CN109270905A - A kind of method that group robot neural network based is looked for food using pheromones communication realization cooperation - Google Patents
A kind of method that group robot neural network based is looked for food using pheromones communication realization cooperation Download PDFInfo
- Publication number
- CN109270905A CN109270905A CN201811229448.5A CN201811229448A CN109270905A CN 109270905 A CN109270905 A CN 109270905A CN 201811229448 A CN201811229448 A CN 201811229448A CN 109270905 A CN109270905 A CN 109270905A
- Authority
- CN
- China
- Prior art keywords
- robot
- food
- pheromones
- neural network
- neuron
- 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
- 239000003016 pheromone Substances 0.000 title claims abstract description 128
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000004891 communication Methods 0.000 title abstract description 12
- 210000002569 neuron Anatomy 0.000 claims abstract description 50
- 230000006399 behavior Effects 0.000 claims abstract description 35
- 230000019637 foraging behavior Effects 0.000 claims abstract description 13
- 230000008569 process Effects 0.000 claims abstract description 12
- 230000008020 evaporation Effects 0.000 claims abstract description 9
- 238000001704 evaporation Methods 0.000 claims abstract description 9
- 238000013461 design Methods 0.000 claims abstract description 8
- 238000003062 neural network model Methods 0.000 claims abstract description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 11
- 230000002431 foraging effect Effects 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 7
- 230000000644 propagated effect Effects 0.000 claims description 7
- 238000007599 discharging Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 230000010429 evolutionary process Effects 0.000 claims description 5
- 230000004888 barrier function Effects 0.000 claims description 4
- 230000032258 transport Effects 0.000 claims description 4
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000005295 random walk Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 abstract 1
- 238000011160 research Methods 0.000 description 13
- 230000003993 interaction Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000007246 mechanism Effects 0.000 description 6
- 230000013016 learning Effects 0.000 description 5
- 238000005457 optimization Methods 0.000 description 5
- 238000012546 transfer Methods 0.000 description 5
- XGWIJUOSCAQSSV-XHDPSFHLSA-N (S,S)-hexythiazox Chemical compound S([C@H]([C@@H]1C)C=2C=CC(Cl)=CC=2)C(=O)N1C(=O)NC1CCCCC1 XGWIJUOSCAQSSV-XHDPSFHLSA-N 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 239000002245 particle Substances 0.000 description 3
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 244000144980 herd Species 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000033001 locomotion Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003997 social interaction Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241000219193 Brassicaceae Species 0.000 description 1
- 241000238631 Hexapoda Species 0.000 description 1
- 206010021703 Indifference Diseases 0.000 description 1
- 241000256602 Isoptera Species 0.000 description 1
- 241000209504 Poaceae Species 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000009326 social learning Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
- G05B19/41885—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by modeling, simulation of the manufacturing system
-
- 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
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- Manufacturing & Machinery (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Manipulator (AREA)
Abstract
The present invention relates to a kind of group robots neural network based, and the method for realizing that cooperation is looked for food is communicated using pheromones, includes the following steps: to establish neural network model, design information element evaporation model, establishes system totality behavior frameworks model.Present invention particularly provides the pheromones evaporation models of group robot cooperation foraging behavior, are defined as Ii (t), i.e. i-th of neuron in formula, attracts pheromones P in the external input of t momentaFor biggish positive value, repulsion pheromone PoWith repulsion pheromone PeFor lesser negative value;When looking for food robot discovery food and being transported back nest, release attracts pheromones Pa;Repulsion pheromone P will be discharged when robot obstacle-avoidingo, repulsion pheromone P will be discharged when robot random search food in the work environmentE,Neural network updates output according to the variation of Ii (t) at any time, and the evolution of neural network makes group robot be able to carry out local communication, and the group behavior of self-organizing is emerged in interactive process.
Description
Technical field
The present invention relates to colony intelligence robot self-organizing behavior emergence, using neural network pheromone model,
Robot emerges swarm intelligence behavior by Local Interaction, belongs to colony intelligence robotic technology field, and in particular to Yi Zhongji
The method for realizing that cooperation is looked for food is communicated using pheromones in the group robot of neural network.
Background technique
The research of multi-robot system starts from later period the 1970s, and researcher is by the multiple agent in artificial intelligence
Theory is applied in multi-robot system, has started the research of the multirobot technology of robot field.The research at initial stage is main
Concentrate on system architecture, motion of multi-robots planning and several aspects such as system is restructural, with distributed artificial intelligence,
The introducing of the theory and method of the research fields such as complication system, sociology, biology, the research of multi-robot system start to inquire into
The theory and technology problem of the key such as system organization form, information interaction approach, evolutionary learning mechanism.Currently, by biology and society
The inspiration that can be learned, the research of multi-robot system critical issue achieve comparable progress, and researcher is by some social phenomenons
Engineering, and apply in the design of robot group behavior, using multi-robot system as a social groups, made for it
Fixed a series of social regulation, robot system just become a more complicated robotic society.Multirobot group row
Between the main research robot of study and robot carries out the mechanism of social interaction with ambient enviroment, in social interaction work
With the group behavior and swarm intelligence for emerging complexity in the process.
Colony intelligence robot is an emerging subject, belongs to the scope of multi-robot system, it is main study robot it
Between and machine human and environment how to pass through limited perception and Local Interaction emerges desired group behavior.With multirobot system
System research deepens continuously, and some social biological self-organizing models provide beneficial open for the research of robot group behavior
Hair, researcher start using Stigmergy (individual realizes indirect communication by changing common environmental) mechanism to multiple robots
Body behavior is modeled and is analyzed.Stigmergy mechanism is put forward by Grasse earliest, for explaining the row of nesting of termite
For Deneubourg etc. has carried out initiative emulation and Physical Experiment to Stigmergy mechanism using " ant robot ".
Swarm robotic systems are the systems imitating social insect or other social biocenose behaviors and being established,
The robot group being made of under complete distributed AC servo system many indifference autonomous robots.Group robot is in engineering
A large amount of exemplary applications, including autonomous driving, transfer robot, autonomous agricultural robot and autonomous storage are produced in practice
Robot etc..How colony intelligence robot makes the limited machine individual human of ability emerge group by Local Interaction if mainly being studied
Intelligence, typical Swarm robotic systems include: The Nerd Herd system, Collective Robotics system and
Swarm system etc..The Nerd Herd system is responsible for out by Maja professor Mataric of University of Southern California of the U.S.
Hair, it is made of multiple identical robots, each robot uses subsumption architecture to carry out Activity design, and Mataric is inquired into
The influence of population density and rule of conduct to system performance, realizes the group behavior that robot team formation passes through door body.This is
System can be used for emulation and the experimental study of robot massive crowd behavior.Collective Robotics system is taken by adding
Big Alberta university exploitation, multi-simple-robot perhaps is formed into a group and completes complicated task.Swarm system be by
The research and development such as American scholar Jin and Beni, the distributed system being made of a large amount of simple autonomous robots.
Cooperation robot system is made of many autonomous robots with certain intelligence, and robot passes through simple
The reciprocation of body behavior completes complicated task, and the typical robot system that cooperates includes: ACTRESS system, CEBOT system
And CESAR Emperor and CESAR Nomads robot system.ACTRESS system is put forward by Asama et al., is
One is used for the multi-robot system of a variety of different type tasks, and system passes through bottom communication structure for robot and peripheral equipment
It connects.Under normal conditions, machine individual human works independently, but if robot can make up when system needs to cooperate
Cooperative group executes specific task.CEBOT system be proposed by professor Fukuda one kind in distributing autonomous robot
The robot largely in system with standalone feature is considered as cell member by the mechanism of dimly visible group collaboration behavior in system, according to
The variation of task or environment, cell member robot can be moved, find and be combined, and self-organizing forms the system of robot of function complexity
System.In self- reconfigurable robots, the reciprocation between robot presents the mobility of cooperation, and robot is according to task need
It physically to work in coordination or interact in a similar way.Doctor Parker of Oak Ridge National Laboratory of the U.S.
And its research group establishes CESAR Emperor and CESAR Nomads experiment porch, carries out in terms of the robotics that cooperates
A large amount of theoretical research and experimental verification, and using ALLIANCE structure demonstrate multirobot target observations.
Currently, it mainly includes particle group optimizing and ant group algorithm that biology, which inspires the realization algorithm of system model,.Population is excellent
Change is a kind of group behavior computation modeling and implementation method based on social influence and social learning, is to the soft imitative of biocenose
It is raw.Pugh etc. proposes a kind of robot group behavior distribution on-line learning algorithm using particle group optimizing method, and visits
Influence of the population size to robot group action learning speed is begged for.Moutarde utilizes particle group optimizing method planing machine
People's recurrent neural network controller, study obtain Group Robots cooperation foraging behavior.During ant group algorithm is looked for food by ant colony
What is presented emerges in large numbers the inspiration of phenomenon, and individual emerges complicated group behavior by the effect of simple Local Interaction.Chiou etc.
A kind of Soccer robot group behavior learning algorithm is proposed based on fuzzy ant group optimization, is advised according to ant colony algorithm for optimization design collision prevention
Then, and using PREDICTIVE CONTROL subsequent time target position is predicted.Chen etc. is realized using ant colony optimization algorithm and affine transformation
Multi-robot order switching is predefined the target position of each robot by affine transformation, is obtained using ant colony optimization algorithm
Obtain robot group collisionless shortest path.
Although researcher achieves a large amount of fruitful research achievements in terms of colony intelligence robotics,
It is some problems due to multirobot action learning theoretically there are no being well solved, as a research field
The theoretical frame and implementation method of colony intelligence robotics are required to further perfect.
Summary of the invention
To solve the above problems, the present invention proposes that a kind of group robot neural network based is realized using pheromones communication
The method looked for food cooperate, and it is an object of the present invention to provide a kind of local communication mode based on pheromones, design robot optimizing decision plan
Slightly, the process of emerging in large numbers for accelerating swarm intelligence behavior, the method for realizing group robot cooperation completion task.
Group robot neural network based of the invention communicates the method for realizing that cooperation is looked for food using pheromones, including such as
Lower step: (1) neural network model is established
Entire neural network forms two dimensional topology by N × N number of neuron, and i-th of neuron corresponds to structure space
I-th of discrete state, the only neuron connection adjacent thereto of each neuron, type of attachment is all identical, has height simultaneously
Capable architecture, all connection weights are all equal, and information bidirectional between neuron is propagated, according to i-th of neuron from
Dissipate time dynamics equation are as follows:
In formula, xi (t+1) and xi (t) are respectively output valve of i-th of neuron at t the and t+1 moment, and N is i-th of mind
Through the neuron number in first neighborhood, Ii (t) is i-th of neuron in the external input of t moment, and neural network is according to Ii (t)
Variation update output at any time, wij is j-th of neuron to the connection weight of i-th of neuron, and f is activation primitive, the activation
Function f selects S type function, is defined as follows:
Neural network generates external input, i-th of mind in evolutionary process, according to mapping of the pheromones in topological structure
External input through member is generated by exploring mapping of the location information in region and pheromone release in neural network topology structure
, it is defined as follows:
In formula, attraction pheromones Pa is biggish positive value, and repulsion pheromone Po and repulsion pheromone Pe are lesser negative
Value;When looking for food robot discovery food and being transported back nest, release attracts pheromones Pa;It will release when robot obstacle-avoiding
Repulsion pheromone Po will discharge repulsion pheromone Pe when robot random search food in the work environment;
The connection weight calculation formula such as following formula:
In formula, | i-j | for the Euclidian distance in structure space between vector x i and xj;
(2) design information element evaporation model
It includes two dynamic processes that pheromones, which develop, i.e. robot passes through medium after some position releases pheromones
It is propagated to surrounding, while pheromones are constantly volatilized to reduce its concentration, and the robot that looks for food is driven to explore new region, pheromones
Evaporation model is defined as follows:
In formula, ρ is volatility, and Δ xj (t) is pheromone concentration variable quantity in i-th of neuron neighborhood;
(3) system totality behavior frameworks model is established
Robot typical behaviour is described as follows in system:
Search: robot carries out random search in entire working region with fixed speed, discharges and repels in search process
Pheromones Pe;
Avoidance: robot avoids if encountering barrier and discharges repulsion pheromone Po;
Wait: robot is if it find that food source then stops search, and waits other robot to carry out near food source
Cooperation, while discharging and attracting pheromones Pa;The robot of other robot discovery wait state then forms cooperative team for food
Carry back nest;
Carry: cooperative team Robot attracts the path pheromones Pa that food is moved back to nest, while discharging more inhale
Draw pheromones Pa;
The group robot cooperation foraging behavior algorithm is as follows:
Neural network output is initialized as zero
xi(t=0)=0
While foraging robot starts to look for food until food exhausts
While exploring robot searching food is until finding food source
Random walk is moved along pheromones path
Repulsion pheromone P is discharged simultaneouslyo orPeAnd update neural network output:
End while exploring
While homing robot transports food back nest
It is moved along pheromones path
It discharges simultaneously and attracts pheromones PaAnd update neural network output:
End while homing
Pheromones volatilization:
End while foraging。
Further, the working space of the neural network and robot topological structure having the same.
Beneficial effects of the present invention are as follows:
1. establishing the neural network model of group robot pheromones propagation, realize that pheromones decline in entire spatial network
It propagates with subtracting: using neural network model of the invention, it, should after when robot, some position releases pheromones in the environment
State, which corresponds to neuron just, has corresponding external input, updates its output according to external input, neuron utilizes part connection
The state in its neighborhood is updated, so that pheromones gradually can damply be propagated in entire working space.
2. devising pheromones evaporation model, neural network is promoted to develop, promotes group robot local communication, accelerate to emerge in large numbers
The group behavior of self-organizing: present invention particularly provides the pheromones evaporation models of group robot cooperation foraging behavior, are defined as
In the external input of t moment, neural network updates output according to the variation of Ii (t) at any time, neural by Ii (t), i.e. i-th of neuron
The evolution of network simulates the release volatilization process of pheromones well, so that group robot is able to carry out local communication, in interaction
The group behavior of self-organizing is emerged in the process.
3. having the function of Automatic Optimal path and behavior: the present invention can make part of the group robot based on pheromones logical
Letter mode has, and effectively selects optimal path and behavior, and other robot is attracted to participate in, and accelerates emerging in large numbers for swarm intelligence behavior
Journey finally realizes the cooperation foraging behavior that all robots can all be carried out along shortest pheromones path.
Detailed description of the invention
Fig. 1 is i-th of neuron neighbour structure schematic diagram.
Fig. 2 is cooperation foraging behavior flow chart.
Fig. 3 is cooperation foraging behavior finite state machine model.
Fig. 4 is that single food source robot looks for food procedural information element distribution schematic diagram.
Fig. 4 (a) pheromones are in t=0 moment distribution map
Fig. 4 (b) pheromones are in t=20 moment distribution map
Fig. 4 (c) pheromones are in t=80 moment distribution map
Fig. 4 (d) pheromones are in t=200 moment distribution map
Fig. 5 is that double food source robots look for food procedural information element distribution schematic diagram.
Fig. 5 (a) pheromones are in t=0 moment distribution map
Fig. 5 (b) pheromones are in t=20 moment distribution map
Fig. 5 (c) pheromones are in t=60 moment distribution map
Fig. 5 (d) pheromones are in t=200 moment distribution map
Fig. 6 is best waiting time τoEfficiency of looking for food when=2s schematic diagram.
Fig. 7 looks for food efficiency schematic diagram when being waiting time τ=5s.
Specific embodiment
Below by way of specific embodiment further illustrate technical solution of the present invention, but technical solution of the present invention not with
Embodiment is limited.
A kind of method that group robot neural network based is looked for food using pheromones communication realization cooperation, including walk as follows
It is rapid:
(1) neural network model is established
Neural network and robot working space's topological structure having the same, each neuron work corresponding to robot
One discrete state in space.All neurons are all only connected with the neuron in its local neighborhood, wherein i-th of neuron
It is as shown in Figure 1 with the type of attachment of neuron in its neighborhood.And its type of attachment is all identical, and entire neural network is by N × N number of
Neuron forms two dimensional topology.Neural network has the architecture of highly-parallel, and all connection weights are all equal, neuron
Between the propagation of information be two-way.Input of the neural network in evolutionary process according to each discrete state updates its neighborhood
Interior state, entire neural network are considered as a discrete time dynamic system.
Neural network is in evolutionary process, according to the release of pheromones and volatilization information in neural network topology structure
Mapping generates the external input of neural network.When looking for food robot discovery food and being transported back nest, release attracts pheromones
Pa, other robot will be gradually attracted on the pheromones track where it;Letter is repelled into release when robot obstacle-avoiding
Element Po is ceased, repulsion pheromone Pe will be discharged when robot random search food in the work environment.S type activation primitive ensure that
Pheromones are gradually damply propagated in working space, and all neuron output values just constitute a curved surface, each on curved surface
The value of a point just represents the pheromone concentration of its corresponding state.
Entire neural network forms two dimensional topology by N × N number of neuron, and i-th of neuron corresponds to structure space
I-th of discrete state, the only neuron connection adjacent thereto of each neuron, type of attachment is all identical, has height simultaneously
Capable architecture, all connection weights are all equal, and information bidirectional between neuron is propagated, when i-th of neuron is discrete
Between kinetics equation are as follows:
In formula, xi(t+1) and xiIt (t) is respectively output valve of i-th of neuron at t the and t+1 moment, N is i-th of mind
Through the neuron number in first neighborhood, IiIt (t) is i-th of neuron in the external input of t moment, neural network is according to Ii(t)
Variation updates output, w at any timeijFor j-th of neuron to the connection weight of i-th of neuron, f is activation primitive, the activation letter
Number f selects S type function, is defined as follows:
Neural network generates external input, i-th of mind in evolutionary process, according to mapping of the pheromones in topological structure
External input through member is generated by exploring mapping of the location information in region and pheromone release in neural network topology structure
, it is defined as follows:
In formula, attract pheromones PaFor biggish positive value, repulsion pheromone PoWith repulsion pheromone PeIt is lesser negative
Value;When looking for food robot discovery food and being transported back nest, release attracts pheromones Pa;It will release when robot obstacle-avoiding
Repulsion pheromone Po, repulsion pheromone P will be discharged when robot random search food in the work environmente;
The connection weight calculation formula such as following formula:
In formula, | i-j | for vector x in structure spaceiAnd xjBetween Euclidian distance;
(2) design information element evaporation model
It includes two dynamic processes that pheromones, which develop, i.e. robot passes through medium after some position releases pheromones
It is propagated to surrounding, while pheromones are constantly volatilized to reduce its concentration, and the robot that looks for food is driven to explore new region, pheromones
Evaporation model is defined as follows:
In formula, ρ is volatility, Δ xjIt (t) is pheromone concentration variable quantity in i-th of neuron neighborhood;
(3) system totality behavior frameworks model is established
System totality behavior frameworks model includes a variety of different typical behaviours: search, avoidance, waiting, carrying etc., cooperation
Finite state machine model look for food as shown in figure 3, each status representative is in the robot quantity of different task, robot is being looked for
The robot of different conditions mutually converts during food, when some robot finds shortest path, more by release pheromone
New neural network is communicated with surrounding machine people, attracts more robots to be added in shortest path, while discharging more information
Element finally realizes that shortest path is all walked by all robots by constantly optimizing.
Robot typical behaviour is described as follows in system:
Search: robot carries out random search in entire working region with fixed speed, discharges and repels in search process
Pheromones Pe;
Avoidance: robot avoids if encountering barrier and discharges repulsion pheromone Po;
Wait: robot is if it find that food source then stops search, and waits other robot to carry out near food source
Cooperation, while discharging and attracting pheromones Pa;The robot of other robot discovery wait state then forms cooperative team for food
Carry back nest;
Carry: cooperative team Robot attracts pheromones PaFood is moved back to nest by path, while discharging more inhale
Draw pheromones Pa;
The group grey relational grade that the present invention is studied is isomorphism system, and system is by one group of identical reaction equation robot
It forms, carries out simple Local Interaction between robot, do not have explicit communication.Robot during group robot cooperation is looked for food
It looks for food by using different strategies.In initial stage all robots not about the priori knowledge of environment, and from nest
Set out random search food source in cave region, needs two robot cooperated to carry to things in foraging mission.Once hair
Existing food searching machine people will wait the cooperation of other robot in next time τ, if without it within the waiting time
His robot finds the robot being waited for, and waits robot that will abandon current task and re-searches for other food sources;
If there is searching machine people to have found that the robot of wait state, two robots that look for food will form team within the waiting time
And food is carried into back nest;Transfer robot re-starts search after food is moved back to nest, and cooperate foraging behavior such as Fig. 2
It is shown.
Group robot of the invention can also continue to optimize behavior and path in the continuous evolution of neural network, find most short
The repulsion pheromone P of the robot release in patheAt least, while attracting pheromones PaIt volatilizees also minimum, attracts more multirobot
Come, as more and more robots are added, pheromones path is continuously available reinforcing, and path of looking for food also is continuously available optimization,
Final all robots that look for food all can carry out foraging behavior along shortest pheromones path.
(4) foraging behavior algorithm of the group robot based on pheromones is as follows:
Neural network output is initialized as zero
xi(t=0)=0
While foraging robot starts to look for food until food exhausts
While exploring robot searching food is until finding food source
Random walk is moved along pheromones path
It discharges repulsion pheromone simultaneously and updates neural network output:
End while exploring
While homing robot transports food back nest
It is moved along pheromones path
It discharges simultaneously and attracts pheromones and update neural network output:
End while homing
Pheromones volatilization:
End while foraging
Embodiment: group robot cooperation is looked for food emulation experiment
In order to illustrate group robot pheromones Realization Method of Communication neural network based, in the movement that laboratory is established
Robot environment's modeling has carried out emulation experiment on exploration software platform.Robot nest of looking for food is located at a left side for working space
In inferior horn rectangle frame, robot begins search for food source from nest, and food source is located at the upper right corner rectangle frame of working environment
Interior, as shown in Fig. 4 (a), grey rectangle is moveable barrier.Searching machine people (circle) is according to equation (1) " i-th of mind
Through first discrete time kinetics equation " release repulsion pheromone PoAnd PeEntire working space is gradually damply traveled to, is such as schemed
Shown in 4 (b), repulsion pheromone PoAnd PeRegion be the region searched of robot, repulsion pheromone will drive other look for
Food companion searches new region, to improve the efficiency for robot discovery food source of looking for food.Irregular shadow region is letter in figure
Cease plain distributed areas.
When searching machine people discovery food source and food is carried back nest by the success that cooperates, robot, and transfer robot is (black
Color dot) in the path release attraction pheromones P from food source to nesta, as shown in Fig. 4 (c), attract pheromones PaEqually by
Gradually damply travel to entire working space.Other robot one will be recruited after food is carried back nest by transfer robot
It rises and looks for food, attract pheromones PaOther robots that look for food constantly are attracted to reach food source and carry food along path of looking for food
Return nest, the attraction pheromones P to look for food on trackaConcentration gradually increases.Final all Robots attract pheromones PaTrack into
Row foraging behavior, so that the colony intelligence behavior of self-organizing has been emerged, as shown in Fig. 4 (d).
In order to verify the validity and reliability of proposed pheromone model, it is arranged other one in the working space lower right corner
A food source, as shown in Fig. 5 (a).Searching robot has found two food sources after foraging mission starts, and food is continuous
It is transported to nest from two food sources, as shown in Fig. 5 (b) and Fig. 5 (c).Transfer robot attracts letter in release on track of looking for food
Cease element PaForm two pheromones tracks from two food sources to nest, grey rectangle obstacle after system operation a period of time
Object moves up 10 grids, and pheromones distribution will be constantly updated, and path of looking for food will also continue to optimize, as shown in Fig. 5 (d).More
It looks for food come the shorter path of more Robot Selections, the attraction pheromones P in shorter pathaConcentration constantly increases, most
All robots abandon longer path of looking for food eventually, along shortest path searching, carry food.It looks for food area information element other than path
It gradually volatilizees, concentration constantly reduces, and finally completely disappears.The present invention is communicated using neural network group robot pheromones
Model has emerged the group behavior of self-organizing during phylogeny by the Local Interaction between robot.
The optimizing decision strategy decision for the system of looking for food that cooperates is in the optimal waiting time of cooperating process, waiting time direct shadow
The efficiency that cooperation is looked for food is rung.Another important restrictions of system are exactly the ability of machine individual human, that is, need how many
Food can be carried back nest by the cooperation of robot, and needing Liang Ge robot to carry out cooperation in the present invention can will eat
Object transports nest back.Robot can obtain food during looking for food, while being also required to energy and time search, carrying food,
The overall goals of system are exactly that most food is obtained with the smallest cost.In order to determine optimizing decision strategy, looked for based on optimal
Food is theoretical to be defined as robot efficiency of looking for food to carry the quantity and the ratio of time of food:
In formula, E is that robot looks for food efficiency, nfIt (t) is the size of food carried in t moment.
Fig. 6 is best waiting time τoEfficiency of looking for food when=2s schematic diagram, the incipient stage, all robots all scanned for,
There is no food to be transported back nest, food curve has smaller slope, and with the progress of time, slope is gradually increased, best fit
Slope of a curve is exactly the average efficiency of foraging behavior, and efficiency of most preferably averagely looking for food is 0.32.
Fig. 7 looks for food efficiency schematic diagram when being waiting time τ=5s, and efficiency of averagely looking for food is 0.23.
Cooperation is realized using pheromones communication to a kind of group robot neural network based provided by the present invention above
The method looked for food is described in detail.The principle of the present invention and embodiment are explained herein by specific embodiment
It states, it is described above to be merely used to help understand method and its core concept of the invention.It should be pointed out that for the art
For those of ordinary skill, without departing from the principle of the present invention, can with several improvements and modifications are made to the present invention,
These improvements and modifications also fall within the scope of protection of the claims of the present invention.
Claims (2)
1. a kind of group robot neural network based communicates the method for realizing that cooperation is looked for food using pheromones, which is characterized in that
Include the following steps:
(1) neural network model is established
Entire neural network forms two dimensional topology by N × N number of neuron, and i-th of neuron corresponds to the i-th of structure space
A discrete state, the only neuron connection adjacent thereto of each neuron, type of attachment is all identical, the body with highly-parallel
Architecture, all connection weights are all equal, and the information bidirectional between neuron is propagated, according to i-th of neuron discrete time
Kinetics equation are as follows:
In formula, xi(t+1) and xiIt (t) is respectively output valve of i-th of neuron at t the and t+1 moment, N is i-th of neuron
Neuron number in neighborhood, IiIt (t) is i-th of neuron in the external input of t moment, neural network is according to Ii(t) variation
Output, w are updated at any timeijFor j-th of neuron to the connection weight of i-th of neuron, f is activation primitive, the activation primitive f choosing
S type function is selected, is defined as follows:
Neural network generates external input, i-th of neuron in evolutionary process, according to mapping of the pheromones in topological structure
External input generated by exploring mapping of the location information of region and pheromone release in neural network topology structure, it is fixed
Justice is as follows:
In formula, attract pheromones PaFor biggish positive value, repulsion pheromone PoWith repulsion pheromone PeFor lesser negative value;When
Look for food robot discovery food and when being transported back nest release attract pheromones Pa;Letter is repelled into release when robot obstacle-avoiding
Cease element Po, repulsion pheromone P will be discharged when robot random search food in the work environmente;
The connection weight calculation formula such as following formula:
In formula, | i-j | for vector x in structure spaceiAnd xjBetween Euclidian distance;
(2) design information element evaporation model
It includes two dynamic processes that pheromones, which develop, i.e., robot passes through medium to week after some position releases pheromones
Propagation is enclosed, while pheromones are constantly volatilized to reduce its concentration, the robot that looks for food is driven to explore new region, pheromones volatilization
Model is defined as follows:
In formula, ρ is volatility, Δ xjIt (t) is pheromone concentration variable quantity in i-th of neuron neighborhood;
(3) system totality behavior frameworks model is established
Robot typical behaviour is described as follows in system:
Search: robot carries out random search in entire working region with fixed speed, and information is repelled in release in search process
Plain Pe;
Avoidance: robot avoids if encountering barrier and discharges repulsion pheromone Po;
Wait: robot is if it find that food source then stops search, and waiting other robot cooperates near food source,
It discharges simultaneously and attracts pheromones Pa;The robot of other robot discovery wait state then forms cooperative team and carries back food
Nest;
Carry: cooperative team Robot attracts pheromones PaFood is moved back to nest by path, while discharging more attraction information
Plain Pa;
(4) group robot cooperation foraging behavior algorithm is as follows:
Neural network output is initialized as zero
xi(t=0)=0
While foraging robot starts to look for food until food exhausts
While exploring robot searching food is until finding food source
Random walk is moved along pheromones path
Repulsion pheromone P is discharged simultaneouslyo or PeAnd update neural network output:
End while exploring
While homing robot transports food back nest
It is moved along pheromones path
It discharges simultaneously and attracts pheromones PaAnd update neural network output:
End while homing
Pheromones volatilization:
End while foraging。
2. group robot neural network based communicates the side for realizing that cooperation is looked for food using pheromones according to claim 1
Method, which is characterized in that the working space of the neural network and robot topological structure having the same.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811229448.5A CN109270905B (en) | 2018-10-22 | 2018-10-22 | A kind of method that group robot neural network based is looked for food using pheromones communication realization cooperation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811229448.5A CN109270905B (en) | 2018-10-22 | 2018-10-22 | A kind of method that group robot neural network based is looked for food using pheromones communication realization cooperation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109270905A true CN109270905A (en) | 2019-01-25 |
CN109270905B CN109270905B (en) | 2019-11-29 |
Family
ID=65194557
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811229448.5A Active CN109270905B (en) | 2018-10-22 | 2018-10-22 | A kind of method that group robot neural network based is looked for food using pheromones communication realization cooperation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109270905B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110196595A (en) * | 2019-05-31 | 2019-09-03 | 武汉科技大学 | A kind of multi-robots system based on infrared ray simulation artificial information's element |
CN110673826A (en) * | 2019-10-25 | 2020-01-10 | 山东大学 | Random search method and system for swarm robots in area search |
CN111882577A (en) * | 2020-07-31 | 2020-11-03 | 中国人民解放军国防科技大学 | Group obstacle avoidance and consistent traveling method and device based on field-of-view perception |
CN112666953A (en) * | 2020-12-29 | 2021-04-16 | 西北工业大学深圳研究院 | Swarm robot lunar surface intelligent building system and method, robot and using method |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2003066639A2 (en) * | 2002-02-08 | 2003-08-14 | Georgetown University Medical Center | Methods for identifying targeting domains and methods and compositions comprising the same |
US20090221316A1 (en) * | 2008-03-03 | 2009-09-03 | Industrial Technology Research Institute | Apparatus and method for transforming signal strength of wireless positioning system |
CN103473598A (en) * | 2013-09-17 | 2013-12-25 | 山东大学 | Extreme learning machine based on length-changing particle swarm optimization algorithm |
CN106295793A (en) * | 2016-08-30 | 2017-01-04 | 吉林大学 | Group robot mixed search algorithm based on biological foraging behavior |
CN106654987A (en) * | 2016-11-18 | 2017-05-10 | 华北电力大学(保定) | Power line multi-robot collaborative inspection method |
CN106779177A (en) * | 2016-11-28 | 2017-05-31 | 国网冀北电力有限公司唐山供电公司 | Multiresolution wavelet neutral net electricity demand forecasting method based on particle group optimizing |
-
2018
- 2018-10-22 CN CN201811229448.5A patent/CN109270905B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2003066639A2 (en) * | 2002-02-08 | 2003-08-14 | Georgetown University Medical Center | Methods for identifying targeting domains and methods and compositions comprising the same |
US20090221316A1 (en) * | 2008-03-03 | 2009-09-03 | Industrial Technology Research Institute | Apparatus and method for transforming signal strength of wireless positioning system |
CN103473598A (en) * | 2013-09-17 | 2013-12-25 | 山东大学 | Extreme learning machine based on length-changing particle swarm optimization algorithm |
CN106295793A (en) * | 2016-08-30 | 2017-01-04 | 吉林大学 | Group robot mixed search algorithm based on biological foraging behavior |
CN106654987A (en) * | 2016-11-18 | 2017-05-10 | 华北电力大学(保定) | Power line multi-robot collaborative inspection method |
CN106779177A (en) * | 2016-11-28 | 2017-05-31 | 国网冀北电力有限公司唐山供电公司 | Multiresolution wavelet neutral net electricity demand forecasting method based on particle group optimizing |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110196595A (en) * | 2019-05-31 | 2019-09-03 | 武汉科技大学 | A kind of multi-robots system based on infrared ray simulation artificial information's element |
CN110196595B (en) * | 2019-05-31 | 2021-12-21 | 武汉科技大学 | Swarm robot system based on infrared ray simulation artificial pheromone |
CN110673826A (en) * | 2019-10-25 | 2020-01-10 | 山东大学 | Random search method and system for swarm robots in area search |
CN110673826B (en) * | 2019-10-25 | 2023-02-28 | 山东大学 | Random search method and system for swarm robots in area search |
CN111882577A (en) * | 2020-07-31 | 2020-11-03 | 中国人民解放军国防科技大学 | Group obstacle avoidance and consistent traveling method and device based on field-of-view perception |
CN111882577B (en) * | 2020-07-31 | 2022-09-09 | 中国人民解放军国防科技大学 | Group obstacle avoidance and consistent traveling method and device based on field-of-view perception |
CN112666953A (en) * | 2020-12-29 | 2021-04-16 | 西北工业大学深圳研究院 | Swarm robot lunar surface intelligent building system and method, robot and using method |
CN112666953B (en) * | 2020-12-29 | 2024-02-20 | 西北工业大学深圳研究院 | Intelligent lunar surface building system and method for group robot, robot and using method |
Also Published As
Publication number | Publication date |
---|---|
CN109270905B (en) | 2019-11-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109270905B (en) | A kind of method that group robot neural network based is looked for food using pheromones communication realization cooperation | |
Patle et al. | A review: On path planning strategies for navigation of mobile robot | |
Watson et al. | Embodied evolution: Embodying an evolutionary algorithm in a population of robots | |
He et al. | Path planning for indoor UAV based on Ant Colony Optimization | |
Song et al. | A novel foraging algorithm for swarm robotics based on virtual pheromones and neural network | |
Kyprianou et al. | Towards the achievement of path planning with multi-robot systems in dynamic environments | |
Qiming et al. | A review of intelligent optimization algorithm applied to unmanned aerial vehicle swarm search task | |
Abarghouei et al. | Advances of soft computing methods in edge detection | |
Floreano et al. | Design, control, and applications of autonomous mobile robots | |
Yin et al. | Multi-AGV task allocation with attention based on deep reinforcement learning | |
Chen et al. | A deep multi-agent reinforcement learning framework for autonomous aerial navigation to grasping points on loads | |
Xiao et al. | An analytical approach to the similarities between swarm intelligence and artificial neural network | |
Shi et al. | The improved Q-Learning algorithm based on pheromone mechanism for swarm robot system | |
Kordon et al. | Swarm intelligence: The benefits of swarms | |
Rozin et al. | The fuzzy ant | |
Unsal | Self-organization in large populations of mobile robots | |
Edwards et al. | Segregation of heterogeneous robotics swarms via convex optimization | |
Elloumi et al. | Ant supervised by PSO | |
Guo et al. | CrowdIM: Crowd-inspired intelligent manufacturing space design | |
Saggu et al. | Applications of swarm intelligence | |
Ohashi et al. | Cooperative transportation of a rod using mobile agents | |
Shi et al. | Adaptive reinforcement q-learning algorithm for swarm-robot system using pheromone mechanism | |
Kumar et al. | Self-sorting in a swarm of heterogeneous agents | |
Wei | Improvement and implementation of best-worst ant colony algorithm | |
Wróbel et al. | The search for beauty: Evolution of minimal cognition in an animat controlled by a gene regulatory network and powered by a metabolic system |
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 |