apmcmlz2302188亚太杯分赛道五岳杯一等奖.pdfVIP

Team#apmcmlz2302188

TeamNumber:apmcmlz2302188

ProblemChosen:A

2023APMCMsummarysheet

ThispaperestablishesaQuantitativeUnaryBinaryOptimization(QUBO)model

basedonbinaryvariablematricestoaddresstheresourceallocationissuesinComputing

PowerNetworks(CPN).Theobjectiveistodistributecomputationalpower,storage,and

algorithmicinformationthroughthenetworkaccordingtouserdemands,involvingend-

users,edge,andcloudservers.Thefocusoftheresearchisonmaximizingthecoverage

ofcomputationalresourceswhileenhancingtheefficiencyofthefacilityundercertain

constraints.

ForProblem1,aQUBOmodelisemployedtomodeltheedgeserverresource

coverageissue.Withina4脳4gridarea,a16-dimensionalbinaryvariablematrix

representsthepossiblelocationsofedgeservers.Theobjectiveistomaximizethetotal

computationaldemandcoverage,withtheconstraintofhavingonlytwoedgeservers.

TheQUBOmatrixisconstructedunderthisconstraint.Thismodelisthenconverted

intoanIsingmodeltorepresenttheHamiltonian,whichissolvedusingaCIMsimulator

(thecoordinatesofthetwoedgeserversare[2,3]and[3,1],withthemaximum

totalcomputationaldemandcoveragebeing419).Tovalidatetheresults,bothclassic

simulatedannealingandtabusearchmethodsareemployed,withtheoptimalsolution

matchingthequantumsimulatedannealingresult.Additionally,underdifferentpenalty

termcoefficients,201quantumsimulatedannealingrunsareconducted,resultinginan

averagerunningtimeof0.596sandasensitivityoftheHamiltonianequationtothe

penaltytermcoefficientchangeof20.080.

ForProblem2,incontrasttoProblem1,thekeyistodesignaQUBOmodelthat

doesnotexceed100quantumbits.Weintroducevariables:the