CN108830373A - The modeling method that the extensive intelligent group of imitative starling cluster flight independently cooperates with - Google Patents
The modeling method that the extensive intelligent group of imitative starling cluster flight independently cooperates with Download PDFInfo
- Publication number
- CN108830373A CN108830373A CN201810587885.8A CN201810587885A CN108830373A CN 108830373 A CN108830373 A CN 108830373A CN 201810587885 A CN201810587885 A CN 201810587885A CN 108830373 A CN108830373 A CN 108830373A
- Authority
- CN
- China
- Prior art keywords
- agent
- rule
- group
- starling
- cluster
- 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.)
- Granted
Links
Classifications
-
- 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]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention proposes the modeling methods that a kind of extensive intelligent group of imitative starling cluster flight independently cooperates with, including initialization population and its parameter, it sets under original state, all Agent are using initial position as personal best particle, and all Agent are using the optimal value of oneself 6 or 7 closest neighbour as local optimum；Fitness function calculates, and sets the neighbours distribution anisotropy around each Agent individual, but its closest 6 or 7 neighbour is isotropism, the interaction between Agent individual depends on topology distance；Selection updates companion, including selecting one of 6 or 7 closest neighbours to update companion as it for certain Agent；Define the interaction relationship between Agent；Carry out the update of Agent speed and position.The present invention will have boundless application prospect in fields such as military affairs, emergency, industry, the medical treatment such as unmanned plane cluster close/intra, large set place crowd's emergency evacuation, large-scale machines crowd's body work compound, transmission control.
Description
Technical field
The invention belongs to the big of artificial intelligence colony intelligence applied technical field more particularly to a kind of flight of imitative starling cluster
The modeling method that scale intelligent group independently cooperates with.
Background technique
Many application fields, such as unmanned plane cluster close/intra, large-scale public place crowd emergency evacuation, industrial machine
Crowd's body work compound etc., needing extensive intelligent body to cooperate jointly could complete.In intelligent group application system, intelligence
The individual capability of energy body (such as sensor, robot, aircraft) is limited, but its group can show efficient cooperative cooperating
Ability and advanced intelligent coordinated level.With the continuous development of the technologies such as computer network, communication, distribution calculating, perhaps
More real application systems often become very huge and complicated, so that knowledge, computing resource etc. of the single intelligent body because of individual
Limitation and it effectively cannot be handled and managed, thus the collaboration of extensive intelligent group plays in numerous applications
Highly important effect.The team unity of intelligent body how is set to reach maximization effect, related intelligent group Synergy is ground
Study carefully be all the time swarm intelligence important topic and key, the purpose is to study dispersion, how autonomous intelligent body to utilize
Collective behavior cooperates, and efficiently, farthest completes the complex task that single intelligent body is difficult to complete jointly, related big
Scale intelligent group systematic research has very important realistic meaning.
In recent years, many scholars at home and abroad have carried out deeply Research on Interactive Problem and have widely studied, from Partial controll strategy
Cluster control problem is studied, by design local rule and control strategy, cluster cooperation is made to emerge desired behavior.To collection
Derived from (1987) such as Reynolds to the analog simulation of flock of birds flight behavior, they propose individual in cluster for the research of group's behavior
Three simple heuristic rules are followed, i.e. aggregation (Cohesion), separation (Separation), alignment (Alignment), building
Aggregation behaviour Boids model.From the point of view of system control, which is substantially a kind of not against central control machine system, is used
Local rule control strategy reaches the thought and strategy of cluster collaboration, although simply, it is largely effective.It is former in Reynolds tri-
On the basis of then, some scholars propose a large amount of crowd hazards model, wherein more classical is Swarm model, Cucker-
Smale model and Cavagna model etc..Swarm is described as the adjacent body of some interactions by Kwong etc. (2003)
Gather, emulation is controlled to Swarm model progress Aggregation behaviour, including aggregation, pitch of the laps move, move, lined up directly around the figure of eight
Linear motion etc..Wang etc. (2005) is then by increasing Aggregation behaviour control of the LOCAL FEEDBACK information to mobile autonomous agent group
System designs cluster local rule model, demonstrates the feasibility of method.Cucker and Smale (2007) are retouched using adjacency matrix
Interaction strength between individual is stated, the interactional model of speed between description individual emphatically is proposed, only accounts in three principles
Alignment principle.But these methods have the shortcomings that it is obvious, the relationship between group's behavior and cluster models parameter be it is unknown,
Need to determine suitable parameter value by a large amount of emulation experiment, at the same collective motion behavior to the value of parameter often excessively
Sensitivity requires very remote from the control of real system.Cavagna etc. (2015) attempts to establish general theory to unify crowd hazards
Model, provides the inertia spin model of collective motion, and when flock of birds flight propagates information similar to the spin in magnetic material outward
Wave.However group system is usually highly complex, Aggregation behaviour is extremely various, is realized by system parameter adjusting method to big quantity set
The limitation that the control of group's behavior has its certain, so relying solely on Partial controll strategy is not able to satisfy large-scale cluster system
Effective control.
Biologist is by having carried out deep observation and research to starling group in recent years.It is every at dusk, some
Regional overhead, tens of thousands of or even hundreds of thousands of starlings flock together flight, and peculiar place is that entire flock of birds is flying
Fully synchronized between individual in the process, fly mechanics are similar to the equilibrium threshold system of the instantaneous transition of snowslide and Crystallization,
But almost instantaneous signal processing speed, this phenomenon cause the broad interest of researcher all over the world.Ballerini etc.
(2008) three-dimensional position that particular individual in starling group is had recorded using computer vision technique, finds individual distribution in flock of birds
There are anisotropy.In huge starling group, individual using topology distance (topological distance) with
Its nearest 6~7 individual interacts, and is not determined by the metric range of individual in population (metric distance).Bode
Deng (2010) observing as a result, developing the fauna based on individual collects motion model according to Ballerini etc., but the model lacks
Weary group's guidance mechanism, entire group movement direction is aimless.Young etc. (2013) gives then by the method for systematology
Gone out the reason of generating this observation result, in perception there are when uncertain factor, intelligent body and around it 6,7 neighbours into
Row interaction helps to optimize the balance between Work team Cohesiveness and individual effort.The newest biology of the above is important to be found to be
Extensive intelligent group synergistic application provides a brand-new thinking.Cavagna (2010) is proposed through polarization come degree
This topological interaction mechanism is introduced population and calculated by the whole order degree for measuring starling group under displacement-speed frame
Method makes it have better adaptivity while with starling flight characteristic, but still without proposing complete realization side
Case.Hereford and Blum (2011) proposes FlockOpt algorithm, is improved Boids model, and starling group is moved
Model is combined with swarm intelligence algorithm, solves the problems, such as that optimal value is found in unimodal search space, but not can solve population for multi-peak searching and ask
Topic.The reflex action of starling collective is introduced particle swarm algorithm to increase population diversity by Netjinda etc. (2015), is realized
Wider search space range, avoids sub-optimum solution.Qiu Huaxin and section beach (2017) have been attempted this flock of birds group
Collection fly mechanics are introduced into the autonomous cluster formation control practical application of unmanned plane, and Preliminary Study Results show to be based on the mechanism
It is combined with each other with swarm intelligence collaborative research with feasibility.
The current newest research results both at home and abroad of summary fail to break through autonomous collaboration of the intelligent group on the order of magnitude.
Summary of the invention
The object of the present invention is to provide what the extensive intelligent group that a kind of starling collective behaviour inspires independently cooperateed with to build
Mould method.It is different from existing particle swarm algorithm, genetic algorithm etc., in conjunction with the collective behaviours such as starling animal ethology and calculates life
Object newest research results parse the internal action of starling biocenose behavioral mechanism and non-stop layer self-organizing and coordinate machine
System establishes mapping mechanism of the starling group to extensive intelligent group synergistic application.
In order to achieve the above object, the extensive intelligent group of imitative starling collective behaviour mechanism provided by the invention is autonomous
The modeling method of collaboration, includes the following steps,
Step S1, initialization population and parameter, including setting original state under, all Agent all using initial position as
Personal best particle, and all Agent are using initial position as individual local optimum position；
Step S2, fitness function calculate, including follow following rule (1) with rule (2), construct topological mechanism of action frame
Frame,
Rule (1), the neighbours distribution anisotropy around each Agent individual, but its closest 6 or 7 neighbour be it is each to
The same sex；
(2), the interaction between Agent individual depends on topology distance to rule；
Step S3, selection update companion, including selecting one of 6 or 7 closest neighbours same as updating for Agent i
With Agent j；
Step S4 defines the interaction relationship between Agent；
Step S5, carries out the update of Agent speed and position, return step S2 arrive at the destination until Agent group or
Cycle-index reaches maximum evolutionary generation.
Moreover, update companion Agent j follows following rule and (3) chooses in step S3, then (4) calculated according to rule suitable
Angle value is answered,
Rule (3), according to pj~1/dijPrinciple, be Agent i around it within the scope of certain radius of view rV from closest
6 or 7 neighbours in selection update companion Agent j,
Wherein, pjFor probability, dijFor the distance between Agent i and Agent j；
(4), to selected update companion Agent j rule is evaluated using fitness function, including according to preset
Fitness function threshold value fthreshold；If Agent j fitness value is greater than fthreshold, then fitness value is poor, Agent j quilt
It eliminates, Agent i keeps oneself original flying method；Otherwise Agent i selects update companion of the Agent j as oneself.
Moreover, defining the interaction relationship between Agent in step S4, implementation is as follows,
Definition repels radius rE, keeps tri- radius parameters of radius rM and Attraction diameter rA, defines opening up between Agent
It flutters relationship and meets following rule,
(5), exclusion rule, Agent i repels other Agent j in its short range to rule, i.e., if dij<RE, then
The heading of Agent i is modified to fly away from the direction of Agent j；
(6) rule, keeps rule, Agent i is immediately following other Agent j within the scope of its moderate distance；
(7) rule, attracts rule, Agent i attracts it compared with other Agent j in far range, i.e., if dij>RA,
The heading for then modifying Agent i is to fly towards the direction of Agent j；
Rule (8), if dij>Any interaction does not occur between Agent i and Agent j then for rA.
Moreover, carrying out the update of Agent speed and position, including (9) introduce polarization according to following rule and make in step S5
With the factor, Agent group is controlled,
(9) rule, measures the whole order degree of group by defining polarization Φ, reflects that the cluster integrally flies
The consistent degree in direction,
Wherein, viIt is the speed of Agent i, | | vi| | to calculate viNorm in its metric space；As Φ=0, table
Bright cluster entirety heading is disorderly and unsystematic；As Φ → 1, show that cluster is whole essentially toward same direction.
Moreover, carrying out the update of Agent speed and position in step S5, including the updated speed of limitation meets rule
(10),
(10), the speed variation of each Agent is limited to [v to rulemin,vmax] in, vmin、vmaxRespectively Agent
Minimum speed, the maximum speed of body.
Moreover, being used for unmanned plane cluster close/intra.
The technical characterstic of the present invention unlike the prior art is as follows：
1, Cavagna (2010) proposes the whole order degree that starling group is measured by polarization, in displacement-speed
It spends under frame, this topological interaction mechanism is introduced into particle swarm algorithm, makes it while with starling flight characteristic
With better adaptivity, but due to the complexity of starling cluster flight, still without proposing actual, complete realization side
Case.The present invention is to be put forward for the first time modeling method that the extensive intelligent group of imitative starling cluster flight independently cooperates with and complete
Implementation and specific steps.
2, the essence of intelligent group collaboration is a large amount of Agent individuals under the stimulation of environment, is loaded jointly certain identical
Rule of conduct generates certain order phenomenon macroscopically by the feedback effect of these rules.The present invention proposes that intelligent group exists
The 10 basic simple rules that should be followed in collaboration, i.e., rule (1)~rule (10).
3, the present invention is using the thought of " non-stop layer self-organizing synergism ", and there is no the overall points of view for Agent individual, it is only closed
Its 6 or 7 closest neighbour of the heart find optimal value, this part thought and current classical particle swarm algorithm in part
The overall point of view be completely different.
4, Hereford and Blum (2011) proposes FlockOpt algorithm, is improved Boids model, by starling
Group's motion model is combined with swarm intelligence algorithm, solves the problems, such as that optimal value is found in unimodal search space, but not can solve multimodal
Search problem.The present invention introduces algorithm under displacement-speed frame, by topological interaction mechanism, and carries out newly to parameter
Setting, make its with starling group flight characteristic while have stronger adaptivity, can not only in unimodal search space,
And optimal value can be found in population for multi-peak searching space.
5, thought proposed by the present invention is based on displacement-speed frame, and the quantity of intelligent group is unrestricted, therefore can transport
Collaboration for extensive intelligent group.
Therefore, the present invention has beneficial effect：
It is of great practical significance to the research of the autonomous synergistic mechanism of extensive intelligent group and application value.From
The intensive flight of starling group, it is mutually coordinated in gain enlightenment, its collective behaviour is introduced into the cooperation of extensive intelligent group, is
Solve intelligent Individual Adaptive regulation, intelligent group non-stop layer self-organizing collaboration two large problems provide brand-new Research Thinking, grind
Study carefully method and calculating mode, it will be in unmanned plane cluster close/intra, large set place crowd's emergency evacuation, large-scale machines people
The fields such as military affairs, emergency, industry, the medical treatment such as multi-agent synergy operation, transmission control all have boundless application prospect.
Detailed description of the invention
Fig. 1 is the realization step of the autonomous collaborative modeling of extensive intelligent group of the imitative starling cluster flight of the present invention.
Specific embodiment
The biological intelligence that starling cluster is flown is introduced into extensive intelligent group synergistic application by the present invention, proposes imitative Europe
The modeling method that the extensive intelligent group of starling collective behaviour mechanism independently cooperates with.Basic thought is：In one grouping of D dimension space
The individual (referred to as Agent) of body scale N gathers { agent1,agent2,…,agentNIn, each Agent has certain fly
Scanning frequency degree, and can determine direction and the distance of its flight.The initial value of the position and speed of Agent is determined by random number.It is all
Agent searches for 6 or 7 neighbours closest around it within the scope of solution space.Each Agent assigns certain memory function,
It can remember searched optimal location.Its fitness value is determined by fitness function, and evaluates its current location accordingly
Superiority and inferiority.After unit time flight, each Agent dynamically adjusts it according to speed renewal equation, location updating equation and flies
The weighting of the local optimum position of its own optimal location and 6 or 7 neighbours will be flown in line direction, speed and position in next step
Center, and pass through the consistency of polarization factor holding cluster heading.
Key is：
1) intelligent group follows basic simple rule in collaboration, sums up 10 primitive rules, i.e., rule (1)~rule
⑽。
2) present invention is using the thought of " non-stop layer opinion ", and there is no the overall points of view for Agent individual, and it is closest that it is only concerned its
6 or 7 neighbours, have locality, it is completely different with the overall point of view of particle swarm algorithm in the prior art.
Referring to Fig. 1, the embodiment of the present invention is for paying close attention to 7 neighbours, and providing realization, specific step is as follows：
Step 1：Initialization population and its parameter
Group and its parameter are initialized, prepares substitution population for subsequent step, including：
1. population size N, maximum evolutionary generation iter is arrangedmax, the initial value k=0 of evolutionary generation is defined, here k<
itermax。
2. maximum speed v is arrangedmax, minimum speed vmin, and define initial velocity for each Agent and beHere rand () is defined as the random number in (0,1) section.
3. spatial dimension is defined as [xmin,xmax], defining initial position for each Agent isHere rand () is defined as the random number in (0,1) section.
4. defining individual history adaptive optimal control degree initial value fi (0)=∞, history local adaptation degree initial value
Under original state, all Agent own using their initial position as its personal best particle
Agent is using initial position as individual local optimum position.
Step 2：Fitness function calculates
Step 2.1：Topological mechanism of action frame
Current newest zoological research shows in starling group in large scale, starling individual with its nearest 6,7
Individual interacts；Neighbours' distribution around each individual is anisotropic, but its 7 closest neighbour is each to same
Property；Interaction between individual depends on topological structure, rather than metric range, therefore, it then follows following primitive rule is (1)
(2) with rule, topological mechanism of action frame, which is constructed, as the basis of the method for the present invention step embodies the theme of collaboration.
Rule is (1)：Neighbours around each Agent individual are distributed as anisotropy, but its closest 7 neighbour be it is each to
The same sex.
Anisotropy refers to the different characteristic in the direction of each individual movement in group；And isotropism refer to it is each
The roughly the same characteristic in the direction of individual movement.
When starting not yet to cooperate with, the individual in entire Agent group flies according to the direction of oneself, as a whole, fortune
Dynamic direction be it is rambling, show as anisotropic feature.After a period of time, Agent individual is most adjacent all in accordance with it
7 close neighbours are adaptively adjusted, and finally on the whole, the direction of motion of group is substantially uniform, show as isotropism
Feature.
Rule is (2)：Interaction between Agent individual depends on topology distance, embodies topology-distance relation, rather than
Measurement-is apart from frame.
Here " topology-distance relation " actually can be understood as nonmetric distance relation, be a kind of relative distance, practical
Distance length it is not important, it is important to the interaction relationship between two Agent can be determined by this topology-distance.
In subsequent step 4.2, illustrates the realization of this interaction, that is, utilize classical " repulsion-holding-attraction "
Three primitive rules, i.e., rule (5)~rule (7), while meet rule (8).
Based on upper topology mechanism of action frame, polarization value Φ is defined in subsequent step 5, it is used to reflect Agent groups
The consistent degree of body entirety heading.The invention proposes autonomous Synergistic method and step, the association realized according to these steps
How same final effect judges, can use the Φ of definition to assess.If Φ shows the cluster integrally side of flight close to 0
To disorderly and unsystematic；Otherwise, show that cluster is whole to fly essentially toward same direction.
Step 2.2：More new individual history adaptive optimal control angle value
When it is implemented, user can according to specific application problem predefine fitness function f, and set its threshold value as
fthreshold。xiFitness function f (xi) be defined as being intended to i-th of private mortgage loan i the current desired positions of target point
Evaluation.In kth time iterative evolution, the current fitness value f of Agent i is calculatedi (k).If fi (k)>fi (k-1),k>1, then more
New individual history optimal locationWhereinThat is the initial optimal location of Agent i,
Step 2.3：Update local optimum fitness value
Using fitness function f (xi) calculate closest 7 neighbours of Agent i local optimum fitness value
IfThen
Update local optimum fitness value
And update local optimum position
Step 3：Selection updates companion
Step 3.1：Obtain closest 7 neighbours of Agent individual
7 neighbours { j closest within the scope of certain radius of view rV around it are obtained for Agent i1,j2,…,j7}。
Step 3.2：Rotating disk roulette wheel selection mechanism calculates, and updates companion for Agent individual choice
It is selected to update companion (update partner) j for Agent i, it then follows rule is (3).
Rule is (3)：According to pj~1/dijPrinciple, be Agent i around it within the scope of certain radius of view rV from closest
7 neighbours { j1,j2,…,j7In selection update companion j.
Wherein, pjFor probability, dijFor the distance between Agent i and Agent j.Apart from smaller, then the probability chosen is got over
Greatly.P is calculated using wheel disc selection mechanismj, i.e.,
diqRefer to the distance between Agent i and Agent q.
Wherein, q be for at a distance from 7 neighbours of Agent i and its sum counting, therefore value be 1 to 7.
Rule is (4)：F (x is used to selected update companion jj) evaluated, it is preferable to ensure to choose fitness value
Companion.According to preset fitness function threshold value fthresholdIf f (xj)>fthreshold, then fitness value is poor, Agent j
It is eliminated, Agent i keeps oneself original flying method；Otherwise Agent i select Agent j as oneself update it is same
Companion.
In the present invention, Agent individual is only concerned its certain 7 closest neighbour within the vision, (3) according to rule
One of them is randomly selected after calculating probability, judges whether this neighbour is fine or not after (4) calculating fitness value further according to rule, this
A neighbours will become " model " of the Agent.
Step 4：Define the interaction relationship between Agent
According to selected update companion j, the interaction relationship between Agent i and Agent j is determined.Specific steps
It is as follows：
Step 4.1：Define radius parameter
Definition repels radius rE, keeps tri- radius parameters of radius rM and Attraction diameter rA.Repel radius be Agent with
It updates the minimum range kept between companion, and mutually collision occurs to avoid the two.Setting keep radius, i.e., Agent with
It is updated within the scope of the moderate distance between companion, and identical motion mode is kept between Agent.Meanwhile in order to keep entire
The cohesiveness of group, Agent are mutually attracted apart from farther away companion, and Attraction diameter is set as the maximum model of the search space Agent
It encloses, i.e. radius of view rV.
Step 4.2：Define the topological relation between Agent
Each Agent embodies topology-distance relation during the motion, and follows classical " repulsion-holding-attraction " three
A primitive rule, i.e., rule (5)~rule (7), while meet rule (8).
Rule is (5)：Exclusion rule, Agent i repels other Agent j in its short range, i.e., if dij<RE, then
The heading of Agent i is modified to fly away from the direction of Agent j.
Rule is (6)：Rule is kept, Agent i is immediately following other Agent j within the scope of its moderate distance.
Rule is (7)：Attract rule, Agent i attracts it compared with other Agent j in far range, i.e., if dij>RA,
The heading for then modifying Agent i is to fly towards the direction of Agent j.
Rule is (8)：If dij>Any interaction does not occur between Agent i and Agent j then for rA.
Step 5：The update of Agent speed and position
When it is implemented, can design after execution updates, return step S2 is until that Agent group arrives at the destination or recycle is secondary
Number reaches maximum evolutionary generation, realizes the renewal process for continuing iteration.
Step 5.1：The polarization factor is introduced, Agent group is controlled
For the problem that conventional model lacks group's guidance mechanism, entire group movement direction is no Objective, to tradition
Model improves, and increase controls the trend of Agent individual each in group, it is instructed to find in unimodal search space
Optimal value.
Rule is (9)：The whole order degree of group is measured by defining polarization Φ, reflects that the cluster integrally flies
The consistent degree in direction, i.e.,
Wherein, viIt is the speed of Agent i, | | vi| | to calculate viNorm in its metric space.As Φ=0, table
Bright cluster entirety heading is disorderly and unsystematic；As Φ → 1, show that cluster is whole essentially toward same direction.
Step 5.2：Calculate the kinetic energy of Agent
The kinetic energy of Agent i is defined as formula (3), i.e.,
Wherein, m is the quality of Agent i, it is assumed that the quality of Agent individual is unit 1.viFor the speed of Agent i.Kind
Group total kinetic energy bePopulation kinetic energy reflects tempo of evolution.Kinetic energy is big, then shows that tempo of evolution is fast, needs to adjust inertia weight
ω reduces it constantly；Conversely, then showing that tempo of evolution is slow, need to increase ω, to guarantee that individual should not concentrations.
Step 5.3：Define speed renewal equation
The pbest variable (optimal location that particle is found) being introduced into particle swarm algorithm (PSO), i.e., particle is found optimal
Position is calculated, i.e., using formula (4)
Pbest=xbest-x (4)
Wherein, x is the current location of Agent individual, xbestIt is the history desired positions of Agent individual.
Under displacement-speed frame, topological interaction mechanism is introduced into algorithm, and carry out new setting to parameter, made
Its with starling group flight characteristic while have stronger adaptivity, can not only in unimodal search space, and
Optimal value can be found in population for multi-peak searching space.Speed renewal equation of the Agent i in k+1 iteration is formula (5), formula
(6)。
Wherein, NiFor the set of 7 neighbours of Agent i, n is for identifying from NiOne of them is taken to be calculated in set,For the average value of the speed of 7 neighbours.ω is inertia weight, is adaptively obtained by formula (7), i.e.,
ω=ωmax-(ωmax-ωmin)×k/itermax+α×rand( )×e-E (7)
K is evolutionary generation, and for the present invention using the parameter of subscript (k) mark kth time iterative evolution, α is controlling elements,
ωmin、ωmaxFor weight dynamic range.E is that the kinetic energy of Agent plays centainly the control of inertia weight ω by the effect of E
Buffer function, can adaptively be adjusted, to the full extent guarantee realize algorithm convergence.
pbestiFor Agent i optimal location experienced, lbestiIt is experienced most for all 7 neighbours of Agent i
Excellent position is local extremum, i.e. lbesti=max { pbestj, j=1,2 ..., 7 }.
Traditional particle swarm algorithm Position Updating formula uses gbestiIndicate global optimum position.With conventional particle group
Algorithm is different, and the present invention is " non-stop layer opinion ", i.e. the principle of 7 neighbours in part, using lbestiThe institute of variable expression Agent i
There are 7 neighbours' optimal locations experienced.
c1、c2For acceleration constant, usual c1、c2=2.
c3Play a part of to keep population diversity as shown in formula (8) for the topology learning factor, enables to realize that algorithm exists
It scans in larger scope.
When ω reduces, population gradually loses diversity, reinforces topology effect at this time, makes Agent individual and other Agent
Information exchange between individual is enhanced, to avoid convergence in population to partial points.In order to guarantee precision, defaulted for 2/3rds generations
Locally optimal solution is found after number and just no longer keeps population diversity, and c is set at this time3=0；Other situation c3=1- ω.
(10) the above updated speed meets rule, i.e.,
Rule is (10)：The speed variation of each Agent is limited to [vmin,vmax] in.Meanwhile during iterative evolution,
If Agent at some speed flight when, position has exceeded boundary position, then by its position restriction be boundary value [xmin,
xmax]。
Step 5.4：Define location updating equation
Position, location updating equation such as formula (9) of the Agent i in k+1 iteration are updated according to the speed of update
It is shown.
Wherein, τ is the unit time, sets τ usually as 1.
2~step 5 of above step is repeated, evolutionary generation k=k+1, Agent i choose its seven nearest neighbour as friendship
Mutual object.By updating speed and the position of Agent individual by constantly iteration with the weighting of the speed of adjacent body, making whole
A Aggregation behaviour reaches unanimity, until group arrives at the destination.Alternatively, reach maximum evolutionary generation if k >=itermax,
Then terminate circulation.
When it is implemented, the automatic running that computer software technology realizes the above process can be used.For reference convenient to carry out
For the sake of, following table is provided：
Table 1：Variable and its description
The modeling method that the extensive intelligent group of imitative starling cluster flight proposed by the present invention independently cooperates with can be used
In multiple fields, especially extensive unmanned plane cluster close/intra practical application.
For example, every frame unmanned plane has certain flight speed in the unmanned aerial vehicle group of certain space range certain amount scale
Degree, and can determine direction and the distance of its flight.It (is defined as between current location and point of destination position by fitness function
The inverse of distance) determine its fitness value, and evaluate the superiority and inferiority of its current location accordingly.All unmanned planes visual field where it
6 or 7 neighbours closest around it are searched in range, it is therefrom selected to update companion.Retained using wheel disc selection mechanism suitable
Answer the preferable companion of angle value or the poor companion of superseded fitness value.According to repelling radius, keeping radius and Attraction diameter, count
It calculates it and updates the interaction force between companion, and determine repulsion-holding-attraction topological relation between the two.Unit
After time flight, every frame unmanned plane dynamically adjusts its heading, speed according to speed renewal equation, location updating equation
And position, fly to the weighted center of its own optimal location and the local optimum position of 6 or 7 neighbours.In flight course, pass through
The polarization factor adjusts tempo of evolution to keep the consistency of unmanned plane cluster heading, and by population kinetic energy.It presses
Form unmanned plane cluster formation mode according to this method, can as extensive starling group autonomous flight.
Under Future Information, networking, confrontation between systems operational environment, lead between dispersion, autonomous extensive unmanned plane
Overstocked cut cooperates, and realizes extensive unmanned aerial vehicle group autonomous formation flight.For single UAV system, unmanned plane collection
Group energy enough using collective behavior cooperate efficiently, farthest common complete single unmanned plane and be difficult to the complexity completed to appoint
Business, and form the advantage of scale.
Claims (6)
1. a kind of modeling method that the extensive intelligent group of imitative starling cluster flight independently cooperates with, it is characterised in that：Including
Following steps,
Under step S1, initialization population and parameter, including setting original state, all Agent are using initial position as individual
Optimal location, and all Agent are using initial position as individual local optimum position；
Step S2, fitness function calculate, including follow following rule (1) with rule (2), construct topological mechanism of action frame, rule
Then (1), the neighbours distribution anisotropy around each Agent individual, but its closest 6 or 7 neighbour is isotropism；Rule
(2), the interaction between Agent individual depends on topology distance；
Step S3, selection update companion, including selecting one of 6 or 7 closest neighbours as update companion for Agent i
Agent j；
Step S4 defines the interaction relationship between Agent；
Step S5 carries out the update of Agent speed and position, and return step S2 is until Agent group arrives at the destination or recycles
Number reaches maximum evolutionary generation.
2. the modeling method that the extensive intelligent group of imitative starling cluster flight independently cooperates with as described in claim 1,
It is characterized in that：In step S3, update companion Agent j follows following rule and (3) chooses, and then (4) calculates fitness according to rule
Value,
Rule (3), according to pj~1/dijPrinciple, be Agent i around it within the scope of certain radius of view rV from closest 6
Or selection updates companion Agent j in 7 neighbours,
Wherein, pjFor probability, dijFor the distance between Agent i and Agent j；
(4), to selected update companion Agent j rule is evaluated using fitness function, including according to preset adaptation
Spend function threshold fthreshold；If Agent j fitness value is greater than fthreshold, then fitness value is poor, and Agent j is washed in a pan
It eliminates, Agent i keeps oneself original flying method；Otherwise Agent i selects update companion of the Agent j as oneself.
3. the modeling method that the extensive intelligent group of imitative starling cluster flight independently cooperates with as described in claim 1,
It is characterized in that：In step S4, the interaction relationship between Agent is defined, implementation is as follows,
Definition repels radius rE, keeps tri- radius parameters of radius rM and Attraction diameter rA, and the topology defined between Agent is closed
System meets following rule,
(5), exclusion rule, Agent i repels other Agent j in its short range to rule, i.e., if dij<RE is then modified
The heading of Agent i is the direction flight away from Agent j；
(6) rule, keeps rule, Agent i is immediately following other Agent j within the scope of its moderate distance；
(7) rule, attracts rule, Agent i attracts it compared with other Agent j in far range, i.e., if dij>RA is then repaired
The heading for changing Agent i is to fly towards the direction of Agent j；
Rule (8), if dij>Any interaction does not occur between Agent i and Agent j then for rA.
4. the modeling method that the extensive intelligent group of imitative starling cluster flight independently cooperates with as described in claim 1,
It is characterized in that：In step S5, carry out Agent speed and position update, including according to it is following rule (9) introduce polarization because
Son controls Agent group,
(9) rule, measures the whole order degree of group by defining polarization Φ, reflects the cluster entirety heading
Consistent degree,
Wherein, viIt is the speed of Agent i, | | vi| | to calculate viNorm in its metric space；As Φ=0, show to collect
The whole heading of group is disorderly and unsystematic；As Φ → 1, show that cluster is whole essentially toward same direction.
5. the modeling method that the extensive intelligent group of imitative starling cluster flight independently cooperates with as described in claim 1,
It is characterized in that：In step S5, the update of Agent speed and position is carried out, including (10) the updated speed of limitation meets rule, rule
Then (10), the speed variation of each Agent is limited to [vmin,vmax] in, vmin、vmaxThe minimum speed of respectively Agent individual
Degree, maximum speed.
6. the extensive intelligent group that the imitative starling cluster as described in claims 1 or 2 or 3 or 4 or 5 is flown independently cooperates with
Modeling method, it is characterised in that：For unmanned plane cluster close/intra.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810587885.8A CN108830373B (en) | 2018-06-08 | 2018-06-08 | Modeling method of large-scale intelligent group autonomous collaboration simulating European 26891bird cluster flight |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810587885.8A CN108830373B (en) | 2018-06-08 | 2018-06-08 | Modeling method of large-scale intelligent group autonomous collaboration simulating European 26891bird cluster flight |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108830373A true CN108830373A (en) | 2018-11-16 |
| CN108830373B CN108830373B (en) | 2022-06-07 |
Family
ID=64143566
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810587885.8A Active CN108830373B (en) | 2018-06-08 | 2018-06-08 | Modeling method of large-scale intelligent group autonomous collaboration simulating European 26891bird cluster flight |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108830373B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109819495A (en) * | 2019-03-15 | 2019-05-28 | 北京科技大学 | A kind of cluster unmanned plane ad hoc network cluster-dividing method |
| CN109901584A (en) * | 2019-03-21 | 2019-06-18 | 南京大学 | A kind of robot team formation method, readable storage medium storing program for executing and robot based on self-organizing |
| CN110058607A (en) * | 2019-04-08 | 2019-07-26 | 北京航空航天大学 | A kind of unmanned plane large-scale cluster super maneuver turning method of imitative starling intelligence |
| CN110058596A (en) * | 2019-05-21 | 2019-07-26 | 吉林大学 | A kind of multi-robot system adaptively disperses cooperative control method and system |
| CN110501905A (en) * | 2019-08-27 | 2019-11-26 | 中国人民解放军国防科技大学 | Multi-agent system self-adaptive method and system based on packing model |
| CN110597059A (en) * | 2019-09-05 | 2019-12-20 | 武汉大学 | Large-leaved dogwood group type intelligent group dynamic network topology construction method facing unmanned system |
| CN112016660A (en) * | 2020-07-03 | 2020-12-01 | 浙江大学 | Physical force and data combined driven three-dimensional cluster behavior simulation method |
| CN112034843A (en) * | 2020-08-10 | 2020-12-04 | 深圳技术大学 | Method, system and storage medium for multi-intelligent-body cooperative object carrying |
| CN112068587A (en) * | 2020-08-05 | 2020-12-11 | 北京航空航天大学 | Man/unmanned aerial vehicle co-converged cluster interaction method based on European 26891bird communication mechanism |
| CN112270398A (en) * | 2020-10-28 | 2021-01-26 | 西北工业大学 | Cluster behavior learning method based on gene programming |
| CN112651574A (en) * | 2020-12-31 | 2021-04-13 | 深圳云天励飞技术股份有限公司 | P median genetic algorithm-based addressing method and device and electronic equipment |
| CN112862055A (en) * | 2021-02-11 | 2021-05-28 | 西北工业大学 | Cluster behavior quantitative analysis method considering cluster consistency and density |
| CN113033756A (en) * | 2021-03-25 | 2021-06-25 | 重庆大学 | Multi-agent control method based on target-oriented aggregation strategy |
| CN113050678A (en) * | 2021-03-02 | 2021-06-29 | 山东罗滨逊物流有限公司 | Autonomous cooperative control method and system based on artificial intelligence |
| CN113311700A (en) * | 2020-02-27 | 2021-08-27 | 陕西师范大学 | UUV cluster cooperative control method guided by non-average mechanism |
| CN113552886A (en) * | 2021-07-23 | 2021-10-26 | 南方科技大学 | Decentralized group robot formation control method and system and electronic equipment |
| CN113867409A (en) * | 2021-11-16 | 2021-12-31 | 中国人民解放军国防科技大学 | Swarm intelligence-based micro unmanned aerial vehicle cluster collision-free control method and device |
| CN114594689A (en) * | 2022-03-15 | 2022-06-07 | 北京理工大学 | Distributed recursive grouping and autonomous aggregation control method of large-scale cluster system |
| CN115220441A (en) * | 2022-03-24 | 2022-10-21 | 华东师范大学 | Unmanned trolley cluster task coordination method based on biological visual perception |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070188489A1 (en) * | 2006-02-15 | 2007-08-16 | Haresh Lalvani | Morphological genome for design applications |
| US20070272797A1 (en) * | 2006-05-23 | 2007-11-29 | Boris Skurkovich | Engine exhaust for modifying a target |
| CN106650915A (en) * | 2016-12-27 | 2017-05-10 | 天津师范大学 | Crowd behavior simulation method based on grid agent |
| CN106843269A (en) * | 2017-01-22 | 2017-06-13 | 北京航空航天大学 | A kind of unmanned plane formation method based on small birds cluster fly mechanics |
-
2018
- 2018-06-08 CN CN201810587885.8A patent/CN108830373B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070188489A1 (en) * | 2006-02-15 | 2007-08-16 | Haresh Lalvani | Morphological genome for design applications |
| US20070272797A1 (en) * | 2006-05-23 | 2007-11-29 | Boris Skurkovich | Engine exhaust for modifying a target |
| CN106650915A (en) * | 2016-12-27 | 2017-05-10 | 天津师范大学 | Crowd behavior simulation method based on grid agent |
| CN106843269A (en) * | 2017-01-22 | 2017-06-13 | 北京航空航天大学 | A kind of unmanned plane formation method based on small birds cluster fly mechanics |
Cited By (33)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109819495B (en) * | 2019-03-15 | 2020-07-31 | 北京科技大学 | Clustering method for self-networking of cluster unmanned aerial vehicle |
| CN109819495A (en) * | 2019-03-15 | 2019-05-28 | 北京科技大学 | A kind of cluster unmanned plane ad hoc network cluster-dividing method |
| CN109901584A (en) * | 2019-03-21 | 2019-06-18 | 南京大学 | A kind of robot team formation method, readable storage medium storing program for executing and robot based on self-organizing |
| CN109901584B (en) * | 2019-03-21 | 2020-04-24 | 南京大学 | Robot formation method based on self-organization, readable storage medium and robot |
| CN110058607A (en) * | 2019-04-08 | 2019-07-26 | 北京航空航天大学 | A kind of unmanned plane large-scale cluster super maneuver turning method of imitative starling intelligence |
| CN110058607B (en) * | 2019-04-08 | 2020-07-07 | 北京航空航天大学 | Simulated European 26891bird intelligent unmanned aerial vehicle large-scale cluster super-maneuvering turning method |
| CN110058596A (en) * | 2019-05-21 | 2019-07-26 | 吉林大学 | A kind of multi-robot system adaptively disperses cooperative control method and system |
| CN110058596B (en) * | 2019-05-21 | 2021-12-10 | 吉林省吉创科豹科技有限公司 | Self-adaptive decentralized cooperative control method and system for multi-robot system |
| CN110501905A (en) * | 2019-08-27 | 2019-11-26 | 中国人民解放军国防科技大学 | Multi-agent system self-adaptive method and system based on packing model |
| CN110501905B (en) * | 2019-08-27 | 2022-02-08 | 中国人民解放军国防科技大学 | Multi-agent system self-adaptive method and system based on packing model |
| CN110597059B (en) * | 2019-09-05 | 2021-05-04 | 武汉大学 | Large-leaved dogwood group type intelligent group dynamic network topology construction method facing unmanned system |
| CN110597059A (en) * | 2019-09-05 | 2019-12-20 | 武汉大学 | Large-leaved dogwood group type intelligent group dynamic network topology construction method facing unmanned system |
| CN113311700A (en) * | 2020-02-27 | 2021-08-27 | 陕西师范大学 | UUV cluster cooperative control method guided by non-average mechanism |
| CN113311700B (en) * | 2020-02-27 | 2022-10-04 | 陕西师范大学 | UUV cluster cooperative control method guided by non-average mechanism |
| CN112016660A (en) * | 2020-07-03 | 2020-12-01 | 浙江大学 | Physical force and data combined driven three-dimensional cluster behavior simulation method |
| CN112016660B (en) * | 2020-07-03 | 2023-11-14 | 浙江大学 | Physical force and data combined driving three-dimensional cluster behavior simulation method |
| CN112068587A (en) * | 2020-08-05 | 2020-12-11 | 北京航空航天大学 | Man/unmanned aerial vehicle co-converged cluster interaction method based on European 26891bird communication mechanism |
| CN112034843A (en) * | 2020-08-10 | 2020-12-04 | 深圳技术大学 | Method, system and storage medium for multi-intelligent-body cooperative object carrying |
| CN112270398A (en) * | 2020-10-28 | 2021-01-26 | 西北工业大学 | Cluster behavior learning method based on gene programming |
| CN112270398B (en) * | 2020-10-28 | 2024-05-28 | 西北工业大学 | Cluster behavior learning method based on gene programming |
| CN112651574A (en) * | 2020-12-31 | 2021-04-13 | 深圳云天励飞技术股份有限公司 | P median genetic algorithm-based addressing method and device and electronic equipment |
| CN112862055A (en) * | 2021-02-11 | 2021-05-28 | 西北工业大学 | Cluster behavior quantitative analysis method considering cluster consistency and density |
| CN112862055B (en) * | 2021-02-11 | 2024-01-12 | 西北工业大学 | Cluster behavior quantitative analysis method considering consistency and density of clustered objects |
| CN113050678A (en) * | 2021-03-02 | 2021-06-29 | 山东罗滨逊物流有限公司 | Autonomous cooperative control method and system based on artificial intelligence |
| CN113033756A (en) * | 2021-03-25 | 2021-06-25 | 重庆大学 | Multi-agent control method based on target-oriented aggregation strategy |
| CN113033756B (en) * | 2021-03-25 | 2022-09-16 | 重庆大学 | Multi-agent control method based on target-oriented aggregation strategy |
| CN113552886A (en) * | 2021-07-23 | 2021-10-26 | 南方科技大学 | Decentralized group robot formation control method and system and electronic equipment |
| CN113867409B (en) * | 2021-11-16 | 2023-10-17 | 中国人民解放军国防科技大学 | Group intelligent-based miniature unmanned aerial vehicle cluster collision-free control method and device |
| CN113867409A (en) * | 2021-11-16 | 2021-12-31 | 中国人民解放军国防科技大学 | Swarm intelligence-based micro unmanned aerial vehicle cluster collision-free control method and device |
| CN114594689B (en) * | 2022-03-15 | 2022-09-27 | 北京理工大学 | Distributed recursive grouping and autonomous aggregation control method of large-scale cluster system |
| CN114594689A (en) * | 2022-03-15 | 2022-06-07 | 北京理工大学 | Distributed recursive grouping and autonomous aggregation control method of large-scale cluster system |
| CN115220441A (en) * | 2022-03-24 | 2022-10-21 | 华东师范大学 | Unmanned trolley cluster task coordination method based on biological visual perception |
| CN115220441B (en) * | 2022-03-24 | 2024-07-12 | 华东师范大学 | Unmanned trolley cluster task cooperation method based on biological visual perception |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108830373B (en) | 2022-06-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108830373A (en) | The modeling method that the extensive intelligent group of imitative starling cluster flight independently cooperates with | |
| Yang et al. | Survey of swarm intelligence optimization algorithms | |
| Yue et al. | Review and empirical analysis of sparrow search algorithm | |
| Neshat et al. | Swallow swarm optimization algorithm: a new method to optimization | |
| Guo et al. | A new improved krill herd algorithm for global numerical optimization | |
| Bi et al. | Self-adaptive bat algorithm with genetic operations | |
| Elloumi et al. | A comparative study of the improvement of performance using a PSO modified by ACO applied to TSP | |
| CN111240356B (en) | Unmanned aerial vehicle cluster convergence method based on deep reinforcement learning | |
| CN112469050B (en) | WSN three-dimensional coverage enhancement method based on improved wolf optimizer | |
| Chaudhary et al. | Swarm bat algorithm with improved search (SBAIS) | |
| Yu et al. | Hawk and pigeon’s intelligence for UAV swarm dynamic combat game via competitive learning pigeon-inspired optimization | |
| CN115185294B (en) | QMIX-based aviation soldier multi-formation collaborative autonomous behavior decision modeling method | |
| CN117434965B (en) | Unmanned aerial vehicle multi-machine collaborative natural pasture intelligent management method and management system | |
| Nedjah et al. | Inspiration-wise swarm intelligence meta-heuristics for continuous optimisation: a survey-part II | |
| Cuevas et al. | An optimization algorithm for multimodal functions inspired by collective animal behavior | |
| Ma et al. | A survey of research on the distributed cooperation method of the uav swarm based on swarm intelligence | |
| Hexmoor et al. | Swarm Control in Unmanned Aerial Vehicles. | |
| Tsai et al. | A new framework for optimization based-on hybrid swarm intelligence | |
| Saggu et al. | Applications of swarm intelligence | |
| Singhal et al. | A comparative study of a class of nature inspired algorithms | |
| Xie | Starling Swarm Algorithm: An Approach to Autonomous Coordination of Intensive Agents | |
| Madamanchi | Evaluation of a new bio-inspired algorithm: krill herd | |
| Xie | A flock-of-starling optimization algorithm with reinforcement learning capability | |
| CN114924587B (en) | Unmanned aerial vehicle path planning method | |
| Revathi et al. | A brief study about nature inspired optimisation algorithms |
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 |