Neighbor Discovery for Cogitive Radio Ad Hoc Networks Zaw Htike Departmet of Computer Egieerig, Kyug Hee Uiversity, 1 Seocheo,Giheug, Yogi, Gyeoggi 449-701 Korea +8-10-561-811 htike@etworkig.khu.ac.kr, Choog Seo Hog Departmet of Computer Egieerig, Kyug Hee Uiversity, 1 Seocheo,Giheug, Yogi, Gyeoggi 449-701 Korea +8-31-01-53 cshog@khu.ac.kr ABSRAC Neighbor discovery is a critical task i cogitive radio ad hoc etworks sice the secodary users operate o available chaels which are dyamically chagig accordig to the primary users activities. I this paper, we propose a eighbors discovery mechaism that provides how a secodary user could fid its eighbor i cogitive radio ad hoc etworks without ay collaboratio. I this mechaism, o prior kowledge of eighborig users is required. I the cogitive radio etworks, the available chael sets are varyig ad users choose eighbor discovery strategies accordig to the umber of available chaels they observed. I other words, eighbor discovery strategies are adaptively chagig i a dyamic eviromet. he strategies are simple ad totally distributed, but these provide miimum time to redezvous (R). Categories ad Subject Descriptors C. [Computer commuicatio etworks]: Wireless commuicatio, Geeral erms Algorithm Keywords Neighbor discovery, time to redezvous (R), ad cogitive radio ad hoc etworks 1. INRODUCION I recet years, may research works have bee ivolved i the developmet of cogitive radio (CR) techology sice it has bee recogized as a ew way to improve the spectral efficiecy of wireless etworks. I cogitive radio etwork, secodary users (SUs) are allowed to utilize free or idle portios of licesed chaels (or spectrum) without causig ay iterferece to primary users (PU) [19]. Geerally, SUs detect the idle chaels ad access the chaels [1]. he chael availability for SUs is varyig accordig to PU activities which chage dyamically i frequecy, space ad time [0]. herefore, the available chaels for each SU might also chage dyamically. However, if a pair of SUs wishes to commuicate with each other, they eed to redezvous o a chael that is commoly available to both of them ad exchage ecessary cotrol iformatio for egotiatio such as request-to-sed (RS)/clear-to-sed (CS) hadshakig of the 80.11 Distributed Coordiatio Fuctio (DCF) [7]. his task is ot trivial i CR etworks sice SUs may operate o differet chaels idepedetly. his is geerally called redezvous or eighbor discovery problem [14]. his research was supported by Next-Geeratio Iformatio Computig Developmet Program through the Natioal Research Foudatio of Korea (NRF) fuded by the Miistry of Educatio, Sciece ad echology (01-000641). Dr. CS Hog is the correspodig author. Permissio to make digital or hard copies of all or part of this work for persoal or classroom use is grated without fee provided that copies are ot made or distributed for profit or commercial advatage ad that copies bear this otice ad the full citatio o the first page. o copy otherwise, or republish, to post o servers or to redistribute to lists, requires prior specific permissio ad/or a fee. ICUIMC(IMCOM) 13, Jauary 17 19, 013, Kota Kiabalu, Malaysia. Copyright 013 ACM 978-1-4503-1958-4 $15.00. (a) Figure 1. (a) Users have differet available chael sets ad (b) users are dwellig o differet chaels idepedetly. 1.1. Problem Statemet he eighbor discovery is more challegig i cogitive radio ad hoc etworks as there is o cetralized cotroller [17]. Figure 1 represets the eighbor discovery problem of cogitive radio ad hoc etworks. As show i Figure 1, there are four chaels for (b)
SUs ad, user x ad y have differet available chael sets as the user y is i the trasmissio rage of a PU, ad users are curretly dwellig o differet chaels. Whe a user, user x, wats to commuicate with its eighbor, user y, these two users eed to redezvous o a chael which is commoly available to them. I this case, the available chaels are sice is occupied by a PU ad, thus, ot commoly available for both SUs. At the very iitial state of the etworks, SUs have o iformatio regardig their eighbors such as the chael that the eighbor curretly dwellig o. However, by performig chael sesig [16], SUs ca achieve local iformatio such as umber of available chaels, PU activities of each chael withi its regio, etc. If a SU eeds to iitiate ay commuicatio, it has to fid its eighbor by usig its ow local iformatio. Moreover, the SU eeds to redezvous with iteded eighbor withi a reasoable time iterval which is, geerally, called time to redezvous (R). 1. Related Works here are two famous approaches to eable eighbor discovery i cogitive radio ad hoc etworks; 1. Usig dedicated commo cotrol chael (CCC) ad. Usig chael hoppig Usig dedicated commo cotrol chael is the simplest way to eable eighbor discovery i CR etworks. Most of the proposed medium access cotrol (MAC) protocols for cogitive radio etworks were desiged by assumig the existece of a CCC ad further assumig that it is available for every secodary user [11]. I fact, this approach origiated from the cocept of MAC protocols for multi-chael wireless etworks [15] [18]. I this approach, the CCC serves as a redezvous chael ad all the ecessary cotrol iformatio is exchaged amog SUs via the CCC. I the CCC approaches, time is divided ito two itervals; egotiatio or cotrol iterval ad data iterval. Whe a SU wats to iitiate commuicatio, it first switches to the CCC durig the egotiatio iterval ad attempts to egotiate with the iteded receiver or eighbor. After egotiatig o the CCC, data commuicatio ca be accomplished durig the data iterval via other available chaels, kow as data chaels [10]. Figure.1 illustrates the ormal operatio of a etwork with a commo cotrol chael. As show i Figure.1, all users attempt to egotiate o the CCC durig the cotrol iterval. After the egotiatio is complete o the CCC, the users move to selected chaels ad perform data commuicatios simultaeously durig the data iterval. Obviously, usig a CCC ca simplify the eighbor discovery process [13]; yet, it is ofte ot feasible or impractical due to lack of CCC availability. he mai drawback usig the CCC approach is it is susceptible to primary user activities. Whe PUs appear o the CCC, all SUs must defer their trasmissios o the CCC ad vacate the chael immediately. Not oly does PUs presece degrade the overall throughput of a CR etwork, but if the trasmissio period of PUs is sigificatly log o the CCC, the presece of the PUs may also block chael access for SUs. Moreover, the available chael sets i CR etworks, icludig the CCC, chage dyamically, hiderig the establishmet of a ever-available cotrol chael for all SUs. hus, i the dyamic eviromet, a ever-available chael for all SUs is ulikely to exist. Figure. Operatio of a protocol with CCC Aother popular solutio to fid the eighbors i CR ad hoc etworks is usig chael hoppig sequeces [1] []. he mai advatage of the chael hoppig sequeces, compare to the CCC approach, is that SUs ca redezvous with the eighbors at ay available chael. herefore, it ca overcome log-term blockig of PU ad a ever-available commo chael is ot required. I this approach, SUs geerate their ow chael hoppig sequeces. Whe a SU (e.g., user x) eeds to commuicate with its eighbor (user y), it switches from oe chael to aother by followig a predefied hoppig sequece util it fids its eighbor. Figure illustrates the operatio of a chael hoppig protocol. As show i figure, user x ad y attempt to fid each other by followig their ow hoppig sequeces. Whe they redezvous o chael, they ca perform commuicatio. Figure. Operatio of a sequece based protocol However, chael hoppig protocols have the followig shortcomigs. a) Lack of etwork status iformatio I sequece-based approaches, SUs geerate chael hoppig sequeces with a assumptio that all SUs use the same chael labels. With this assumptio, sequece-based approaches provide upper boud of time to redezvous. However, this is a big assumptio sice SUs have o iformatio regardig the eighbors at the iitial state of the etwork. b) Require Sychroizatio Most of the sequece-based approaches eed sychroizatio amog users [3][4] which is hard to achieve i ad hoc etworks. For example, i [3], time is divided ito fixedtime itervals i which each represets oe of the available
chaels. At the begiig of the time slot, every ode i the etwork must switch to the correspodig chael for egotiatio. c) Complexity he ext difficulty i sequece-based protocols is overcomig the complexity of geeratig chael hoppig sequeces. Desigig the chael hoppig algorithm is a great challege because whe users geerate chael hoppig sequeces, ay pair of these sequeces should overlap at least oce withi a sequece period, so that ay pair of users which eeds to commuicate ca redezvous [6]. Moreover, the R values betwee ay pair of sequeces should be reasoable ad, obviously, it is determied by chael hoppig algorithm. I [3], chael hoppig sequeces were created i a roud robi fashio ad, as metioed above, this protocol requires tight sychroizatio amog SUs which is difficult to achieve i ad hoc eviromet. I [8], the authors proposed biased pseudo-radom sequeces. hese sequeces do ot eed tight sychroizatio, but the average R may ot be bouded. he authors of [] proposed permutatio-based chael hoppig sequeces. I their proposal, the expected time to redezvous was bouded by a quadratic fuctio of the umber of available chaels. A quorum-based scheme was proposed i [9], ad the authors claimed that redezvous betwee ay pair of users ca occur at least oce withi time slots, where is the umber of available chaels. I this paper, we propose a alterative way of eablig eighbor redezvous i cogitive radio ad hoc etworks. Our proposed mechaism uses either dedicated commo cotrol chael or chael hoppig. So, the proposed mechaism does ot require geeratig the predefied hoppig sequeces. I our proposed eighbor discovery mechaism, (1) SUs eed oly local iformatio, such as umber of available chaels, to perform eighbors discovery. herefore, it ca be applied i a distributed maer. () he algorithms are simple ad (3) these ca provide less expected time to redezvous tha sequece-based approaches. Moreover, this mechaism is immue from some irratioal assumptios such as (1) all chaels are idexed with the same labels by secodary users ad () tight sychroizatio amog users. We preset our proposed mechaism i sectio.. I sectio, we describe the eighbor discovery strategies ad discuss how to update the strategies accordig to the umber of available chaels that SUs observed. he, the comparisos of the umerical results are described i sectio 3. Sectio 4 cocludes the paper.. NEIGHBORS DISCOVERY MECHANISM We assume that eighbor discovery process ca be accomplished betwee two SUs if they redezvous o the same chael ad exchage ecessary cotrol iformatio (i.e. eighbors discovery message (DOV) ad ackowledgmet (ACK)). DOV is just a probe message ad ay SU, which receives DOV, ca simply reply the ACK. he goal of every SU is to fid its eighbors with miimum delay or R. We also assume that the PU activities o the chaels are dyamically chagig. hus, SUs updates the eighbor discovery strategies every roud accordig to the umber of available chaels they observed..1 Neighbor Discovery Strategies Whe a SU eeds to perform eighbor discovery, it performs chael sesig first ad create a available chael set. All free chaels (the chaels that are ot curretly used by PUs) will be icluded i the available chael set. he SU chooses oe of the followig two strategies; Strategy oe: Switch oe available chael after aother without repeatig ad fid a eighbor. Strategy two: Radomly select oe of the available chaels ad wait for a eighbor o selected chael. SUs perform eighbor discovery util they redezvous with their eighbors. First, the SU (user x) that chooses strategy oe selects a available chael radomly. he it switches to the selected chael ad seses for the presece of PU. If the SU seses the chael is free, it will broadcast DOV ad waits for the ACK. All message trasmissios follow the priciple of Distributed Coordiatio Fuctio (DCF) of IEEE 80.11 [6] ad the basic procedure of packets trasmissio ca be see i Figure.3. Figure 3. Procedure of packets trasmissio he time iterval of the whole process is defied as oe time slot, slot, ad it ca be estimated as, ACK DOV DIFS SIFS B slot, (1) rate where, rate is trasmissio rate ad B represets the radom back-off. While the SU is waitig the ACK, it may receive DOV from other eighbors. If it receives DOV istead of ACK, it will simply reply ACK ad eighbor discovery has completed for these two SUs. If user x does ot receive ay ACK or DOV, it shall switch to aother available chael ad broadcast the DOV agai. his process is repeated ad oe roud of eighbor discovery for user x is over whe it has received the ACK from oe of its eighbors or after it has visited all available chaels. he ecessary time iterval for oe roud of eighbor discovery ca be expressed as roud., () slot where, is the total umber of available chaels. he SU that chooses strategy two, user y, just selects a radom chael from the available chael list ad waits its eighbor for oe roud. If it receives DOV from its eighbor, let say from user x, it will reply ACK ad the eighbor discovery has bee successfully accomplished betwee these two SUs, x ad y. Oe roud of user y is over after it has received DOV ad replied ACK or oe roud has expired. If user y does ot redezvous with ay of its eighbor withi a roud, it will perform chael sesig agai ad update its strategies for ext roud. Choosig
strategies for ext roud is idepedet of previous roud but o the available chaels it seses (the reaso of why decisio makig depeds o the available chaels ca be foud i sectio.3).. Expected Payoffs he expected payoffs for differet evets are represeted with R. Accordig to eighbor discovery strategies, ay pair of SUs ca be i oe of the followig three evets. Evet A: Both SUs choose the same strategy, strategy oe. Evet B: SUs choose differet strategies. Evet C: Both SUs choose the same strategy, strategy two. If evet A occurs, both SUs try to fid each other by switchig from oe chael to aother without revisitig the chaels. I this evet, the probability of meetig these two users o is he, the probability of ot meetig at all withi a roud becomes he probability of at least oe redezvous occurs withi a roud is (3) (4). (5) he, the expected time slots for this evet ca be estimated as: 1 E [ ] (1 (1 ) A R ). (6) If evet B occurs, SU that chooses strategy oe (user x) switches from oe chael to aother without repeatig ad tries to fid its eighbors. Neighbor SU (user y) that chooses strategy two selects a chael radomly ad waits its eighbor for oe roud. he probability of meetig these two SUs o is 1. If user x does ot redezvous with its eighbor, it will switch to aother chael while user y is waitig o the selected chael. herefore, he probability of meetig o ext chael ( ) is. Similarly, the probability of meetig o this evet is becomes 3. he, the expected time to redezvous for i 1... 1. (7) i 1 EB[ R] If evet C occurs, obviously these two SUs will ot meet each other withi a roud. Oe roud is wasted i this situatio ad the expected time slots for this evet ca be expressed as E [ R]. (8) C roud Suppose p ad q (= 1 - p) are probabilities of choosig strategy oe ad two respectively, the we ca describe the overall expected time slots for oe roud with the followig expressio. 1 E[ R] p(1 p)( 1) p (1 ( ) ) (1 p). (9).3 Optimal Strategies Obviously the aim of SUs is to miimize the expected payoff. We take derivative to (6) ad we get optimal value of p as 1 p. (10) 1 (1 ( ) ) It is clear that the optimal strategy profiles (p, q) are the fuctio of umber of available chaels,. herefore (as metioed i previous sectio), each SU updates its strategies every roud accordig to the umber of available chaels it seses. 3. NUMERCIAL RESULS It is obvious that SUs do ot wat evet C. he authors of [7] claimed that optimal values of p ad q are 0.75 ad 0.5 respectively as. I proposed eighbor discovery mechaism, SUs are more aggressive to choose strategy oe, as, p 0.8. Whe we compare the results, proposed mechaism provides less occurreces of evet C ad it ca achieve miimum E[R] as show i able 1. able 1. he E[R] agaist differet optimal values. Optimal q E[R] Proposed Proposed Ref [7] ND Ref [7] ND 3 4 5 6 7 8 9 10 50 0.5000 0.3333 0.30 0.301 0.914 0.84 0.791 0.753 0.7 0.51 0.475 0.0000 0.1000 0.1351 0.1531 0.1641 0.1714 0.1767 0.1807 0.1838 0.043 q 0.000.0000.6667 3.5683 4.3793 5.133 6.041 6.8716 4.3793 8.5300 8.5300 1.5000.1000.707 3.306 3.9101 4.514 5.1184 5.77 6.370 30.5060 We also compare the results of proposed eighbor discovery (ND) mechaism ad that of radom algorithm (RA) ad orthogoal sequece-based algorithm (OSA) from [14]. Radom Algorithm is similar to evet A. Nodes (SUs) try to fid each other by switchig from oe chael to aother util they
redezvous oe of their eighbors o a commo chael. he expected time to redezvous of the radom algorithm is, (11) where, is the umber of available chaels ad is the probability of a successful hadshake. OSA algorithm is a alterative of radom algorithm. he differece is, i OSA, odes switch chael by followig a predefied hoppig sequece like sequece-based approaches ad the expected R value OSA algorithm is E OSA 4 6 3 [ R]. (1) 3 ( 1) Figure.4 shows compariso of the results i terms of E[R]. As show i the figure, the proposed eighbor discovery mechaism provides less R tha that of OSA ad radom algorithms. 4. CONCLUSION We have preseted a eighbors discovery mechaism for cogitive radio ad hoc etworks. I this mechaism, the eighbor discovery strategies are simple ad totally distributed. It is flexible with dyamic ature of CR etworks as SUs update their strategies accordig to the umber of available chaels. Moreover, the umerical results cofirm that it provides less expected time to redezvous tha previous works. Figure 4. E [R] of OSA, RA ad proposed eighbor discovery mechaism REFERENCES [1] J. Shi, D. Yag, ad C. Kim, A chael redezvous scheme for cogitive radio etworks, IEEE COMMUNICAIONS LEERS, vol. 14, o. 10, pp. 954 956, November 010. [] L. DaSilva ad I. Guerreiro, Sequece-based redezvous for dyamic spectrum access. IEEE DySpa, 008, pp. 1 7. [3] Y. Kodareddy ad P. Agrawal, Sychroized MAC protocol for multihop cogitive radio etworks, IEEE ICC, 008. [4] C.-F. Shih,. Y. Wu, ad W. Liao, DH-MAC: A dyamic chael hoppig MAC protocol for cogitive radio etworks. IEEE ICC, 010, pp. 1 5. [5] K. Bia, J.-M. Park, ad R. Che, Cotrol chael establishmet i cogitive radio etworks usig chael hoppig, IEEE Joural o Selected Areas i Commuicatios, vol. 9, o. 4, pp. 689 703, 011. [6] G. Biachi, Performace aalysis of the IEEE 80.11 distributed coordiatio fuctio, IEEE Joural o Selected Areas i Commuicatios, vol. 18, o. 3, pp. 535 547, March 000. [7] E. Aderso ad R. Weber, he redezvous problem o discrete locatios, Joural of Applied Probability, vol. 7, o. 4, pp. 839 851, 1990. [8] Cormio ad K. Chowdhury, Commo cotrol chael desig for cogitive radio wireless ad hoc etworks usig adaptive frequecy hoppig, Ad Hoc Networks, vol. 8, o. 4, pp. 430 438, Jue 010. [9] K. Bia, J.-M. Park, ad R. Che, A quorum-based framework for establishig cotrol chaels i dyamic spectrum access etworks,. MobiCom 009, 009, pp. 5 36. [10] X. Zhag ad H. Su, CREAM-MAC: cogitive radioeabled multichael MAC protocol over dyamic spectrum access etworks, IEEE Joural of Selected opics i Sigal Processig, vol. 5, o. 1, pp. 110 13, 011. [11] C. Cordeiro ad K. Challapali, C-MAC: A cogitive MAC protocol for multi-chael wireless etworks. IEEE DySpa, 008, pp. 147 157. [1] M. immers, S. Polli, A. Dejoghe, L. Va der Perre, ad F. Catthoor, A distributed multichael mac protocol for multihop cogitive radio etworks, IEEE rasactios o Vehicular echology,, vol. 59, o. 1, pp. 446 459, 010. [13] B. F. Lo, A survey of commo cotrol chael desig i cogitive radio etworks, Physical Commuicatio, vol. 4, o. 1, pp. 6 39, March 011. [14] N. C. heis, R. W. homas, ad L. A. DaSilva, Redezvous for cogitive radios, IEEE RANSACIONS ON MOBILE COMPUING, vol. 10, o., pp. 16 7, February 011. [15] Duc Ngoc Mih Dag, Mui Va Nguye, Choog Seo Hog, Sugwo Lee ad Kwagsue Chug, "A Eergy Efficiet Multi-chael MAC Protocol for Wireless Ad hoc Networks", IEEE Globecom 01, Dec 3-7, 01, Aaheim, Califoria, USA [16]. Yucek ad H. Arsla, A survey of spectrum sesig algorithms for cogitive radio applicatios, IEEE Commuicatios Surveys & utorials, vol. 11, o. 1, pp. 116 130, First Quarter 009. [17] I. F. Akyildiz, W.-Y. Lee, ad K. R. Chowdhury, CRAHNs: Cogitive radio ad hoc etworks, Ad Hoc Networks, vol. 7, o. 5, pp. 810 836, July 009. [18] Jugmi So, Niti H. Vaidya. Multi-Chael MAC for Ad hoc Networks: Hadlig Multi-chael Hidde ermials Usig A Sigle rasceiver, Proceedigs of the 5th ACM iteratioal symposium o Mobile ad hoc etworkig ad computig (Mobihoc 04), Pages -33, 004.
[19] I.F. Akyildiz, W. Y. Lee, M. C. Vura, ad S. Mohaty, NeXt geeratio / dyamic spectrum access / cogitive radio wireless etworks: A survey, Computer Networks Joural (Elsevier), Volume 50, Issue 13, 15 September 006 Pages 17 159 [0] I. Akyildiz, W. Lee, M. Vura, ad S. Mohaty, A survey o spectrum maagemet i cogitive radio etworks, IEEE Commuicatios Magazie, April 008, pp. 40-48. [1] A. De Domeico, E. C. Striati, ad M.-G. Di Beedetto, A Survey o MAC Strategies for Cogitive Radio Networks IEEE Commuicatios Surveys & utorials, First Quarter 01, Page(s): 1-44