A Cognitive Radio Approach for Usage of Virtual Unlicensed Spectrum Danijela abri Shridhar - PDF document

Figure 1. Spectrum utilization measurement at BWRC ation measurement at BWRC Frequency [Hz] Pool,whichcouldcover from tens of MHz to several GHz creating a virtual unlicensed bandŽ. It is not necessarily a contiguous frequency range and Spectrum Pools of differentSU Groups may overlap which impliesthatSUGroupswillcompete for the available resources. EachSpectrumPoolwillbe further divided into N Sub-Channels, which will bethebasic resolution used for sensing and transmission.Figure2showsthe principle idea of a Spectrum Poolingsystem in CORVUS. Primary Users own differentpartsofthespectrum but may not be active at a particular time. The shaded frequency bands indicate that the PU iscurrentlyusingits spectrum and consequentlythis frequency band cannot be used by any SU. Thefigurealso shows three different activeSecondary User communications. For each communication a pair of SUs picked a patternofsub-channelstoforma(SU Link). The number of sub-channelsin a SU link may vary dependingonthequalityofthesub-channels, the bandwidth of a singlesub-channelandQoSrequirement for that connection.Sub-channels selected to create a SU Link should be scatteredovermultiple PU frequency. This principle has a doublesignificance.Onone hand it limits the interferenceimpact of a SU on a re-appearance of a PU, while on theotherhandif a PU appears during the lifetime of a SU Link it wouldimpact very few (preferable one) of theSub-Channelsusedbythe SU Link. The communication peers usingthatlinkwouldhave to immediately clear the affected sub-channel and find a new free sub-channel. In order to maintain QoS, SUsshouldalways have a redundant number of sub-channelsfortheirSUWithinCORVUS, SUs use dedicated logical channels for the exchange of control and sensing information. We envisiontwo different kinds of logical control channels,aControl Channel and Group Control ChannelsControl Channel is globally unique and has to be knowntoeverySUoperating in the relevant frequency bands, sinceaccess to that channel is pre-requisite for initiating communications.Themain purpose of the Universal ControlChannelis to announce existing groups and to give therelevanttransmissionparameters to enable newly arrivingusers to join a group. Additionally SUs, which want to create a new group can request the local primaryuserfootprintsonthatchannel.Although globally unique the communicationrangeshouldbelocally limited as SU Groups are limited to a local area. In addition to the Universal Control Channel each group has one logical Group Control Channel for theexchange of group control and sensing information.ited load of low-bitratea)dedicated spectrum for this purposeb)an unlicensed band such as the ISM/UNII bands c)unlicensed UWB (Ultra Wide Band)We believe the UWB option is especiallyattractiveifweareconsidering use of the 3-10GH band. UWB controlchannelswouldbeunlicensed but with low impact on othertypes of communication and withthepossibilitytooperateindependently using different spreading codes. Thereareseverelimitationsonthepower of UWB emissions limiting its range, but the controlchannelrequires very low data rates, so spreading gain will increase the range to be adequateformostapplications (more than 10,000 times lowerdataratethanthecommercialUWBsystems being envisaged in this band). NotethattheUniversal Control Channel and the Group ControlChannels are logical concepts, which might even bemappedOur system design only covers theISO/OSIlayersone(physical layer) and two (link layer). Higherlayerswillimplement standard protocols not specific to cognitive radiosandthusare not relevant to our discussion. Figure 3 shows themain building blocks for the deployment of a Cognitive Radiosystem.Weidentify six systems functions and two controlchannels that will implement the core functionality.A.Physical Layer FunctionsThe main function of the physicallayeristo sense the spectrum over all available degrees of freedom (time, frequency and space) in order toidentifysub-channelscurrently available for transmission. From thisinformation, SU Links can be formed from a composition of multiple sub-channels. This will require the ability toprocessawide bandwidth of spectrum and then perform a widebandspectral,spatialand temporal analysis. Sub-Channelscurrently used for transmission by SUs have tobesurveyedatregularintervals … at least every … to detect Primary Users activity on those Sub-Channels (reclaiming the usage of theirSub-ChannelsŽ) and if there is activitythenthoseSub-Channels must be given up.It will be necessary for the SUs to exchange and mergetheir local sensing information in order to optimallydetecthidden terminal problemcooperationbetween SUs within a communicating group will be important to realizeadequate accuracy of interference activity. Spectrum sensing is best addressedasacross-layerdesignproblemsince sensitivity can be improved byenhancing radio RF front-end sensitivity, exploitingdigitalsignalprocessing gain for specific primary user signal, and network cooperation where users share their spectrum sensingmeasurements [9]. Channel Estimation:In order to set up thelink,channel sounding is used to estimatethequalityofsub-channelsbetween SUs that want to communicate. The transmission parameters (transmit power, bit rate,coding,etc.)are determined based on the channel sounding results.Afterthe setup, the physical layer continuously estimates the qualityof sub-channels analyzing thedata packets received during ongoing communication.CRsoptimally uses the availablespectrum as determined by the spectrum sensingandchannelestimation functions.Therefore it should have the ability to operateat variable symbol rates, modulation formats (e.g. lowtohigh order QAM), differentchannel coding schemes, power levels and be able to use multiple antennas forinterferencenulling,capacity increase (MIMO) or range extension (beamforming). One possible strategy would be based on an OFDM-likemodulationacrosstheentire bandwidth in order to mosteasily resolve the frequency dimension with subsequentspatial and temporal processing.B.Link Layer FunctionsWe assume that any secondarystation will belong to a SU Group. A newly arriving usercaneither join one of the existing groups or createanewone covers the setup of a link inorderto enable the communication between two SUs and afterwardsthemaintenance of this SU Link for the duration of thecommunication. The link layer will initially choose a set ofSub-Channels in order to create a complete SU linksubjecttoMedium Access Control (MAC):As long as itcanbeassured that all Sub-Channels are used exclusively, i.e. all Sub-Channels used by one SU Link cannotbeusedbyanyother SU Link this problem comes down toasimpletoken-passingalgorithm ensuring that only one of the twocommunication peers is using the link. However, when consideringamulti-group,multi-user system, which may not becentrally organized, making the assumption of exclusivelyusedSub-Channelsisnotrealistic. So the MAC has to providemeansto concurrently access a SU Link by SUs or even to manage the concurrent access of individual Sub-Channels byIn this paper we presenttheCORVUS system concepts to harness unoccupied frequency bands for the creation of virtualunlicensed spectrum. The motivation for this approach comesfrom the enormous success of unlicensedbandsandtherealizationthat the present strategy of allocation has resultedin much under-utilized spectra.le of sensing their spectralenvironment and locating free spectrumresources.InCORVUS, these radios perform local spectrum sensing butPrimaryUserdetectionand channel allocation is performed inacoordinatedmanner. This collaborative (either centralized ordistributed) effort greatly increases thesystemsabilityinidentifying and avoiding Primary Users. In the CORVUS architecture, a group of Cognitive Radios forms a Secondary User Grouptocoordinatetheircommunication. Each member of this group senses theSpectrum Pool, which is divided into sub-channels. A pairofSecondary Users picks a set of sub-channels spread over multiple Primary User frequency bands toformaSecondaryUser Link. Sub-channels are picked basedonestimatedchannelgain of a sub-channel and the users QoS requirements. Furthermore, chosen sub-channels are scatteredoverthe frequency bands of multiple Primary Users to reducedisruption when a Primary User reappears. For group management a number of underlay control channels exists.AUniversal Control Channel is used by allgroupsforcoordination and separateGroupControl Channels are used bymembersof a group to exchange sensing information and [1] http://www.fcc.gov/oet/info/database/spectrum/[2]FCC. Et docket no. 02-155. Order, May 2002.[3]FCC. Et docket no. 03-322. Notice of Proposed Rule Making and Order,December 2003. [4]JosephMitola III. Cognitive radio for flexible mobile multimediacommunications. InSixth International Workshop on Mobile Multimedia Communications (MoMuC99)[5]Joseph Mitola III. Cognitive Radio An Integrated Agent ArchitectureforSoftware Defined Radio. PhD thesis, KTH Royal Institute of Technology, Stockholm, Sweden, 2000.[6]Timo Weiss and Friedrich Jondral. Spectrumpooling:Aninnovativestrategy for the enhancement of spectrum efficiency., 42:S8…S14, March 2004. [7]Fatih Capar, Ihan Martoyo, TimoWeiss,andFriedrichJondral.Comparison of bandwidth utilization for controlled and uncontrolledchannel assignment in a spectrumpoolingsystem.InIEEE 55th Vehicular Technology Conference VTC Spring20021069…1073, Birmingham (AL), 2002.[8]AdaS.Y. Poon, Robert W. Brodersen and David N. C. Tse. Degrees of Freedom in Spatial Channels:ASignal Space Approach. submitted to [9]D.Cabric,S. M. Mishra, R.B. Brodersen, Implementation Issues inSpectrum Sensing for Cognitive RadiosŽ, Annual Asilomar, November 2004