NASH BARGAINING BASED BANDWIDTH ALLOCATION IN COGNITIVE RADIO FOR DELAY CRITICAL APPLICATIONS

Transcription

1 ISSN: (ONLINE) DOI: /jct ICTACT JOURNAL ON COUNICATION TECHNOLOGY, DECEBER 015, VOLUE: 06, ISE: 04 NASH BARGAINING BASED BANDWIDTH ALLOCATION IN COGNITIVE RADIO FOR DELAY CRITICAL APPLICATIONS Kalyan Kulkarn 1 and Bharat Chaudhar 1, Department of Electroncs and Telecommuncaton Engneerng, aharashtra Insttute of Technology, Inda Emal: 1 kalyankul@gmal.com, bharat.chaudhar@mtpune.edu.n Abstract In order to effectvely regulate the exstng resources, dynamc spectrum access n cogntve rado needs to adopt the effectve resource allocaton strateges. ultmeda applcatons ure large banddth and have to meet the delay constrants hle mantanng the data qualty. Game theory s emergng as an effectve analytcal tool for the analyss of avalable resources and ts allocaton. Ths paper addresses resource allocaton schemes employng barganng game model for ult-carrer CDA based Cogntve Rado. Resource allocaton scheme s desgned for transmsson of vdeo over cogntve rado netorks and am to perform banddth allocaton for dfferent cogntve users. Utlty functon based on barganng model s proposed. Prmary user utlty functon ncludes the prcng factor and an upbeat factor that can be adjusted by observng the delay constrants of the vdeo. Allocated banddth to the secondary user can be adjusted by changng the upbeat factor. Throughput n the proposed scheme s ncreased by % as compared to other reported prcng based resource allocaton schemes. The edge PSNR of reconstructed vdeo obtaned as 3.6dB resultng to optmum decodng of the vdeo at the recever. The study also shos upbeat factor can be used to enhanced capacty of the netork. Keyords: Resource Allocaton, Cogntve Netorks, Game Theory, Utlty Functon, Vdeo 1. INTRODUCTION Demand for reless multmeda transmsson has been constantly ncreasng because of the despread deployment of hgh data rate reless netorks and the mprovements n vdeo compresson technologes. Delay-senstve multmeda communcatons servces have brought profound changes to human socety. ore and more people have found ther lves beng enrched and facltated by vdeo applcatons such as vdeo telephony, onlne vdeo streamng, vdeo conferencng, vdeo gamng, and moble TV broadcastng [1]. Dgtal vdeo has already become the man traffc payload for Internet and major reless netorks. Despte the ncreasng demand, reless multmeda communcatons, especally real-tme vdeo applcatons, stll suffer from number of problems. The reless envronment s much dfferent from the Internet that t usually leads to performance degradaton by drectly applyng vdeo transmsson technques that are used n the current Internet envronment. In reless netorks along th ad hoc netorks, a reless lnk usually has a hgh transmsson error rate because of shadong, fadng, and nterferences from other transmttng users. An endto-end path n reless netorks has an even hgher error rate snce t s the concatenaton of multple reless lnks. oreover, user moblty makes the netork topology causng fuent change. An end-to-end route may only exst for a short perod of tme. The fuent lnk falures and route changes cause packet losses, thus degradng the receved vdeo qualty. Strngent banddth resource sunable to fulfll Qualty of Servce (QoS) urements of vdeo over reless. Cogntve Rado (CR) s a promsng technology hch extends the softare-defned rado concept to mprove the spectrum utlzaton. Secondary user () n a cogntve rado netork (CRN) s able to operate n the lcensed band by adjustng ts transmsson parameters. The Cogntve term as ntroduced n [] th the ne communcaton system that can observe and learn from the surroundng rado envronment as ell as can adapt ts on transmsson parameters by keepng user urements n ve. FCC Spectrum Polcy Task Force reported that a large amount of spectrum s under-utlzed [3]. In order to mprove the utlzaton of the spectrum, a secondary system must coexst th the prmary system (lcensed netork). Ths secondary system must bound the nterference caused to the prmary system. Ths mples that resource allocaton (RA) s the key challenge n the successful mplementaton of cogntve rado technology. The ntroducton of CR technology poses ne resource allocaton (RA) problems that need to be solved. Compared to conventonal reless communcaton systems, to ne ssues arse, namely, the nterference poer to the prmary user bands should be kept belo a certan threshold and optmum Qualty of Servce (QoS) should be provded to CRs n spte of the tme-varyng nature of the avalable spectrum. To make unlcensed sharng of the lcensed spectrum a realty, PU operaton must not be compromsed. Thus, CRs should montor and keep the generated nterference to PU bands to an acceptable level. The FCC Spectrum Polcy Task Force has recommended the use of nterference temperature for assessng the level of nterference. Specfcaton of an nterference temperature lmt for a PU corresponds to a maxmum alloed level of nterference poer. CRs can use PU fuency bands as long as the total generated nterference poer to the PUs s kept belo ths lmt. In a fadng envronment, a CR sgnal may undergo deep fadng and receved th very lttle poer at the PU recever. As a result, apart from the spectrum holes, CRs can opportunstcally share PU actve fuency bands, as long as the total generated nterference poer at the PU recever s belo the specfed nterference poer threshold. Specfcally CRs are ured to fnd the spectrum holes n the spectral band and to decde f the spectrum allocaton meets the QoS urements of dfferent users. Game theory s a mathematcal tool for analyzng the nteracton beteen to or more decson makers [4]. It has been used n a varety of felds such as economcs, poltcal scence, and bology. A strategc game conssts of manly three components: a set of players, a strategy set for each player and a utlty (payoff) functon for each player hch measures the degree of happness of the player. Game theory s proved to be very sgnfcant n the telecommuncatons, partcularly reless 1167

2 KALYANI KULKARNI AND BHARAT CHAUDHARI: NASH BARGAINING BASED BANDWIDTH ALLOCATION IN COGNITIVE RADIO FOR DELAY CRITICAL APPLICATIONS communcaton. User s nteracton n a reless netork can be modeled as a game n hch user s termnals are the players n the game competng for netork resources (.e. banddth and energy). Any acton taken by a user affects the performance of other users n the netork. Resource allocaton can be modeled as a game that deals largely th ho ratonal and ntellgent ndvduals nteract th each other n an effort to acheve ther on goals. A reless netork can be analyzed th dfferent types of games. Ths ncludes cooperatve and non-cooperatve games. The non-cooperatve game theory focuses on the analyss of compettve decson-makng nvolvng several players. The players may have partally or totally conflctng nterests over the outcome of the decson process hch s affected by ther actons. Furthermore, a game can be th complete nformaton or ncomplete nformaton. In a game th complete nformaton, each player s aare of the denttes of all other players, ther strateges, and pay-offs. In ths ork, game theory s used to solve problem of banddth allocaton. The remander of ths paper s organzed as follos. Secton covers the related ork hereas n secton 3 e formulate the system model th sngle prmary user. Secton 4 and secton 5 presents formulaton of prcng-based utlty functon and utlty functon for the secondary users respectvely. Exstence of Nash Equlbrum s dscussed n secton 6. Smulaton and dscusson s presented n secton 7 hle paper s concluded n secton 8.. RELATED WORK CR technology can greatly mprove spectrum effcency by allong unlcensed s to opportunstcally obtan spectrum resources from lcensed PUs, and thus can effectvely allevate the ever-ncreasng netork pressure due to the rapd groth of reless multmeda servces [4]. Cao and Zheng [5] consdered cooperatve local barganng to provde both spectrum utlzaton and farness. Local barganng s performed by constructng local groups accordng to a poverty lne that ensures a mnmum spectrum allocaton to each user. Jang et al. [6] proposed a renforcement-learnng-based spectrum-sharng scheme. CR users can learn from the nteracton beteen themselves and the envronment to assess the success level of a partcular acton. Zheng and Cao [7], unlke the aforementoned references, consdered non cooperatve ntra-netork spectrum sharng, n hch an opportunstc spectrum management scheme as proposed. Users allocate channels based on ther observatons of nterference patterns and neghbors. In [8], the authors propose a dynamc game model beteen multple prmary users n cogntve rado netorks. Prmary users usng Bertrand model game each other and then acheve the best prce ultmately. In lterature [9], authors present secondary users utlty functon and compete for banddth through noncooperatve game.utlty functon [10] s defned n terms of system throughput and acheves the maxmum system throughput ultmately through prce and spectrum competton. ethod of jont poer and rate control mechansm s proposed n [11]. Based on the control of secondary users transmt rate, the method lmts ther poer reasonable to reduce nterference on the prmary user. In [1], a poer nterference threshold of secondary users s set. Wthn alloable nterference range, the secondary users game mutually, such that the fnal utlty functon s maxmum. Authors n [13], propose method based on mcroeconomcs. It ntroduces layerng the users accordng to dfferent servce, and then allocates the spectrum dynamcally on the bass of dfferent herarchy. Sub-layer users can perceve upper levels users spectrum, and sharng banddth th them. Ths method not only ensures the upper levels traffc needs, but also mproves the spectrum effcency. Dfferent methods and strateges dscussed above do not explctly consder the secondary user s traffc characterstcs. Vdeo servces such as vdeo conferencng, Internet TV, etc gradually ncrease. Compared to tradtonal data servces, these vdeo servces are sgnfcantly dfferent, such as that vdeo servce need consder the delay senstve, user s subjectve vsual experence and so on. So the vdeo traffc cannot smply use the conventonal method of data servce to allocate the banddth. Cross-layer optmzaton strateges have been proposed as a soluton for mprovng the performance of vdeo over reless applcatons. These solutons nclude jont PHY- AC, APP-PHY, AC-APP layer optmzatons for robust vdeo over reless transmsson. Snce vdeo over cogntve netorks ure seamless communcaton n the dynamc spectrum access envronment, game-theoretc technques are more sutable for the spectrum allocaton. Ths paper proposes barganng-based utlty functons for prmary and secondary users. Prmary user s utlty functon ntroduces to factors termed as upbeat factor and penalty factor. These factors contrbute for releasng more banddth by the PU and also applyng penalty to n order to protect PU s QoS respectvely. We propose a utlty functon for the s hch ncorporate the delay-senstve characterstcs of the multmeda transmssons. 3. SYSTE ODEL WITH SINGLE PRIARY USER CRN under consderaton comprses one PU and multple s. Banddth allocated to PU s W Hz hereas the mnmum banddth ured to PU to carry hs on traffc s consdered as B. There are secondary users and try to share banddth of PU. Prmary user calculates the penalty factor nformaton accordng to ther utlty maxmzaton prncple, and then, nforms the secondary users about the value of penalty factor. Secondary users bargan mutually untl they reach the satsfactory barganng soluton. 3.1 VIDEO RATE-DISTORTION ODEL Vdeo transmsson s subjected to some degree of dstorton due to the compresson. Ths factor s consdered n the Ratedstorton model. Vdeo Rate-Dstorton model descrbes the relatonshp beteen compresson rate and dstorton of the vdeo. Dstorton model s gven by [10], D C R e (1) here, D s vdeo dstorton and C s the vdeo compresson rate. and β represent the specfc parameters of vdeo hch are dfferent from dfferent vdeo content. Peak Sgnal to Nose Rato (PSNR) s used to descrbe the qualty of the vdeo as gven by, 55 PSNR 10log10. () D 1168

3 ISSN: (ONLINE) ICTACT JOURNAL ON COUNICATION TECHNOLOGY, DECEBER 015, VOLUE: 06, ISE: 04 In vdeo transmsson, Edge Peak Sgnal to Nose Rato (EPSNR) s sgnfcant n descrbng QoS parameters. It gves the average of the dfferences beteen the edge pxels of the source vdeo sequence and the correspondng pxels of the processed vdeo sequence. It can be consdered as the edge mean squared error of the processed vdeo sequence. EPSNR s for the peak value P of mage and edge mean squared error SE edge, EPSNR s gven by, P EPSNR 10log. (3) 10 SE edge 4. FORULATION OF PRICING-BASED UTILITY FUNCTION (PRIARY USER) Utlty functon for the prmary user n ths spectrum sharng game model protects QoS parameters of prmary user by ntroducng penalty factor for s. Thus utlty functon ll be governed manly by penalty to s, self enthusasm factor hch n the orst case ll be the entre banddth of PU. Consderng above factors, e can defne the prmary user utlty functon as follos: U pu gp B 1 q (4) B 1 here, g s the rent per unt banddth hen the prmary user s banddth leases to secondary users; p s a upbeat factor of the prmary user. Large value of p ll facltate secondary users to utlze ncreased amount banddth. Penalty factor s to punsh the second users and ncreases th the occuped banddth of secondary users. Total number of the banddth of each secondary user occupes ; 1 Tradeoff parameter q controls the banddth allocaton. PU guarantees hs on QoS by keepng banddth B reserved. Prmary user s alloed to release entre banddth B. Thus, utlty functon represents benefts obtaned by leasng the banddth to secondary users and the nterference from secondary users to prmary user durng banddth sharng process. Second term counts for the cost of prmary users. axmum utlty functon s obtaned by determnng γ. Takng the partal dervatves of over the utlty functon: then, U B 1 pu gp 0 (5) B B 1. (6) gpb The Eq.(6) presents that the penalty factor ncrease th the secondary users banddth occupyng factor FORULATION OF UTILITY FUNCTION FOR THE SECONDARY USERS As the vdeo transmsson s banddth-demand system, as the occuped banddth reaches toards the avalable banddth, there s congeston n the netork. Eventually average delay n the netork ll be ncreased. Consderng netork congeston Z n the secondary user utlty functon, Z B B 1 1 here, s the transmsson rate of user. Hence, 1 represents the total traffc n the netork and denomnator represents the current avalable netork banddth. Utlty functon for the secondary users s gven by, U (7) 1 ln PSNR k g (8) B 1 Compresson rate s gven as [9], R = (9) = log (1 + K ) (10) 1.5 K (11) log BER T here, s secondary user that receves sgnal to nose rato.the target bt error rate s BER T. If the secondary users sgnal to nose rato and target bt error rate s knon, ρ can be determned. Smplfyng, U 1 ln k. (1) B 1 Vdeo s encoded at dfferent rates. The rate dstorton curve s of sgnfcant mportance n determnng vdeo-specfc parameters α and β of the vdeo transmsson. In ths functon, (ε = ln55 lnα ) s a parameter related to the transmsson of vdeo content of the frst user. Frst term n Eq.(1) refers to revenue obtaned from a secondary user for the transmsson of vdeo. As the secondary users occupy more banddth causng delay they are punshed hch s gven by second term of Eq.(7). As the banddth occuped by the secondary users ncreases, netork congeston s controlled by ncreasng the factor k hch s the mpact factor of netork congeston. It ndcates that traffc s very senstve to netork latency. Factor k can be adjusted to acheve the average netork delay urements. Prmary user determnes penalty factor γ hch can reduce the nterference to the prmary user s on traffc. The mnmum transmsson rate R mn sgnfes the mnmum transmsson rate that each vdeo ures to be dstngushed. In order to mprove the utlzaton of resources, each vdeo s assgned by a maxmum transfer rate R max. The ured mnmum and the maxmum transmsson rate correspondng to the 1169

4 KALYANI KULKARNI AND BHARAT CHAUDHARI: NASH BARGAINING BASED BANDWIDTH ALLOCATION IN COGNITIVE RADIO FOR DELAY CRITICAL APPLICATIONS mnmum and maxmum ured banddth ( (mn) and (max)) can be decded n the case of certan spectral effcency. Game model of secondary user s gven as, 1 max U ln k g (13) B 1 s.t. (mn) (max) (14). 1 B B (15) For a statc game, t s assumed that each secondary user knos the current banddth allocaton of other secondary users. In order to maxmze the utlty functon, partal dervatve of the utlty functon U s taken over and set the dervatve to 0, that s, U B B j1, j B kb 1 j EXISTENCE OF NASH EQUILIBRIU (16) We consder there are secondary users hch bargan for the spectrum. Game odel can be expressed as, G = {,{u },{U }} ( N) (17) here, {u } ndcates strategy space of s.e. amount of banddth they got through competton and U represents ther utlty functons. Each user maxmzes ts utlty functon n the game of banddth allocaton. Nash Equlbrum exsts hen the strategy space allocaton polces are satsfed th U{u *, u -} U{u, u -}. The strategy combnaton {u 1, u,..., u N} s called the Nash Equlbrum. For secondary users Nash Equlbrum exsts f follong condtons are met [10]: Condton 1: Wth person partcpant the Game, ever user s all feasble strategy space u s non-empty and compact convex sets on the R m. Proof: In the proposed game model, each secondary user has banddth lmts, ( (mn), (max)), and ndcates a nonempty and convex set. Condton : The user s utlty functon s contnuous quasconcave functons. Proof: Partal dervatve of Eq.(1) over, gves, U B B j1, j W kb 1 and partal dervatve of Eq.(18) over gves, j (18) U B kb (19) pb Therefore, user s utlty functon U s a concave functon n terms of that satsfes the condton. Ths ndcates that there exsts a Nash Equlbrum n ths game. 7. SIULATION RELTS AND DISCUSSION H.64/PEG4-AVC vdeo standard s consdered for the smulaton scenaro, hch s a dely used as an ndustral standard that offers better compresson effcency and greater flexblty n compressng, transmttng and storng vdeo. Compared th standards such as PEG- and PEG-4 Vsual, H.64/PEG4-AVC can delver better mage qualty at the same compressed bt-rate and a loer compressed bt-rate for the same mage qualty. The total banddth of prmary user s consdered as 3Hz. The transcever uses 18 sub-carrer C-CDA for communcaton. The mnmum ured banddth for prmary user to protect hs on transmsson, B s 1Hz. The dynamc learnng factor used n to secondary user s game s u 1 = u = 0.1. Receved Sgnal to Nose Rato (SNR) s consdered asequal for to secondary users.target bt error rate BER T s taken as Parameters of to secondary user s vdeo shon as Table.1. Table.1. Parameters of Vdeo for s Secondary User R(mn) (kbps) R(max) (kbps) User User The Fg.1 shos the effect of banddth released by PU on avalable banddth for s. Banddth released to s s a functon of the factor k hch s the mpact factor of netork congeston. In actual vdeo transmsson, average delay urements of the netork are acheved by adjustng value of mpact factor. As k s ncreased, netork congeston s ncreased and traffc has more strct urements n the netork delay. Ths ures to ncrease the punshment to the s by ncreasng the penalty factor. Ths ncreased value of penalty factor ll reduce the banddth allocated to s. For k = 1, banddth accessed by 1 s 1.4Hz hereas hen k s ncreased, ndcatng congeston and netork delay, and hence penalty factor reduces the banddth accessed by 1. Thus, value of factor of netork congeston and penalty factor acheves the delay urements n the vdeo transmsson. 1170

5 ISSN: (ONLINE) ICTACT JOURNAL ON COUNICATION TECHNOLOGY, DECEBER 015, VOLUE: 06, ISE: Banddth accessed by (Hz) , k=1 1, k=1000, k=1000, k=1 Blockng Probablty of s p=0.5 p= Banddth of PU (Hz) Fg.1. Effect of Penalty Factor on 's Banddth Allocaton In Fg., e analyze the throughput of the netork as a functon of ncrease rate of s. Compared th [15], throughput n the proposed scheme for 5 s s ncreased by %. It s also observed that upto certan ncrease n call arrval rate, throughput ncreases but later on t gets saturated because of non avalablty of addtonal resources to s. Throughput Throughput-Proposed Scheme Throughput-[15] Increase Rate of s Fg.. Performance of s at Dfferent Arrval Rate The blockng probablty as a functon of arrval rate of s s shon n Fg.3. It s observed that the blockng probablty ncreases th the ncrease n arrval rates of s. In the proposed scheme, blockng probablty can be reduced by ncreasng the upbeat factor hch releases more banddth for s. As ndcated n Fg.3, blockng probablty for p = 0.5, s less as compared to ts value for p = Increase Rate of s Fg.3. Blockng Probablty of s for Dfferent Arrval Rate The Fg.4 shos the effect of ncrease rate of PUs on the mean PSNR. It depcts that th the ncreasng arrval rate of PUs, the transmsson qualty of s ncreases. Hoever, the stll acheves better qualty than other 1 due to ts hghest prorty as mplemented n barganng game. Introducton of upbeat factor n the utlty functon also mproves the capacty of netork. In Fg.5, e ndcate ths mprovement n capacty as functon of upbeat factor for a specfed value of SNR. It s observed SNR of 10dB and upbeat factor equal to 0.3, capacty s 4 bts/s/hz hereas t s mproved to 8bts/s/Hz for upbeat factor of 0.8. The EPSNR of the proposed scheme s 3.6dB hch ndcates the optmum decodng of the vdeo at the recever. ean PSNR (db) Increase Rate of PUs 1 Fg.4. ean PSNR of Vdeo vs Increase Rate of PUs 1171

6 KALYANI KULKARNI AND BHARAT CHAUDHARI: NASH BARGAINING BASED BANDWIDTH ALLOCATION IN COGNITIVE RADIO FOR DELAY CRITICAL APPLICATIONS Capacty (bt/s/hz) Fg.5. Capacty of Netork as a Functon of Up-beat Factor The results dscussed above reflect that the method proposed approaches to better farness beteen dfferent secondary users. In [14], the authors use the result of banddth multpled by the unt banddth proft and banddth effcency as the gan of secondary users thout consderng the vdeo contents. Thus, as shon n Fg.1, accordng to the proposed method that take nto account the dfferent vdeo content, the s allocated more banddth because of complex content hereas results n [15] shos that the to secondary users are allocated the same banddth. The nnovatve utlty functons for prmary and secondary users proposed n ths paper consder dfferent features of the vdeos and banddth s allocated after the far barganng of the users n order to maxmze ther on utlty functons. 8. CONCLUSION The major contrbuton of ths paper s toards the banddth allocaton for vdeo transmsson over C-CDA based cogntve rado netorks by reducng the latency n netork by consderng the netork traffc. Proposed utlty functons are desgned on barganng-based non-cooperatve game theory. The upbeat factor ll release more banddth hle the netork delay s reduced by applyng penalty factor to the secondary users. The scheme also mproves PSNR and EPSNR durng the vdeo transmsson. Whle the proposed algorthms are for statc game model, banddth allocaton can be further mproved by ntroducng dynamc modelng. Our results demonstrate that the proposed game-theoretc barganng model sgnfcantly mprove the vdeo transmsson performance. REFERENCES p = 0.3 p = 0.5 p =0.8 p = SNR (db) [1] G. Chandra Sekhar, Shreyans Parakh and Adtya K. Jagannathan, Optmal 4G OFDA Dynamc Subcarrer and Poer Aucton-based Allocaton toards H.64 Scalable Vdeo Transmsson, Defence Scence Journal, Vol. 63, No. 1, pp. 15-4, 013. [] J. tola, Cogntve Rado: An Integrated Agent Archtecture for Softare Defned Rado, Ph.D Thess, Royal Insttute of Technology (KTH), 000. [3] S. Haykn, Cogntve Rado: Bran-Empoered Wreless Communcaton, IEEE Journal on Selected Areas n Communcatons, Vol. 3, No., pp. 01-0, 005. [4] Zhu Han, Dust Nyato, Wald Saad, Tamer Basar and Are Hjorungnes, Game Theory n Wreless and Communcaton Netorks, Cambrdge Unversty Press, 01. [5] L. Cao and H. Zheng, Dstrbuted Spectrum Allocaton va Local Barganng, Proceedngs of nd Annual IEEE Communcatons Socety Conference on Sensor and Ad Hoc Communcatons and Netorks, pp , 005. [6] T. Jang, D. Grace and Y. Lu, Performance of Cogntve Rado Renforcement Spectrum Sharng Usng Dfferent Weghtng Factors, Proceedngs of 3 rd Internatonal Conference on Communcatons and Netorkng n Chna, pp , 008. [7] H. Zheng and L. Cao, Devce-centrc Spectrum anagement, Proceedngs of IEEE Internatonal Symposum on Ne Fronters n Dynamc Spectrum Access Netorks, pp , 005. [8] D. Nyato and E. Hossan, Optmal Prce Competton for Spectrum Sharng n Cogntve Rado: A Dynamc Gametheoretc Approach, Proceedngs of IEEE Global Telecommuncatons Conference, pp , 007. [9] N. Ne and C. Comancu, Adaptve Channel Spectrum Etquette for Cogntve Rado Netorks, Proceedngs of 1 st IEEE Internatonal Symposum on Ne Fronters n Dynamc Spectrum Access Netorks, pp , 005. [10] Y. Chen, B. Wang and K.J.R. Lu, ultuser Rate Allocaton Games for ultmeda Communcatons, IEEE Transactons on ultmeda, Vol. 11, No. 6, pp , 009. [11] Y. Chun-gang, L. Jan-Dong and L. We-yng, Jont Rate and Poer Control based on Game Theory n Cogntve Rado Netorks, Proceedngs of 4 th Internatonal Conference on Communcatons and Netorkng n Chna, pp. 1-5, 009. [1] P. Zhou, W. Yuan, W. Lu and W. Cheng, Jont Poer and Rate Control n Cogntve Rado Netorks: A Gametheoretcal Approach, Proceedngs of IEEE Internatonal Conference on Communcatons, pp , 008. [13] Lars Berlemann and Stefan angold, Cogntve Rado and Dynamc Spectrum Access, John Wley & Sons, 009 [14] Dust Nyato and Ekram Houssan, Compettve Spectrum Sharng n Cogntve Rado Netorks: A Dynamc Game Approach, IEEE Transactons on Wreless Communcatons, pp , 008. [15] Tgang Jang, Honggang Wang and A.V. Vaslakos, QoE- Drven Channel Allocaton Schemes for ultmeda Transmsson of Prorty-Based Secondary Users over Cogntve Rado Netorks, IEEE Journal on Selected Areas n Communcatons, Vol. 30, No. 7, pp ,