PinPoint精确定位射频干扰源

PinPoint:LocalizingInterferingRadios

KiranJoshi,StevenHong,SachinKatti{krjoshi,hsiying,skatti}@stanford.edu

Abstract

ThispaperpresentsPinPoint,atechniqueforlocalizingrogueinterferingradiosthatadheretostandardproto-colsintheinhospitableISMbandwithoutanycooper-ationfromtheinterferingradio.PinPointisdesignedtobeincrementallydeployedontopofexisting802.11WLANinfrastructure,andusedbynetworkadministra-torstoidentifyandtroubleshootsourcesofinterferencewhichmaybedisruptingthenetwork.PinPoint’skeycon-tributionisanovelalgorithmthataccuratelycomputesthelineofsightangleofarrival(AoA)andcyclicsig-nalstrengthindicator(CSSI)ofthetargetinterferingsig-nalatallAPs,evenwhenthelineofsight(LoS)com-ponentisburiedbystrongermultipathcomponents,inter-ferenceandnoise.PinPointleveragesthisalgorithmtodesignanoptimizationtechnique,whichcanlocalizein-terferingradiosandsimultaneouslyidentifythetypeofin-terference.Unlikeseverallocalizationtechniqueswhichrequireextensivepre-deploymentcalibration(e.g.RF-Fingerprinting),PinPointrequiresverylittlecalibrationbythenetworkadministrator,andusesanovelalgorithmtoself-initializeitsbearings,evenifthelocationsofsomeAPareinitiallyunknownandareorientedrandomly.WeimplementPinPointonWARPsoftwareradiosandde-ployinanindoortestbedspanninganentire oorofourdepartment.WecomparePinPointwiththebestknownpriorRSSI[8,11]andMUSIC-AoAbasedapproachesandshowthatPinPointachievesamedianlocalizationerrorof0.97meters,whichisaroundthreetimeslowercomparedtotheRSSI[8,11]andMUSIC-AoAbasedap-proaches.

tackledindoorlocalization.However,noneofitisappli-cabletolocalizinginterferingradios.First,mostofthemareRSSIbasedandworktypicallywithWiFi,i.e.theymeasuretheRSSIoftheWiFisignalfrommultiplevan-tagepoints,andthenleveragepropagationmodelsandtri-angulationtechniquestolocalize.However,whenlocal-izinginterference,thesourcecouldoftenbeanon-WiFiradio.Further,itsunlikelywecangetagoodestimateoftheinterferingsignalsRSSIbecausetherecouldbemul-tiplesignalspresentfromconcurrenttransmittingradios.AnotherclassofRSSItechniquesrequiresextensiveRF ngerprintingoftheindoorenvironment.HoweverthesetechniquesdonotworkunderinterferenceeithersincetheRSSI ngerprintswillbedistortedwhentherearemultipleconcurrenttransmissions.Further,thesetech-niquesareexpensivetodeploysincetheyrequirecon-stantandrecurringsite ngerprinting.Anotherclassoftechniques[16,17,18,19,24]usenon-RSSIbasedtech-niquessuchasrange- ndingandtimeofarrival,howeverallofthemrequiremodi cationstoandcooperationfromtheclientradio(e.gintheformofspecialbeaconinghard-ware),whichisuntenablewhenwearetryingtolocalizeaninterferernotunderourcontrol.

InthispaperwepresentPinPoint,asystemthatcom-putesthenatureaswellasthelocationoftheinterferingradio(s)withsub-meteraccuracy.PinPointisrobust,itcanlocalizeeachinterferingradioevenwhenmultiplein-terferingradiosmaybetransmittingconcurrently.Fur-thermore,PinPoint’saccuracyisatleasttwotimesbetterascomparedtoRSSIbasedtechniquesevenwhennoin-terferenceispresent.HenceeventhoughPinPoint’sde-signismotivatedbythescenariooflocalizinginterfer-ence,itprovidesageneralindoorlocalizationtechniquethatworksacrossawidevarietyofscenarios.ThesystemconsistsofanindoorAPinfrastructurewithPinPointca-pability,ofwhichasmallsubset(3 5per oorofalargedepartmentbuilding)areanchorAPsthatalreadyknowtheirabsoluteindoorlocation.ThePinPointAPsworktogethertodetectandlocalizeinterferingradios.Pin-Pointassumesnoco-operationfromtheinterferingradio,workswithlegacyclientradios,anditdoesnotassumeanyknowledgeoftheprotocol,powerorthespectrumatwhichinterferingradiosaretransmitting.Further,Pin-Pointdoesnotrequireanyexpensivecalibrationorsur-veying,eitheratinstallationorinsubsequentoperation.Webelievethiscombinationofaccuracy,robustnessandgeneralityisa rst.

PinPoint’skeycontributionisanovelalgorithmthatac-1

1Introduction

Interferenceisthenumberonecauseforpoorwirelessperformance.Allofushavehadanecdotalexperiences,where,eventhoughtheAPisquiteclose,weexperiencepoorperformanceandmoreoftenthatnot,interferenceistoblame.Yet,inspiteofthesepervasiveproblems,weof-tenknowverylittleaboutwherethisinterferenceiscom-ingfrom.Wedonotknowthenatureoftheinterferingra-dio(e.g.whetheritisanotherWiFinetwork,BluetoothorZigbee),neitherdoweknowwhereitislocated.Withoutsuchlocalization,troubleshootingperformanceproblemsbecomeshard.

Onemightimaginethatwecouldleveragetheextensivepriorwork[4,8,11,15,20,21,22,23,24,25]thathas