资源描述：
《EvolvedMachines Shed Light on Robustness and Resilience.pdf》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、INVITEDPAPEREvolvedMachinesShedLightonRobustnessandResilienceThispaperdiscussesthetwo-wayinteractionbetweenbrainsandbodies,andtheconsequencesforadaptivebehavior,alongwithreviewingresearchthatbuildsoninsightsfromtheneurobiologyoftheseinteractionstoinformthedesignofevolvingandadaptiverobots.ByJos
2、hBongard,MemberIEEE,andHodLipson,MemberIEEEABSTRACT
3、Inbiomimeticengineering,wemaytakeinspira-I.INTRODUCTIONtionfromtheproductsofbiologicalevolution:wemayEmbodiedcognition[16],[65](Fig.1)isaparticularinstantiatebiologicallyrealisticneuralarchitecturesandalgo-paradigmforaddressingthechallengeofcr
4、eatingauton-rithmsinrobots,orwemayconstructrobotswithmorpholo-omousmachines,aswellasunderstandinghowanimalsgiesthatarefoundinnature.Alternatively,wemaytakeandhumansproduceadaptivebehavior.Itdictatesthatinspirationfromtheprocessofevolution:wemayevolvecognitioncannotbeunderstoodVorrealizedinmachi
5、nesVpopulationsofrobotsinsimulationandthenmanufacturewithoutgroundingitinbehavior.Indeedcognitioncanbephysicalversionsofthemostinterestingormorecapableviewedasacollectionofparticularlysubtlewaysofrobotsthatevolve.Ifwefollowthislatterapproachandevolveexploitingone’sinteractionwiththeenvironment;
6、anboththeneuralandmorphologicalsubsystemsofmachines,interactionthatrequiresneuralcomputationbutalsousewecanperformcontrolledexperimentsthatprovideuniqueofabodytocarryoutthatinteraction.Itfollowsfromthisinsightintohowbodiesandbrainscanworktogethertothattheremayexistalargenumberofpairingsofpartic
7、ularproduceadaptivebehavior,regardlessofwhethersuchbodiesmechanicalconstructsandneuralarchitecturesthat,whenandbrainsareinstantiatedinabiologicalortechnologicalcombined,giverisetothebehaviorofinterest.Thisraisessubstrate.Inthispaper,wereviewselectedprojectsthatuseseveralengineeringchallenges.su
8、chmethodstoinvestigatethesynergiesandtradeoffsFirst,ifaparticularmorphologyischosen,thedesignerbetweenneuralarchitecture,morphology,action,andadaptivemayunintentionallyconstrainhimselftoasetofbehavior.unsatisfyingneuralsystems.For
