into subtasks of different levels according to the existing scene information (mainly the geographic position coordinates and the income degree of the known target).
Step three: subtask priority automatic ordering module — harmony population generation.
A harmony population is generated using a harmony search algorithm. The step is the first step of nesting the subtask priority automatic sequencing module and the resource scheduling module in the step four. The nested module realizes the functions as follows: and generating a plurality of harmony individuals by using a sub-task priority automatic ordering harmony search algorithm, wherein each harmony individual represents a sub-task priority ordering mode, and entering each sub-task priority ordering mode into a multi-dimensional dynamic list planning algorithm of the resource scheduling module in the step four for resource scheduling, wherein the obtained resource utilization rate and task completion time are harmony evaluations of the harmony individuals. The nested module has three steps to five steps.
(1) Initial parameters
The following parameters need to be initialized:
and size of acoustic memory bank HMS: is the size of the harmonic population.
And the value probability HMCR of the harmony memory bank: the probability of taking a harmony from the existing population (HM and sound bank).
Pitch fine adjustment probability PAR: the probability of taking out the harmony is fine-tuned.
Tone fine tuning bandwidth BW: the magnitude of the fine tuning.
Number of creations Tmax: the number of adjustments is the number of iterations.
(2) Initializing and acoustic memory banks
Randomly generating HMS harmony (harmony is understood as individual, HMS harmony is understood as population) from solution space by vector X1,X2,…,XHMSIndicating that each harmony represents a sort of subtask priority. Put into harmony memory bank and record correspondenceAnd f (x), the harmony library is in the form of:
Figure GDA0002596710500000061
where n represents the number of subtasks, of the ith harmonic Xi
Figure GDA0002596710500000062
Representing the priority size of the jth sub-task in the harmony entity. (x) represents the harmony assessment of the harmony individual (i.e., the subtask priority sequence).
Step four: resource scheduling Module-harmony evaluation
The step is a resource scheduling module in a nested algorithm, and the function of the resource scheduling module is to calculate the HM and f (X) in the harmony database, namely harmony assessment.
For each harmony individual, line X in HMiThe resource scheduling of tasks for a given priority is implemented using a multidimensional dynamic list programming (MDLS) algorithm.
(1) Task priority
When all predecessor tasks of a task (i.e. tasks that have to be completed before the task is processed) have been processed, the task enters a READY task set READY where the allocable resource task is located. Selecting priority in READY set
Figure GDA0002596710500000063
And the largest subtask j firstly carries out platform resource scheduling and serves as a task to be processed.
(2) Free platform set
All the callable platforms without processing tasks are put into the FREE platform set FREE, and the selection of the platform is directly selected from the FREE.
(3) Platform set selection
The platform group selection is to select a platform group for executing a task to be processed, and is a key part of the MDLS algorithm for multi-dimensional dynamic list planning. Here, the platform group is selected by calculating the priority P for each platform, i.e. the applicability of the platform to the task, and the formula is as follows:
P＝Time+∑R*T
in the formula, Time represents the Time (normalization processing) when the platform moves to a task destination, and Σ R × T represents the sum of the processing capacity of the capacity required by each task of the platform;
and then, arranging the capabilities of the superposed resources from large to small according to the priority P sequence, checking whether the superposed capabilities exceed the resource requirements of the tasks, stopping selecting the platform once the requirements of the tasks are met, and then starting pruning the platform group to remove redundant platforms.
After the platform group selection is completed, the task allocation platform is finished. And removing the required platform from the FREE platform set FREE, removing the task from the prepared task set READY, and putting the task into a new set ALREADY, namely the allocated task set.
(4) Time updating
And when the resource capacity required by the task to be processed is larger than the sum of all the platform capacities in the FREE platform set (FREE), updating the time. The time begins to lapse until the task in the distributed task set ALREADY is completed, the task immediately after the task is completed is placed into the prepared task set READY, the platform used by the completed task is released and placed into the FREE platform set FREE, the time used by each platform is recorded, the jth platform is recorded as tj. And (4) returning to the step (1) for task priority to continue the algorithm until no task is distributed. At this time, the resource scheduling is completed, and the time T to which the time is pushed at this time is recorded.
(5) Harmony evaluation
And the last step of the resource scheduling module is to perform harmony evaluation on the current resource scheduling after the multidimensional dynamic list planning algorithm is performed.
First, when the algorithm first proceeds to this point, for X at that pointiIts harmony score was calculated as follows:
the total time T (i) of the task at the moment obtained in the step (4) and the time t of each platform being usedj(i) And respectively calculating a task completion time reference value TP and a platform utilization rate reference value PP.
TP＝T
Figure GDA0002596710500000071
Wherein T represents the total time of the first assignment of task, TjRepresenting the time that each platform was allocated for use for the first time and N representing the total number of platforms.
Therefore, the platform utilization rate reference value is used for carrying out normalization processing on the task completion time and the platform utilization rate to obtain a time priority coefficient T when the algorithm is carried out each timeP(i) And PP(i)
Figure GDA0002596710500000072
Figure GDA0002596710500000081
Wherein T (i) represents the total time of the ith task, tj(i) Represents the time that the platform is used in the ith allocation, and N represents the total number of platforms.
For this approach, it is desirable to have an action plan that minimizes the time to complete the task and maximizes platform utilization efficiency. The openness parameter beta is set, and the size of the beta determines the tendency of the task time to be completed. The larger beta represents a higher requirement of the process user for shortening the time to complete the task.
The comprehensive priority coefficient is:
Figure GDA0002596710500000082
the harmony score is the reciprocal of the integrated priority coefficient,
Figure GDA0002596710500000083
then every f (X)i) Returning to the HM, the harmony library HM is completed.
Step five: subtask priority automatic ordering Module-Harmony memory Bank update
(1) Generating a new harmony
In [0,1 ]]A random number r is generated and compared with the harmony memory bank value probability HMCR, if r is<And the HMCR randomly takes out a harmony variable from the harmony memory library, otherwise, randomly generates a harmony variable from the solution space. From the above, a harmony variable is obtained, and if the harmony variable is obtained from the harmony library memory, the harmony variable needs to be fine-tuned, at [0,1 ]]A random number s is generated in between. If s<PAR (tone fine tuning probability), adjusting the obtained harmony variable according to the fine tuning bandwidth BW to obtain a new harmony variable; otherwise, no adjustment is made. Finally obtaining new harmony Xnew。
(2) Harmony evaluation
Mixing XnewAnd substituting the harmony assessment into the step four.
(3) Updating and acoustic memory banks
To XnewEvaluation was carried out, i.e. f (X)new) If it is better than the worst one of the function values in the harmony library HM, i.e. f (X)new)<f(Xworst) Then X will benewReplacing harmony X with the worst function value in harmony library HMworst(ii) a Otherwise, no modification is made.
(4) Repeating the steps (1) to (3) until the repetition times (adjustment times) reach Tmax。
Step six: and outputting an excellent scheme.
The number of the repetition times of the step five reaches TmaxThen, in the obtained harmony sound library HM, f (X) maximum harmony sound individual X is detectediThen the ith subtask priority order is the optimal order. In the four-resource scheduling module of the step XiThe corresponding resource scheduling is the optimal resource scheduling, i.e. the optimal action scheme.
The invention discloses an automatic generation method of an unmanned aerial vehicle group action scheme, which has the advantages and effects that:
1. the automatic generation method of the full-automatic unmanned aerial vehicle group action scheme is provided, the action scheme can be automatically generated according to task cores without human participation, and high automation, high response rate and intelligent cooperation are realized.
2. The generated scheme not only ensures the excellence, but also ensures the diversity, and can be sorted according to the excellence degree for selection.
3. The subtasks decomposed by the subtask decomposition module are classified according to the comprehensive income degree, and the generated action scheme is higher in execution efficiency and more definite in purpose.
4. By using the harmony search algorithm, since the singing individuals (i.e. the priority of a single task) in a single harmony can turn down/up the vocal part according to the suggestions by themselves, the efficiency is higher during searching, and an excellent solution can be achieved without traversing all the arrangements.
5. The subtask priority module is combined with the resource scheduling module, optimal scheduling and evaluation of the optimal priority sequence and the sound individuals are simultaneously carried out, the accuracy of time sequence and resource scheduling is ensured, and the calculation efficiency is increased.
6. The three modules used in the method have higher modular algorithm compatibility.
Drawings
FIG. 1 is a basic flow diagram of the method of the present invention.
FIG. 2 is a basic flow diagram of a subtask decomposition genetic algorithm.
FIG. 3 is a nested flow diagram of a subtask decomposition module and a resource scheduling module.
Fig. 4 is a gantt chart form of the output result.
Detailed Description
The method of the present invention will now be described in further detail with reference to the accompanying drawings.
The basic flow of an automatic generation method of an unmanned aerial vehicle group action scheme is shown in fig. 1, and the main part of the method is a subtask decomposition module, a subtask priority ordering module and a resource scheduling module. The following is a detailed description of the specific embodiments according to this scheme.
The examples are as follows: when a fire breaks out in a forest, 10 fire points are to be extinguished. And at present, 20 fire-extinguishing unmanned aerial vehicles are dispatched to the forest for fire fighting. The task core is to extinguish the flame of all ignition points and prevent further diffusion. Firstly, a scene model is established according to task cores, firstly, clustering decomposition is carried out in a subtask decomposition module according to the distance between an unmanned aerial vehicle and a fire point and the size of fire, the subtasks are divided into a plurality of subtasks, secondly, a plurality of priority sequences are generated by using a harmony search algorithm for waiting for resource scheduling, thirdly, unmanned aerial vehicle resource scheduling is carried out on each harmony individual priority sequence of the harmony search algorithm, and resource scheduling is carried out by using a multidimensional dynamic list planning algorithm according to the parameters such as fire control capability, fire extinguishing capability and unmanned aerial vehicle speed. And finally, outputting the optimal scheduling corresponding to the optimal harmony individual of the harmony search algorithm as an optimal action scheme. The specific process is as follows:
s1, determining task core and scene information:
from the task core, a final task goal may be determined. When the method is used, the information and the quantity of the platform and the information and the quantity of the target are judged by combining the specific scene information, and modeling is carried out. Firstly, according to the scene characteristics, the resource requirements, such as battlefield detection requirements, defense requirements, bombing requirements and the like, which need to be eliminated or achieve the target in the task are determined, and the resource capabilities, such as anti-detection capability, anti-explosion capability and anti-bombing capability, of the platform corresponding to the requirements are determined. And corresponding the requirements and the capabilities, and modeling the scene.
In this example, the task is central to extinguishing all fires and preventing further diffusion. From this task core, scenarios are modeled as shown in table 2 and table 3 below.
Number of ignition point 1 2 3 4 5 6 7 8 9 10
The size of the fire 10 8 6 8 10 8 6 8 12 12
Ability to spread fire 12 6 16 10 12 6 16 10 8 8
Distance from base 10 24 8 12 10 24 8 12 20 10
TABLE 2
Figure GDA0002596710500000101
Figure GDA0002596710500000111
TABLE 3
S2, the subtask decomposition module:
the basic flow chart of the subtask decomposition genetic algorithm is shown in FIG. 2. The module uses a genetic algorithm to break down the task core into a number of subtasks, each subtask being a set of several fire point objectives. In this step, the method user is required to give basic parameters of the genetic algorithm according to his own needs. The specific parameter size is selected according to the conventional genetic algorithm convention. In this embodiment, the population size is 30, the crossover probability is 0.97, the mutation probability is 0.01, and the maximum number of iterations is 500.
1. Chromosome representation
The chromosome representation of the genetic algorithm is in the form of binary coding, each chromosome is composed of a matrix F, wherein the number of known targets is 10, and is decomposed into 5 subtasks, based on the context information determined in step one. The value of each element in the F matrix is determined by the following equation, and the encoding condition of the F matrix is shown in table 4.
Figure GDA0002596710500000112
Figure GDA0002596710500000113
TABLE 4
After the target allocation scheme matrix F is determined, each enemy target under the battlefield environment can be respectively allocated to subtasks of different levels according to the F matrix result. Each chromosome of the genetic algorithm is an F matrix.
2. Fitness function
When the task planning is carried out, a 'comprehensive income index' weight of the target is obtained according to the position and income degree of the target by weighting:
weight＝location×α+gain×(1-α)
the weight of the coefficient alpha is given by a method user, the weight value variance of each subtask in the chromosome is calculated, then the variances of all subtasks are added, and if the sum of the variances is smaller, the fitness of the individual is considered to be higher. Thus, the concrete form of the fitness function is:
Figure GDA0002596710500000121
in the formula, weight is a comprehensive income index of an element (namely an enemy target) in a chromosome, E is an expectation of a subtask in the chromosome, and fitness is reciprocal of variance, so that the maximum value of the fitness is reached when a genetic algorithm is converged, and thus an excellent decomposition scheme ensures that the subtasks cluster targets with different comprehensive income indexes together.
3. Selection operation of subtask decomposition algorithm
Adopting a wheel disc method, namely adopting a fitness proportion selection mode to select individuals, wherein the selection probability of each individual is in direct proportion to the fitness of the individual; the population scale is set as M, and the fitness of the w (w ═ 1, 2.. multidot.M) th individual in the population is set as fwThe selection probability p of the w-th individualwThe following formula shows:
Figure GDA0002596710500000122
4. interleaving of subtask decomposition algorithms
Since the object operated by the subtask decomposition algorithm and the output result are target allocation schemes, that is, the number of "1" in each individual is equal to the allocated target number, if the structure of the original individual is changed in the crossing process to make the number of "1" in the individual become more or less, the problem that the target is idle or redundant target appears in the task boundary scheme given by the algorithm will be caused, which is obviously not the expected result.
In order to ensure that the number of '1' in each row of individuals is constant after the crossover operation, an adjustment measure is provided to improve the basic crossover operation of the genetic algorithm. The method comprises the following specific steps: if the s-th position needs to be crossed and the two positions of As and Bs are different, the two positions are not exchanged for the first time, but are put into the stack for storage, then the subsequent positions of A and B are searched continuously, if the pair of different positions of Ai and Bi is found, and Ai and Bi are also different, the two groups of gene positions of As, Bs, Ai and Bi are exchanged simultaneously, and thus the number of '1' in the A and B individuals can be ensured to be fixed after the crossing operation.
The specific operation of judging whether the intersection is needed is as follows: the specific operation method is to generate a random number x (x is more than or equal to 0 and less than or equal to 1), and if x is less than cross probability PjThen carry out crossover operation if x > crossover probability PjAnd no processing is performed.
5. Mutation operation of subtask decomposition algorithm
Similarly, it is not desirable that the mutation operation of the genetic algorithm destroys the stability of the individuals in the original scheme, and a special mutation method is required to ensure that the number of '1's in each individual is constant after the mutation operation. The method comprises the following specific steps: and (3) determining whether a certain bit s of the gene sequence of an individual in a row is mutated according to a given mutation probability, if so, selecting another random bit i in the same row, and if the two bits are also different, directly exchanging the bit s and the bit i, otherwise, not exchanging the bit s and the bit i, thereby achieving the effect of mutation.
The specific operation of judging whether the mutation is needed is as follows: the specific operation method is to generate a random number x (x is more than or equal to 0 and less than or equal to 1), and if x is less than the variation probability PmThen, mutation operation is performed, if x > mutation probability PmAnd no processing is performed.
6. Genetic termination conditions
And stopping the genetic algorithm when the specified genetic algebra is reached.
Through the improved genetic algorithm, the task core can be decomposed into subtasks of different levels according to the existing scene information (mainly the geographic position coordinates and the income degree of the known target).
In this embodiment, the subtask allocation situation is as follows:
subtask 1: target 9, target 10.
Subtask 2: target 1, target 5.
Subtask 3: target 2, target 6.
Subtask 4: target 3, target 7.
Subtask 5: target 4, target 8.
S3, subtask priority automatic ordering module-Harmony population Generation:
the 10 target fires are divided in step S2 to generate 5 subtasks, in which the 5 subtasks are prioritized.
The following modeling is first performed. The Harmony Search (HS) algorithm is used, each harmony individual comprises five singers and represents five subtasks, and the singing melody represents the priority level of the subtask. And randomly generating 10 harmony individuals from the solution space, wherein the harmony individuals represent 10 seed task priority ordering modes.
The following parameters are given by the method user: and size of acoustic memory bank HMS: 10. and the value probability HMCR of the harmony memory bank: 0.5. pitch fine adjustment probability PAR: 0.1. tone fine tuning bandwidth BW: 0.1. number of creations Tmax：100。
(2) Initializing and acoustic memory banks
Randomly generating 10 harmony words from solution space by vector X1,X2,…,X10Indicating that each harmony represents a sort of subtask priority. Putting the harmony database into a harmony memory library, and recording corresponding f (X), wherein the harmony database is as follows:
Figure GDA0002596710500000141
s4, resource scheduling Module-harmony assessment
And scheduling resources for each harmony individual in the current harmony population, namely a group of subtask priority sequencing modes. The resource scheduling is performed by using a multidimensional dynamic list programming (MDLS) algorithm with the model established in step S1. And after the resource scheduling is finished, the optimal scheduling corresponding to each sound individual is obtained. Harmony evaluation is performed on the optimal scheduling.
(1) Task priority
When all predecessor tasks of a task (i.e. tasks that have to be completed before the task is processed) have been processed, the task enters a READY task set READY where the allocable resource task is located. Selecting priority in READY set
Figure GDA0002596710500000142
And the largest subtask j firstly carries out platform resource scheduling and serves as a task to be processed.
(2) Free platform set
All the callable platforms without processing tasks are put into the FREE platform set FREE, and the selection of the platform is directly selected from the FREE.
(3) Platform set selection
The platform group selection is to select a platform group for executing a task to be processed, and is a key part of the MDLS algorithm for multi-dimensional dynamic list planning. Here, the platform group is selected by calculating the priority P for each platform, i.e. the applicability of the platform to the task, and the formula is as follows:
P＝Time+∑R*T
in the formula, Time represents the Time (normalization processing) when the platform moves to a task destination, and Σ R × T represents the sum of the processing capacity of the capacity required by each task of the platform;
and then, arranging the capabilities of the superposed resources from large to small according to the priority P sequence, checking whether the superposed capabilities exceed the resource requirements of the tasks, stopping selecting the platform once the requirements of the tasks are met, and then starting pruning the platform group to remove redundant platforms.
After the platform group selection is completed, the task allocation platform is finished. And removing the required platform from the FREE platform set FREE, removing the task from the prepared task set READY, and putting the task into a new set ALREADY, namely the allocated task set.
(4) Time updating
And when the resource capacity required by the task to be processed is larger than the sum of all the platform capacities in the FREE platform set (FREE), updating the time. And (3) the time begins to be shifted until the tasks in the distributed task set ALREADY are completed, placing the tasks immediately after the tasks are completed into a prepared task set READY, releasing the platforms used by the tasks to be completed, placing the platforms into a FREE set, and recording the time used by each platform, wherein the jth platform is recorded as tj. And (4) returning to the step (1) for task priority to continue the algorithm until no task is distributed. At this time, the resource scheduling is completed, and the time T to which the time is pushed at this time is recorded.
(5) Harmony evaluation
And the last step of the resource scheduling module is to perform harmony evaluation on the current resource scheduling after the multidimensional dynamic list planning algorithm is performed.
First, atThe first time the algorithm proceeds to this point, for X at this pointiIts harmony score was calculated as follows:
the total time T (i) of the task at the moment obtained in the step (4) and the time t of each platform being usedj(i) And respectively calculating a task completion time reference value TP and a platform utilization rate reference value PP.
TP＝T
Figure GDA0002596710500000161
Wherein T represents the total time of the first assignment of task, TjRepresenting the time that each platform was allocated for use for the first time and N representing the total number of platforms.
Therefore, the platform utilization rate reference value is used for carrying out normalization processing on the task completion time and the platform utilization rate to obtain a time priority coefficient T when the algorithm is carried out each timeP(i) And PP(i)
Figure GDA0002596710500000162
Figure GDA0002596710500000163
Wherein T (i) represents the total time of the ith task, tj(i) Represents the time that the platform is used in the ith allocation, and N represents the total number of platforms.
For this approach, it is desirable to have an action plan that minimizes the time to complete the task and maximizes platform utilization efficiency. The openness parameter beta is set, and the size of the beta determines the tendency of the task time to be completed. The larger beta represents a higher requirement of the process user for shortening the time to complete the task.
The comprehensive priority coefficient is:
Figure GDA0002596710500000164
the harmony score is the reciprocal of the overall priority coefficient, which is
Figure GDA0002596710500000165
Then every f (X)i) Returning to the HM, the harmony library HM is completed.
Calculated harmonic scores were:
Figure GDA0002596710500000171
s5, subtask priority automatic ranking Module-and harmonic library memory update:
(1) generating a new harmony
In [0,1 ]]A random number r is generated and compared with the harmony memory bank value probability HMCR, if r is<And the HMCR randomly takes out a harmony variable from the harmony memory library, otherwise, randomly generates a harmony variable from the solution space. Obtaining a harmony variable from the above, and if the harmony variable is obtained from the harmony memory bank, fine-tuning the harmony variable at [0,1 ]]A random number s is generated in between. If s<And PAR (pitch fine tuning probability), adjusting the obtained harmonic variable according to the fine tuning bandwidth BW to obtain a new harmonic variable. Otherwise, no adjustment is made. Finally obtaining new harmony Xnew。
(2) Harmony evaluation
Mixing XnewAnd substituting the harmony assessment into the step four.
(3) Updating and acoustic memory banks
To XnewEvaluation was carried out, i.e. f (X)new) If it is better than the worst one of the function values in the harmony library HM, i.e. f (X)new)<f(Xworst) Then X will benewReplacing harmony X with the worst function value in harmony library HMworst(ii) a Otherwise, no modification is made.
(4) Repeating the steps (1) to (3) until the repetition times (adjustment times) reach Tmax。
The resulting cohort was as follows:
Figure GDA0002596710500000181
steps S3-S5 are the nesting of the subtask priority automatic ordering module and the resource scheduling module, and the flow chart is shown in FIG. 3.
S6, action plan generation
What is obtained at this point after the convergence algebra is reached by step S5 are 10 excellent harmony individuals, i.e., subtask prioritization, and corresponding optimal assignments for each individual. The allocation scheme may be represented in gantt charts, such as fig. 4. The combination of the scheduling scheme and the sub-task decomposition scheme in step S2 is the output result of the automatic generation method for the action scheme of the present unmanned aerial vehicle fleet.

Claims (3)

1. An automatic generation method for an action scheme of an unmanned aerial vehicle group is characterized by comprising the following steps: the method comprises the following steps:
the method comprises the following steps: determining task core and scene information;
step two: the subtask decomposition module is used for decomposing the task core into a plurality of subtasks by using a genetic algorithm;
step three: subtask priority automatic ordering module — harmony population generation;
generating a harmony population using a harmony search algorithm: generating a plurality of harmony individuals by using a subtask priority automatic ordering harmony search algorithm, wherein each harmony individual represents a subtask priority ordering mode, and entering each subtask priority ordering mode into a multidimensional dynamic list planning algorithm of the resource scheduling module in the step four for resource scheduling, wherein the obtained resource utilization rate and task completion time are harmony evaluation of the harmony individual;
step four: resource scheduling Module-harmony evaluation
For each harmony individual, line X in HMiThe multi-dimensional dynamic list planning MDLS algorithm is used for realizing the resource scheduling of tasks with given priority;
step five: subtask priority automatic ordering Module-Harmony memory Bank update
Step six: outputting an excellent scheme;
detecting and evaluating the maximum harmony individual X (X) in the harmony database HM obtained after the update repetition times of the harmony memory database reach Tmax in the step fiveiIf so, the ith subtask priority is ordered into the optimal ordering; in the four-resource scheduling module of the step XiThe corresponding resource scheduling is the optimal resource scheduling, namely the optimal action scheme;
the specific process of the third step is as follows:
(1) initial parameters, the following parameters need to be initialized:
and size of acoustic memory bank HMS: is the size of the harmony population;
and the value probability HMCR of the harmony memory bank: the probability of taking a harmony from the existing population, namely the HM and the sound library;
pitch fine adjustment probability PAR: probability of fine tuning the extracted harmony sound;
tone fine tuning bandwidth BW: the magnitude of the fine tuning;
number of creations Tmax: the number of times of adjustment is the number of times of iteration;
(2) initializing and acoustic memory banks
Randomly generating HMS harmony from solution space, wherein the harmony is understood as individual, the HMS harmony is understood as population, and vector X is used for1,X2,…,XHMSRepresenting, each harmony represents a sorting mode of the subtask priority; putting the harmony library into a harmony memory library, and recording the corresponding f (X), wherein the harmony library is in the form of:
Figure FDA0002653028260000021
where n represents the number of subtasks, the ith harmony XiIn (1)
Figure FDA0002653028260000022
Represents the jth subsidiary in the harmony objectThe priority size of the service; (x) harmony measures representing the harmony individuals, i.e., the subtask priority sequences;
the specific process of the step four is as follows:
(1) task priority
When all the preceding tasks of a certain task, namely the tasks which need to be completed before the task is processed, are processed, the task enters a READY task set READY where the task with the allocable resources is located; selecting priority in READY set
Figure FDA0002653028260000023
The largest subtask j is firstly subjected to platform resource scheduling and used as a task to be processed;
(2) free platform set
All the callable platforms without processing tasks are put into a FREE platform set (FREE), and the selection of the platform is directly selected from the FREE;
(3) platform set selection
The platform group selection is to select a platform group for executing a task to be processed, specifically, to calculate the priority P for each platform, that is, the applicability of the platform to the task, and the formula is as follows:
P＝Time+∑R*Time
in the formula, Time represents the Time when the platform moves to a task destination, and Σ R Time represents the sum of the processing capacity of the capacity required by each task of the platform;
secondly, arranging the capabilities of the superposed resources from large to small according to the priority P sequence, checking whether the superposed capabilities exceed the resource requirements of the tasks, stopping selecting the platform once the requirements of the tasks are met, and then starting pruning the platform group to remove redundant platforms;
after the platform group selection is completed, the task distribution platform is finished; removing a required platform from an idle platform set FREE, removing the task from a prepared task set READY, and putting the task into a new set ALREADY, namely an allocated task set;
(4) time updating
When the resource capacity required by the task to be processed is larger than nullUpdating time when the sum of all platform capabilities in the FREE platform set FREE is obtained; the time begins to lapse until the task in the distributed task set ALREADY is completed, the task immediately after the task is completed is placed into the prepared task set READY, the platform used by the completed task is released and placed into the FREE platform set FREE, the time used by each platform is recorded, the jth platform is recorded as tj(ii) a Then, returning to the step (1) to continue the algorithm in the task priority until no task can be allocated; at this time, the resource scheduling is finished, and the time T to which the time is pushed is recorded;
(5) harmony evaluation
First, for XiIts harmony score was calculated as follows:
the total time T (i) of the task at the moment obtained in the step (4) and the time t of each platform being usedj(i) Respectively calculating a task completion time reference value TP and a platform utilization rate reference value PP;
TP＝T
Figure FDA0002653028260000031
wherein T represents the total time of the first assignment of task, TjRepresenting the time for which each platform is used for the first time, and N representing the total number of platforms;
using the platform utilization rate reference value to carry out normalization processing on the task completion time and the platform utilization rate to obtain a time priority coefficient T when the algorithm is carried out each timeP(i) And PP(i)
Figure FDA0002653028260000032
Figure FDA0002653028260000033
Wherein T (i) represents the total time of the ith task, tj(i) Represents the ith allocationThe time each platform was used, N represents the total number of platforms;
setting an openness parameter beta, wherein the size of the beta determines the time tendency for completing the task; the larger beta is, the higher the requirement of a method user for shortening the task completion time is;
the comprehensive priority coefficient is:
Figure FDA0002653028260000034
the harmony score is the reciprocal of the overall priority coefficient, which is
Figure FDA0002653028260000035
Then every f (X)i) Returning to the HM, the harmony library HM is completed.
2. The automated drone swarm behavior scheme generation method of claim 1, wherein: the specific process of the second step is as follows:
(1) genetic algorithm chromosome representation
The chromosome representation of the genetic algorithm adopts a binary coding form, and each chromosome is composed of a matrix F based on the scene information determined in the step one, wherein n is the number of subtasks, and m is the number of known targets; the value of each element in the F matrix is determined by the following equation:
Figure FDA0002653028260000041
after the target allocation scheme matrix F is determined, each enemy target under the battlefield environment can be respectively allocated to subtasks of different levels according to the F matrix result; each chromosome of the genetic algorithm is an F matrix;
(2) fitness function of subtask decomposition genetic algorithm
Firstly, when the existing perception target of an enemy is subjected to task planning, a comprehensive income index weight is obtained according to the position location and the income gain degree of the target:
weight＝location×α+gain×(1-α)
wherein the weight of the coefficient alpha is given by a method user, meaning the priority of the method user for the position index and the income index; arranging the comprehensive income indexes matched with each target and the subtasks into a matrix W; performing dot multiplication on the matrix W and the matrix F to obtain a distributed comprehensive income index matrix A;
calculating the variance of each row in the comprehensive income index matrix A, then adding the variances of the rows in the comprehensive income index matrix A, and if the sum of the variances is smaller, considering that the fitness of the individual is higher; thus, the concrete form of the fitness function is:
Figure FDA0002653028260000042
in the formula, weight (f)ab) Elements of the row a and the column b in the comprehensive income indicator matrix A are shown; e (weight)x) The expectation of a row of data in the comprehensive income index matrix A;
(3) selection operation of subtask decomposition algorithm
Adopting a wheel disc method, namely adopting a fitness proportion selection mode to select individuals, wherein the selection probability of each individual is in direct proportion to the fitness of the individual; setting the population scale as M and the fitness of the w-th individual in the population as fwThe selection probability p of the w-th individualwWherein, w is 1,2, said, M, as shown in the following formula:
Figure FDA0002653028260000051
(4) interleaving of subtask decomposition algorithms
In order to ensure that the number of '1' in each row of individuals is fixed after cross operation, an adjustment measure is provided to improve the basic cross operation of a genetic algorithm, and the specific steps are as follows: assuming that A and B are two parents to be crossed and a last-in first-out stack is arranged, As and Bs represent values on the s-th bits of chromosomes of the two parents, if the two bits of As and Bs are different, the two bits are not exchanged temporarily, but are put into the stack for storage, then the subsequent bits of A and B are searched continuously, if the pair of different bits of Ai and Bi is found, and Ai and Bi are also different, the two groups of gene bits of As, Bs and Ai and Bi are exchanged simultaneously, so that the number of '1' in the individuals can be ensured to be constant after the crossing operation of the two individuals A and B;
(5) mutation operation of subtask decomposition algorithm
Similarly, the mutation operation of the genetic algorithm is not expected to damage the stability of the individuals in the original scheme, and a special mutation method is required to be adopted in order to ensure that the number of '1' in each individual is fixed after the mutation operation; the method comprises the following specific steps: determining whether a certain bit s of a row of individual gene sequences has mutation according to a given mutation probability, if so, selecting another random bit i in the same row, and if the two bits are also different, directly exchanging the bit s and the bit i, otherwise, not exchanging the bit s and the bit i, thereby achieving the effect of mutation;
(6) genetic termination conditions
And stopping the genetic algorithm when the specified genetic algebra is reached.
3. The automated drone swarm behavior scheme generation method of claim 1, wherein: the concrete process of the step five is as follows:
(1) generating a new harmony
In [0,1 ]]A random number r is generated and compared with the harmony memory bank value probability HMCR, if r is<HMCR, randomly taking out a harmony variable from a harmony memory library, otherwise, randomly generating a harmony variable from a solution space; from the above, a harmony variable is obtained, and if the harmony variable is obtained from the harmony library memory, the harmony variable needs to be fine-tuned, at [0,1 ]]Generating a random number s; if s<The tone fine-tuning probability PAR, the obtained harmony variable is processed according to the fine-tuning bandwidth BWAdjusting to obtain a new harmony variable; otherwise, no adjustment is made; finally obtaining new harmony Xnew；
(2) Harmony evaluation
Mixing XnewSubstituting into the step four to carry out harmony evaluation;
(3) updating and acoustic memory banks
To XnewEvaluation was carried out, i.e. f (X)new) If it is better than the worst one of the function values in the harmony library HM, i.e. f (X)new)<f(Xworst) Then X will benewReplacing harmony X with the worst function value in harmony library HMworst(ii) a Otherwise, no modification is made;
(4) repeating the steps (1) to (3) until the repetition times reach Tmax。
CN201810988692.3A 2018-08-28 2018-08-28 Automatic generation method for unmanned aerial vehicle group action scheme Active CN109242290B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810988692.3A CN109242290B (en) 2018-08-28 2018-08-28 Automatic generation method for unmanned aerial vehicle group action scheme

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810988692.3A CN109242290B (en) 2018-08-28 2018-08-28 Automatic generation method for unmanned aerial vehicle group action scheme

Publications (2)

Publication Number Publication Date
CN109242290A CN109242290A (en) 2019-01-18
CN109242290B true CN109242290B (en) 2020-11-27

Family

ID=65068045

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810988692.3A Active CN109242290B (en) 2018-08-28 2018-08-28 Automatic generation method for unmanned aerial vehicle group action scheme

Country Status (1)

Country Link
CN (1) CN109242290B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109947130A (en) * 2019-04-03 2019-06-28 广东电网有限责任公司 A kind of transmission of electricity multiple aircraft formation tour management method and system based on grid-based management
CN113050688B (en) * 2021-03-22 2022-03-29 中国人民解放军国防科技大学 Planning method for multi-unmanned aerial vehicle collaborative search path in key target sealing control
CN113553777B (en) * 2021-09-18 2021-11-30 中国人民解放军国防科技大学 Anti-unmanned aerial vehicle swarm air defense deployment method, device, equipment and medium
CN115239204B (en) * 2022-09-19 2023-02-14 中国电子科技集团公司第十四研究所 Collaborative task planning method for multi-platform unmanned aerial vehicle-mounted radio frequency system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001350799A (en) * 2000-06-09 2001-12-21 Sanyo Electric Co Ltd Concurrent design supporting system
CN103279812A (en) * 2013-06-19 2013-09-04 河海大学 Task and resource allocation optimizing method capable of eliminating priority conflicts
CN103631758A (en) * 2013-11-21 2014-03-12 陕西理工学院 Method for solving non-linear programming and absolute value equation through improved harmony search algorithm
CN104698838A (en) * 2014-12-23 2015-06-10 清华大学 Discourse domain based dynamic division and learning fuzzy scheduling rule mining method
US20150170093A1 (en) * 2013-12-17 2015-06-18 Asd Inc. Break Management in Business Center or the Like
CN106600147A (en) * 2016-12-15 2017-04-26 合肥工业大学 Resolvable task oriented task assigning method and apparatus for multiple unmanned aerial vehicles

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001350799A (en) * 2000-06-09 2001-12-21 Sanyo Electric Co Ltd Concurrent design supporting system
CN103279812A (en) * 2013-06-19 2013-09-04 河海大学 Task and resource allocation optimizing method capable of eliminating priority conflicts
CN103631758A (en) * 2013-11-21 2014-03-12 陕西理工学院 Method for solving non-linear programming and absolute value equation through improved harmony search algorithm
US20150170093A1 (en) * 2013-12-17 2015-06-18 Asd Inc. Break Management in Business Center or the Like
CN104698838A (en) * 2014-12-23 2015-06-10 清华大学 Discourse domain based dynamic division and learning fuzzy scheduling rule mining method
CN106600147A (en) * 2016-12-15 2017-04-26 合肥工业大学 Resolvable task oriented task assigning method and apparatus for multiple unmanned aerial vehicles

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
Application of harmony search to vehicle routing;Geem Z W etal.;《American Journal of Applied Sciences》;20051231;第2卷(第12期);第1552-1557页 *
基于遗传算法的多学科设计优化分解方法;陈云霞 等;《北京航空航天大学学报》;20090630;第35卷(第6期);第673-677页 *
战役任务计划的数学描述与求解算法研究;阳东升 等;《系统工程理论与实践》;20060131(第1期);第26-34页 *
混合变邻域和声搜索的独立任务调度问题研究;姜华 等;《计算机工程与设计》;20131031;第34卷(第10期);第3516-3519页 *

Also Published As

Publication number Publication date
CN109242290A (en) 2019-01-18

Similar Documents

Publication Publication Date Title
CN109242290B (en) Automatic generation method for unmanned aerial vehicle group action scheme
Tsai et al. Parallel cat swarm optimization
CN103279793B (en) A kind of unmanned vehicle formation method for allocating tasks determined under environment
CN109409773B (en) Dynamic planning method for earth observation resources based on contract network mechanism
CN114545975A (en) Multi-unmanned aerial vehicle system task allocation method integrating multi-target evolution algorithm and contract network algorithm
CN112561395A (en) Unmanned aerial vehicle cooperation method, system, device, electronic equipment and storage medium
CN115755954B (en) Routing inspection path planning method, system, computer equipment and storage medium
US20150170052A1 (en) Method of reducing resource fluctuations in resource leveling
CN109961130A (en) A kind of method and apparatus that target object position is determined based on particle swarm algorithm
CN109523178A (en) A kind of O&amp;M method and device towards power communication scene
CN112862380A (en) Project type product assembly workshop personnel scheduling method and device based on hybrid algorithm and storage medium
CN116166048A (en) Unmanned aerial vehicle group fault-tolerant task planning method
CN117077981B (en) Method and device for distributing stand by fusing neighborhood search variation and differential evolution
CN111831421B (en) Task allocation method and terminal equipment
Alexandrescu et al. A genetic algorithm for mapping tasks in heterogeneous computing systems
CN117582652A (en) Weiqi man-machine playing method, device, computer equipment and storage medium
Zhou et al. A novel mission planning method for UAVs’ course of action
CN116050086A (en) Multi-machine air combat task allocation method based on genetic algorithm
CN116339973A (en) Digital twin cloud platform computing resource scheduling method based on particle swarm optimization algorithm
RU80604U1 (en) AUTOMATED RESOURCE DISTRIBUTION SYSTEM FOR OPTIMUM SOLUTION OF TARGET TASKS
CN106709572A (en) Data processing method and equipment
Mazurek et al. Fleet mix computation using evolutionary multiobjective optimization
Rezaeian et al. Scheduling of a flexible flow shop with multiprocessor task by a hybrid approach based on genetic and imperialist competitive algorithms
CN110059806A (en) A kind of multistage weighted network community structure detection method based on power-law function
Jingfang et al. The way of solving traveling salesman problem the research on scheduling in AS/RS

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