A New Throughput Analysis in Cognitive Radio Networks Using Slotted CSMA

Transcription

1 A Ne Throughut Analyi in Cognitive Radio Netork Uing Slotted CSMA Afono Joé de Faria and Joé Marco Câmara Brito Intituto Nacional de Telecomunicaçõe - INATEL Santa Rita do Saucaí, Brazil faria.afono@mtel.inatel.br, brito@inatel.br Abtract Cognitive Radio Netork (CRN) are a ne area of interet for reearcher and a ne technology for the next generation irele. Multile acce rotocol i an imortant iue to define the erformance. In thi aer, the throughut of a rimary and econdary i analyzed conidering the Cature Effect in both ytem and the Packet Error Rate (PER) due to the interference beteen rimary and econdary tation, conidering that the lotted Aloha rotocol i rooed for the licened and Slotted Carrier Sene Multile Acce (CSMA) rotocol i ued in the econdary. Keyord- Cognitive Radio; Multile Acce; Throughut; Performance analyi I. INTRODUCTION The radio frequency ectrum i a natural reource [2] and it i artitioned into everal band that are generally attributed to licened holder through long-term agreement [3][4]. Inide frequency ectrum there are ome unlicened ortion reerved for indutrial, cientific or medical (ISM) uroe and they are commonly ued for data communication in maller [3]. The tudie and meaurement erformed by the Sectrum Policy Tak Force (SPTF), linked to the Federal Communication Comion (FCC), concluded that certain ectrum band are heavily ued by licened or unlicened uer (ISM uer), hile other ectrum fraction are ued occaionally or rarely, deending on geograhic location and time [2][5]. Alo according to [6], in the future there could be carcity of thi reciou reource due to increaing demand oered by a variety of factor, like the raid econoc groth of the telecommunication ector and the convenience offered by them, the emerging ervice and alication, the increaing of the human mobility and the aearance of ne technologie [6]. Change in the olicy for the frequency ectrum that become more flexible the acce to thi reource, by uing dynac ectrum acce (DSA), and the imrovement in the ectrum management rocedure ould imrove the efficiency in the uing of thi natural reource and ould avoid it oible carcity [2] [3] [4] [6]. According to [5], the DSA become viable by the emerging technology of the cognitive radio (CR). It i a ne concet in the develoment of irele communication ytem enabling more efficient ue of radio ectrum and, therefore, it i a trong candidate a a technological olution for the future irele, o-called NeXt Generation (xg) or cognitive radio. In [5], CR i defined a a radio that can change it tranion arameter baed on the interaction ith the oerating environment and it main goal are to rovide reliable and eamle communication and to enable the efficient ectrum utilization. The cognitive radio mut alo be able to reconfigure their communication arameter raidly and in the real time [2]. The cognitive radio can be defined a that can dynacally alter their functionality and/or toology in accordance ith the changing need of it uer, taking into account current environmental condition. Thi dynac modification i done in accordance ith alicable buine rule and regulatory olicie [7]. The CRN architecture i formed by to grou: the rimary and the econdary. Thee grou can coexit in the ame geograhic region and they can oerate in the ame ectrum band. According to [4], the rimary i an exiting here rimary uer (PU) have licene to oerate in a ecific frequency band. Licened uer have higher riority in channel acce. The econdary, or cognitive are thoe that do not have licene to oerate in the deired band [4] and the ocalled econdary uer (SU) oerate in uch. According to [8], the econdary uer have loer tranion riority and they exloit the frequency ectrum in an oortunitic fahion, through the ectrum hole and ithout cauing harmful interference to licened uer tranion. According to [], in the rimary, the rotocol for medium acce control (MAC) are imortant to organize acce of the different rimary uer to the channel. In the econdary, the MAC rotocol are reonible for organizing acce of the econdary uer to the free rimary channel, avoiding or making the interference accetable in the rimary. The throughut i affected by the cature effect. In [], the erformance of cognitive radio (CRN) i analyzed for everal MAC rotocol, including an analyi that conider uing Slotted Aloha in the rimary and the lotted carrier ene multile acce rotocol (CSMA) in the econdary. In thee analyze the Cature Effect i alo taken into account in rimary and econdary : if the difference beteen the oer level of a concerned acket ignal in relation to other interfering acket i higher than a threhold called cature ratio (R), then the concerned acket can be detected by the receiver, herea all other fail in medium acce [9]. Hoever, the analyi introduced in [] doe not conider the oible error due to interference during the acket detection. In thi aer, e extend the analyi of [], taking Coyright (c) IARIA, 23. ISBN:

2 into account the acket error rate due to multile acce cheme and their effect on the throughut. The remainder of thi aer i organized a follo. In Section II, e reent the original model ued in [] for erformance analyi; a ne ytem model i introduced in Section III and the throughut i evaluated conidering the PER; Section IV introduce and comare the analytical reult for both model; and our concluion are hon in Section V. II. THE ORIGINAL SYSTEM MODEL AND ITS PERFORMANCE ANALYSIS The Fig. ho the architecture analyzed in []. The rimary ue Slotted Aloha a multile acce rotocol to acce the medium and in the unlicened i conidered Slotted-CSMA rotocol. The rimary acce oint (PAP) and the econdary acce oint (SAP) rovide ervice for rimary and econdary reectively. All rimary uer can be vieed by SAP and vice vera. In the rimary there are N rimary uer (PU) and among thee, I tation are attemting to trant their data acket during a time lot. On other hand, the econdary ha N econdary uer (SU). During a time lot, there are J unlicened uer attemting to end their acket []. The rimary uer have riority to trant their data acket and, therefore, the SU ene the channel to avoid interference ith PU and to identify clearly the ectrum hole that occur hen a time lot of lotted aloha i idle. The Fig. 2 ho the tructure of time lot for lotted aloha and lotted CSMA. In the rimary, each time lot can be buy or idle, deending on tranion tate of PU during a time lot. If there i no rimary uer attemting to trant acket at the beginning of a time lot, then it i conidered idle. Thi ectrum oortunity can be exloited by the econdary uer []. Netork Licened Netork Slotted Aloha SU-7 PU-4 PU-7 PU- PAP SU-3 SU-5 PU-3 PU-5 PU-6 SU- SU-2 SAP Netork Cognitive Netork Slotted CSMA PU-2 SU-4 Figure. Original model architecture; Slotted Aloha in the rimary and Slotted CSMA in the cognitive (lotted Aloha) Slotted CSMA... Slotted CSMA Figure 2. Slot tructure of Slotted Aloha for rimary uer and Slotted CSMA for econdary uer A. Netork Analyi According to the traffic model introduced in [], any PU that i not in a retranion tate can generate a ne acket ith robability σ. Therefore, the robability that a PU doe not generate any acket i (-σ ). If a ne acket i generated in the, it i trantted immediately in the next time lot. If the acket i not uccefully trantted during a time lot, it i retrantted ith robability σ in the folloing time lot until that acket i uccefully trantted. Uer in the retranion tate cannot generate ne data acket []. ) Fading Model in the Netork: let x be the intantaneou oer of a concerned acket ignal in the rimary and be y i the intantaneou oer of the interfering acket ignal generated by the other PU during a time lot. The fading model conidered in [] i a Rayleigh fading channel ith the folloing exonential ditribution, x X x ( x ) e () X y ( yi ) e (2) Yi here X and Y i are the average oer of concerned and interfering acket ignal, reectively. In [], X =Y i. 2) Cature Effect in the Netork: according to [] and [9], the ignal arriving at the receiver have different oer level due to tranion oer racticed by the uer, fading or hadoing. In thi cae, hether the oer of the concerned data acket from a PU i greater than the um of the oer level of all interfering acket in thi and atifie a given threhold, o-called cature ratio (R), the concerned acket can be detected by PAP and all other interfering uer fail in acce medium. Thu, the robability of cature (P ca->pap ) can be calculated a [], I x P ca PAP ( I ) Pr R. (3) I R y i i Where I rereent the total number of imultaneou trantting rimary uer at a given time lot []. 3) Netork Throughut: according to [], the rimary throughut, S o, can be calculated a belo, y i Yi (lotted Aloha) t-2 t- t Coyright (c) IARIA, 23. ISBN:

3 S o N i i N i i N i. i (4) R B. Cognitive Netork Analyi for the original model In thi, Slotted CSMA rotocol i ued for medium acce and it rooed lot tructure i reented in the Fig. 3 []. In the Fig. 3, the firt and third time lot of the rimary are buy, i.e., they are ued by PU. The econd and fourth one rereent ectral oortunitie and they can be exloited by SU. Each of idle time lot of Slotted Aloha i ubdivided into ni-lot. So, the channel i time lot baed for rimary and ni-lot baed for econdary uer. The duration of each ni lot i equal to the maximum roagation delay () found in the rimary and econdary and correond to the ditance from oint a to b in the Fig. 3 []. There are to kind of ni-lot: () fe are deigned for carrier ening eriod (S ), and (2) the mot are aimed for acket tranion (T ) of the SU []. According to the Fig. 3, the maximum ening eriod alloed i from oint a, i.e., the beginning of an idle time lot, to oint c and the ening oint i et to haen at the beginning of each ni-lot. The ditance beteen oint c and oint e i ecified a the maximum length of the data acket (T ) from the econdary in term of the number of nilot. Therefore, the acket length of the econdary i horter than the acket length of the rimary due to the carrier ening eriod []. ) Traffic model for the econdary : in the econdary, uing Slotted CSMA a rotocol to acce the channel, each SU can generate a ne acket ith robability (σ ) during a ni-lot. Conequently, the robability of a SU doe not generate a ne acket i (- σ ). Whether an unlicened uer i in the retranion tate, it cannot generate a ne acket []. During an idle or buy time lot in the rimary, if a ne acket i generated by a SU ithin carrier ening eriod of a ni-lot, it ene the channel in the folloing ening oint of the carrier ening eriod. If the channel i idle, it acket i trantted immediately. If the channel i buy, the SU give u and tart ening the channel ith robability σ during each ening oint of the remaining carrier ening eriod in the current time lot, i.e., the oint c in Fig. 3. And hether the channel remain buy during thi carrier ening eriod, the roce continue ith robability σ during each ening oint in the folloing time lot until the channel i idle and the acket i uccefully trantted. If a ne acket i generated outide the deignated carrier ening eriod, the ne acket i tored and the tation begin to ene the channel ith robability σ during each ening oint in the folloing carrier ening eriod until the channel become idle and the ne acket i uccefully trantted. According to [], a acket tranion of a SU can tart from any ening oint of the carrier ening eriod that channel i ened idle []. (lotted Aloha) Sened Slotted CSMA Tranion Period Uer - Packet lenght Figure 3. Time lot tructure of Slotted CSMA for econdary uer Oberving the cae () and (2) in the Fig. 3, one can oberve that if the channel i ened idle before the end of the carrier ening eriod in the current time lot (oint c in the Fig. 3), then after the end of acket tranion remain ome unued ni-lot []. Finally, in the model rooed in [], SU hould be able to ene the channel and deterne if it i buy or not. The idle time lot rereent ectral oortunitie that are diuted by the econdary uer ithin an environment for cooeration beteen themelve. 2) Fading Analyi and Cature Effect for the Netork: let x and z j be intantaneou oer level of the concerned acket ignal and the interfering acket ignal originated in thi, reectively. In [] i conidered a Rayleigh fading channel ith the folloing exonential ditribution [], x X Px ( x ) e (5) X j (lotted Aloha) a b c d e τ Tranion Period ni lot Carrier Sening eriod Uer - Packet lenght Sened a b c d e τ Carrier Sening eriod Tranion Period ni lot Sened Uer - Packet lenght a b c e τ τ i the roagation delay Carrier Sening eriod z j Z j Slotted CSMA Cae Cae 2 Cae 3 Pz ( z j ) e (6) Z here X an Z j are the mean oer level of that ignal. In Slotted CSMA, the cature robability can alo be calculated a belo [], J z j P casap ( J ) Pr R. (7) J R z j j Where J denote the number of SU are attemting acket tranion during an idle time lot and R i the Cature Ratio []. 3) The Netork Throughut: according to [], the econdary throughut i defined a the acket length in term of number of ni-lot divided by the total number of ni-lot that are ent in the roce of acket tranion, including in thi cae both buy and idle lot. Then, the econdary throughut, S o, i comuted by [], Coyright (c) IARIA, 23. ISBN:

4 S o N j ( T N j j N S j T S ) R j N j. N C. Overall Netork Throughut The overall throughut, S oto, i the um of rimary throughut and econdary throughut, a calculated belo [], S S S (9) III. oto o o. THE PROPOSED NEW SYSTEM MODEL In thi ection, e rooe an extenion to the original model conidering the influence of PER in the calculation of throughut. Tranion error occur during the acket detection due to the interfering ignal. Thi aroach become a more realitic model, ince the acket received ith error are dicarded and retrantted in the mot of data communication alication. The ytem architecture hon in Fig. i alo ued for the ne model, a ell a the tructure of time lot and ni-lot introduced in the Fig. 2 and Fig. 3 in the original model. In thi ne model, the US alo ue Slotted CSMA rotocol to acce the medium and they can ene the channel. Thu, during a acket tranion in a given, rimary or econdary, in the SIR calculating are conidered only the interfering ignal generated by uer of that, i.e., SU cannot trant acket hen there are PU attemting to trant their acket. The Uer from cognitive only can trant acket hen a time lot of the rimary i idle. A. Packet Error Rate The knoledge of the acket error rate in communication ytem i imortant, ince in mot of thee ytem, data are trantted in acket rather than bit tream. Moreover, their erformance i deterned by PER intead of bit error rate (BER) or ymbol error rate (SER) []. The PER i deendent of the ignal-to-interference lu noie ratio (SNIR) in the conidered channel. Hoever, a in [], the additive noie i negligible in interference-lited channel. Therefore the model called ignal-to-interference ratio (SIR) i ued in thi aer. Reference [2], [3], [4], [5], [6] and [7] introduce erical or aroximate method for PER calculating. Their concluion i that uch calculation i quite comlex, imrecie and cannot be generalized to the real alication. All rooed method above to calculate or etimate the PER model the communication channel according to a Markov chain, here the ignal-to-noie ratio (SNR) i artitioned into a finite number of tate that can range beteen to and everal. The difficultie of orking ith Markovian model are in to et the tranition robabilitie of tate to reflect the real channel behavior. In [8], [9] and [2] method for PER calculating are analyzed and i rooed to tudy uch (8) behavior by collecting the real tatitical information or even by uing uitable imulation tool on comuter. Due to the exoed above, in thi aer, e chooe to ork ith the methodology introduced in [], hich allo the calculation of PER a a function of SIR, in a direct and imle fahion by uing a highly accurate uer bound for the ytem analyzed. Conidering that the SU can liten the channel and they cannot caue interference to PU, e can obtain the exected value for SIR in the rimary, Δ, a belo, X. () ( I ) Yi ( I ) Taking into account that the SU comete for idle time lot from the rimary, then the average SIR for the unlicened, Δ, i given by, X. () ( J ) Z j ( J ) No, let f(δ) be a function that link the PER ith the intantaneou SIR at recetion (δ) in a channel ith additive hite Gauian noie (AWGN) and (δ) i the robability denity function of SIR in the receiver, ith exonential ditribution. According to [], the average PER, rereented by P ave (Δ), can be calculated by the folloing integral, P ave ( ) f ( ) ( ) d (2) Pave ( ) f ( ) e d. (3) Conidering the modulation technique emloyed, acket length and the coding cheme ued, the reolution of (3) for the general cae i quite difficult. Then in [] an aroximation to calculate the PER by uer bound i rooed, according to the folloing inequality, P ave( ) e. (4) The uccefully trantted acket rate (PSR) i then given by the equation belo,. PSR e (5) Where i a contant value for the Rayleigh fading channel and can be obtained through the integral belo [], f ( ) d. (6) And f(δ) can be obtained a follo [], n f ( ) { [ b( )] }. (7) Where b(δ) i the BER in AWGN channel. For a modulation technique a the binary hae hift keying (BPSK) ith coherent detection, ithout uing the channel coding cheme and conidering acket of n bit, b(δ) can be calculated according to the equation given belo [], b( ) erfc( ). (8) 2 Reference [] reent an analytical reolution of the uer bound and the correonding imulation for the exected value for PER a a function of average SNR. From the reult reented in [], e can oberve that the uer Coyright (c) IARIA, 23. ISBN:

5 bound rovide an accurate value for the PER under ome aumtion, e.g., hen coherent BPSK i emloyed a a technique for modulation, ith or ithout a channel coding cheme and by uing ome acket length (greater than or equal to 27 bit hen channel coding i not ued). In our analyi e aume coherent BPSK modulation ithout a channel coding cheme and ith acket length of 27 [bit]. According to the conideration above, in thi aer, e conider that the PER and the PSR are obtained in an aroximate fahion by the equation reented belo, P ( ) e (9) ave PSR ( ) e. (2) Uing MATLAB and comaring to the reult reented in [], the value obtained for are hon in the Table I. B. The Netork Throughut for the Ne model The rimary throughut i defined a the total number of acket trantted by the licened uer and received correctly by PAP during a time lot []. In the ne ytem model, a acket i conidered uccefully trantted hen it i catured by the receiver and it doe not have any error due to interference reent in the. In thi cae, the rimary throughut, S n, i aroximately given by: S S PSR( ) S e. (2) n o o ( I ) Where S o i the rimary throughut for the original model and PSR(Δ ) i the uccefully trantted acket rate. C. The Netork Throughut for the Ne Model Referring to the econdary, the throughut i defined a the length of acket in term of ni-lot divided by the total number of ni-lot ued in the tranion roce, including both buy and idle lot []. When one conider only the acket received ithout error due to interference of the, the econdary throughut, S n, i given aroximately by, ( J ) S n So PSR( ) So e. (22) Where S o i the econdary throughut for the original model and PSR(Δ ) i the uccefully trantted acket rate. D. The Overall Netork Throughut for the Ne Model The overall throughut for the ne ytem model, S otn, i the um of the rimary and econdary throughut, a belo: S S S (23) otn n n. in [], i.e., N = 2 (uer), N = 2 (uer), R = 3 (db), =, S = (ni lot), T = (ni lot) e σ = σ. TABLE I. VALUES OF W CONSIDERING COHERENT BPSK MODULATION Packet length in n (bit) Uncoded 27 (bit) Uncoded 23 (bit) The grah of Fig. 4, Fig. 5 and, Fig. 6 ho that the effect of the PER on the rimary and econdary tranion, due to interference caued by their tation, cannot be diregarded. When one conider the PER, there i a ignificant reduction in the rimary and econdary throughut and alo in the overall throughut. Figure 4. Cognitive throughut (N =N =2, S =, R=3 db) IV. NUMERICAL RESULTS The Fig. 4, Fig. 5 and, Fig. 6 hon the analytical reult obtained for throughut of rimary and econdary throughut and overall throughut ytem. To comare the reult beteen the ne model and the original model, in the folloing grahic are ued the ame arameter introduced Figure 5. throughut (N =N =2, S =, R=3 db) Coyright (c) IARIA, 23. ISBN:

6 Figure 6. Overall throughut (N=N=2, S=, R=3 db) V. CONCLUSION In thi aer, e rooe a ne model to comute the throughut in a cognitive radio, conidering Slotted Aloha in the licened and Slotted CSMA in the cognitive. Thi rooed model i an extenion of the model analyzed in []. The ne model rooed conider the interference beteen the tation of the and their effect over the throughut of each. It i verified that the interference increae the acket error rate and reduce the rimary throughut, the econdary throughut and overall throughut. Therefore, it i concluded that the PER, due to interference beteen radio tation of both, cannot be neglected, a haen in []. A uggetion for future tudy, it i rooed to invetigate mechanim to reduce the acket error rate on the and thereby imrove the throughut of each and the overall throughut. REFERENCES [] Z. Yang, "Invetigation of multile acce rotocol in cognitive radio,"ph.d. diertation, Det. of Elect. and Comut. Eng., Steven Int. of Technology, Hoboken, NJ, 2. [2] S. Haykin, "Cognitive radio: brain-emoered irele communication," IEEE J. Sel. Area Commun., vol. 23, no. 2, Feb. 25, [3] E. Hoain, D. Niyato, and Z. Han, "Introduction to cognitive radio," in Dynac Sectrum Acce and Management in Cognitive Radio Netork, t ed., Ne York, NY: Cambridge Univerity Pre, 29, ch.2, [4] I.F. Akyildiz, W.Y. Lee, M.C. Vuran, and S. Mohanty, "A urvey on ectrum management in cognitive radio," IEEE Commun. Mag., vol. 46, iue:4, Ar. 28, [5] I.F. Akyildiz, W.Y. Lee, M.C. Vuran, and S. Mohanty, "NeXt generation/dynac ectrum acce/cognitive radio irele : a urvey," Comut. Netork, vol. 5, iue:3, May 26, [6] P. Kolodzy et al, "Sectrum Policy Tak Force eek ublic comment on iue related to comion ectrum olicie," The Federal Communication Comion (FCC), Indianaoli, IN, ET Docket No. 2-35, 22. [7] J. Straner, "The role of autonoc aetorking in cognitive," in Cognitive Netork: Toard Self-Aare Netork, t ed., Ne York, NY: Wiley, 28, ch. 2, [8] T. Yucek and H. Arlan, "A urvey of ectrum ening algorithm for cognitive radio alication," IEEE Commun. Survey & Tutorial, vol., no., Mar. 29, [9] Y. Onozato, J. Liu, and S. Noguchi, "Stability of a lotted Aloha ytem ith cature effect," IEEE Tran. on Veh. Technology, vol. 38, no., Feb. 989, [] Y. Xi, A. Burr, J. Wei, and D. Grace, "A general uer bound to evaluate acket error rate over quai-tatic fading channel," IEEE Tran. on Wirele Commun., vol., no. 5, May 2, [] S. Kucera, S. Aia, and S. Yohida, "Adative channel allocation for enabling target SINR achievability in oercontrolled irele," IEEE Tran. on Wirele Commun., vol. 9, no. 2, Feb. 2, [2] H. Bai and M. Atiquzzaman, "Error modeling cheme for fading channel in irele communication: a urvey," IEEE Communic. Survey & Tutorie, vol. 5, no. 2, Dec 23, [3] J.R. Yee and E. J. Weldon, "Evaluation of the erformance of error-correcting code on a Gilbert channel," IEEE Tran. on Commun., vol. 43, no. 6, Aug. 995, [4] J.P. Ebert and A. Willy, "A Gilbert-Elliott bit error model and the efficient ue in acket level imulation," TKN Techn. Re., Tech. Univerity Berlin, Berlin, re. 99-2, 999. [5] H. Bichl and E. Lutz, "Packet error rate in the noninterleaved Rayleigh channel," IEEE Tran. on Commun., vol. 43, no. 2/3/4, Feb./Mar./Ar. 995, [6] G. Sharma, A. Dholakia, and A. Haan, "Simulation of error traing decoder on a fading channel," In. Proc IEEE Veh. Technology Conf. 996, Atlanta, GA, vol. 2, [7] H. Labiod, "Performance of Reed Solomon error-correcting code fading channel," in 999 IEEE Int. Conf. on Peronal Wirele Commun.,Jaiur, India, [8] B.D. Fritchman, "A binary channel characterization uing artitioned Markov chain," IEEE Tran. on Inform. Theory, vol. 3, no. 2, Ar. 967, [9] J. Arauz, P. Krihnamurthy, and M.A. Labrador, "Dicrete Rayleigh fading channel modeling," Wirele Commun.and Mobile Comuting, vol. 4, iue:4, jun. 24, [2] C.C. Tan and N.C. Beaulieu, "On firt-order Markov modeling for the Rayleigh fading channel," IEEE Tran. on Commun., vol. 4, no. 2, Dec. 2, Coyright (c) IARIA, 23. ISBN: