30023

ارايه يك رويكرد يادگيري ماشين براي ارتباطات دستگاه-به-دستگاه همياري شده با راديوشناختي در شبكه‌هاي 5G

كارشناسي ارشد

مهندسي كامپيوتر- گرايش شبكه هاي كامپيوتري

1398

1402/2/13

دكتر ابوالفضل ديانت

مهندسي كامپيوتر

-هﺎﮕﺘﺳد تﺎﻃﺎﺒﺗرا رد ار يرﺎﯿﺴﺑ يﺎﻫﺖﯿﻠﺑﺎﻗ ،ﯽﻃﺎﺒﺗرا ﻪﺘﻓﺮﺸﯿﭘ يﺎﻫيروﺎﻨﻓ زا ﯽﮑﯾ ناﻮﻨﻋ ﻪﺑ (5G) ﺞﻨﭘ ﻞﺴﻧ هاﺮﻤﻫﻦﻔﻠﺗ يﺎﻫﻪﮑﺒﺷ ﺎﺑ ﻢﯿﻘﺘﺴﻣ طﺎﺒﺗرا ﻪﺑ زﺎﯿﻧ نوﺪﺑ ﺮﮕﯾﺪﮑﯾ ﺎﺑ ﺪﻧردﺎﻗ ﺎﻫﻪﮑﺒﺷ ﻦﯾا .ﺪﻨﻨﮐﯽﻣ ﻢﻫاﺮﻓ ﯽﺘﺧﺎﻨﺷﻮﯾدار دﺮﮑﯾور زا هدﺎﻔﺘﺳا ﺎﺑ (D2D) هﺎﮕﺘﺳد-ﻪﺑ ﺮﺑ ﯽﻨﺘﺒﻣ 5G يﺎﻫﻪﮑﺒﺷ رد ﯽﺘﺧﺎﻨﺷﻮﯾدار دﺮﮑﯾور .ﺪﻧراﺬﮕﺑ كاﺮﺘﺷا ﻪﺑ ار تﺎﻋﻼﻃا و ﺪﻨﻨﮐ راﺮﻗﺮﺑ طﺎﺒﺗرا ﻢﯿﻘﺘﺴﻣ ترﻮﺻ ﻪﺑ ،ﻪﯾﺎﭘ هﺎﮕﺘﺴﯾا يﺎﻫزﺎﯿﻧ و ﯽﮑﯾﺰﯿﻓ ﻂﯿﺤﻣ ﻪﺑ ﻪﺟﻮﺗ ﺎﺑ .ﺪﻨﺘﺴﻫ ﯽﻟﺎﺧ يﺎﻫﺲﻧﺎﮐﺮﻓ ﻒﯿﻃ ﯽﯾﺎﺳﺎﻨﺷ ﻪﺑ ردﺎﻗ ﺎﻫهﺎﮕﺘﺳد ﻦﯾاﺮﺑﺎﻨﺑ .ﺖﺳا ﺎﯾﻮﭘ ﻒﯿﻃ ﺺﯿﺼﺨﺗ ياﺮﺑ ﺎﻫنآ زا و ﺪﻨﻫد ﺺﯿﺨﺸﺗ ار هدﺎﻔﺘﺳاﻼﺑ ﺲﻧﺎﮐﺮﻓ ﺲﭙﺳ .ﺪﻨﻨﮐ ﻞﯿﻠﺤﺗ و ﻪﯾﺰﺠﺗ ار ﺎﻫﺲﻧﺎﮐﺮﻓ ﻒﯿﻃ ﺪﻨﻧاﻮﺗﯽﻣ ﺎﻫهﺎﮕﺘﺳد ،ﯽﻃﺎﺒﺗرا .ﺪﻨﻨﮐ هدﺎﻔﺘﺳا طﺎﺒﺗرا يراﺮﻗﺮﺑ ﻪﻠﺻﺎﻓ .ﺖﺳا لﺎﻧﺎﮐهﺮﻬﺑ و ﺮﯿﺴﻣ ﺖﻓا ،ﻪﻠﺻﺎﻓ ﻞﻣﺎﺷ نآ ﯽﻠﺻا ﺮﺻﺎﻨﻋ ﻪﮐ ،ﺖﺳا ﻞﺧاﺪﺗ ندﺮﮐ ﻞﻗاﺪﺣ ﻪﻟﺎﺴﻣ رﺎﺘﺷﻮﻧ ﻦﯾا ﻢﻬﻣ ﺶﻟﺎﭼ ﻪﺑ ﺮﯿﺴﻣ ﺖﻓا .ﺪﺷﺎﺑ ﻪﺘﺷاد هﺎﮕﺘﺳد-ﻪﺑ-هﺎﮕﺘﺳد طﺎﺒﺗرا دﺮﮑﻠﻤﻋ و ﺖﯿﻔﯿﮐ ﺮﺑ ﯽﯾﻻﺎﺑ ﺮﯿﺛﺎﺗ ،ﺪﻧاﻮﺗﯽﻣ طﺎﺒﺗرا رد يﺎﻫهﺎﮕﺘﺳد ﻦﯿﺑ ﯽﮑﯾﺰﯿﻓ ﯽﮑﯾ .ﺪﺷﺪﻫاﻮﺧ دﺎﺠﯾا ﯽﮑﯾﺰﯿﻓ و ﯽﻄﯿﺤﻣ صاﻮﺧ ﻞﯿﻟد ﻪﺑ ﺖﻓا ﻦﯾا .دراد هرﺎﺷا ﺎﻫهﺎﮕﺘﺳد ﻦﯿﺑ ﯽﻃﺎﺒﺗراﺮﯿﺴﻣ لﻮﻃ رد لﺎﻨﮕﯿﺳ ﻒﯿﻌﻀﺗ ﺮﺻﺎﻨﻋ ﻪﯾﺎﭘهﺎﮕﺘﺴﯾا ﮏﯾ و ﻪﯾﻮﻧﺎﺛناﺮﺑرﺎﮐو ﻪﯿﻟواناﺮﺑرﺎﮐزا ياﻪﻋﻮﻤﺠﻣ .ﺖﺳا ناﺮﺑرﺎﮐ ياﺮﺑ ﻞﺧاﺪﺗ ،هﺪﺷﻪﺘﺧادﺮﭘ نآ ﻪﺑ ﻪﮐ ﻂﯾاﺮﺷ ﻦﯾازا .ﺪﻨﺘﺴﻫ لﺪﻣ ياﺮﺑ تﺎﻃﺎﺒﺗرا ﯽﻠﺻا زا ﻪﮐ ﯽﻧاﺮﺑرﺎﮐ.ﺪﻧﻮﺷﯽﻣ هﺪﯿﻣﺎﻧ ﻪﯿﻟوا ناﺮﺑرﺎﮐ.ﺪﻨﺷﺎﺑ ﻪﺘﺷاد ﻢﯿﻘﺘﺴﻣ ﯽﺳﺮﺘﺳد ﻪﻣﺎﻨﯿﻫاﻮﮔ ياراد ﻒﯿﻃ ﻪﺑ ﺪﻧراد هزﺎﺟا ﻪﮐ ﯽﻧاﺮﺑرﺎﮐ ﻪﺑ مﺰﻠﻣ ﻪﯾﻮﻧﺎﺛ ناﺮﺑرﺎﮐ .ﺪﻧﻮﺷﯽﻣ هﺪﯿﻣﺎﻧ ﻪﯾﻮﻧﺎﺛ ناﺮﺑرﺎﮐ،ﺪﻨﻨﮐﯽﻣ هدﺎﻔﺘﺳا ﯽﺼﺨﺸﻣ ﻂﯾاﺮﺷ ﮏﯾ ﺖﺤﺗ ،ﻪﯿﻟواناﺮﺑرﺎﮐ هدﺎﻔﺘﺳاﻼﺑ ﻒﯿﻃ ﻪﯿﻟوا ناﺮﺑرﺎﮐ ﻪﺑ ﻪﯾﺎﭘ هﺎﮕﺘﺴﯾا ﻂﺳﻮﺗ ﻪﻣﺎﻨﯿﻫاﻮﮔ ياراد ﻒﯿﻃ .ﺪﻨﺘﺴﻫ ﻪﯿﻟواناﺮﺑرﺎﮐ ﺖﺳاﻮﺧرد ﺾﺤﻣ ﻪﺑ هﺪﺷ لﺎﻐﺷا ﻒﯿﻃ نﺪﻧادﺮﮔﺮﺑ و ﻞﺧاﺪﺗ ﺶﻫﺎﮐ ياﺮﺑ يزﺎﺳﻪﻨﯿﻬﺑ ﻪﻟﺎﺴﻣ ﮏﯾ .ﺪﺷﺪﻫاﻮﺨﻧ هدﺎﻔﺘﺳا ﻞﻣﺎﮐ ترﻮﺻ ﻪﺑ ﻪﯿﻟواناﺮﺑرﺎﮐ ﻂﺳﻮﺗ ﻒﯿﻃ ﻦﯾا .ﺪﺑﺎﯾﯽﻣ صﺎﺼﺘﺧا يروأﻊﻤﺟ ياﺮﺑ نآ ﺞﯾﺎﺘﻧ زا و هﺪﺷ ﻞﺣ ﮏﯿﺳﻼﮐ شور ﻪﺑ ﻪﻟﺎﺴﻣ ﻦﯾا .ﺖﺳاهﺪﺷ ﻒﯾﺮﻌﺗ ﺶﻟﺎﭼ ﻦﯾا ﻞﺣ ياﺮﺑ هدﺎﻔﺘﺳاﻼﺑ ﻒﯿﻃ يروهﺮﻬﺑ .ﺪﺷﺪﻫاﻮﺧ هدﺎﻔﺘﺳا هدادﻪﻋﻮﻤﺠﻣ ﺖﻓﺎﯿﻫر زا هدﺎﻔﺘﺳا ﺎﺑ ﻦﯾاﺮﺑﺎﻨﺑ .ﺖﺳا ﻻﺎﺑ ﯽﻧﺎﻣزﻪﻨﯾﺰﻫ مﺰﻠﺘﺴﻣ ﺮﯿﻐﺘﻣ ناﺮﺑرﺎﮐ داﺪﻌﺗ و ﯽﻟاﻮﺘﻣ يﺎﻫراﺮﮑﺗ ياﺮﺑ ﻪﻟﺎﺴﻣ ﮏﯿﺳﻼﮐ ﻞﺣ ﻪﺘﺧادﺮﭘ يزاﻮﻣﻪﮑﺒﺷ ﺪﻨﭼ ﻪﻠﯿﺳﻮﺑ ﯽﺒﺴﭼﺮﺑﺪﻨﭼيﺪﻨﺑﻪﻘﺒﻃ هﺪﯾا ﺎﺑ ﻒﯿﻃ يروﺮﻬﺑ ياﺮﺑ ،ﻞﺧاﺪﺗ ﺶﻫﺎﮐ ﻪﻟﺎﺴﻣ ﻞﺣ ﻪﺑ ﻖﯿﻤﻋيﺮﯿﮔدﺎﯾ ﯽﺟوﺮﺧ و يدورو يﺎﻫﻪﯾﻻ ﺖﺳرد ﻢﯿﻈﻨﺗ ﺎﺑ ﻪﮐ هﺪﺷدﺎﻬﻨﺸﯿﭘ رﻮﮐﺬﻣ هﺪﯾاﻞﺣ ياﺮﺑ ،(MLP) ﻪﯾﻻﺪﻨﭼ يﺎﻫنوﺮﺘﭙﺳﺮﭘ ﻢﺘﯾرﻮﮕﻟا .ﺖﺳاهﺪﺷ ﺖﺳﺪﺑ ﻪﻨﯿﻬﺑ تﺎﻄﻘﻧ ﻪﺑ ﯽﯾﻻﺎﺑ ﺖﻗد ﺎﺑ هﺪﯾا ﻦﯾا ،داد نﺎﺸﻧ ﻪﮐ ﺪﺷ ﯽﺠﯾﺎﺘﻧ ﻪﺑ ﺮﺠﻨﻣ،بﻮﻠﻄﻣ ﻪﻨﯾﺰﻫ ﻊﺑﺎﺗ و ﺐﺳﺎﻨﻣ زﺎﺳلﺎﻌﻓ بﺎﺨﺘﻧا و .ﺖﺳا ﺮﺗبﻮﻠﻄﻣ رﺎﯿﺴﺑ ﮏﯿﺳﻼﮐ شور ﻪﺑ ﺖﺒﺴﻧ ﯽﻧﺎﻣز ﻪﻨﯾﺰﻫ ﺪﻌﺑ زا و ﮏﯾدﺰﻧ ﮏﯿﺳﻼﮐ شور ﻪﺑ هﺪﺷ ﻞﺣ يزﺎﺳﻪﻨﯿﻬﺑ ﻪﻟﺎﺴﻣ هﺪﻣآ

1402/08/13

A Machine Learning approach toward cognitive radio assisted D2D communication in 5G networks

5/2/2024 12:00:00 AM

5G mobilenetworks,asoneoftheadvancedcommunicationtechnologies,providemanycapabilitiesindevice- to-devicecommunicationsusingacognitiveradioapproach.Thesenetworkscanestablishdirectcommunication and shareinformationwitheachotherwithouttheneedfordirectconnectiontoabasestation.Thecognitive radio approachin5Gnetworksisbasedondynamicspectrumallocation.Therefore,devicesareabletoidentify available frequencyspectra.Dependingonthephysicalenvironmentandcommunicationrequirements,devices can analyzefrequencyspectra,detectunusedfrequencies,andutilizethemforcommunication. A significantchallengeinthiscontextisminimizinginterference,whichincludesfactorssuchasdistance, pathloss,andchannelcapacity.Thephysicaldistancebetweencommunicatingdevicescanhaveasignificant impactonthequalityandperformanceofdevice-to-devicecommunication.Pathlossreferstosignalattenuation along thecommunicationpathbetweendevices,whichoccursduetoenvironmentalandphysicalproperties. One oftheseconditionsaddressedistheinterferenceamonguserpaths.Asetofprimaryusers,secondaryusers, and abasestationarethekeycomponentsofthecommunicationmodel. Primaryusersareauthorizedtohavedirectaccesstothespectrum.Secondaryusersarethosewhousethe unusedspectrumofprimaryusersunderspecificconditionsandareobligatedtoreleasetheoccupiedspectrum uponrequestfromprimaryusers.Thespectrumwithalicenseisallocatedtoprimaryusersbythebasesta- tion. Thisspectrumwillnotbefullyutilizedbyprimaryusers.Anoptimizationproblemisdefinedtoreduce interferenceandimprovetheefficiencyofunusedspectrumutilizationtosolvethischallenge.Thisproblemis solvedusingclassicalmethods,andtheresultswillbeusedfordatasetcollection. Solving theclassicalproblemforconsecutiveiterationsandvariablenumbersofusersincurshightimecosts. Therefore,adeeplearningapproachhasbeenappliedtosolvetheinterferencereductionproblemandimprove spectrumefficiencyusingtheideaofmulti-labelclassificationthroughmultipleparallelnetworks.TheMulti- LayerPerceptron(MLP)algorithmisproposedtosolvethementionedideabyappropriatelyconfiguringthe input andoutputlayers,selectingsuitableactivationfunctionsanddesiredlossfunctions.Theresultshave shownthatthisideaachieveshighaccuracyinobtainingoptimalpointsforthesolvedoptimizationproblem, closely approachingtheclassicalmethod,andsignificantlyreducingtimecostscomparedtotheclassicalap- proach.

راديوشناختي , ارتباطات دستگاه-به-دستگاه , يادگيري عميق , شبكه هاي موازي , طبقه بندي چندبرچسبي , شبكه هايارتباطاتسلولي نسل

CognitiveRadio , Device-to-DeviceCommunication , DeepLearning , ParallelNetworks , Multilable Classification , FifthGenerationCellularNetworks

Shima Torabi

Dr. Abolalfazl Diyanat