HYDRA:FromCellularBiologyto
Shape-ChangingArtefacts
EsbenH.Østergaard1,enn1,PeterEggenberger2,Tim
Taylor3,1
1AdapTronicsgroup,MaerskMc-KinneyMollerInstituteforProduction
Technology,UniversityofSouthernDenmark
2ArtificialIntelligenceLaboratory,DepartmentofInformationTechnology,
UniversityofZ¨urich
3MobileRoboticsRearchgroup,InstituteofPerception,ActionandBehaviour,
UniversityofEdinburgh
RAworkprovidesinsightintotheexploitationof
holisticbehaviouralandmorphologicaladaptationinthedesignofnew
entialofthenewdesignprinciplehasbeenexemplified
throughtheconstructionofroboticsystemsthatcanchangemorphology.
Twoprototypebuildingblocksystemshasbeendeveloped,HYDRONfor
afluidscenario,YDRON
ca,theindividualmodulecanperform3Dmotionandisabletoar-
rangeinclustersofspecificformationwithoutthenecessityofphysical
TRONca,themodulesareindividuallysimpler,
attachthroughphysicalconnections,andperform3Dmotionsbycol-
lmechanismsidentifiedfromcellularbiologyhas
beensuccessfullytransferredtothephysicalbuildingblocks.
1Introduction
TheHYDRAprojectfocusonthedesignofbuildingblocksforlf-reconfigu-
ldingblocksallowrobustandefficientmorphologicalde-
velopmentofartefacts,inordertoallowend-urstodesignnewartefactsinan
edbybiologicalprinciples,theHYDRAprojectrealingi-
neeringstructureswiththepropertiesofdifferentiationandlf-reconfiguration.
Investigationsofbiologicalprinciplesrevealthatthecellisanappropriate
basisforthiswork,ads
tocontrolmechanismsbadoninspirationfromcellularmechanismssuchas
celldivision,cellmotion,celldeath,celladhesion,changeofcellshape,andcell
differentiationandinduction[6].
Byexploringdifferentpossiblebuildingblocksinsoftwareandhardwarede-
velopment,theprojectdefinesphysicalbuildingblocksthatallowdevelopmentof
systemscomprisinghundredsofbasicbuildingblocksthatexhibitlf-asmbly,
lf-repair,entialofthenewdesignstandardhasbeen
exemplifiedthroughtheconstructionoftworoboticsystemsthatcanchange
morphology.
Thecontrollingmechanismfoundinsimulationhasbeenmodified,imple-
ally,the
abstractgradient-badcontrolmechanismsfromthesimulationshasproven
:uralgenes(AandI),arecontrolledby
responsiveelements(RE).Aishereatranscriptionfactorregulatingitsownsynthesis
:Interactionsofgenes,morphogensandcellular
alstrainsofcellsarecontrolledbythetwochemicalgradients,shown
ina)andb).a)aloneproducestheshapeind),whiletheircombinationproducesthe
shapeshownine).From[3].
RAhardwareincludes20HY-
DRONmodulesforaquaticuand100ATRONmodulesforterrestrialu,
bothsystemscapableoflf-reconfigurationin3D.
2CellularBiology
Thebiologicalmechanismsoflf-asmblyandlf-repairwereinvestigatedand
modeledwithinanartificialevolutionarysysteminthecontextofcellularsys-
cprinciplesidentifiedfromthedevelopmentalmechanismswere
udtoimplementcontrolalgorithmsforthetwoHYDRAhardwareplatforms.
Havingchonthecellasourlevelofabstraction,ourmaintaskwasto
identifythodevelopmentalprocess,whichallowedanartificialevolution-
arysystemtomimicgrowthandregeneration[4]s
purpo,differentdevelopmentalprocessweresimulatedandexploredbyevo-
basiccellularmechanismswasidentified,whichcouldbeudto
simulateawiderangeofhigherlevelmechanismssuchaspatterngenerationor
co-evolutionofmorphologyandbehaviour.
Allthebiologicalmechanismsthatareesntialfordevelopment,celldivi-
sion,growth,differentiation,patternformationandmorphogenesisaremediated
teitherdirectlyoranzymestoproduceother
roteinsareencodedbygenes,sodevelopmentiscontrolled
ternofgeneexpressioninthe
embryodetermineswhere,whenandinwhatquantityparticularproteinsare
roteinsplaysuch
animportantrolefordevelopment,themechanismligand-receptorinteractions
::CellularAdhesionControlin2Dofa
gingtheattractiveandrepulsive
forcebetweentypesofcells,differentconfigurationscanbeachieved.
wasimplemented,whichcanmimicmanyspecificinteractionsamongproteins
torisusuallyalargeprotein,foldedinawaytobeableto
recognizespecifigand(asignalingmolecule)to
beufulitmustactlectivelyonparticulartargetssuchasgeneregulatorsor
ansthatareceptorwillonlyrecognizeligandsofacertain
incipleofbinding-site
andligandspecificitycanbefoundalmosteverywhereinmulticellularorganisms.
Thiscomplementaryspecificity,whichisbadontheveryexactmolecularrecog-
nitionpropertiesofmolecules,iscentraltoexplainingmanyofthephenomena
ofdevelopmentalbiology.
Thebasicmechanismsidentifiedincells-celldivision,celldeath,celladhe-
sion,expressionofreceptors,andproductionofsignalingmolecules-wereud
toevolveandsimulatehigher-levelprocesssuchascelldifferentiation,pattern
generation,morphogenesis,growthofneuralnetworkswithinter-neuronalcom-
1providesan
overviewofthesimulatedcellularmechanisms.
3TheHYDRONModule
AHYDRONunitisshowninfiitisroughlyspherical,with
adiameterofapproximately11cm,suspendedinwater,andactuatedinthe
Nunithasfournozzleswhichexpelwaterdrawn
throughanimpelleratthebottomoftheunitwhenactivated,andwhichare
gedrawsorexpelswaterthroughthebottom
oftheunittocontrolunitbuoyancy,andtherebyactuatetheunitalongthe
itshullalsosupportsasmalltofswitchableoptical
l
nsorsandtransmitterswerechonbecautheyprovideasimpleandflexible
underwatercommunicationmechanism.
SimulationworkonaCellularAdhesionMolecule(CAM)badcontrolap-
proach[8]showshowthissimple,biologically-inspiredapproachtodecentralized
multi-robotcontrolcanbeudforformingavarietyofspatialpatterns,as
showninfinbeachievedinmoduleswithlimitedcapacityfor
ulationresultsarealsoconsistentwith
thepredictionsofSteinberg’sDifferentialAdhesionHypothesisforsortingof
biologicalcellularaggregates[9].Theapproachisflexibleandrobust,andthe
designofthecontrollerpermitasytransfertotherealHYDRONrobots.
ThecombinationoftheCAMcontrollerwiththeGeneticRegulatoryNet-
work(GRN)controller[10]showsthattheGRNcanbeevolvedtoproduce
time-varyingexpressionofCAMsonarobots(virtual)membraneinorderto
achievespecifiallyformorecomplextasks(suchasreact-
ingtoanexternalsignal,orproducingdifferentiatedbehaviourfromaninitially
homogeneouscluster),theevolutionarypowerofthecombinedGRN-CAMcon-
trollerisabletoproducebetterperformancethancouldbeachievedbyeither
oftheprimitivecontrollersindividually.
4TheATRONModule
TheATRONmodules,showninfigure3andfurtherdescribedin[5],arelattice
badlf-reconfigurablerobotmodulesfor3Doperationinlandenvironments.
Greatlysimplified,anATRONmoduleiscompodoftwohemispheresjoined
togetherbyarotationmechanism
ATRONmodulescanconnectusingmechanicalhookswhichattachtoanar-
rangementofbarsonaneighbour,similarlyadhesionproteinsbindtoligandson
halfmodule,therearetwofemale(bars)
andtwoactuatedmaleconnectors(hooks).Thenovelmechanicalconnectorde-
sign,emaycommunicatewith
neighbouringmodulesthroughIRcommunication.
Whenplacedinthesurface-centredcubiclatticestructure,themodulescan
lf-reconfiguretoachievediff
shapeallowsonemoduletomovetoanadjacentholeinanotherwifullypacked
structure(withoutcollidingwithothermodules).Indeed,thedesignwasguided
byconsiderationsonhowtoreducecontrolcomplexityoflf-reconfiguration,
r,comparedtobiologicalcellsand
theHYDRONmodule,-
ityrelatedrestrictionsincludestaticstabilityoftheconfiguration,notexceeding
motortorquelimits,andobeyingstructuralstiff,careshouldbetaken
duringlf-reconfigurationtoavoidmodulecollisionsandtomaintainstructural
temalsohassimilaritieswithcellularsystems,inthata
questionishowtheprinciplesfrombiologyshouldbetransferredtosuitthehard
constraintsoftheATRONhardware.
In[1]wedescribeaGRN-stylesystem,whereeachmoduleiscontrolledby
simplerulesoftheform{precondition,action},badonthelocalneighbour-
ivationofrulesisdeterminedbythe
hormonegradients,suchthattheeffortofeachindividualmoduleisorchestrated
ongthelineshaveshownthat
suchsimplerulesbadonlocalmorphologycangeneratescalableandrobust
Fig.3.a)ExplodedviewoftheATRONCADdrawing,andb)aphotoofthefinal
hardware.c)ExperimentontheATRONplatformshowingthreewhitemodulesasa
meta-moduleintheprocessofmigratingonasubstrateofmodules.d)Thesimulated
migrationofATRONmeta-modulespermitsapproximationofthetargetshapedefined
asattractionpoints.
behavioratorganismlevel,suchasclusterwalk,obstacleavoidanceandterrain
following[7].
Chemicalgradientsareknowntoguidemigratingcells[6].Inordertoachieve
adesiredlf-reconfiguration,thismechanismhasbeentransferredtotheATRON
platform[2].ATRONmodulescanutheirIR-badneighbour-to-neighbour
ion
isachievedusingameta-moduleconsistingofthreemodules,e3c).Sucha
meta-modulehastheabilitytomoverelativelyfreelyonthesubstrateofother
-modulemergefromthestructureofmodules,migratebadon
cuesfromitnvironment,anddiewhenreachingtheirtarget,onceagainbe-
proachissimilartothedivisionanddeath
binationofgradientsandmigratingmeta-modules
enablethesystemtochangeitsshape,efigure3d),andtherebyadapttothe
requiredfunctionalityofthesystem.
5Conclusion
TheHYDRAworkprovidesinsightintotheexploitationofholisticbehavioural
andmorphologicaladaptationinthedesignofnewartefacts.
Investigationsofbiologicalprincipleshaverevealedthatthecellisanappro-
priatelevelofabstraction,ads
tocontrolmechanismsbadoninspirationfromcellularmechanismssuchascell
division,cellmotion,celldeath,celladhesion,changeofcellshape,andcelldiffer-
imulationworkshowshowsuchmechanisms
trollingmech-
anismfoundinsimulationhasbeenmodified,implementedandtestedascontrol
tion,gradient-badcontrolmechanisms
havebeenabstractedfromthesimulations.
20HYDRONmodulesand100ATRONmoduleshavebeenproduced,and
oticsystem’s
abilitytoreconficessfuloutcome
oftheprojecthasbeenvideotapedforprentationandthephysicalmodules
demonstrationshavebeenshowcadatvariousPReventsandscientificevents
worldwide.
Acknowledgment
ThisworkwasfundedbyEU’sInformationSocietyTechnologiesProgramme,
IST-2001-33060,HYDRA,FutureandEmergingTechnologies.
References
andE.H.Øoursubdivisionandgeneralizationof
rulesinrulebadcontroloftheatronlf-reconfieedings
ofInternationalSymposiumonRoboticsandAutomation(ISRA),pages67–74.
SecretariadeEducaci´onP´ublica,2004.
enn,E.H.Østergaard,dulecontrolforthe
ATRONlf-reconfieedingsoftheThe8thCon-
ferenceonIntelligentAutonomousSystems(IAS-8),pages685–692,Amsterdam,
2004.
-physicsinteractionasanewconcepttoreducethenumber
ofgeneticparametersinartifieedingsofTheIEEE2003
CongressonEvolutionaryComputation(CEC2003),2003.
ingregenerativemechanismsfoundinflatwormsbyartificial
ressonEvolu-
tionaryComputation(CEC),Canberra,2004.
5.M.W.Jørgenn,E.H.Østergaard,rATRON:Mod-
ulesforalf-reconfieedingsofIEEE/RSJInternational
ConferenceonIntelligentRobotsandSystems(IROS),pages2068–2073,2004.
forpatternformationofhypostome,tentacles,andfootin
hydra:howtoformstructuresclotoeachother,howtoformthematadistance.
DevelopmentalBiology,157(2):321–333,June1993.
7.E.H.Østergaard,,butedmetamorphosisof
eedingsof6thInternationalSymposiumon
DistributedAutonomousRoboticSystems(DARS),Toulou,France,June23-26,
pages161–170,2004.
owardlf-reconfigurablerobotsystemsby
,,ni,
a,¨o,editors,ProceedingsoftheThe8thConferenceonIntel-
ligentAutonomousSystems(IAS-8),pages720–728,Amsterdam,
Press.
e,141:410–
408,1962.
icregulatorynetwork-inspiredreal-timecontrollerforagroupof
,,ni,a,¨o,
editors,ProceedingsoftheThe8thConferenceonIntelligentAutonomousSystems
(IAS-8),pages403–412,Amsterdam,ss.
本文发布于:2023-01-02 12:00:21,感谢您对本站的认可!
本文链接:http://www.wtabcd.cn/fanwen/fan/90/77558.html
版权声明:本站内容均来自互联网,仅供演示用,请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系,我们将在24小时内删除。
留言与评论(共有 0 条评论) |