Algorithms for Dynamic Frequency Selection for Femto-cells of Different Operators

1. “Algorithms for Dynamic Frequency Selection for Femto-cells of Different Operators” R. Pomposini, M. Petracca, F. Mazzenga, R. Giuliano, F. Vatalaro University of Rome Tor Vergata

2. Outline Introduction 3GPP Femtocells Interferenceproblems in femtocellnetworks DynamicFrequencySelection (DFS) algorithms Scenario description Simulationenvironment Performance results Conclusions Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 2

3. 3GPP Femtocells Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 3 Small cellular base stations (BS) designed for use in a home or small business environment Work on licensed band(s) Typically support 2 to 4 active mobile phones Allow service providers to extend service coverage indoors Extend BS to allow a simpler, self-contained deployment Provides improvements to both coverage and capacity

4. The self-installation nature offemtocellssharing the samefrequency band can leadtoharmfulinterferencelevels. Macro-to-femto (cross-tier) interference can beavoidedbyallocatingseparatedchannels …butlowerspectrumefficiencyforoperators. Femto-to-femto (co-tier) interference …evenif a dedicatedfrequency band isallocatedto the femtocell network. harmful in highly dense HNB’s environments! can compromise the opportunityofenjoying high data rate services. Interference problems in femtocell networks Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 4

5. DynamicFrequencySelectionAlgorithms WeproposedsimpleDFS algorithmsaimedtoredistribute the availablespectrumresourcesbelongingtodifferent network operatorsamongfemtocells. Algorithmsbased on: localinterferencemeasurements (Cognitive Radio). arrangementsamongoperators (mutualfrequencyexchange) In a multi-operator scenario, thisapproachoffersadvantagestoalloperators in termsofachievable network capacity and QoSprovidedtocustomers. Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 5

6. Twodifferentapproachesforfemtocells GreedyDFS algorithm(GDFS): femtocells are absolutelygreedy and aimtoimmediatelymaximizetheirSignal-to-InterferenceRatio (SIR) carelessofwhichoperatoroffers the lessinterferedchannel Operator-orientedDFS algorithm (ODFS): eachfemtocellusesitsown band until the QoSconditionisverified (SIR≥ρ0) DFS algorithmsdescription Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 6

7. GDFS algorithmdescription Eachfemtocellselects the channelthatmaximizesits SIR regardlessofwhichoperatorhaslicenseforthatfrequency band. Each node performs these tasks: Interference measurement on f1,f2,…,fN Select the i-thfrequence | SIRi = max(SIR1,…,SIR2)) if SIRi> threshold Select the operator that correspond to fi Else Sensing mode Sensing mode: programmed standby featureenvisaged in ordernottodamage the otherfemtocells. Since the interferenceconditionsprevent the femtocellfromtransmitting, ittemporarilydisablestransmission and just performsspectrumsensing. Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 7

8. ODFS algorithmdescription The operatorassign a dedicatedchannelforfemtocells Each node performs these tasks: Interference measurement on f1,f2,…,fN If SIRop> threshold Select the i-th frequency (belonging to own operator) Else if SIRext > threshold Select the j-th frequency (belonging to a different operator) Else Sensing mode Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 8 Eachfemtocelltakesinto account itssubscriptionto the specificoperator and attempttouseitsown band until the SIR isabove the requiredthreshold

9. DFS algorithmswithPowerControl Eachfemtocellperformthesetasks: If SIR > threshold Decrease Ptxin order to reach the target quality level Else IncreasePtx up toPmax IfPtx>Pmax DFS algorithms If Ptx required to reach the threshold is greater than Pmax Sensing mode Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 9 The femtocell, given the SIR on the frequency band on which it is operating, adjust its transmission power in order to reach the target quality level

10. Reference Scenario 3 network topologies Random Regular grid Perturbedgrid Pmax = 20 dBm DFBS-Terminal = 10 m Outdoor pathlossexponent = 3.2 Indoor pathlossexponent = 2.8 Downlinktransmission direction isconsidered Greedy data source SINR threshold : 9.4 dB F = 1800 MHz Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 10

11. Network topologies Random Femtocells are randomdistributed in a area of 100 x 100 m2 Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 11 Regular grid Baselineforcomparing the behavior and performance of DFS algorithms Perturbedgrid Femtocells are positionedover a regular gridwithrandom 2D displacementaroundtheiroriginalpoint

12. Interference Pathloss indoor Pathloss outdoor SIR in the case of regular gridtopologywithoptimalfrequencyassignment Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 12

13. Results (1) Regular gridwithoptimalfrequencyassignment Break ofequilibrium Femtocellssimultaneouslyfallbelow the threshold SIR distributioncouldbeapproximatedas a delta function Femtocells at the borders spread the SIR distribution Regular gridrandom network Largervariance Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 13 Empiricalprobability density function (PDF) of SIR

18. Regular Grid (dashed line)

20. Thanks for the attention! Remo Pomposini - pomposini@ing.uniroma2.it PIMRC – September 27th 2010 19