基于多Agent的网络控制模型

《基于多Agent的网络控制模型》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、东南大学博士学位论文基于多Agent的网络控制模型姓名：卞正皑申请学位级别：博士专业：计算机应用技术指导教师：罗军舟20080707东南大学博1：学位论文ABSTRACTCurrentnetworkmanagementofIntemetisbasedonpointtopointmanagementmodel．Themanagementmodelhasthesefollowingfeatures：protocolsanddevicesexposetheirinternaldetailstothemanager,man

2、ageraccomplishesitsdesirebydeployingalotofmanualconfigurations，andtheinformationpanelsofnetworkelementsareindependenteachother．Themanagementmodelshowstheflawsofpoormanageabilityasthenetworkevolves：forexamples，perceptiondiffersfromreality，error-proneconfigurati

3、on，andlackofdependencymaintenance，theaboveflawswillbringthedifficultyofnetworkmanagement．SosomeresearcherspresentedtheideaofnewmanagementpartitioninthenextgenerationIntemetresearch(suchastheworkof4Dproject)．Theydetachedthecommonmanagementdecisionfunctionsfromt

4、hevariousconfigurationprofilesandcomposedthemasanindependencedecisionplane．Iftheideaisfeasible，thenetworkmanagementsystemwillactasaclose-loopcontrolsystem．Thedeciscionplaneadoptsstrategiesaccordingtotherelationshipsamongthenetworkelementstoachievetheobjectstat

5、es．ItisevidentthattheideaCanadapttothedemandsofmanagementtocomplexservicesifitdoesworks．Buttherearemanyissuesarisefromthisidea．Theyincludethedefinitionofdynamiccoordinationrelationshipamongdifferentdecisionentities，thestabilityofcoordinationrelationshipandthec

6、ollaborationamongdifferententities．Inordertosolvetheseissues，aflexiblesoftwaretechnologyisneeded．SoagenttechnologyWasintroducedandsomesolutionsweredescribedwithinthispaper．ThechiefaimofrecentworkWastoinvestigatethefeaturesofthenetworkcontrolsystemandextendedth

7、emeritofagenttechnologytothestudyofnetworkmanagementandcontr01．AbasicstatespaceequationWaspresentedforhierarchicaldecisionagentcoordinationmechanismbasedontheabstractmodelsofindividualagentandinteractionbehaviorsamongmulti—agent．Weagainproposedthebasicprincipa

8、lsofmulti—agentsystem；theyarecontrollabilityandobservabilityofsystem，optimalityandstabilityofgoalstate，reachabilityofcontrolinformationandoptimalcollaborationmechanism．Controllabil