CovenantUniversity

7/7/17 1

CalibratingtheStandardPathLossModelforUrban

EnvironmentsusingField

MeasurementsandGeospatialData

S.I.PopoolaA.A.AtayeroN.FarukC.T.CalafateE.AdetibaV.O.Matthews

CovenantUniversity,Ota,Nigeria

Presentedat

The2017InternationalConferenceofWirelessNetworksWorldCongressonEngineeringLondon,UnitedKingdom

7thJuly,2017.2

Introduction

Thesusceptibilityofthewirelesschanneltobothsignalfadingandinterferencehasmadeefficientradionetwork

planningavitalpartofthepre-deploymentprocessincellular

systems[1].

Thereliabilityoftheradionetworkplanning,towardoptimalcoverage,largely

dependsontheaccuracyofthepathlosspredictionmodels that

areemployed[2].

Introduction(Cont’d)

Findingsfrompreviousresearch[3-7]showedthatthepredictionresultsofwidelyusedempiricalpathlossmodelsdonotmatch,

ingeneral,thefieldmeasurementdatacollectedonthelocalterrainsinNigeria.

Hence,theneedformodelcalibration thatwillimprovetheaccuracyofthemodelssoastotrulyrepresenttheactualsignalpropagationbehaviorofthetargetlocalenvironmentsof

Nigeria.

ResearchAim

•ThisresearchworkisaimedatimprovingthepredictionaccuracyoftheStandardPropagationModel(SPM)tothespecificcaseofLagos,Nigeria,byadequatelyaccountingfortheeffectoflocalgeographicfeatureswithintheurbanpropagationenvironments.

5

MaterialsandMethod

•Extensivemeasurementcampaignswereconductedinthedensesuburban(LocationsAandB) anddenseurban (LocationsCandD)areasofLagostocollectReceivedSignalStrength(RSS)dataoverGSMnetworksoperatingat1800MHz.

MaterialsandMethod(Cont’d)

• Inordertoreducethepredictionerror,automaticmodelcalibrationwasperformedinATOLLradionetworkplanningtoolbasedonthegeospatialdata:

§Clutterheight;§Clutterclasses;§Altitude;and§Elevation oftheareaunderstudy.

7

MaterialsandMethod

(Cont’d)

• FieldmeasurementsdatacollectedatLocationsA andCwereusedformodelcalibrationwhileLocationsBandDservedasthemodelverificationsites.• TheperformanceoftheSPMandthecalibratedSPMwerecomparedusingthefollowingstatisticalperformancemetrics:§MeanAbsoluteError(MAE);§RootMeanSquareError(RMSE);§StandardDeviation(SD).

ResultsandDiscussion

Generally,theRSSdecreasedasthedistancebetweenthe

Basestationandthemobilestation

increased,asexpected.

TheRSSLvariesrandomly

between-43dBmand-100dBm.

Thecalibrationprocesssignificantlychangedthemodelparameters,asshowninTable1.• Infact,thecalibratedSPMaccountedforadditionalpathlossesof25dBintheurbanenvironments.

Results& Discussion (Cont’d)

Table1:CalibrationResultofSPMforUrbanEnvironments

ResultsandDiscussion(Cont’d)

11Figure1:PathLossPredictionsinLocationA(DenseSuburban)

12Figure2:PathLossPredictionsinLocationB(DenseSuburban)

ResultsandDiscussion(Cont’d)

13Figure3:PathLossPredictionsinLocationC(DenseUrban)

ResultsandDiscussion(Cont’d)

14Figure4:PathLossPredictionsinLocationD(DenseUrban)

ResultsandDiscussion(Cont’d)

15

Table2:PerformanceEvaluationoftheModelCalibrationforDenseSuburbanAreas.

Location A (Calibration Site)

Location B (Verification Site)

SPM Calibrated SPM SPM Calibrated SPM

MAE (dB) 18.28 4.49 18.60 5.56

RMSE (dB) 21.16 6.17 21.81 7.03

SD (dB) 10.65 4.24 11.40 4.31

ResultsandDiscussion(Cont’d)

16

Location C(Calibration Site)

Location D(Verification Site)

SPM Calibrated SPM SPM Calibrated SPM

MAE (dB) 18.15 5.49 18.26 6.07

RMSE (dB) 21.49 6.96 21.75 7.46

SD (dB) 11.51 4.29 11.83 4.33

ResultsandDiscussion(Cont’d)Table3:PerformanceEvaluationoftheModelCalibrationforDenseUrbanAreas.

Conclusion

• OurfindingsshowthatthecalibratedSPMprovidesasignificantlybetterfitnesswiththemeasureddatacollected.• TheaverageMAE,RMSEandSDEacrossalltheroutesare5.40,6.90and 4.29,respectively.• ThesevaluesaremuchlowerwhencomparedwithdefaultvaluesobtainedwithSPM(18.32dB,21.55dBand 11.34dB,respectively)predictionresultsinthestudyarea.

17

Conclusion• In detail, we found that the calibrated SPMaccounted for additional path losses of 25 dB in theurban environments.•Thus, a generic and optimized model equation wasprovided for path loss as given by equation (1):

18

𝑃𝐿 𝑑𝐵 = 51.4 + 20𝑙𝑜𝑔 𝑑 + 20𝑙𝑜𝑔 ℎ1 − 4.97𝑙𝑜𝑔 𝑑 x𝑙𝑜𝑔 ℎ1(1)

SelectedReferences

[1] J.D.Parsons,“TheMobileRadioPropagationChannel”,SecondEdition,JohnWileyandSonsLtd.,2000.

[2] A.J.Mishra,‘AdvancedCellularNetworksPlanningandOptimization2G/2.5G/3G…Evolutionto4G,JohnWileyandSonsLtdISBN13978-0-470-01471-4,2007,pp1-12

[3] S.I.PopoolaandO.F.Oseni.PerformanceEvaluationofRadioPropagationModelsonGSMNetworkinUrbanAreaofLagos,Nigeria.InternationalJournalofScientificandEngineeringResearch,vol.5(6),June,2014,pp.1212-1217.

[4] S.I.PopoolaandO.F.Oseni.EmpiricalPathLossModelsforGSMNetworkDeploymentinMakurdi,Nigeria.InternationalRefereedJournalofEngineeringandScience,vol.3(6),June,2014,pp.85-94.

[5] O.F.Oseni,S.I.Popoola,R.O.Abolade,O.A.Adegbola,“ComparativeAnalysisofReceivedSignalStrengthPredictionModelsforRadioNetworkPlanningofGSM900MHzinIlorin,Nigeria”,InternationalJournalofInnovativeTechnologyandExploringEngineering(IJITEE),volume4,issue3,pp.45-50.

19

SelectedReferences

[6] N.Faruk,A.A.AyeniandY.AAdediran,"ErrorBoundsof

EmpiricalPathLossModelsatVHF/UHFBandsinKwaraState,Nigeria"inProc.IEEEEUROCONconference,CroatiaJuly1st-4th,

2013,pp602– 607.

[7] N.Faruk,Y.A.AdediranandA.A.Ayeni,"Onthestudyofempirical

pathlossmodelsforaccuratepredictionofTVsignalforsecondaryusers"ProgressinElectromagneticResearch(PIER)B,USA,Vol.49,pp

155- 176,2013.

[8] M.Hata“EmpiricalFormulaforPropagationLossinLand–MobileRadioServices”,IEEETransactionsonVehicularTechnology,vol.29,No

3,1980,pp317– 325.

[9] COST231,“UrbanTransmissionLossModelsfor

MobileRadiointhe900&1800MHzband”,COST231TD(90)119Rev2,1991.TheHague,Netherlands.

[10] Lagos,availableonhttp://www.latlong.net/place/lagos-

nigeria-2286.html

[11] Forsk,ATOLL3.2.0RadioPlanning&OptimizationSoftwareUserManual,France.Available

online:www.forsk.com.

ThankYou

21

CovenantUniversity