分布式網(wǎng)絡管理中多Agent任務分解與調(diào)度.pdf

1、東南大學博士學位論文分布式網(wǎng)絡管理中多Agent任務分解與調(diào)度姓名：劉波申請學位級別：博士專業(yè)：計算機應用技術(shù)指導教師：羅軍舟20061201ABSTIUCTABSTRACTTherapidgrowthofnetworkinformntionandnetworkheterogeneitypromotethedevelopmentofthedistributednetworkmanagementAgenttechnologyhasplay

2、edallimportantroleindistributednetworkmanagementInagentbaseddistributednetworkmanagementthemainproblemsmtaskdecompositionandmultiagentschedulingTaskdecompositiontechnologyintheamofdistributedcomputinghasdrawnmuchattentio

3、nwhichrangesfromserialprogramdecompositiontoprogramslicingbutthesemethodsCallnotapplytOthecomplicatedandlargenetworkmanagementtaskAgentschedulingisacriticalissueindistributednetworkmanagementwhichhasgreatinfluenceontheef

4、ficiencymultiagenttasksystem。Researchershavedevotedthemselvesintostudyingvariousschedulingmethodssuchasaddingrestrictionsorsimplifyingtheslnlcn鵬oftasksButthesemethodsarenotadaptivewellforcomplicatedtasksCenteredwithagent

5、taskdecompositionandschedulingweproposedaseriesofnewalgorithmsandmethodsinthisdissertationtosolvetheproblemsmentionedaboveFirstly，basedontheresearchonagent’Sabstractmodelendinteractivemodel，adistributednetworkmanagementf

6、rameworkandamultiagentschedulingframeworkarepresentedwhichintroduceanetworkmanagementscenariotosupportdynamicschedulingdecisionsDuringthetaskprocessaXMLbasedtasktreeviewdescriptionlanguage(TTVDL)isusedtodescribetherelati

7、onshipamongthevariousrolesBasedOIltaskrepresentationaseriesofalgorithmsafepresentedtodecomposethewholenetworkmanagementtaskintoseveralgroupsofsubtasksDuringthecourseofdecompositiondifferentprioritiesalflgassignedtosubtas

8、ksPriority—rankedsubtasks撇groupedaccordingtotheirinterdependencesSubtaskswitllthes啪epriorityareputintothesamegroupendtheyCallbeperformedinparallelmannerThealgorithmcanimprovetheefficiencyoftaskexecutionendsavenetworkband

9、widthButnexLhowtomakenetworkmanagermanagethenetworkefficientlyandrespondquicklyendexactlyisanotherimportantproblemEffectivedynamicschedulingforagentscanmanagewiththisissueforacertaindegreeSoasemionlinemultiagentschedulin

10、galgorithmbasedondependencesofsubtasksisproposedThecompetitiveanalysisandproofrevealthattheexistingfullonlineschedulingalgorithmisaspecialca靶ofsemionlineschedulingalgorithmWhenconfrontedwithcomplicatedtaskscomprisingboth

11、serialandparallelrelationships，thesemionlineschedulingalgorithmisbelterthentheexistingfullonlineschedulingalgorithmWhenconfrontedwithparalleltasksthesemi—onlineschedulingalgorithmisequaltothefullonlineschedulingalgorithm

12、HowtomakethemostoftheexistingagentsisanotheroptimizationaspectSOaloadbalancingstrategybasedonprobabilityisproposedTovalidatetheperformanceoftheschedulingmechanismendthecharacteristicoftheschedulingmodel，coloredPetrinetis