Ichimuraetal.SpringerPlus2014,3:712 http://www.springerplus.com/content/3/1/712 a SpringerOpen Journal RESEARCH OpenAccess Emergence of altruism behavior in army ant-based social evolutionary system TakumiIchimura1*†,TakuyaUemoto1†,AkiraHara2†andKennethJMackin3† Abstract Armyantsperformthealtruismbehaviorthatanantsacrificesitsownwell-beingforthebenefitofanotherants.They buildbridgesusingtheirownbodiesalongthepathfromafoodtothenest.Wedevelopedthearmyantinspired socialevolutionarysystembyusingSwarmlibrary.Thesystemhas2kindsofantagents,‘Majorant’and‘Minorant’. Theycommunicatewitheachotherviapheromones.Armyantcanrecognizethemasthesignalsfromtheotherants. Thepheromonesevaporatewiththecertainratioanddiffusedintothespaceofneighborsstochastically.Ifthe optimalbridgeisfound,thepaththroughthebridgeistheshortestroutefromthefoodtothenest.Wedefinethe probabilityforananttoleaveabridgeastothenumberofneighboringants.Theconstructingmethodoftheoptimal routehasbeenproposed.Inthispaper,thebehaviorsofantundertheenvironmentwithtwoormorefeedingspots wereobserved.Someexperimentalresultsshowthebehaviorsofgreatinterestwithrespecttoaltruismofants.The knowledgediscoveryofsocialevolutionaryprocessfromsomecomputersimulationresultsisdescribedinthispaper. Keywords: Artificiallifesimulationsystem;Armyant;Altruismbehavior Introduction thedaytocaptureinsects,spiders,andsoon.Thestation- In animal societies, self-organization is the theory of aryphasestartswhenthelarvaepupateforafewweeks. how minimal complexity in the individual can generate Moreover, army ants build a living nest with their bod- greater complexity at the population. The rules specify- iesinsteadofbuildinganestlikeotherants.Eachantwill ingtheinteractionsamongthecomponentsinthesystem hold on to the other legs and form a linked chain or a are implemented by using only local information with- ballstructure.Thisbehaviorisknownasabivouac.This out global information. In the study of social evolution, allowsthebridgingofanemptyspace.Inordertoaddress armyantperformsaltruismasonebehaviorofcomplex- theself-assembledstructureasaparticulartypeofaggre- ities, where each individual reduces its own fitness but gation, Deneubourg et al. defined the probability of an increasesthefitnessofotherindividualsinthepopulation. antenteringorleavingchainin(Deneubourgetal.2002). Such behaviors seem to be involved acts of self-sacrifice Moreover,theyshowedthatthegregariousbehaviorfacili- in order to aid the others. In evolutionary biology, such tatescooperationbyBlattellagermanicainsheltersduring abehavioriscalledreciprocalaltruism.Theconceptwas therestingperiod.Theprobabilitytoleavetheshelterwas initiallydevelopedtoexplaintheevolutionofcooperation defined. asmutuallyaltruisticacts(Trivers1971).Thebasicideais Ishiwata et al. (2011) developed the simulation sys- closetothestrategyof“equivalentrelation”inthestudyof tem for the foraging behavior and the altruism of army strategicdecisionmaking. ants by using Swarm library, Swarm-2.2 (Lancaster et al. Army ants are characterized by their two different 2002).(Theoriginalwebsitewww.swarm.orgisinthepro- phases of activities, a nomadic phase and a stationary cessofbeingrebuilt.)Theprobabilitiestoformthechain phase.Duringthenomadicphase,armyantsmoveduring defined in (Lioni et al. 2001) was used in their simula- tionexperiments.Thenumberofneighboringactiveants *Correspondence:[email protected] isconsideredastheconditionforaltruisticbehavior.Their †Equalcontributors 1FacultyofManagementandInformationSystems,PrefecturalUniversityof simulationresultsshowamimicaltruisticbehavior. Hiroshima,1-1-71,Ujina-Higashi,Minami-ku,734-8558Hiroshima,Japan By inspiring Ishiwata’s study, Ichimura et al. devel- Fulllistofauthorinformationisavailableattheendofthearticle oped the multi-agent simulation system to execute more ©2014Ichimuraetal.;licenseeSpringer.ThisisanOpenAccessarticledistributedunderthetermsoftheCreativeCommons AttributionLicense(http://creativecommons.org/licenses/by/4.0),whichpermitsunrestricteduse,distribution,andreproduction inanymedium,providedtheoriginalworkisproperlycredited. Ichimuraetal.SpringerPlus2014,3:712 Page2of12 http://www.springerplus.com/content/3/1/712 realistic altruistic behavior where two or more kinds of Simulationenvironment agentsrealizethedivisionofrolesinarmyants(Ichimura Theswarmisthebasicunitofsimulationforacollection and Douzono 2012). According to the environment in ofagentsexecutingascheduleofactions.TheSwarmpro- (Ichimura and Douzono 2012), the simulation results videsobjectorientedlibrariesofreusablecomponentsfor reportedthattheoptimalpathfromthefoodtothenest buildingmodelsandanalyzing,displaying,andcontrolling cannot be always found, because two or more chains in experiments on those models. We executed the altruism theenvironmentwereformed.Althoughmoreemergence simulationsystembypheromoneevaporationanditsdif- ofaltruisticbehaviorswasobserved,butthecapabilitiesof fusion in army ant multi agent systems. The developed formingchainwasdispersed.Asaresult,theperformance system is depicted in 100 × 100 2D space as shown in of foraging decreases and some ants took a circuitous Figure 1. The solid-filled rectangle, which consists of 3 route.Onthecontrary,Ichimuraetal.definedtheevap- kindofbars:‘leftbar’,‘rightbar’,and‘centerbar’,dividesthe oration rate of pheromone dues to normal distribution spaceinto2parts.Theinnerpartisthenestregionandthe probability and the probability to leave from the chain bottompartundertherectangleisthefoodsourcepart. when the ants in its neighbor region depart gradually in The 4 coordinates (x,y) of leftbar, rightbar, and center- (Ichimura and Douzono 2012). The altruism simulation barare{(30,30),(33,30),(30,70),and(33,70)},{(70,30), resultsarereportedtofindmoreoptimalpathsfromfood (73,30),(70,70),and(73,70)},and{(30,70),(73,70),(30, tothenest. 73),and(73,73)},respectively.Eachbarrepresentsaditch In this paper, we observed the behaviors of ant agents and the width of ditch is 3. The center rectangle of the under the multi feeding spots in the same environment space is ‘nest’ and the bottom rectangle under the bar is of(IchimuraandDouzono2012).Someexperimentswith ‘foodsource.’Sinceanantcannotcrosstheditchbyitself, different ratio of feed size were investigated. In general, someantsbeginaltruisticbehaviortocooperatewitheach antagentstakeanactiontobeconcentratedinthelargest other. The two hypotheses were proposed as the judg- feedingspot.Theshortestpathfromthespottothenest mentcriteriaforaltruisticactivity,Model1:Basedonthe is constructed and the ants bring feed to the nest. Then, PresenceofNeighboringAntsandModel2:Basedonthe thefeedingspotswillbedisappearedintheorderoflarger Presence of Pheromone (Ishiwata et al. 2011). In Model spot. However, it has turned out that there is a certain 1,anantwillstartformationoflivingbridgeoveragully tendencywithoutregardtothesizeoffeed.Thealtruism only when neighboring ants are present. Hypothetically, behaviordoesnotworkwellandthebridgewillbebroken, thisapproachwillbemoreefficientcomparedtoforming if enough ant agents are not gathered into the ditch. As abridgeblindly.InModel2,theplaceswherepheromone aresult,thefoodatthespotsremainstotheendofsim- concentrationsarehigherthanafixedlevelaretheloca- ulation.Moreover,thegroupofagentswasautomatically tionsthatmanyantshavepassedand/orwillpassthrough constructedduringsearch.Theexperimentalresultswith infuture. the different number of agents show the building group Figure2showstheareaofactivitiesandthevisualfield fortheeffectivesearch. byanarmyant.Inthispaper,antagentscanmoveinthe Suchbehaviorsarealsoseeninthecollaborativesocial diagonal direction, but the scattered pheromone dimin- networks. Especially, a research framework for studying ishescomparedtotheadjacentpositionsonupanddown, social systems uses agent based modeling and simula- leftandright.Thedistancefromapositiontotheneighbor tion. Madey et al. (2002, 2003) describe the simulation is defined Distance. For example, the distance to ‘A’ and results for the collaborative social network composed of ‘C’inFigure2a√re1Distanceand2Distance,respectively. opensourcesoftware(OSS)developersandprojects.The Practically,‘B’is 2Distance.However,forthesakeofease, obtained knowledge in this paper will be useful for the wedefine‘B’as1Distanceinthediagonaldirection. collaborativesocialnetworks. The remainder of this paper is organized as follows. Agentbehaviors ThesectionSimulationenvironmentdescribesthesimu- The actions include foraging for foods and transport of lationenvironmentwithSwamlibrary.ThesectionAgent them and communications with neighboring ants using behaviors defines the behaviors of agents such as search pheromone.Thepheromoneisreleasedbyanagentwhen phase, homing phase (return to the nest), and altruism it finds food. Once the pheromone is attenuated and phase. The section Proposed method describes the pro- is dispersed, the information about the food position is posedmethodrelatedtopheromoneandtheleavingprob- disseminatedamongtheants. ability from chain. Experimental results for simulations The system has 2 kinds of ant agents, ‘Major ant’ and aredescribedinthesectionExperimentalresultsforaltru- ‘Minorant’andtheantscommunicatewitheachothervia ismandthesectionExperimentalresultsfortheformation pheromones. Major ant scatters pheromones and moves of group. In the section Conclusive discussion, we give throughoutintheenvironment.Ontheotherhand,Minor somediscussionstoconcludethispaper. ant makes a mimic altruistic behavior to foraging and Ichimuraetal.SpringerPlus2014,3:712 Page3of12 http://www.springerplus.com/content/3/1/712 Figure1Environmentinarmyantsimulationsystem. transporting. Ichimura et al. shows the numerical supe- 4. Searchaditch riority in case of the 2 variants of ants (Ichimura and Searchaditchwithin1Distance.Ifthereisaditch, Douzono2012).Inthispaper,theexperimentalsimulation transittoAltruismState.otherwisegoto6). related to the altruistic behaviors has the 2 kinds of ant 5. Move agents. Movetotheotherpositionaccordingtothe Aminorantagentaimstofindafoodsourceandthen scatteredpheromonedescribedinthe toreturntothenest.Ifthereisaditchinthepathamong section“Pheromoneupdate”. them,theantwillbuildabridge.The3kindofstatesare 6. Randomselectionofwalkingdirection defined according to the behavior of ants (Ishiwata et al. Checkwhethertheotheragentstaysoraditchexists 2011). atthenextpositionexceptgoingstraightahead.The nextpositionisselectedwithanarbitraryprobabilityα. Searchstate Ifthepositionisempty,gobackto1).Otherwise, Search state is an initial condition of agents to seek the selectanotherposition.Moreover,iftheagentis food source by random walk. Once an agent reaches the surroundedbyotheragentsoraditch.itstaysatthe food,itmovesintoReturnstate.Theanttakesafoodon samepositionuntiltheneighborbecomesempty. the way back to the nest until the food source becomes empty.Belowshowsthesearchalgorithm. Returnstate 1. Searchoffoodsecure In Return state, an agent comes back to the nest carry- First,searchafoodsourcewithin1Distance.Ifthe ingthefood.Afterreachingthenest,theagentmovesinto agentfindsagoodatthedestination,movinginto Searchstate.Belowshowsthereturnalgorithm. Returnstate. 2. Perceptionofpheromone 1. Currentposition Perceivingpheromonewithin2Distances. Checkthecurrentpositionofanagent.Itisinthe 3. Searchtheotheragents nest,gotoSearchstate.Otherwise,itgoestonext Checkingtheotheragentswithin2Distances.Ifthe steptomovetothenest. otheragentsstay,goto4).Otherwisegoto6). 2. Searchaditch Ichimuraetal.SpringerPlus2014,3:712 Page4of12 http://www.springerplus.com/content/3/1/712 C B A Figure2Theareaofactivities. Ifthereisaditchwithin1Distance,goto3). Pheromoneupdate Otherwise,itmovesanextpositiontothenextand In many works related ant systems, the ants commu- goto1). nicate with each other via the pheromone dissemina- 3. Randomwalk tion.However,theresearchershavediscussedonlyabout Checkwetherthetheotheragentstaysorditch the pheromone on the ground. We consider that the existsattherandomlyselectednextpositionexcept pheromone evaporates and spreads into the space. The goingstraightahead.Ifthepositionisempty,go ant in this study can recognize the volatilization of backto1).Otherwise,selectanotherposition. pheromoneinthespace,butnotknowthepheromoneon Moreover,iftheagentissurroundedbyotheragents the ground. Based on such an idea, pheromone update oraditch,itstaysatthesamepositionuntilthe processisexecutedbyEq.(1)andEq.(3). neighborbecomesempty. Altruismstate space(cid:3)(x,y)(t) = rA×sp⎛ace(x,y)(t) ⎞ (cid:4) Some agents stop walking before a ditch and come togetherasflock.Twoormoreagentswillbuildabridge. +rB×⎝ space(ip,jp)(t)−4space(x,y)(t)⎠ belowshowsthealtruismalgorithm. ⎛ p ⎞ (cid:4) 1. Sifetahrecrhetahreeonthaegrenatgsenwtithin2Distances,theagent +rC ×⎝ space(iq,jq)(t)−4space(x,y)(t)⎠ q stayswithanarbitraryprobability1-P describedin 1 thesection“Themodelofarmyant”andcontinues (x ,y ) = {(x,y+1),(x,y−1),(x+1,y),(x−1,y)} tocheckitssurroundedsituation.Otherwise,goto ip jp 2)withtheprobabilityP1. (xiq,yjq) = {(x+1,y+1),(x+1,y−1), 2. Gotosearchstate (x−1,y+1),(x−1,y−1)} (1) Selectapositionwithin1Distanceinthepartof chain.Ifthepositionisempty,gotoposition,and space(x,y)(t+1)=space(cid:3)(x,y)(t)+re∗groundx,y(t) (2) thenmakethetransitiontoSearchstate.Otherwise, theantisembeddedinthechain. ground(x,y)(t+1)=ground(x,y)(t)−re∗ground(x,y)(t) (3) Ichimuraetal.SpringerPlus2014,3:712 Page5of12 http://www.springerplus.com/content/3/1/712 where space(x,y) means the amount of pheromone in the Table1Thesizeoffeedingspots space over the position (x,y) in Eq. (1). rA is the decay Env. Agents A:B:C Env. Agents A:B:C ratio. r is the diffusion rate in the direction of up and B 1-1 2:1:1 2-1 2:1:1 down,leftandright.r isthediffusionrateinthedirec- C tion of the diagonal. ground (t) means the pheromone 1-2 Major:3 1:2:1 2-2 Major:3 1:2:1 amountonthegroundatthexp,yosition(x,y)inEq.(2).r is 1-3 Minor:100 1:1:2 2-3 Minor:50 1:1:2 e theevaporationrate. 1-4 4:2:1 2-4 4:2:1 Multifeedingspots Figure3showstheenvironmentwithmultifeedingspots toextendthesimulationsystem.AsshowninTable1,we pheromone in simulation environment partially. The sit- investigate the behaviors of ant for some ratios of food uationincreasestheagentsswarmingaroundthem.Asa size. result,itbecomeseasytoenterAltruismStateandtwoor more bridges are built without the shortest path from a Proposedmethod foodsourcetothenest. The simulation system mainly focuses two parts, In order to avoid such a situation, the ratio of ‘PheromoneEvaporationandItsDiffusion’and‘Probabil- pheromone evaporation is defined based on the nor- ity for leaving from chain’. In this paper, the pheromone mal distribution probability as shown in Figure 4. evaporation method and a new probability for leaving Figure 5 shows the transverse plane of Figure 4. The fromchainaredefinedasfollows. rate of pheromone in 3D space is set to ‘r ’: ‘r ’: A B ‘r ’=0.788:0.043:0.010inFigure5. C Pheromoneevaporationanditsdiffusion As for the former part, Pheromone Evaporation and Its Themodelofarmyant Diffusion, the method in (Ichimura and Douzono 2012) Theprobabilitiesofanantenteringachain(P )orleaving e assumedtheimproperrateofevaporationanddiffusionof a chain (P) are depending on the size of the chain. The l pheromoneintheagentanditsbehavior.Theparameter chain is a small part of constructing bridge. Lioni et al. settingcausesabiasfortheflock.Thatis,thereismuch (2001)definedtheseprobabilitiesasfollows. Figure3Theenvironmentformultifeedingspots. Ichimuraetal.SpringerPlus2014,3:712 Page6of12 http://www.springerplus.com/content/3/1/712 Figure4Thedistributionofpheromone. C X that of the initial group. ν is a parameter of the growing P =C + e1 i (4) e e0 C +X rateofgroupduringtheconstructionofchain.Thevalues e2 i oftheseprobabilitiesshouldbeintherangeof0and1. C Pl =Cs0+ C +s1Xν (5) The function Pe expresses the idea that the probabil- s2 i ityforananttojointhechaingrowsthenumberofnest whereXiisthenumberofantsinthechaini.Ce0,Ce1,and matesalreadypresentedandreachesaplateauvalueequal Ce2 are parameters for entering the chain. Cs0, Cs1, and toCe0+Ce1.Ce0isthevalueofspontaneoushangingwhen Cs2 are parameters for leaving the chain. The size of the Xi = 0. The function Pl expresses the probability for an agent group when they leave a chain will be larger than anttoleavethechaindecreaseswithX. i r r B C r A Figure5Theevaporationrateofpheromone. Ichimuraetal.SpringerPlus2014,3:712 Page7of12 http://www.springerplus.com/content/3/1/712 The ant in the chain does not always stay in the same sectionAltruismStateforenteringachainandEq.(6)for chain. A certain probability for leaving from the chain leavingthechain. isrequiredtorealizeAltruismStatus.Duetointeraction between ants, the probability decreases with the num- Experimentalresultsforaltruism berofcon-specificsinthechain.Thephenomenaisruled Thebehaviorofarmyantsateachenvironmentasshown byempiricalequationverysimilartothatforOecophylla inTable1wasobserved.Parametersettingsareasfollows: (Deneubourgetal.2002).Theprobabilityforleavingfrom n = 2, a = 0.4, and re = 0.05. At each environment chainisgivenbyEq.(6). inthispaper,10trialsforeachsetwereexecutedandthe behaviorofantswererecordedasthemotionvideo.There a are 2 kinds of ants at each environment, Major ants to Pi = 1+bX2 (6) makearandomsearchandMinorantstofollowthescat- i tered pheromone. In this paper, for almost trials we can where X is the number of ants in the chain i. a is observe the following simulation results. During the ini- i the probability of leaving a chain under a disregard tialphaseasthesearchofanareaforprey,themovement for the number of other agents. b is the parameter of of Major agent will be a key in the change of course to depending the amount of pheromone in chain i: b = searchthesubspace,becausetheMajoragentscattersthe min{η(log(space(x,y) + 1)) + (cid:5),1}. A theoretical model pheromone while moving in a space. After the discovery suggests that these basic mechanisms account for the offood,theMinoragentcatchesthefoodandscattersthe clustering of insects (Deneubourg et al. 2002; Lioni and pheromoneonthewayfromthespottothenest.Thepath Deneubourg2004;Lionietal.2001). becomes congested since there is an obstacle of a ditch. Ichimuraetal.(2012)reportstheagentsinaltruismsit- Such situation causes the construction of the bridge on uation perform the shortest path by construction of the theditch,sincetheantssearchtheshortestpathfromthe bridge.Figure6showstheconstructedbridgeontheway feeding spot to the nest. Moreover, the agent swarming fromthefeedingspottothenest.Inthispaper,weinves- around the food increases with the size of food, because tigatethealtruismsituationbyusingtheconditioninthe morepheromoneisscatteredwhiletheagentbringsafood Figure6Thediscoverofshortestpath. Ichimuraetal.SpringerPlus2014,3:712 Page8of12 http://www.springerplus.com/content/3/1/712 tothenest.Thatis,thelargerthefoodintheenvironment IntheexperimentastoFigure8(A),agentsgatheredto has,themorepheromonetheagentsscattered. area B too much. As a result, the wide bridge was con- structed between area B and the nest. Because almost Transitionof100agents agentswereinaltruismstateandonlyafewagentscould Figures 7 show the number of agents in each area, A, B, carry the food to the nest, it took a long time to finish andC,with100agentsintheenvironment.Incaseof100 theworkinareaBandtheycouldnotproceedtothearea agents,theyaredividedintosomegroupsandeachgroup A. Some agents in area B moved to area C according to cansearchthearea,respectively.Figure7(A),Figure7(B), pheromonescatteredbyMajorants.Aftertheagentsfin- Figure7(C),andFigure7(D)showthetransitionofagents ishedtheworkinareaC,theycouldmovetoareaAatlast. withA : B : C = {2 : 1 : 1, 1 : 2 : 1, 1 : 1 : 2, 4 : 2 : When ant agents concentrated superfluously, the search 1},respectively.Thetransitionofagentsasshowninthese mightnotprogress. figuresemerged7or8timesof10trialsineachsituation. IntheexperimentastoFigure8(B),agentsgatheredto We can observe that ant agents carry the food from the area B and the bridge between area B and the nest was nearestfoodspotorthelargestsizeoffoodspot.Notethat constructed.SomeagentsinareaBwereattractedbythe evenifthefoodsizeofareaAisthelargest,thefoodofarea largestsizeofareaA.Asaresult,becausetherewerenot AremainsuntilareaBorareaC isdisappeared.Because almost agents in area C, the bridge was not constructed antagentsfindareasBorCandbringallfoodfromareaB in area C and the agents avoided the ditch to carry the orareaC,respectively. food. On the contrary, Figure 7(C) (A : B : C = 1 : 1 : 2), Wesummarythecharacteristicbehaviorsunderthe100 Figure 8(A) (A : B : C = 2 : 1 : 1), and Figure 8(B) agentsasfollows. showthattheresultsarebeyondourexpectations.Inthe experimentastoFigure7(C),moreagentsgatheredtothe 1. Mostagentsgathertotheleftareaandtheright areasBandCthantheareaAattheinitialphase.Andthen areanearfromthenest.Theymovetothe agentsmovedtoareaClargerthanareaBandtheymadea bottomareaaftertheyfinishtheworkateach bridgeintheareaC.AfterthefoodwaslostinareaC,the area. antsinareaC movedtoareaA.WeexpectedareaBthan 2. PheromonescatteredbyMajoragentsattractsthe area A will be disappeared, because some agents visited minoragentsandtheymovetooneoftheleftarea area B and they knew the existence of area B. However, andtherightarea.Theymovetothebottomarea most agents moved to A, because the pheromone at the aftertheyfinishedtheworkatthearea. initialphasewasdisappeared.Therefore,thefoodofBwas 3. Someagentsmakeabridgeonthewayfromthenest carriedtothenestafterareaAwaslost. tothefoodspot. Figure7Transitionof100Agents.(A)A:B:C=2:1:1,(B)A:B:C=1:2:1,(C)A:B:C=1:1:2and(D)A:B:C=4:2:1. Ichimuraetal.SpringerPlus2014,3:712 Page9of12 http://www.springerplus.com/content/3/1/712 Figure8Transitionof100Agents(othercase).(A)A:B:C=2:1:1and(B)A:B:C=4:2:1. Transitionof50agents nest,howeververyfewagentscannottakeallofthem.The Figures 9 show the number of agents in each area, A, B, scattered pheromone was smaller than the evaporated and C, with 50 agents in the environment. The situation pheromone.Therefore,thepheromonearoundthebridge has50agentsintheenvironment,theycannotsearchthe disappears and then the bridge was destroyed, because spacesufficiently.Theantsdidnotdivideintosomesearch the agents in altruism situation depart the bridge in the groupandthenthesearchofareawasprocessedsequen- conditionoflesspheromone.Theagentleavingfromthe tiallysuchas A → B → C.Figure9(A)andFigure9(D) bridgemovestoareaA,butthereareonlyafewagentsin showthetransitionofagentsincaseofA : B : C = {2 : the area. From such results, only few agents in the envi- 1:1,4:2:1},respectively.Asshowninthesefigures,we ronment with large size of food cannot get the altruism can observe general behaviors in the search space. That situationeasily. is,thefeedingspotsaredisappearedintheorderoflarger Wesummarythecharacteristicbehaviorsunderthe50 spot.Onthecontrary,Figure9(B)(A : B : C = 1 : 2 : 1) agentsasfollows. and Figure 9(C) (A : B : C = 1 : 1 : 2) show that the resultisbeyondourexpectations.IncaseofFigure9(B), 1. Mostagentsgathertotheleftareaandtherightarea theconstructedbridgeswerenotintheshortestpathon nearfromthenest.Theymovetothebottomarea the way to the nest as shown in Figure 8(B). In case of aftertheyfinishtheworkateacharea. Figure9(B),itisaninterestingcase,andalmostagentswill 2. Ifthefoodsizeisthelargestinthebottomarea,the makebridgesplacetoplaceintheditchfromareaCtothe agentsgatherthereregardlessofthedistancefrom nest.Theremainingagentsshoulddeliverthefoodtothe thefoodspottothenest.Theyseparatetomovethe Ichimuraetal.SpringerPlus2014,3:712 Page10of12 http://www.springerplus.com/content/3/1/712 Figure9Transitionof50Agents.(A)A:B:C=2:1:1,(B)A:B:C=1:2:1,(C)A:B:C=1:1:2,and(D)A:B:C=4:2:1. Figure10Theconstructionof3bridgessimultaneously.
Description: