Study on Power and Rate Control Algorithm for Cognitive Wireless. Networks

Transcription

1 WSEAS TANSATONS on OMMUNATONS Qian Hu Zhenzhou Tang Study on Power and ate ontrol Algorithm or ognitive Wireless Networs QAN HU and ZHENZHOU TANG ollege o Physics and Electronic normation Wenzhou University Wenzhou 35 HNA mrtangzz@gmailcom huqian@wzueducn Abstract:-ate control and ower control are two imortant issues in cognitive radio networs n this aer three rate control strategies are irst introduced namely Dirty Paer oding (DP) strategy Noise strategy and Oortunistic ntererence ancellation (O) strategy The achievable rates o the three strategies are also comared Then otimal ower control algorithms are roosed each corresonding to one rate control strategy The algorithms achieve the maximum transmit rate or the cognitive user by aroriately controlling the transmit ower on each subchannel under the intererence temerature constraint o the rimary user Simulation results show that the roosed algorithms can dramatically imrove the transmit rate o cognitive user ey-words:-ognitive radio ntererence temerature Power control ate control NTODUTON Wireless sectrum is quite a scarce resource and as the raid growth o the wireless standards and the oulation o wireless users the scarcity o radio sectrum resource is becoming a serious issue which imoses increasingly severe restrictions on the develoment o wireless communication industry n recent years a new wireless communication technology called ognitive adio () which enables multi-radio system to share sectrum oortunistically is emerging as a romising technique to deal with this increasingly tense situation This technology imroves the eiciency o sectrum utilization and to some extent solves the roblem o the scarcity o radio sectrum resource []-[3] Power control and rate control are two ey technologies in cognitive radio system n cognitive wireless networs the rimary user (authorized user) has riority in accessing the licensed sectrum; cognitive user detects the usage o sectrum and oortunistically access the unoccuied channel n this aer rimary user shares sectrum with cognitive user and in order to rovide a rotection or rimary user's communication the transmit ower o cognitive user is strictly controlled On the other hand cognitive user adjusts its transmit rate based on channel condition and the intererence rom rimary transmitter in order to achieve the otimal utilization o sectrum resource There are several literatures concerning the above roblems rom the ersective o inormation theory On the roblem o cognitive wireless channel caacity: Devroye et al [4] regard cognitive channel as a secial intererence channel with the entire nowledge o intererence and obtain the caacity limit o cognitive user by using Gel'and-Pinser's SSN: ssue 4 Volume 9 Aril

2 WSEAS TANSATONS on OMMUNATONS Qian Hu Zhenzhou Tang coding Jovicic et al [5] urther oint out that in order to guarantee the transmit rate o rimary user cognitive user should divide its transmit ower into two arts: one art relays rimary user's inormation and the other art transmits its own inormation Also they assume that the rimary user's inormation is nown to the cognitive user so the cognitive transmitter can emloy Dirty Paer oding (DP) to cancel the intererence caused by the rimary user at the cognitive receiver n [6] and [7] the authors discuss the achievable rate region o cognitive multi-access channel and the methods to achieve the caacity On the roblem o ower control: Ghasemi et al [8] study the caacity limit o cognitive user and the otimal strategies o ower allocation and rate control under the constraint o the intererence temerature Literature [9] rooses an oortunistic ower control strategy in ading wireless channel which can maximize the ergodic caacity o cognitive user under the remise o maintaining the rimary user's outage robability On the roblem o rate control: Poovsi et al [] roose the Oortunistic ntererence ancellation (O) strategy where the transmitter emloys suerosition coding while the receiver alies successive decoding This strategy can signiicantly imrove the transmit rate o cognitive user n this aer we jointly consider the technologies o rate control and ower control in cognitive radio networs First three commonly used rate control strategies are reviewed and comared: Dirty Paer oding (DP) strategy Noise strategy Oortunistic ntererence ancellation (O) strategy Then we ocus on OFDM based multi-channel cognitive wireless system orresonding to the above rate control strategies three otimal ower control strategies are roosed Aiming at maximizing the transmit rate o cognitive user these strategies determine the otimal transmit ower on each subchannels under the constraint o average intererence ower at the rimary user Simulation results show that the roosed ower control strategies can signiicantly imrove the transmit rate o cognitive user SYSTEM MODEL Fig hannel model o cognitive radio networs hannel model o cognitive radio networs is shown in Figure The rimary receiver communicates with the rimary transmitter using the licensed sectrum The cognitive user communicates using the same sectrum The rimary user and cognitive user both emloy the OFDM modulation technique to transmit inormation Assume that there are subchannels in the system and or the -th subchannel g denotes the channel gain between cognitive transmitter and rimary receiver h denotes the channel gain between cognitive transmitter and cognitive receiver denotes the channel gain between rimary transmitter and cognitive receiver n addition we also assume that the cognitive system obtains the channel gain inormation through certain methods and adjust the transmit ower and rate on each subchannel based on the channel inormation The received signal on the -th subchannel o the cognitive receiver can be described as: where y = h x x z () denote the transmit ower o cognitive user and rimary user on the -th subchannel resectively symbols and ower x x denote the transmit z denotes the Gaussian noise with SSN: ssue 4 Volume 9 Aril

3 WSEAS TANSATONS on OMMUNATONS Qian Hu Zhenzhou Tang n cognitive radio networs the rimary user ignores the resence o cognitive user n other words rimary user s transmit ower transmit rate and are determined by only rimary user and are uncorrelated with g this aer we regard and h and So in as nown arameters On the other hand in order to avoid the intererence at the rimary user caused by the cognitive user the average intererence ower at rimary receiver is strictly constrained Let Q be the maximum average ower that the rimary user can tolerate at its receiver thus Note that = g Q () Q in the above inequality is determined by the intererence temerature o the rimary receiver 3 ATE ONTOL The roblem o cognitive user rate control can be described as: given h and z what is the maximum achievable rate o cognitive user on the -th subchannel? From () we can see that the received signal at the cognitive receiver contains the intererence signal caused by the rimary user so the ey roblem to deal with is how we can cancel the rimary user s intererence at the cognitive receiver At resent there are three oular strategies on rate control o cognitive user A Dirty Paer oding (DP) Strategy n [4] and [5] the authors emloy this strategy to cancel the intererence caused by the rimary user the cognitive transmitter has the nowledge o x and the receiver can comletely cancel the rimary user intererence by using Dirty Paer oding n this case the transmit rate o cognitive user is: DP ( ) = (3) h where ( x) = log( x) = n act the Dirty Paer oding strategy requires that the cognitive transmitter should have rior nowledge o the rimary user s intererence which is merely not oerational in actual systems B Noise Strategy Noise strategy reers to the simle but widely used strategy o rate control [8] [9] egarding the intererence as noise the transmit rate o cognitive user can be described as: where = Noise = (4) Oortunistic ntererence ancellation (O) strategy Oortunistic ntererence ancellation (O) strategy is roosed in [] n this strategy the cognitive channel is regarded as a multi-access channel (MA) which contains the rimary user s intererence signal and the cognitive user signal as shown in igure n the igure the decoding order at L is: irst decode cognitive user signal and then decode rimary user s intererence signal The rate air is: (5) ( ) = ( ) Decoding order at L c is just the oosite: irst decode rimary user s intererence signal and then decode the cognitive user signal The rate air is: (6) ( ) = ( ) SSN: ssue 4 Volume 9 Aril

4 WSEAS TANSATONS on OMMUNATONS Qian Hu Zhenzhou Tang ( ) intererence signal at rate a satisies the equation: a ( ) Fig The caacity region o MA n O strategy the transmit rate o cognitive user is relevant to the transmit rate o cognitive user = (9) a the intererence signal o the rimary user can be decoded correctly At last decode = () a ( ) According to (9) we can get a a x () at rate / β = Given in dierent regions the maximum whereβ satisies = ( β ) So the transmit achievable rates o the cognitive user have dierent exressions as described below: > ( ) This case corresonds to egion in igure The intererence signal o the rimary user can not be decoded correctly in this region So we have to treat the intererence signal as noise and then the achievable rate can be described as: O ( ) < = (7) This case corresonds to egion in igure The cognitive transmitter divides the transmit signal into two arts by using suerosition coding []: x = (8) () () ( a ) x a x This strategy resembles the ate Slitting roosed in [] and [] The decoding order at the cognitive receiver is: irst decode rate () () ( a ) x at ( a ) = and then decode the a rate o the rimary user is: O = () () 3 = log β = M () This case corresonds to egion in igure The cognitive receiver can correctly decode intererence signal caused by the rimary user irst and then decode its own signal n this case the cognitive user can achieve a rate o: O ( ) = () n conclusion the maximum achievable rate o the cognitive user in O strategy can be exressed as ollows: O > ( ) = log < β ( ) ( ) () SSN: ssue 4 Volume 9 Aril

5 WSEAS TANSATONS on OMMUNATONS Qian Hu Zhenzhou Tang This strategy is oortunistic because the maximum achievable rate o the cognitive user deends on the rimary user s transmit rate as well as the instantaneous channel side inormation between rimary transmitter and cognitive receiver When the transmit rate o the rimary user is high (case ) the cognitive receiver cannot correctly decode the rimary user's intererence signal thus has to treat the intererence signal as noise n this case the O is identiied with Noise strategy When the transmit rate o the rimary user is quite low (case 3) the rimary user's intererence signal can be decoded correctly So the intererence can be cancelled comletely n this case the O strategy is identiied with DP strategy Note that in case and 3 the inlection oint is at = / β ate(bit/hz) DP O Noise Fig3 Transmit rate vs Figure 3 gives the numerical illustration o relations between transmit rates and strategies where Because ( ) =bit/hz in dierent = the transmit rate o cognitive user is determined by equations () and () Observed rom the igure there exists an inlection oint on the curve o O beore which the O curve behaves lie the DP curve (Equ) and ater which the O curve erorms as an uer shit M o the curve in Noise strategy The inlection oint is at = / β = Figure 4 shows the relations between the transmit rate o cognitive user and where = From the igure we can see that as to Noise strategy the transmit rate decreases with when is low ( ) As to O strategy > the rimary user's intererence signal cannot be decoded So the achievable rate is the same as the rate in Noise strategy With the increase o the rimary user's intererence signal can be decoded and cancelled gradually and the achievable rate increases gradually When ( ) β the rimary user's intererence signal can be cancelled comletely and the achievable rate curve coincides with the curve in DP strategy ate(bit/hz) 5 DP O = O = Noise 4 POWE ONTOL n Section 3 we studied on the rate control strategies or determining and the main ocus is how to Fig4 Transmit rate vs cancel the intererence caused by the rimary user n this section we will research on ower control strategies o cognitive user The objective is to SSN: ssue 4 Volume 9 Aril

6 WSEAS TANSATONS on OMMUNATONS Qian Hu Zhenzhou Tang maximize the total achievable rate o cognitive users under the constraint o intererence temerature at rimary receiver g First let l = ecall that = h h so the constraint inequality () can be transormed as: l Q Then consider that the achievable rate is = l an increasing unction o inequality can be urther simliied: so the constraint l = Q = l onsequently the otimization roblem can be described as: max st = ( ) l = Q = l A Dirty Paer oding (DP) Strategy For this strategy Lagrange's method o multiliers is alied To ind the otimal solution o the ower allocation we orm the Lagragian: o Let J = J log ( ) λl = = we derive: where ( x ) (3) = onsidering the nonnegativity = (4) λl x = x> The water level is x determined by the ollowing equation: = l λ = Q (5) B Noise Strategy For this strategy we can get the otimal solution o the ower allocation using the same aroach: Similarly λ satisies = λl = l l = λ (6) Q Oortunistic ntererence ancellation (O) strategy For O strategy when ( ) rate control unction ) is a iecewise analytic unction o ( which is nondierentiable at the inlexion oint So Lagrange's method o multiliers can not be alied and we have to transorm the otimization roblem Let ot be the otimal solution o the ower allocation According to subchannels can be classiied into our sets: { ( > )} = ot { ( ) } > ot the = ot { ( ) = } = 3 ot { ( ) < } = corresonding to the 4 subchannels in egion egion nlexion Point and egion resectively Then the objective unction ( ) can be transormed as: = M ( ) ( ) 3 4 Emloy Lagrange's method o multiliers to derive ot : SSN: ssue 4 Volume 9 Aril

7 WSEAS TANSATONS on OMMUNATONS Qian Hu Zhenzhou Tang λl ot = 3 (7) 4 λl where λ satisies ot l = Q Then the original = otimization roblem transorms to a new roblem: to ind the otimal value o λ and the division o sets 3 4 Note that 3 4 are relevant to ot but is not So we roose the ollowing iterative algorithm to solve the roblem: Ste: nitialize = { > ( )} { ( )} = 3 = 4 = Φ Ste: Find λ which satisies l l l l λ 3 λ 4 = Q Ste3: the inequality λ l holds brea rom the loo; otherwise udate: = 4 < 3 λl 4 = 3 λl 3 Ste4: eeat ste The rincile o this iterative algorithm is as ollows: irst assume that the transmit ower on each subchannel is high and the subchannels belong to set or Based on this assumtion ind the water level λ the inequality λ l is whose ower are below the water level the initial division o the sets is correct and the algorithm converges; otherwise adjust the division based on the result o ower control and search or the water level until the algorithm converges Ater each iteration the water level λ declines so art o subchannels in 3 transer to 3 or 4 When the algorithm converges satisy: )For the otimal solution and 3 4 will the division o the sets 3 4 is determined; )The otimal solution satisies equation (7) under the division o the sets 3 4 Thereore the algorithm converges to the otimum solution o ower control Also the comlexity o the roosed algorithm is not quite high n the worst case the algorithm iterates times and only one subchannel transers to other sets each time 5 SMULATON ESULTS onsider a cognitive system in the ollowing scenario The system bandwidth B is 5MHz and the number o subchannels =64 The noise ower on each cognitive sunchannel is equal and normalized to ie = The transmit ower on each rimary subchannel is equal to db ie =db and the transmit rate on each rimary subchannel is equal to =bit/hz Furthermore g h and are subject to ayleigh distribution with arameters g h resectively n the simulation or simlicity we ix satisied by all subcarriers belonging to set g = db h =db We name the algorithm SSN: ssue 4 Volume 9 Aril

8 WSEAS TANSATONS on OMMUNATONS Qian Hu Zhenzhou Tang roosed in Section 4 as the Otimal Power ontrol (OP) strategy n order to evaluate the erormance o OP we also roose a simle ower control algorithm namely Equal Power ontrol (EP) which assumes that the intererence ower is equal on each subchannel ie = Q /( g ) to comare with it Fig5 shows the erormance o dierent joint ower and rate control algorithms when = db As shown in the igure the achievable rate o DP is maximum while the achievable rate o Noise strategy is minimum and the achievable rate o O alls somewhere in between On the other hand OP has a better erormance than EP However with the increase o Q the dierence between the two strategies becomes narrower When Q =db there is no signiicant dierence The result indicates that EP aroximates to OP when high intererence temerature is tolerable ate(bit/hz) OPDP EPDP OPNoise EPNoise OPO EPO ntererence temerature(db) Fig5 Transmit rate vs intererence temerature ate(bit/hz) DP Noise O (db) Fig6 Transmit rate vs Figure 6 lots the curve o achievable rates o dierent algorithms versus Note that Q =db and OP is adoted in the comutation From the igure we can see that the erormance o DP is indeendent o However the achievable rate o Noise strategy decreases sharly with the intererence o because rimary user increases with For O the achievable rate irst decreases but then increases with increases and aroximates to OP gradually or the same reason as in Fig4 6 ONLUSON n this aer we irst introduced three rate control strategies in cognitive radio networ: Dirty Paer oding strategy Noise strategy and Oortunistic ntererence ancellation strategy orresonding to these rate control strategies we roosed three otimal ower control algorithms or cognitive user with multi- subchannels Simulation results showed that the ower control strategies we roosed can signiicantly imrove the transmit rate o cognitive user SSN: ssue 4 Volume 9 Aril

9 WSEAS TANSATONS on OMMUNATONS Qian Hu Zhenzhou Tang eerences: [] J Mitola et al ognitive radio: Maing sotware radios more ersonal EEE Personal ommunications vol 6 no Aug 999 [] J Mitola ognitive radio: An integrated agent architecture or sotware deined radio Doctoral dissertation oyal nst Technol (TH) Stocholm Sweden [3]S Hayin ognitive adio: Brain-Emowered Wireless ommunications EEE Journal on Selected Areas in ommunications vol 3 no Feb 5 [4] N Devroye P Mitran and V Taroh Achievable ates in ognitive hannels EEE Transactions on normation Theory vol 5 no May 6 [5] A Jovicic and P Viswanath ognitive adio: An normation-theoretic Persective in roc EEE ST July 6 [6] N Devroye P Mitran and V Taroh ognitive Multile Access Networs in roc EEE ST Set 5 [7] Peng heng Guanding Yu Zhaoyang Zhang et al On the Achievable ate egion o Gaussian ognitive Multile Access hannel EEE ommunications Letters vol no May 7 [8] A Ghasemi and E S Sousa Fundamental Limits o Sectrum-Sharing in Fading Environments EEE Transactions on Wireless ommunications vol 6 no Feb 7 [9] Yan hen Guanding Yu Zhaoyang Zhang et al On ognitive adio Networs with Oortunistic Power ontrol Strategies in Fading hannels EEE Transactions on Wireless ommunications acceted Aril 7 [] P Poovsi H Yomo Nishimori Taranto ate Adatation or ognitive adio under ntererence rom Primary Sectrum User EEE Journal on Selected Areas in ommunications submitted March 7 [OL]htt://arxivorg/d/757 [] B imoldi and Urbane A rate-slitting aroach to the Gaussian multile-access channel EEE Transactions on normation Theory vol 4 no March 996 [] J ao and E M Yeh Asymtotically Otimal Multile-Access ommunication Via Distributed ate Slitting EEE Transactions on normation Theory vol 53 no Jan 7 SSN: ssue 4 Volume 9 Aril