downloads.hindawi.comdownloads.hindawi.com/journals/specialissues/824361.pdfeditorial board...
TRANSCRIPT
-
Cooperative Communications for Wireless Ad Hoc and Sensor Networks in 2013Guest Editors: Yong Sun, Shukui Zhang, Hongli Xu, and Shan Lin
International Journal of Distributed Sensor Networks
-
Cooperative Communications for WirelessAd Hoc and Sensor Networks in 2013
-
International Journal of Distributed Sensor Networks
Cooperative Communications for WirelessAd Hoc and Sensor Networks in 2013
Guest Editors: Yong Sun, Shukui Zhang, Hongli Xu,and Shan Lin
-
Copyright ยฉ 2014 Hindawi Publishing Corporation. All rights reserved.
This is a special issue published in โInternational Journal of Distributed Sensor Networks.โ All articles are open access articles distributedunder the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, pro-vided the original work is properly cited.
-
Editorial Board
Miguel Acevedo, USASanghyun Ahn, KoreaMohammod Ali, USAJamal N. Al-Karaki, JordanHabib M. Ammari, USAJavier Bajo, SpainPrabir Barooah, USAAlessandro Bogliolo, ItalyRichard R. Brooks, USAJames Brusey, UKErik Buchmann, GermanyJian-Nong Cao, Hong KongJoaฬo Paulo Carmo, PortugalJesuฬs Carretero, SpainLuca Catarinucci, ItalyHenry Chan, Hong KongChih-Yung Chang, TaiwanPeriklis Chatzimisios, GreeceAi Chen, ChinaPeng Cheng, ChinaJinsung Cho, KoreaKim-Kwang R. Choo, AustraliaChi-Yin Chow, Hong KongWan-Young Chung, KoreaMauro Conti, ItalyDinesh Datla, USAAmitava Datta, AustraliaDanilo De Donno, ItalyIlker Demirkol, SpainDer-Jiunn Deng, TaiwanChyi-Ren Dow, TaiwanGeorge P. Efthymoglou, GreeceFrank Ehlers, ItalyMelike Erol-Kantarci, CanadaGiancarlo Fortino, ItalyLuca Foschini, ItalyDavid Galindo, FranceWeihua Gao, USADeyun Gao, ChinaAthanasios Gkelias, UKIqbal Gondal, AustraliaJayavardhana Gubbi, AustraliaCagri Gungor, TurkeySong Guo, JapanAndrei Gurtov, Finland
Qi Han, USAZ. Hanzalek, Czech RepublicTian He, USAJunyoung Heo, KoreaZujun Hou, SingaporeBaoqi Huang, ChinaChin-Tser Huang, USAYung-Fa Huang, TaiwanXinming Huang, USAJiun-Long Huang, TaiwanWei Huangfu, ChinaMohamed Ibnkahla, CanadaTan Jindong, USAIbrahim Kamel, UAELi-Wei Kang, TaiwanRajgopal Kannan, USASherif Khattab, EgyptLisimachos Kondi, GreeceMarwan Krunz, USAKun-Chan Lan, TaiwanYee W. Law, AustraliaYoung-Koo Lee, KoreaKyung-Chang Lee, KoreaYong Lee, USAJongHyup Lee, KoreaSungyoung Lee, KoreaSeokcheon Lee, USAJoo-Ho Lee, JapanShijian Li, ChinaMinglu Li, ChinaShuai Li, USAShancang Li, UKYe Li, ChinaZhen Li, ChinaYao Liang, USAJing Liang, ChinaWeifa Liang, AustraliaWen-Hwa Liao, TaiwanAlvin S. Lim, USAKai Lin, ChinaZhong Liu, ChinaMing Liu, ChinaDonggang Liu, USAYonghe Liu, USAZhigang Liu, China
Chuan-Ming Liu, TaiwanLeonardo Lizzi, FranceGiuseppe Lo Re, ItalySeng Loke, AustraliaJonathan Loo, UKJuan Antonio Loฬpez Riquelme, SpainPascal Lorenz, FranceKingShan Lui, Hong KongJun Luo, SingaporeJose Ramiro Martinez-de Dios, SpainNirvana Meratnia, The NetherlandsShabbir N. Merchant, IndiaMihael Mohorcic, SloveniaJoseฬ Molina, SpainV. Muthukkumarasamy, AustraliaEduardo Freire Nakamura, BrazilKamesh Namuduri, USAGeorge Nikolakopoulos, SwedenMarimuthu Palaniswami, AustraliaAi-Chun Pang, TaiwanSeung-Jong J. Park, USASoo-Hyun Park, KoreaMiguel A. Patricio, SpainWen-Chih Peng, TaiwanJanez Per, SloveniaDirk Pesch, IrelandShashi Phoha, USAAntonio Puliafito, ItalyHairong Qi, USANageswara S.V. Rao, USAMd. Abdur Razzaque, BangladeshPedro Pereira Rodrigues, PortugalJoel J. P. C. Rodrigues, PortugalJorge Sa Silva, PortugalMohamed Saad, UAESanat Sarangi, IndiaStefano Savazzi, ItalyMarco Scarpa, ItalyArunabha Sen, USAXiao-Jing Shen, ChinaWeihua Sheng, USALouis Shue, SingaporeAntonino Staiano, ItalyTan-Hsu Tan, TaiwanGuozhen Tan, China
-
Shaojie Tang, USABulent Tavli, TurkeyAnthony Tzes, GreeceAgustinus B. Waluyo, AustraliaYu Wang, USARan Wolff, IsraelJianshe Wu, ChinaWen-Jong Wu, TaiwanChase Qishi Wu, USA
Bin Xiao, Hong KongQin Xin, Faroe IslandsJianliang Xu, Hong KongYuan Xue, USATing Yang, ChinaHong-Hsu Yen, TaiwanLi-Hsing Yen, TaiwanSeong-eun Yoo, KoreaNing Yu, China
Changyuan Yu, SingaporeTianle Zhang, ChinaYanmin Zhu, ChinaT. L. Zhu, USAYi-hua Zhu, ChinaQingxin Zhu, ChinaLi Zhuo, ChinaShihong Zou, China
-
Contents
Cooperative Communications for Wireless Ad Hoc and Sensor Networks in 2013, Yong Sun,Shukui Zhang, Hongli Xu, and Shan LinVolume 2014, Article ID 274378, 2 pages
Spectrum Leasing in Cognitive Radio Networks: A Survey, Aqeel Raza Syed and Kok-Lim Alvin YauVolume 2014, Article ID 329235, 22 pages
Partial Relay Selection with Feedback Delay and Cochannel Interference: Performance Analysis andPower Optimization, Xuanxuan Tang, WeifengMou, Fang Fang, Yueming Cai, Weiwei Yang, and Tao ZhangVolume 2014, Article ID 629348, 6 pages
Cognitive Wireless Sensor Network Platform for Cooperative Communications, Agustฬฤฑn Tena,Guillermo Jara, Juan Domingo, Elena Romero, and Alvaro AraujoVolume 2014, Article ID 473905, 8 pages
Improved Message Diffusion Model for Node Coverage Problem of Ad Hoc Network Based on NodeVisit Times, Zhe Yang, Lingzhi Li, Shukui Zhang, Yong Sun, and Yanqin ZhuVolume 2013, Article ID 264383, 11 pages
Distributed Testbed for Coded Cooperation with Software-Defined Radios, Changcai Han and Si LiVolume 2013, Article ID 325301, 9 pages
Energy Efficient Power Allocation for Bidirectional Relaying with Imperfect Channel Estimation,Song Li, Yanjing Sun, Hongli Xu, Zhi Sun, and Wenjuan ShiVolume 2013, Article ID 725459, 8 pages
Hierarchical Spatial Clustering in MultihopWireless Sensor Networks, Zhidan Liu, Wei Xing,Yongchao Wang, and Dongming LuVolume 2013, Article ID 528980, 11 pages
Joint Relay Ordering and Linear Finite Field Network Coding for Multiple-Source Multiple-RelayWireless Sensor Networks, Yulun Cheng and Longxiang YangVolume 2013, Article ID 729869, 9 pages
Application of Baro-Altimeter Sensor in Emergency Positioning and Navigation Based on CompassGEO Satellites, Jianwei Zhan, Jing Pang, Guozhu Zhang, and Gang OuVolume 2013, Article ID 974132, 18 pages
Throughput-Optimal Scheduling for Cooperative Communications inWireless Ad Hoc Networks,Thong Huynh, Won-Joo Hwang, and Suk-Hwan LeeVolume 2013, Article ID 376028, 10 pages
A Service Model for 6LoWPANWireless Sensor Networks, Xiaonan Wang and Haili HuangVolume 2013, Article ID 692735, 8 pages
-
Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2014, Article ID 274378, 2 pageshttp://dx.doi.org/10.1155/2014/274378
EditorialCooperative Communications for Wireless Ad Hoc andSensor Networks in 2013
Yong Sun,1 Shukui Zhang,1 Hongli Xu,2 and Shan Lin3
1 Computer Science and Technology Institute, Soochow University, Suzhou, Jiangsu 215006, China2 School of Computer Science and Technology, University of Science and Technology of China, Hefei, Anhui 215011, China3Department of Computer and Information Sciences, Temple University, 324 Wachman Hall, 1805 North Broad Street,Philadelphia, PA, USA
Correspondence should be addressed to Yong Sun; [email protected]
Received 29 December 2013; Accepted 29 December 2013; Published 19 March 2014
Copyright ยฉ 2014 Yong Sun et al.This is an open access article distributed under theCreative CommonsAttribution License, whichpermits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The concept of cooperative communications for wireless adhoc and sensor networks (WAHSNs) has recently attractedconsiderable attention. Being different from conventionalpoint-to-point communications, cooperative communica-tions allow different users or nodes in a WAHSN to shareresources to create collaboration through distributed trans-mission. It realizes a new form of spatial diversity to combatthe effects of severe fading; thus, it can be used widely inWAHSNs for the sake of resource minimization. However,the impact and challenges of cooperative communication inWAHSNs are not well understood yet. Some fundamentalaspects requiring immediate studies include theoretical toolsfor cooperative networks, effective incentive mechanismsfor cooperation, and new protocol design for cooperativenetworks. This issue tries to collect cutting-edge researchachievements in this special area.
The paper entitled โImproved message diffusion model fornode coverage problem of ad hoc network based on node visittimes,โ by Z. Yang et al., analyzes the causes of the inaccuracyproblems of random samplingmodel and solves the problemsby specially introducing the factors such as node degreeand visit times. As for the E-R random network topology, itvalidates the effectiveness of the model proposed herein incontrast to the simulation experiment results.
In the paper entitled โDistributed testbed for coded coop-eration with software-defined radios,โ by C. Han and S. Li,the authors design and implement a distributed hardwaretestbed using software-defined radios for cooperative com-munication, and the performance of two coded cooperation
schemes with turbo codes is evaluated in the physical layer.Furthermore, a distributed node synchronization scheme isimplemented and the source node and relay node work in thetime division protocol without any centralized controlling.
The paper entitled โEnergy efficient power allocation forbidirectional relaying with imperfect channel estimation,โ byS. Li et al., investigates the power allocation problem tominimize the total transmit power subject to constraints ontwo source nodesโ received signal-to-noise-ratios (SNRs).Thebest relay that minimizes the total transmit power is selectedin a multiple relay network.They also present outage analysiswhen the proposed power allocation is adopted and a close-form approximation of outage probability is obtained byshrinking the integral interval.
The paper entitled โHierarchical spatial clustering in mul-tihop wireless sensor networks,โ by Z. Liu et al., considers theproblem of spatial clustering for approximate data collectionthat is feasible and energy efficient for environment monitor-ing applications. They propose a clustering algorithm namedHSC to group the most similar sensor nodes in a distributedway. HSC runs on a prebuilt data collection tree and thusgets rid of some extra requirements such as global networktopology information and rigorous time synchronization.
The paper entitled โJoint relay ordering and linear finitefield network coding for multiple-source multiple-relay wirelesssensor networks,โ by Y. Cheng and L. Yang, proposes a relayordering algorithm based on finite field NC (FFNC). In thescheme, the relays who initially fail to decode from sourcesare kept listening and searching for the opportunity to decode
-
2 International Journal of Distributed Sensor Networks
the signals from other relays, so as to recover the failure links.Moreover, the scheme is proved to own the merit of diffusioneffect, which makes the diversity improvement more efficientby simply increasing the relay number in the network.
The paper entitled โApplication of baro-altimeter sensor inemergency positioning and navigation based on compass GEOsatellites,โ by J. Zhan et al., presents a low-cost high-resolutionMS5534B barometric (baro-) module which applies to BDS.Firstly, the principle of emergency positioning based on abaro-altimeter sensor and its performance such as the accu-racy are elaborated. Then the effects of baro-altimeter sensormeasurement error on positioning are analyzed. Finally, afteranalyzing the limitation of the conventional algorithms, anew high-accuracy emergency positioning algorithm withbaro-altimeter sensor aiding is proposed, which is not limitedby the integration and userโs altitude.
The paper entitled โThroughput-optimal scheduling forcooperative communications in wireless ad hoc networks,โ by T.Huynh et al., introduces the relay selection schemes that cancontrol the interference at the relay to prevent the relay thatmay harm other pairs. Then, they propose the throughput-optimal scheduling that takes into account error probabilityin decision andmaximizes throughput, that is, the amount ofpackets transmitted without error in network.
The paper entitled โA servicemodel for 6LoWPANwirelesssensor networks,โ by X. Wang and H. Huang, proposesa 6LoWPAN service model based on the IPv6-based k-Anycast communication model. This model is extended into6LoWPAN service model so that the data-centric servicesof WSN can be achieved efficiently in the address-centric6LoWPAN.
Yong SunShukui Zhang
Hongli XuShan Lin
-
Review ArticleSpectrum Leasing in Cognitive Radio Networks: A Survey
Aqeel Raza Syed and Kok-Lim Alvin Yau
Department of Computer Science and Networked System, Sunway University, No. 5 Jalan Universiti, Bandar Sunway,46150 Petaling Jaya, Selangor, Malaysia
Correspondence should be addressed to Aqeel Raza Syed; [email protected]
Received 8 July 2013; Accepted 6 December 2013; Published 13 February 2014
Academic Editor: Shukui Zhang
Copyright ยฉ 2014 A. R. Syed and K.-L. A.- Yau.This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.
Cognitive Radio (CR) is a dynamic spectrum access approach, in which unlicensed users (or secondary users, SUs) exploit theunderutilized channels (or white spaces) owned by the licensed users (or primary users, PUs). Traditionally, SUs are oblivious toPUs, and therefore the acquisition of white spaces is not guaranteed. Hence, a SUmust vacate its channel whenever a PU reappearson it in an unpredictablemanner, whichmay affect the SUsโ network performance. Spectrum leasing has been proposed to tackle theaforementioned problem through negotiation between the PU and SU networks, which allows the SUs to acquire white spaces for aguaranteed period of time.Through spectrum leasing, the PUs and SUs enhance their network performances, and additionally PUsmaximize their respective monetary gains. Numerous research efforts have been made to investigate the CR, whereas the researchinto spectrum leasing remains at its infancy. In this paper, we present a comprehensive review on spectrum leasing schemes inCR networks by highlighting some pioneering approaches and discuss the gains, functionalities, characteristics, and challenges ofspectrum leasing schemes along with the performance enhancement in CR networks. Additionally, we discuss various open issuesin order to spark new interests in this research area.
1. Introduction
Cognitive Radio (CR) network, which is the next-generationwireless network, aims to improve the efficiency of spectrumutilization through dynamic spectrum access. There aretwo categories of users, namely, primary users (PUs) andsecondary users (SUs). Traditionally in CR networks, thePUs are the licensed users, and they have exclusive right touse their respective channels, while SUs are the unlicensedusers, and they use the underutilized channels (or whitespaces) opportunistically whenever PUs are not transmittingany packets. Hence, PUs are oblivious of the presence of SUs.There are twomain challenges associated with the traditionalCR Networks (CRNs) that adopt the opportunistic channelaccess approach. Firstly, the unpredictable PUsโ activitiesat any given time can significantly degrade SUsโ networkperformance (e.g., throughput and end-to-end delay) [1โ4].Secondly, channel sensing [1], which is one of the main func-tions in the traditional CRNs, may require SUs to exchangechannel sensing outcomes among themselves, and this incurs
high amount of communication overhead resulting in higherenergy consumption and packet latency [5]. In addition to thetraditional CRNs [2, 3], there have been research activitiesin the area of CR sensor networks [4]. CR sensor networksare the next-generation wireless sensor networks that exploitwhite spaces through dynamic spectrum access.
Spectrum leasing is a dynamic spectrum access techniquein which PUs and SUs form a partnership formutual benefits.In spectrum leasing, the SUs negotiate with PUs and acquiretheir white spaces [6], while the PUs lease their channels andreceive rewards in the form ofmonetary gain or network per-formance enhancement through packet forwarding by SUs[7].Hence, PUs are fully aware of the presence of SUs. Figure 1presents a taxonomy of spectrum leasing, which covers itsadvantages, functionalities, characteristics, and challenges.Further descriptions about the taxonomy are found in the restof this section, as well as Sections 2, 3, and 4, respectively.Generally speaking, with the use of spectrum leasing, PUsand SUs receive the following advantages represented by (A1)and (A2) (see Figure 1), respectively.
Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2014, Article ID 329235, 22 pageshttp://dx.doi.org/10.1155/2014/329235
http://dx.doi.org/10.1155/2014/329235
-
2 International Journal of Distributed Sensor Networks
(F1.1)Determination of the cost of white spaces
(F1.2)Determination
of PUโs and SUโs channel access time
(F1.3) Relay selection
(F1.4)
transmission
(F2.1)Collaborator
selection
(F2.2)Determination
of SUโs channel access
time
(F2.3)SUโs packet transmission
(H1)Enhancing PUโs and
SUโs network performance
(H2)Inferring the channel characteristics of SU
relay nodes (H3)
Continuous
white spaces beingleased to SUs by PUs
(F1)PUโs functions
(F2)SUโs functions
(C2)Intracooperative
mode
(C3)Intercooperative
mode
(C1)Networking
topology
A Gains
FFunctions
CCharacteristics
HChallenges
Spectrum leasing
(A1)PUโs gain
(A2)SUโs gain
(A1.2)Network
performanceenhancement
(A1.1)Monetary
(A2.1)Dedicated
channel access(C1.1)
Centralized
(C1.2)Distributed
(C2.1)Intraco-
operative
(C2.2)Nonintraco- operative
(C3.1)Interco-
operative
(C3.2)Noninterco- operative
(H1.1)Increasing the amount of white spaces being
leased by PUs
(H1.2)Selecting an optimal channel with white
spaces by SUs
monitoring of
PUโs packet
gain
Figure 1: Taxonomy of spectrum leasing in CRNs.
(A1) PUโs Gain
(A1.1) Monetary Gain. PUsmay lease its licensed chan-nels during idle periods for financial rewardor revenue. For instance, Jayaweera et al. [6]propose a PUโs utility function based on itsmonetary gain (e.g., the price set by PUs ofwhitespaces).
(A1.2) Network Performance Enhancement. The PUlinks may deteriorate due to shadowing andinterference. Through spectrum leasing, one ormore SUs form an alternative route and relayPUsโ traffic, and this enhances the PUsโ networkperformance, such as successful transmissionrate, throughput, end-to-end delay, and energyefficiency [8].
(A2.1) Dedicated Channel Access. The SUs access whitespaces allocated by PUs. Subsequently, thisenhances the SUsโ throughput performance.Since spectrum leasing enhances the through-put performance of PUs (A1.2), it reduces thetransmission time of PUs, therefore leavingmore white spaces and transmission opportuni-ties to SUs for dedicated access [9].
The advantages motivate PUs and SUs to participatein spectrum leasing. For instance, in [5], spectrum leasingmaximizes a weighted sum of PUsโ and SUsโ throughputperformance.
This paper provides an extensive survey on existingspectrum leasing schemes in CRNs. The purposes are toestablish a foundation and to spark new interests in thisresearch area covering new kinds of CR networks such asCR sensor networks [4]. Our contributions are as follows.Sections 2, 3, and 4 present the functionalities, character-istics, and challenges, respectively. Section 5 presents vari-ous spectrum leasing schemes in CRNs. Section 6 presentsperformance enhancement achieved by spectrum leasingschemes. Section 7 presents open issues. Finally, we presentconclusions.
2. Functionalities of Spectrum Leasing inCognitive Radio Networks
This section discusses the functionalities of PUs and SUsfor spectrum leasing in CRNs. Generally speaking, spectrumleasing is comprised of the following functionalities.
(F1) PUโs Function
(F1.1) Determination of the Cost of White Spaces. PUsdetermine the cost (e.g., monetary price) ofwhite spaces to be imposed on SUs.
(F1.2) Determination of PUsโ and SUsโ Channel AccessTime. PUs are the rightful owners of the licensedspectrum, and so the PU Base Station (BS)may determine suitable channel access time fortransmission opportunities for both PUs and
-
International Journal of Distributed Sensor Networks 3
SUs. For instance, in centralized networks, thePUhosts send their respective information (e.g.,idle time) to PU BS. Subsequently, the PU BSallocates transmission opportunities for PU andSUnetworks. In other words, the PUs determinethe amount of white spaces to be leased toSUs. The objective is to maximize the networkperformance (e.g., throughput) of PUs and SUs[10, 11].
(F1.3) Relay Selection. PUs select the SUs that providethe highest gain (e.g., PU-SU linkswith the best-known signal-to-noise ratio (SNR)) as relays inorder to maximize throughput performance.
(F1.4) PUsโ Packet Transmission. PUs transmit theirown packets to destination in order to enhancetheir network performance.
(F2) SUโs Function
(F2.1) Collaborator Selection. SUs select the suitablePUs to collaborate with. This covers the eval-uation of the gain (e.g., the amount of whitespaces with sufficient SNR) and cost (resourcesrequired to relay PUsโ traffics, such as energyconsumption).
(F2.2) Determination of SUโs Channel Access Time. SUsdetermine the amounts of white spaces, whichincrease with channel access time, to requestfrom PUs based on the cost imposed by thePUs. For instance, in a Time-Division MultipleAccess (TDMA) system, SUs must determinethe optimal time duration in which they mustinvolve as relay to transmit PU packets and totransmit their own packets [8].
(F2.3) SUsโ Packet Transmission. SUs transmit pack-ets, and this involves two phases. Firstly, theSUs relay PU packets. To ensure continuouscollaboration with PUs, the SUs must achieve acertain level of network performance enhance-ment while relaying the PUsโ packets. Secondly,the SUs transmit their own packets. Spatialreuse is possible, and so the SUs must mini-mize interference among themselves [12]. Forinstance, in centralized networks, SU BS andhostsmay serve as relays to transmit PUpackets,and subsequently the SU BS allocates the whitespaces offered by PUs to its SU hosts fairly [10,13].
Spectrum leasing involves several steps and messagehandshaking, and we describe a general procedure inFigure 2. Consider two centralized PU and SU networks,which are collocated in the same area. Several PUhosts (or SUhosts) are associated with a PU BS (or SU BS).The procedureis as follows.
Step 1. ThePUhosts send information on their respective idleperiods (or white spaces) to PU BS.
J PU hosts PU BS SU BS K SU hostsStep 1
Step 2
Step 5
Step 3
Step 4
Step 6Step 7
Step 8Step 9
Step 10 Step 10
Figure 2: A general spectrum leasing procedure.
Step 2. The PU BS determines the cost (F1.1) and duration(F1.2) of white spaces. There are ๐ฝ PU hosts to be leased toSUs.
Step 3. The PU BS sends the cooperation information (e.g.,the cost and duration, as well as SNR of the white spaces) toSU BS.
Step 4. The SU BS broadcasts the cooperation information toits SU hosts.
Step 5. The SU hosts determine the optimum transmissionand relaying strategies (i.e., (F2.2) and (F2.3)) using thecooperation information. If auction mechanism is applied,the SU hosts may determine bid values.
Step 6. The SU hosts send their respective decisions (e.g.,strategies and bid values) to SU BS.
Step 7. The SUBS decides to accept the lease or not and selectthe suitable PUs to collaborate with (F2.1).
Step 8. The SU BS sends its decisions to PU BS.
Step 9. The PU BS decides to lease or not and select thesuitable SUs as relays (F1.3).
Step 10. Finally, based on the lease, the PU BS transmits itspackets (F1.4) directly through a single hop, or indirectlythrough SU relay nodes, to the PU BSโs destination node.TheSU BSmay divide the white spaces and assign the access timeof each white space to each SU hosts (F2.2).The SUs transmitpackets accordingly (F2.3).
3. Characteristics of Spectrum Leasing inCognitive Radio Networks
This section discusses the characteristics of spectrum leasingin CRNs. There are three characteristics as follows.
-
4 International Journal of Distributed Sensor Networks
(C1) Network Topology: Centralized (C1.1) and Distributed(C1.2). In centralized networks (C1.1), a central entitywhich is usually referred as Base Station (BS) isresponsible for communications between PU andSU networks [14], whereas, in distributed networks(C1.2), BS does not exist, and PUs and SUs sharetheir information through a common control channel[14]. For instance, in [5], a centralized network (C1.1)topology is used, in which PUs are leaders andresponsible to select the most appropriate SU forcooperative communication and hence the SUs arefollowers.
(C2) Intracooperative Mode: Intracooperative (C2.1) andNonintracooperative (C2.2). The PUs may cooper-ate among themselves through an intra-cooperativeapproach in order to achieve the advantages (A1.1)-(A1.2) and (A2.1). Likewise, the SUs may adoptthe same approach. In Figure 3, the intracooperative(C2.1) mode is shown in (a) and (c) and fromthe SUโs perspective, the SUs may cooperate amongthemselves and jointly improve network-wide perfor-mance such as throughput performance, as well asto reduce the monetary and nonmonetary spectrumleasing costs imposed by PUs. In other words, a groupof SUs may lease a channel and subsequently sharethe channel among themselves in order to reducespectrum leasing costs. In Figure 3, the nonintraco-operative (C2.2) mode is shown in (b) and (d) andfrom the PUโs perspective, each PUmay competewitheach other to lease their respective white spaces andhence each PU may set a competitive price based onthe demand of channel access from SUs. From theSUโs perspective, the SUs may also compete with eachother to acquire the white spaces through auction-based mechanisms [15]. For instance, in [5], each SUoptimizes its power allocation in the transmission ofPU packets in order to fulfill the packet transmissionrequirements of PUs. This helps each SU to remaincompetitive in order to obtain white spaces in theupcoming auctions and this has been shown toimprove SU throughput performance.
(C3) Intercooperative Mode: Intercooperative (C3.1) andNonintercooperative (C3.2). PUs and SUsmay cooper-ate with each other in order to achieve the advantages(A1.1)-(A1.2) and (A2.1). In Figure 3, the intercoop-erative (C3.1) mode is shown in (c) and (d) andthe PUs and SUs cooperate with each other, and sothis improves the overall network-wide performancesuch as throughput performance. In Figure 3, thenonintercooperative (C3.2) mode is shown in (a) and(b) and the PUs and SUs are referred to as selfishusers, and they do not cooperate with each other. Forinstance, in [16], the PUs attempt to maximize theirprofit or reward out of the white spaces, while the SUsattempt to reduce their cost.
4. Challenges of Spectrum Leasing inCognitive Radio Networks
This section discusses the challenges associated with spec-trum leasing in CRNs. There are three challenges as follows.
(H1) Increasing the Monetary Gain of PUs. PUs aim toincrease their monetary gain through spectrum leas-ing. This encourages the PUs to participate in spec-trum leasing by increasing the amount of white spacesavailable to SUs. Subsequently, this increases PUsโand SUsโ throughput performance [10]. The PUs maycooperate or compete with each other to lease theirwhite spaces. As an example, in [10], PUs cooperatewith each other, and linear programming is appliedto set the optimal price of the white spaces in order toincrease their monetary gain. As another example, in[16], PUs compete with each other, and game theoryis applied to set the optimal price of the white spacesin order to increase their monetary gain.
(H2) Selecting an Optimal Channel with White Spaces bySUs. SUs aim to access the licensed channel or whitespaces in order to increase their network performance(e.g., throughput). So, this encourages the SUs toparticipate in spectrum leasing and subsequentlyincreases PUsโ and SUsโ network performance [17].However, the access to white spaces by SUs requiresmonetary cost, and so there is a need to find anoptimal channel that provides the best possible net-work performance while incurring the least possiblecost. For instance, Cao et al. [5] propose a spectrumsharing policy in which white spaces are being leasedto SUs, in order to increase the network capacity of SUnetwork.
( H3) Scheduling theChannel Access of PUs and SUs.ThePUsschedule the time for the transmissions of PUsโ andSUsโ packets in order to enhance their respective QoSperformance (e.g., throughput). The time allocationfor SUsโ links must be sufficiently higher comparedto that of PUsโ links in order to reap the benefitsof spectrum leasing [9]. Otherwise, the queue sizeat SU relay nodes may grow and eventually becomeinsufficient to accommodate new packets from bothPUs and SUs leading to packet loss. However, thewhite spaces being leased to SUsmay not be sufficientto cater for PUsโ and SUsโ packets. For instance,Huanget al. [18] propose a coalition game to allocate asuitable fraction of channel access time among PUsand SUs, so that SUs transmit PUsโ packets as well astheir own packets.
(H4) Continuous Monitoring of White Spaces Being Leasedto SUs by PUs. Upon negotiation, the PUs and SUsmay need to monitor the white spaces (e.g., amountand channel quality) and the Quality of Service (QoS)of packet transmission in order to make sure that
-
International Journal of Distributed Sensor Networks 5
SU plane
PU plane
(b) Non-intracooper-
ative and non-
intercooper-ative
(a) Intra-cooperative
and non-intercooper-
ative
(c) Intra-cooperativeand inter-
cooperative
(d) Non-intracooper-
ative and intercooperative
Primary user (PU)Secondary user (SU)
Figure 3: Mode of cooperation between PU and SU network.
each party follows suit. However, the continuousmonitoring of SUs requires more intelligence to beincorporated into the PU network. For instance, in[15], PUs additionally acts as an online auctioneer tomonitor the SUs activities. Likewise, in [19], PUs needto ensure that the interference caused by SUs is lessthan the acceptable interference level. Furthermore,SUs also need tomonitor the SUsโ signal level in orderto reduce interference with PUs [20].
5. Spectrum Leasing Schemes inCognitive Radio Networks
This section presents existing work on spectrum leasingschemes in CRNs. The schemes are categorized with respectto the challenges (see Section 4) and on the basis of adoptedapproaches (e.g., game theoretic approaches and nongametheoretic approaches) to address the challenges. The gametheoretic approaches, such as Stackelberg game [21], are usedto achieve the equilibrium state (e.g., Nash equilibrium [22])and it involves PUs and SUs as players of the game. Examplesof the nongame theoretic approaches are reinforcementlearning [23] and convex optimization [24]. Table 1 presentsthe gains, functions, and characteristics of the spectrumleasing schemes. The performance enhancement achieved byeach scheme is shown in Table 2 (see Section 6).
5.1. Increasing the Monetary Gain of PUs. There are six spec-trum leasing schemes that focus on addressing the challengeof increasing the monetary gain of PUs that motivates thePUs to participate in spectrum leasing. These schemes havebeen shown to increase the monetary gain of PUs, as well asto enhance PUsโ or SUsโ QoS performance (e.g., throughput).
5.1.1. SchemesThat Use GameTheoretic Approaches. Alptekinand Bener [16] propose one PU F(1) and one SU F(2)functionalities, namely, determination of the cost of whitespaces (F1.1), as well as collaborator selection (F2.1) inorder to increase PUsโ monetary gain (A1.1) and to providededicated channel access to SUs (A2.1) in centralized (C1.1)SU networks. The purpose is to maximize the PUsโ profitas seller in terms of its utility function ๐๐, which helps tosatisfy theQoS parameters (e.g., jitter) of SUs as buyers, in thepresence of ๐ฝ PUs and๐พ SUs.The functionalities aremodeledand solved using game theory and the Nash equilibrium in anonintracooperative (C2.2) mode and non-intercooperative(C3.2) mode, respectively. The cumulative utility function of๐ฝ PUs is defined as
๐๐ =
๐ฝ
โ
๐=1
๐๐๐ โ ๐๐๐ โ ๐๐๐ โ ๐๐๐, (1)
where ๐ = {1, 2, . . . , ๐ฝ}, ๐๐๐ is the price that PU ๐ imposes onSU ๐, ๐๐๐ is the demand factor (i.e., SU ๐โs expectation onQoSrequirement including jitter and throughput from PU ๐), and๐๐๐ is the cost associatedwith the channel leased to SU ๐whichmust be paid by PU ๐ to regulatory authorities (e.g., FederalCommunications Commission (FCC)).The PU ๐ determinesthe cost of white spaces (F1.1) and on that basis selects SU ๐if the difference between price ๐๐๐ and cost ๐๐๐ in PU utilityfunction is positive, which indicates a monetary gain for PU๐. The SU ๐ selects a PU collaborator (F2.1) to achieve its QoSlevel as indicated in the demand factor ๐๐๐ while paying thePU ๐ at the specified price ๐๐๐. It has been shown that PUs aremore likely to fulfill the SUsโ QoS demandwith the incrementof price ๐๐๐ (i.e., monetary gain).
Lin and Fang [25] propose one PU F(1) and one SU F(2)functionalities, namely, determination of the cost of whitespaces (F1.1), as well as SUsโ packet transmission (F2.3) inorder to increase PUsโ monetary gain (A1.1) and to provide
-
6 International Journal of Distributed Sensor Networks
Table1:Gains,fun
ctions,and
characteris
ticso
fthe
spectrum
leasingschemes.
Challeng
esRe
ferences
(A)G
ains
(F)F
unctions
(C)C
haracteristics
(A1)PU
sGains
(A2)
SUs
Gains
(F1)PU
sFun
ctions
(F2)
SUsF
unctions
(C1)Networking
topo
logy
(C2)
Intra-op
erativem
ode
(C3)
Inter-op
erativem
ode
(A1.1)
Mon
e-tary
gain
(A1.2
)Network
perfo
rmance
enhancem
ent
(A2.1)
Dedi-
cated
chan-
nel
access
(F1.1)
Deter-
mina-
tionof
the
costof
white
spaces
(F1.2
)Deter-
mina-
tionof
PUsโ
and
SUsโ
chan-
nel
access
time
(F1.3
)Re
lay
selection
(F1.4
)PUsโ
Packet
transm
ission
(F2.1)Col-
labo
rator
selection
(F2.2)
Deter-
mina-
tionof
SUโs
chan-
nel
access
time
(F2.3)
SUsโ
Packet
transm
ission
(C1.1)
Centralized
(C1.2
)Distrib
uted
(C2.1)Intra-
coop
erative
(C2.2)
Non
-intracoo
perativ
e(C
3.1)Inter-
coop
erative
(C3.2)
Non
-intercoo
perativ
e
(H1)
Increasin
gthe
mon
etary
gain
ofPU
s
(Alptekinand
Bener,2009)
[16]
รร
รร
รร
ร
(Lin
andFang
,2008)[25]
รร
รร
รร
ร
(Yietal.,2010)
[10]
รร
รร
รร
รร
รร
(Kim
andSh
in,
2009)[26]ร
รร
รร
(Son
gandLin,
2009)[13]
รร
รร
รร
ร
(Wuetal.,
2008)[7]
รร
รร
รร
(H2)
Selectingan
optim
alchannel
with
white
spaces
bySU
s
(Chanetal.,
2011)[17]
รร
รร
รร
รร
(Vazqu
ez-Vilar
etal.,2010)[20]
รร
รร
รร
รร
(Cao
etal.,
2012)[5]
รร
รร
รร
ร
(Jayaweera
etal.,2011)
(centralized
approach)[8]
รร
รร
รร
รร
(Jayaweera
etal.,2011)
(distrib
uted
approach)[8]
รร
รร
รร
รร
(Muraw
skiand
Ekici,2011)[27]
รร
รร
รร
รร
(Torou
jeni
etal.,2011)[28]
รร
รร
รร
รร
-
International Journal of Distributed Sensor Networks 7
Table1:Con
tinued.
Challeng
esRe
ferences
(A)G
ains
(F)F
unctions
(C)C
haracteristics
(A1)PU
sGains
(A2)
SUs
Gains
(F1)PU
sFun
ctions
(F2)
SUsF
unctions
(C1)Networking
topo
logy
(C2)
Intra-op
erativem
ode
(C3)
Inter-op
erativem
ode
(A1.1)
Mon
e-tary
gain
(A1.2
)Network
perfo
rmance
enhancem
ent
(A2.1)
Dedi-
cated
chan-
nel
access
(F1.1)
Deter-
mina-
tionof
the
costof
white
spaces
(F1.2
)Deter-
mina-
tionof
PUsโ
and
SUsโ
chan-
nel
access
time
(F1.3
)Re
lay
selection
(F1.4
)PUsโ
Packet
transm
ission
(F2.1)Col-
labo
rator
selection
(F2.2)
Deter-
mina-
tionof
SUโs
chan-
nel
access
time
(F2.3)
SUsโ
Packet
transm
ission
(C1.1)
Centralized
(C1.2
)Distrib
uted
(C2.1)Intra-
coop
erative
(C2.2)
Non
-intracoo
perativ
e(C
3.1)Inter-
coop
erative
(C3.2)
Non
-intercoo
perativ
e
(H3)
Schedu
ling
thec
hann
elaccessof
PUsa
ndSU
s
(Chenetal.,
2011)[29]
รร
รร
รร
รร
(Huang
etal.,
2011)[18]
รร
รร
รร
รร
ร
(Wangetal.,
2010)[30]
รร
รร
รร
รร
รร
(Stano
jevetal.,
2008)[31]
รร
รร
รร
รร
(Wangetal.,
2010)[32]
รร
รร
รร
รร
(Zhang
etal.,
2010)[33]
รร
รร
รร
รร
(Zhu
etal.,
2012)[34]
รร
รร
รร
(Asadu
zzam
anetal.,2011)[35]
รร
รร
รร
รร
ร
(Khalil
etal.,
2011)[36]
รร
รร
รร
รร
ร
(Zho
uetal.,
2011)[11]
รร
รร
รร
รร
(H4)
Con
tinuo
usmon
itorin
gof
white
spaces
being
leased
toSU
sbyPU
s
(Jayaweera
etal.,2010)[6]
รร
รร
รร
(Jayaweera
and
Li,200
9)[19
]ร
รร
รร
ร
(Hakim
etal.,
2010)[37]
รร
รร
รร
(Sod
agarietal.,
2011)[15]
รร
รร
รร
ร
-
8 International Journal of Distributed Sensor Networks
Table2:Perfo
rmance
enhancem
entsachieved
bythes
pectrum
leasingschemes.
Challenges
References
Perfo
rmance
Enhancem
ent
(P1)Lo
wer
outage
prob
ability
(P2)
Higher
outage
capacity
(P3)
Bette
rQoS
level
(P4)
Higher
energy
efficiency
(P5)
Higher
mon
etarygain
(P6)
Balanced
tradeoffbetween
spectrum
cost
andmon
etary
gain
(P7)
Balanced
tradeoffbetween
PUsโandSU
sโchannelaccess
time
(P8)
Bette
rsecuritylevel
(P9)
Lower
PUsโ
interfe
rencelevel
(H1)Increasin
gthem
onetary
gain
ofPU
s
(Alptekinand
Bener,2009)[16]
รร
ร
(Lin
andFang
,2008)[25]
ร
(Yietal.,2010)
[10]
รร
(Kim
andSh
in,
2009)[26]
ร
(Son
gandLin,
2009)[13]
รร
(Wuetal.,2008)
[7]
รร
(H2)
Selectingan
optim
alchannel
with
whitespaces
bySU
s
(Chanetal.,2011)
[17]
รร
รร
(Vazqu
ez-Vilare
tal.,2010)[20]
รร
(Cao
etal.,2012)
[5]
ร
(Jayaweera
etal.,
2011)[8]
รร
ร
(Muraw
skiand
Ekici,2011)[27]
ร
(Torou
jeni
etal.,
2011)[28]
รร
(H3)
Schedu
ling
thec
hann
elaccessof
PUsa
ndSU
s
(Chenetal.,2011)
[29]
รร
(Huang
etal.,
2011)[18]
รร
(Wangetal.,
2010)[30]
รร
(Stano
jevetal.,
2008)[31]
รร
(Wangetal.,
2010)[32]
รร
ร
(Zhang
etal.,
2010)[33]
รร
ร
(Zhu
etal.,2012)
[34]
ร
(Asadu
zzam
anet
al.,2011)[35]
รร
ร
(Khalil
etal.,
2011)[36]
ร
(Zho
uetal.,2011)
[11]
ร
-
International Journal of Distributed Sensor Networks 9
Table2:Con
tinued.
Challenges
References
Perfo
rmance
Enhancem
ent
(P1)Lo
wer
outage
prob
ability
(P2)
Higher
outage
capacity
(P3)
Bette
rQoS
level
(P4)
Higher
energy
efficiency
(P5)
Higher
mon
etarygain
(P6)
Balanced
tradeoffbetween
spectrum
cost
andmon
etary
gain
(P7)
Balanced
tradeoffbetween
PUsโandSU
sโchannelaccess
time
(P8)
Bette
rsecuritylevel
(P9)
Lower
PUsโ
interfe
rencelevel
(H4)
Con
tinuo
usmon
itorin
gof
whitespaces
beingleased
toSU
sbyPU
s
(Jayaweera
etal.,
2010)[6]
รร
ร
(Jayaweera
and
Li,2009)
[19]
รร
(Hakim
etal.,
2010)[37]
รร
ร
(Sod
agarietal.,
2011)[15]
รร
ร
-
10 International Journal of Distributed Sensor Networks
dedicated channel access to SUs (A2.1) in distributed (C1.2)SU networks. The purpose is to maximize the PUsโ and SUsโutility functions ๐๐ and ๐๐ , respectively, while taking intoaccount the mutual benefits of PUs (or sellers) and SUs(or buyers). The functionalities are modeled in the presenceof ๐ฝ PUs and ๐พ SUs and solved using a two-level gamethat is split into PU-level game and SU-level game in anon-intracooperative (C2.2) mode and non-intercooperative(C3.2) mode, respectively. In this hierarchy of games, PUscompete with each other to lease their spectrum to SUs byadjusting their price of white spaces in order to maximizetheir respective utility functions; each SU attempts to leasea certain amount of white spaces from PU that provides theoptimal quality white spaces. The PUsโ ๐ โ ๐ฝ utility functionis defined as
๐๐,๐ =
๐พ
โ
๐=1
๐ต๐๐ {๐๐๐ โ ๐๐} , (2)
where ๐ต๐๐ is the bandwidth (or white spaces) that PU ๐allocates to SU ๐, ๐๐๐ is the price that PU ๐ imposes on SU๐, and ๐๐ is the cost associated with the channel leased to SU๐ which must be paid by PU ๐ to regulatory authorities (e.g.,FCC). A PU decides to play a game if price ๐๐๐ is greater thancost ๐๐ of the leased channel (F1.1).The SUsโ utility function isdefined as
๐๐ = {log2(1 + ๐ ๐ ,๐) Case-I
log2(1 + ๐
MAX๐ ,๐
) Case-II,(3)
Where ๐ ๐ ,๐ and ๐ MAX๐ ,๐
are the transmission rate, as well as itsmaximum value, of SU ๐. In Case-I, PU allocates lesser whitespaces to SU ๐ than it demands, while in Case-II, PU allocateshigher bandwidth to SU ๐ than it demands. The higher theamount of white spaces provided by PU to SU, the higher isthe transmission rate of SU ๐ (F2.3). It has been revealed thatthe number of SUs increases with the price of white spacesthat PUs impose to SU.
Yi et al. [10] propose three PU F(1) and two SU F(2)functionalities, namely, determination of the cost of whitespaces (F1.1), relay selection (F1.3) and PUsโ packet trans-mission (F1.4), as well as determination of SUโs channelaccess time (F2.2), and SUsโ packet transmission (F2.3) inorder to increase PUsโ monetary gain (A1.1) and to providededicated channel access to SUs (A2.1) in centralized (C1.1)SU networks. The purpose is to maximize the PUsโ and SUsโnetwork utility functions, ๐๐ and ๐๐ , respectively. The PUsand SUs are rational and selfish in nature. The functionalitiesare modeled and solved using Stackelberg game, in whichthe PU is the leader and the SU is the follower in an intra-cooperative (C2.1) mode and inter-cooperative (C3.1) mode,respectively. The Nash equilibriummaximizes both PUsโ andSUsโ utility functions, ๐๐ and ๐๐ . The PUsโ utility function isdefined as
๐๐ = ๐ข๐ + ๐ข๐, (4)
where ๐ข๐ and ๐ข๐ are revenues. Revenue ๐ข๐ is dependent onthe ratio of total PUsโ packet transmissions, which include
successful packet transmissions through direct transmissions(i.e., fromPUhost to PUBS) and relaying through SUs to totaltraffic demand of all PU hosts. Revenue ๐ข๐ is derived from thewhite spaces being leased to SUs. The SUsโ utility function isdefined as
๐๐ = ๐ข๐ โ ๐ข๐, (5)
where ๐ข๐ is derived from the total SUsโ packet transmissionsfrom all SU hosts. Both๐๐ and๐๐ take into account the SNRof the channels. There are two main steps in the Stackelberggame. Firstly, the PU BS (or leader) determines its strategycomprised of a set of potential SU relaying nodes (F1.3) andthe costs (i.e,. the price of white spaces per unit access time)to be imposed on SUs (F1.1) and sends the PUsโ strategy to SUBS. Using the fixed leaderโs strategy, the SU BS (or follower)determines the amount of white spaces to request from PUsbased on the costs (F2.1); hence, higher cost may reduce theamount of white spaces to request. The SU BS sends the SUstrategy to PU BS. Secondly, using a fixed followerโs strategy,the PU BS selects relay nodes and finalizes the costs and startpacket transmissions (F1.4). Similarly, the SU BS allocates theleased white spaces amongst SUs for their respective packettransmission (F2.3). The spectrum leasing scheme has beenshown to increase PUsโ and SUsโ utility functions,๐๐ and ๐๐ ,as well as to increase the amount of white spaces being leased.This scheme also decreases the price of white spaces per unitaccess time.
5.1.2. Schemes That Use Nongame Theoretic Approaches. Kimand Shin [26] propose one PU F(1) function, namely, deter-mination of the cost of white spaces (F1.1) in order to increasePUsโ monetary gain (A1.1) in distributed (C1.2) SU networks.The purpose is to maximize the PUsโ profit by controllingthe SUsโ admission and eviction strategies. The admissionstrategy allows the SUs to utilize PUsโ channels on the basis ofthe requested amount of white spaces, which basically yieldsthe PUsโ profit. Hence, if SUs demands a small amount ofwhite spaces, then PUs may reject their admissions due tothe less monetary gain. This is because the PUs are interestedto allocate white spaces to SUs that request larger amountof white spaces in order to maximize their monetary gain,whereas the eviction strategy is set so that SUs evacuate thechannel immediately if PUsโ activities reappear. The functionis modeled and solved using semi-Markov decision processand linear programming in a non-intracooperative (C2.2)mode and non-intercooperative (C3.2) mode, respectively.The PUs allocates their underutilized channels to a group of๐ SUs. The expected revenue of PUs is defined as
๐๐ = โ
๐โ๐พ
๐๐๐๐๐พ, (6)
where ๐๐ is the price that ๐ SUs pay to PU in return of itsQoS demand ๐๐, while ๐พ is the number of SUs in the group.Higher PUsโ revenue, which comes with higher price of whitespaces (F1.1), indicates higher QoS demand from SUs. It hasbeen shown that PUsโ revenue increases with the amount ofwhite spaces. However, the PUsโ revenue decreases when thewhite spaces become oversupplied.
-
International Journal of Distributed Sensor Networks 11
Song and Lin [13] propose one PU F(1) and one SUfunctionalities, namely, determination of the cost of whitespaces (F1.1), as well as SUsโ packet transmission (F2.3) inorder to increase PUsโ monetary gain (A1.1) and to providededicated channel access to SUs (A2.1) in distributed (C1.2)SU networks. The purpose is to maximize the profit ofPUs while allocating the white spaces to SUs. The functionis modeled and solved using auction-based property-rightsmodel mechanism in a nonintracooperative (C2.2) modeand nonintercooperative (C3.2) mode, respectively. In aproperty-rights model, SUs are divided into non-overlappinggroups and a leader is elected from each group. The auctionmechanism is divided into time windows, and each windowis further divided into two phases, namely, auction andcommunication. There are four main purposes in regard tothe auctionmechanism. Firstly, it maximizes the overall spec-trum utilization. Secondly, it maximizes the number of SUwinners (or SU groups that gain a channel). Thirdly, it fulfillsthe bandwidth requirement of SUs. Note that the channelsare heterogeneous and each channel has different amount ofbandwidth (or white spaces). Fourthly, it maximizes the PUsโrevenue. In a round of bidding, each SU leader determines abid value based on hungry degree, which takes into accountthe amount of white spaces required by its group of SUs.During the auction phase, the PU auctions off ๐ channelswith white spaces to ๐ SU leaders in two phases. Each SUleader uses an auction phase, which is based on its bandwidthrequirement, to bid for a leasing channel. Higher value ofhungry degree leads to higher bid value. During the firstphase of auction, in order to meet the first, second, and thirdpurposes, the PU grants channels to as many groups of SUsas possible to meet their respective minimum requirementon the amount of white spaces. During the second phaseof auction, in order to achieve the fourth purpose, the PUallocates the channels with white spaces to SU leaders thatoffer higher bid values (F1.1). During the communicationphase (F2.3), the SUs transmit packets and the PU keeps trackof availablewhite spaces for auctions in the next timewindow.The spectrum leasing scheme has been shown to increasethroughput performance in regard to vacant channels.
Wu et al. [7] propose one PU F(1) function, namely,determination of the cost of white spaces (F1.1) in order toincrease PUsโ monetary gain (A1.1) and to provide dedicatedchannel access to SUs (A2.1) in centralized (C1.1) SU net-works. The purpose is to maximize the PU monetary gainand SUs network utility function ๐๐ , while preventing thecollusive SUs to access the PUsโ white spaces. The collusiveSUs form a coalition and deliberately decrease the priceof white spaces offered by PUs. The function is modeledand solved using binary linear programming and convexoptimization in an intra-cooperative (C2.1) mode and non-intercooperative (C3.2) mode, respectively. Binary linearprogramming is a mathematical method to determine theoptimal results that comprises binary integers (i.e., 0 and1). The PU sells white spaces to ๐พ SUs with the assistancefrom a third-party spectrum broker. Upon the reception ofbid values ๐๐ = {๐1, ๐2, . . . , ๐๐พ} from ๐พ SU, the spectrumbroker announces the winning SUs by defining the channelallocation ๐ฅ๐ = {๐ฅ1, ๐ฅ2, . . . , ๐ฅ๐พ} and the associated price
๐ = {๐1, ๐2, . . . , ๐๐พ} for ๐พ SUs. For the winning SUs, thechannel allocation ๐ฅ๐ is set to one (i.e., ๐ฅ๐ = 1), whichindicates that the channel has been allocated to winner SU๐. The gain of each winning SUs is ๐๐, which lead to anefficient channel allocation which is used to compute theutility function of SUs; that is,๐๐ = โ
๐พ
๐=1๐๐ โ ๐ฅ๐. Higher values
of ๐๐ indicate higher number of winning SUs in the auctionfor white spaces. It has been shown that, as the number ofwinning SUs increases, the price of the white spaces imposedby the PUs as sellers also increases.
5.2. Selecting an Optimal Channel with White Spaces by SUs.There are six spectrum leasing schemes that focus onmotivat-ing the SUs to participate in spectrum leasing by increasingthe amount of white spaces for SUs. These schemes havebeen shown to enhance PUsโ or SUsโ QoS performance (e.g.,throughput).
5.2.1. Schemes That Use Game Theoretic Approaches. Chanet al. [17] propose two PU F(1) and one SU F(2) func-tionalities, namely, determination of PUsโ and SUsโ channelaccess time (F1.2) and relay selection (F1.3), as well as SUsโpacket transmission (F2.3) in order to enhance the networkperformance of PUs (A1.2) and to provide dedicated channelaccess to SUs (A2.1) in centralized (C1.1) SU networks. Thepurpose is to maximize the spectrum utilization of PUand SU networks by adopting the cooperation strategiesin between of ๐ฝ PUs and ๐พ SUs in the form of PUs andSUs utility functions, ๐๐ and ๐๐ , respectively. In separatecooperation, PU ๐ and SU ๐ form a one-to-one collaborativerelationship with each other, while in grand cooperation, PUsand SUs form a coalition that comprises of many one-to-oneand one-to-many collaborative relationships with each other.The functionalities are modeled and solved using canonicalcoalition game theoretic framework and convex optimizationproblem in a non-intracooperative (C2.2) mode and inter-cooperative (C3.1)mode, respectively.ThePUutility functionis defined as
๐๐ = ๐ข (๐ ๐) + ๐๐ ๐๐โ ๐ฟ (๐๐) , (7)
where ๐ข(โ ) and ๐ฟ(โ ) are concave function that maps the PUachievable transmission rate ๐ ๐ as utility gain and PU cost ๐๐as utility loss, while ๐๐ ๐๐ is the price of white spaces that PU ๐imposes on SUs. The SU utility function is defined as
๐๐ = ๐ (๐ ๐ ) + ๐๐ ๐๐, (8)
where ๐(โ ) is (โ ) concave function that projects SU achievablerate ๐ ๐ as revenue and ๐๐ ๐๐ is the price that PUs imposes onSU ๐ in order to lease its channel. It has been shown that thegrand cooperation strategy produces higher optimal utilityvalue than individualsโ cooperation.
Vazquez-Vilar et al. [20] propose two PU F(1) and oneSU F(2) functionalities, namely, relay selection (F1.3), andPUsโ packet transmission (F1.4), as well as SUsโ packet trans-mission (F2.3) in order to enhance the network performanceof PUs (A1.2) and to provide dedicated channel access toSUs (A2.1) in centralized (C1.1) SU networks. The purpose
-
12 International Journal of Distributed Sensor Networks
is to maximize the PUsโ and SUsโ utility functions ๐๐ and๐๐ in the presence of a PU communication node pair inorder to minimize the SUsโ interference to PUs by reducingtheir power consumption.The PU determines the maximumallowable interference that PU can tolerate from SUs ๐ผ๐max,while the SUs aim to reduce their transmission power in orderto fulfill the requirement ๐ผ๐max. The function is modeled andsolved using Stackelberg game in an intracooperative (C2.1)mode and intercooperative (C3.1) mode, respectively. In thisscheme, the PU is the leader and the SU is the follower. Tofoster collaboration with SUs, the PU maximizes its utilityfunction, and it is defined as
๐๐ = ๐ข๐ (๐ผ๐
๐, ฮ๐๐
๐) , (9)
where ๐ข๐ increases with the increment of interference fromSU ๐ (or ๐ผ๐
๐โค ๐ผ๐
max) and decreases with the increment of PUsโtransmission powerฮ๐๐๐ . To foster collaborationwith PU, theSU maximizes its utility function, and it is defined as
๐๐ = ๐ข๐ (๐ ๐
๐, ๐ผ๐
๐) , (10)
where ๐ข๐ increases with the increment of the SU transmissionrate๐ ๐
๐พand decreases with interference fromSU ๐ผ๐
๐. Note that,
๐ ๐
๐(๐๐) and ๐ผ
๐
๐(๐๐) increase with the SU transmission power
๐๐. Maximizing๐๐ helps tomaximize the SU ๐โs power vector๐๐ผ๐max๐พ
= argmax๐{๐ข๐ (๐ ๐
๐, ๐ผ๐
๐)}. This has led to computing the
overall utility function of PUs and SUs on the basis of ๐ผ๐max.ThePU selects a SU relay node ๐ โ ๐พ (F1.3) that has the lowesttransmission power for transmission of PU packets (F1.4),as well as SUsโ packets (F2.3) among the other SUs. It hasbeen shown that the proposed scheme achieves higher utilityfunction for both PUs and SUs compared to the traditionalscheme.
5.2.2. SchemesThat Uses NongameTheoretic Approaches. Caoet al. [5] propose twoPUF(1) and one SUF(2) functionalities,namely, relay selection (F1.3) and PUsโ packet transmission(F1.4), as well as SUsโ packet transmission (F2.3) in orderto enhance the network performance of PUs (A1.2) incentralized (C1.1) SU networks. The purpose is to maximizethe spectrum utilization of PU and SU networks, wherethe PU and SU BSs operate in an intracooperative (C2.1)mode and intercooperative (C3.1) mode, respectively.The PUsource node ๐ selects the best available SU relay node ๐, andestablishes communication with the PU destination node ๐.The SU relay is used to transmit PU and SU packets usinga quadrature modulation scheme, which depends on twofactors, namely, power allocation factor 0 โค ๐น๐
๐๐,๐โค 1 and
weight factor 0 โค ๐ค๐๐๐,๐
โค 1. The power allocation factordetermines the transmission of packets through SU relaynode. Note that the SU relay node transmits PU packets onlyif ๐น๐๐๐,๐
= 1, the SU packets only if ๐น๐๐๐,๐
= 0, and both PUsโ andSUsโ packets if 0 < ๐น๐
๐๐,๐< 1, whereas the weight factor deter-
mines the respective throughputs of PU and SU network,respectively. The selected SU relay node ๐ transmits PU andSU packets simultaneously using transmission power ๐๐
๐๐,๐
in two orthogonal channels (i.e., in-phase and quadraturechannels) exploited using a quadraturemodulation approach.The SU relay node relays PU packets using transmissionpower ๐น๐
๐๐,๐โ ๐๐
๐๐,๐using in-phase channel and sends SU packets
using transmission power (1 โ ๐น๐๐๐,๐) โ ๐๐
๐๐,๐in quadrature
channel. The throughput of PUs and SUs is represented bya weighted sum throughput ๐๐, which is defined as
๐๐ = (1 โ ๐ค๐
๐๐,๐) โ ๐๐ + ๐ค
๐
๐๐,๐โ ๐๐ , (11)
where ๐๐ and ๐๐ represent PUsโ and SUsโ throughput, respec-tively. Note that ๐๐ = ๐๐ if ๐ค
๐
๐๐,๐= 0 and ๐๐ = ๐๐ if ๐ค
๐
๐๐,๐= 1,
while๐๐ and๐๐ achieve a balance if๐ค๐
๐๐,๐= 1/2. A primal-dual
subgradient algorithm, including Lagrange multipliers andthe Karush-Kuhn-Tucker conditions, is used to optimize ๐น๐
๐๐,๐
and ๐๐ ๐๐,๐
in order to optimize the weighted sum throughput๐๐. The PU selects a SU only if it improves throughputperformance (F1.3), while the selected SU transmits the PUand SU packets simultaneously (F1.4), or the SU packetsonly (F2.3) when the PU is inactive. Through achievingbalanced throughputs ๐๐ and ๐๐ , the scheme has been shownto maximize ๐๐, and this is due to the dependence of ๐๐ and๐๐ on power allocation factor ๐น
๐
๐๐,๐and weight factor ๐ค๐
๐๐,๐.
Jayaweera et al. [8] propose two PU F(1) and one SU F(2)functionalities, namely, relay selection (F1.3) and PUsโ packettransmission (F1.4), as well as SUsโ packet transmission (F2.3)in order to enhance the network performance of PUs (A1.2)and to provide dedicated channel access to SUs (A2.1) incentralized (C1.1) and distributed (C1.2) SU networks. Thepurpose is to maximize the PUsโ and SUsโ utility functions๐๐and ๐๐ , respectively, in terms of power savings of PUs whenthey collaborate with SUs in the presence of ๐ฝ PUs and๐พ SUs.For centralized CRNs, the functionalities are modeled andsolved using reinforcement learning in an intracooperative(C2.1) mode, and intercooperative (C3.1) mode, respectively,whereas for distributed CRNs, the functionalities are mod-eled and solved using reinforcement learning in a nonintra-cooperative (C2.2) mode and inter-cooperative (C3.1) mode,respectively. The PU ๐ โ ๐ฝ utility function is defined as
๐๐ =๐๐,๐ โ ๐๐ (๐๐(๐),๐)
๐๐,๐(๐ ๐ (๐ผ๐) โ ๐ ๐,min) , (12)
where ๐๐,๐ is the maximum transmission power of PU ๐through direct PU-PU transmission without using a SU relaynode, ๐๐(๐๐(๐),๐) is the PU ๐ transmission power through PU-SU-PU transmission using SU ๐ as a relay node where ๐๐(๐)is the transmission power for SU ๐ to relay the PUsโ packetsto its destination, and ๐ ๐(๐ผ๐) and ๐ ๐,min are the achievabletransmission rate of PU ๐ after allocating ๐ผ๐ of white spacesto SUs and the minimum transmission rate of PU ๐ for directtransmission, respectively. The SU ๐ โ ๐พ utility function isdefined as
๐๐ ,๐ = ๐ผ๐๐๐ log (1 + SNR๐,๐) (BER๐๐,min โ BER๐๐,(๐๐(๐))) ,(13)
-
International Journal of Distributed Sensor Networks 13
where ๐๐ is the bandwidth used by SU to transmit its ownsignal, SNR๐,๐ is the signal-to-noise ratio of SU ๐, whileBER๐๐,min and BER๐๐,(๐๐(๐)) are the minimum and observedBit Error Rate (BER) values of SU ๐ while relaying PU jโspackets. It has been shown that the transmission power of PUdecreases with increasing the transmission power of SU.
Murawski and Ekici [27] propose two PU F(1) andone SU F(2) functionalities, namely, relay selection (F1.3)and PUsโ packet transmission (F1.4), as well as SUsโ packettransmission (F2.3) in order to enhance the network per-formance of PUs (A1.2) and to provide dedicated channelaccess to SUs (A2.1) in distributed (C1.2) SU networks. Thepurpose is to maximize the throughput of PUs and SUsin an intra-cooperative (C2.1) mode and inter-cooperative(C3.1) mode, respectively. The network considers a singlePU source node that communicates with a PU destinationnode through direct PU-PU transmission or indirect PU-SU-PU transmission via SU relay node. The PU destinationnode transmits Request to Send (RTS), while the SU replieswith Request to Cooperate (RTC) composed of channel stateinformation upon receiving RTS from the PU. Subsequently,the PUdestinationnode selects the suitable SUs as relay nodesusing the channel state information.The criterion adopted byPU for selecting a suitable SU relaying node is based on thebasis of higher throughput value of a given PU-SU-PU linkwith respect to the throughput value of PU-PU direct link.The PU destination node sends clear to coordinate (CTC)message to a selected SU relay node, which indicates that agiven PU-SU-PU link offers higher throughput than the PU-PU direct link; whereas, if the throughput being offered bythe PU-SU-PU link is lower than the PU-PU direct link, thenthe PU destination node sends clear to send (CTS) messageto the SU relay node, which indicates that the direct link ofPU-PU communication can take place. For the calculationof expected throughput value either from PU-SU-PU link orfrom PU-PU direct link, abackoff mechanism of distributedcoordination function [38] is used. The expected throughputvalue is dependent on the probability of successful packettransmission ๐๐ , packet transmission time ๐กpacket, collisiondetection time ๐กcollide, and the expected size of PU packets๐ธpacket size. Furthermore, for attaining a higher throughputgain, adaptive modulation schemes (e.g., BPSK, QPSK, and16-QAM) is used with respect to the SNR of the channels.It has been shown that, higher throughput can be achievedby changing the adaptive modulation scheme from BPSKto QPSK, and from QPSK to 16-QAM. Additionally, higherthroughput of PUs can be achieved by reducing the numberof SUs as relaying nodes which reduces the communicationoverheads.
Toroujeni et al. [28] propose two PUF(1) and one SUF(2)functionalities, namely, relay selection (F1.3) and PUsโ packet,transmission (F1.4), as well as SUsโ packet transmission (F2.3)in order to enhance the network performance of PUs (A1.2)and to provide dedicated channel access to SUs (A2.1) indistributed (C1.2) SUnetworks.The purpose is to increase thelink reliability by maximizing the transmission rate of a PUcommunication node pair and๐พ SUs.The functionalities aremodeled and solved using Orthogonal Frequency Division
Multiplexing (OFDM) [39] symbols in an intra-cooperative(C2.1) mode and inter-cooperative (C3.1) mode, respectively.There are a total of ๐๐ + ๐๐๐ OFDM symbols, in which ๐๐๐symbols are dedicated for a PU-PUcommunication node pairfor direct transmission, and the ๐๐ symbols are dedicatedfor PU-SU and SU-SU transmissions, respectively. The PUselects the maximum transmission link ๐ ๐ either from PU-PU direct link ๐ ๐๐ or from PU-SU-PU relayed link ๐ ๐๐ ๐, andit is defined as
๐ ๐ = max {๐ ๐๐, ๐ ๐๐ ๐} . (14)
Each SU ๐ โ ๐พ chooses the best channel to relay the packetsfrom PU source node to PU destination node as well as itsown packets to another SU. The SU cooperates with PU ifSU-SU transmission rate ๐ ๐ ๐ is equal to the price ๐๐, whichis charged by the PU, times the SU-PU transmission rate ๐ ๐ ๐,and it is defined as
๐ ๐ ๐ =
๐พ
โ
๐=1
๐๐ โ ๐ ๐ ๐. (15)
Higher value of ๐ ๐ ๐ indicates higher achievable transmissionrate between SU relay node and PU destination node. Ithas been shown that as the distance increases between PUsource node and SUs, it decreases the number of selected SUsas relaying nodes. Furthermore, higher cost being incurredby SUs reduces the achievable transmission rates of PUsalthough it increases the achievable transmission rates of SUs.
5.3. Scheduling the Channel Access of PUs and SUs. Thereare ten spectrum leasing schemes that focus on schedulingof channel access time in between of PUs and SUs for theirrespective transmission. These schemes have been shown toenhance PUsโ and SUsโ QoS performance (e.g., throughput).
5.3.1. Schemes That Use Game Theoretic Approaches. Chenet al. [29] propose two PU F(1) and one SU F(2) func-tionalities, namely, determination of PUsโ and SUsโ channelaccess time (F1.2) and relay selection (F1.3), as well as SUsโpacket transmission (F2.3) in order to enhance the networkperformance of PUs (A1.2) and to provide dedicated channelaccess to SUs (A2.1) in distributed (C1.2) SU networks.The purpose is to maximize the PUsโ and SUsโ networkutility functions ๐๐ and ๐๐ in the presence of ๐ฝ PUs and๐พ SUs. The functionalities are modeled and solved using athree-tier game in a non-intracooperative (C2.2) mode andnonintercooperative (C3.2) mode, respectively. The PU andSU network communicate with each other using a controlchannel protocol in order to participate and achieve a gameequilibrium. Both PUs and SUs are rational in nature. ThePU selects the suitable SUs as relay nodes to transmit PUโspackets in order to increase its transmission rate and the SUsin return achieve a portion of channel access time set bythe PU to maximize their transmission rate. The PU dividesthe transmission period into three phases. The first phase isfor primary transmission (PU-PU and PU-SU) during whichthe PUs transmit their packets to other PUs and SUs. Thesecond phase is for relayed transmission (SU-PU) during
-
14 International Journal of Distributed Sensor Networks
which the SUs help the PUs to relay PUsโ packets, whereasthe third phase is for secondary transmission (SU-SU) duringwhich the SUs transmit their own packets. The length of theprimary transmission phase is๐ผ, the relay nodes transmissionphase is (1โ๐ผ)(1โ๐ฝ), and the secondary transmission phaseis (1 โ ๐ผ)๐ฝ. Higher value of ๐ผ indicates that PUs is willing tolease its spectrum to SUs while higher value of ๐ฝ encouragesSUs to collaborate and relay PUsโ packets. Thus, the PU mustdetermine optimal values of ๐ผ and ๐ฝ (F1.2) that maximize itsown and SUsโ transmission rate. The PU ๐ utility function isdefined as
๐๐,๐ = min {๐ผ๐ ๐๐ ,๐, (1 โ ๐ผ) (1 โ ๐ฝ) ๐ ๐ ๐,๐} , (16)
where ๐ ๐๐ and ๐ ๐ ๐ are the maximum transmission ratethrough SU relay nodes (F1.3). The SU ๐ utility function isdefined as
๐๐ = ๐ฝ๐ ๐ ๐ โ (1 โ ๐ผ) ๐๐ ๐๐ , (17)
where ๐๐ is the cost of per unit power ๐๐ consumed by SU๐ as relay node to transmit PU source node packet to PUdestination node. Therefore, the utility function of SU ๐ isthe difference between its revenue in terms of achievable rate๐ ๐ ๐ (F2.3) and the cost of power which SU ๐ must bear inorder to relay the PUโs packets. It has been shown that asthe distance increase between the PU and SUs, their utilityfunctions increase until a certain limit which then decrease.
Huang et al. [18] propose three PU F(1) and one SUF(2) functionalities, namely, determination of the cost ofwhite spaces (F1.1), determination of PUsโ and SUsโ channelaccess time (F1.2), and relay selection (F1.3), as well as SUsโpacket transmission (F2.3) in order to enhance the networkperformance of PUs (A1.2) and to provide dedicated channelaccess to SUs (A2.1) in centralized (C1.1) SU networks. Thepurpose is to maximize the PUsโ and SUsโ utility functions๐๐and๐๐ in the presence of ๐ฝPUs and๐พ SUs.The functionalitiesare modeled and solved using canonical coalition game inan intracooperative (C2.1) mode and intercooperative (C3.1)mode, respectively. The PU divides a unit time slot intothree subslots for primary transmission (PU-PUandPU-SU),relayed transmission (SU-PU), and secondary transmission(SU-SU), respectively.The length of the primary transmissionsubslot is 1โ๐ผ, the relay nodes transmission subslot is ๐ฝ, andthe secondary transmission subslot is ๐ผ โ ๐ฝ. Higher value of๐ผ indicates that PUs are willing to lease their spectrum toSUs while higher value of ๐ฝ encourages SUs to collaboratemore and relay PU packets. Thus, the PU must determinethe optimal values of ๐ผ and ๐ฝ that maximize its own as wellas SUsโ transmission rate. The PU ๐ โ ๐ฝ utility function is๐๐ = ๐น(๐ ๐), where ๐น(โ ) is an increasing concave functionthat represents PUsโ gain and ๐ ๐ is the minimum achievabletransmission rate, which can be either from PU-SU or fromSU-PU, and dependent on transmitter power ๐๐ก, channelgain ๐บ, and noise level ๐2. The SUsโ utility function is ๐๐ =๐บ(๐ ๐ ) โ ๐๐ , where G(โ ) is an increasing concave function thatrepresents SUsโ gain and ๐๐ is the price that SU needs to payin order to lease channels from PUs. It has been shown thatas the SUsโ channel access time increases, the transmission
rate of SUs increases significantly, which increases the PUsmonetary gainwhile decreasing its transmission rate since SUuses more power to transmits its own packets.
Wang et al. [30] propose three PU F(1) and two SUF(2) functionalities, namely, determination of the cost ofwhite spaces (F1.1), determination of PUsโ and SUsโ channelaccess time (F1.2), and relay selection (F1.3), as well asdetermination of SUโs channel access time (F2.2) and SUsโpacket transmission (F2.3) in order to enhance the networkperformance of PUs (A1.2) and to provide dedicated channelaccess to SUs (A2.1) in centralized (C1.1) SU networks. Thepurpose is to maximize the PUsโ and SUsโ utility functions๐๐and ๐๐ , respectively, in the presence of a PU communicationnode pair and ๐พ SUs. The functionalities are modeled andsolved using Stackelberg game in an intra-cooperative (C2.1)mode and inter-cooperative (C3.1) mode, respectively. ThePU divides the transmission period into three phases. Thefirst phase is for primary transmission (PU-PU and PU-SU) during which the PUs transmit their packets to otherPUs and SUs. The second phase is for relayed transmission(SU-PU) during which the SUs help the PUs to relay PUsโpackets whereas the third phase is for secondary transmission(SU-SU) during which the SUs transmit their own packets.The length of the primary transmission phase is (๐ โ ๐ก๐ )/2,the relay nodes transmission phase is (๐ โ ๐ก๐ )/2, and thesecondary transmission phase is ๐ก๐ . The PU utility functionis defined as
๐๐ = ๐บSNR (SNR๐๐ + SNR๐๐ ๐)๐ โ ๐ก๐
2๐, (18)
where ๐บSNR is the channel gain per unit SNR and SNR๐๐ andSNR๐๐ ๐ are the SNR values of PU-PU direct link and PU-SU-PU relayed link whereas, the SUsโ utility function is definedas
๐๐ = ๐บ๐ก๐ โ ๐
๐ โ ๐ก๐
2{(SNR๐๐ + 1) ๐โ ๐ก๐ โ ๐
2
(SNR๐๐ โ ๐๐ก๐ )๐บ๐๐} , (19)
where ๐ is the cost per unit energy consumption, ๐ is theprice that SUs needs to bear in order to buy white spaces fromPUs, and ๐2 is the noise variance. It has been shown that asthe distance increase between the PU and SUs, their utilityfunctions increase until a certain limit which then decrease.
Stanojev et al. [31] propose two PU F(1) and one SUF(2) functionalities, namely, determination of PUsโ and SUsโchannel access time (F1.2) and relay selection (F1.3), aswell as SUsโ packet transmission (F2.3) in order to enhancethe network performance of PUs (A1.2) and to providededicated channel access to SUs (A2.1) in distributed (C1.2)SU networks. The purpose is to maximize the PUsโ trans-mission rate and the SUsโ utility function. The PU divides aunit time slot into three subslots for primary transmission(PU-PU and PU-SU), relayed transmission (SU-PU), andsecondary transmission (SU-SU), respectively. The length ofthe primary transmission subslot is (1 โ ๐ผ) โ ๐กslot, the relaynodes transmission subslot is ๐ผ โ ๐ฝ โ ๐กslot, and the secondarytransmission subslot is ๐ผ โ (1 โ ๐ฝ) โ ๐กslot. Higher value of ๐ผ andlower value of ๐ฝ encourage SUs to collaborate, and so the PU
-
International Journal of Distributed Sensor Networks 15
must determine optimal values of ๐ผ and ๐ฝ, while maximizingits own transmission rate. The functionalities are modeledand solved using Stackelberg game in a nonintracooperative(C2.2) mode and intercooperative (C3.1) mode, respectively.In this scheme, PU is the leader and SU is the follower. Thegame aims to foster collaboration between PUs and SUs bymaximizing the PUsโ transmission rate and enhancing theSUsโ utility function. The PU source node ๐ chooses a setof SU relay node ๐ that provides an optimum value of PUtransmission rate, which is dependent on the transmissionrate from PU source node ๐ to SU relaying node ๐, or ๐ ๐๐
๐๐,๐,
while SU relaying node ๐ calculates the transmission ratefrom SU relay node ๐ to PU destination node ๐, or ๐ ๐ ๐
๐๐,๐, as
well as ๐ฝ. Hence, the value of ๐ฝ must be chosen carefully toencourage collaboration between PU and SU. The choice of๐ฝ must maximize the SU-PU transmission rate (๐ผ โ ๐ฝ โ ๐กslot) โ ๐ ๐ ๐
๐๐,๐, on the other hand, the choice of ๐ผ must maximize
the SU-SU transmission rate {(๐ผ โ (1 โ ๐ฝ)) โ ๐กslot} โ ๐ ๐ ๐
๐. The
optimal value of ๐ฝ is ๐ฝ = argmax๐ฝโ[0,1]
๐ฝ โ ๐ ๐ ๐
๐๐,๐and ๐ฝ is
applied in the calculation of ๏ฟฝฬ๏ฟฝ = ๐(1/๐ฝ).The PU source nodeselects a suitable SU relay node (F1.3) to transfer its packetsto PU destination node (F1.4) if SU relay node provideshigher transmission rate; otherwise, it chooses PU-PU directlink. The PU calculates channel access time for PUs andSUs (F1.2). It has been shown that, as the number of SUrelay nodes increases, the outage probability of PU decreasesand the transmission rate of SUs increases. The SUs aim tomaximize their utility function in order to transmit its ownpackets (F2.3). The SUs utility function is ๐ข๐ ๐
๐๐,๐โ๐, where ๐๐
is the transmission power of SU relaying node ๐, and ๐โ๐ isa vector of the transmission power of the SU nonrelayingnodes. The PU adjusts ๐ฝ to determine the time distribu-tion among PUsโ and SUsโ (F1.2) transmissions, and this isfollowed by the selection of the best available SUs as relaynodes (F1.3) for possible communication between a PU nodepair. It has been shown that the PUsโ and SUsโ throughputperformances can be increased by increasing the number ofSU relay nodes ๐ and decreasing the distance between PU andSU.
Wang et al. [32] propose two PU F(1) and one SU F(2)functionalities, namely, determination of PUsโ and SUsโ chan-nel access time (F1.2) and relay selection (F1.3), as well as SUsโpacket transmission (F2.3) in order to enhance the networkperformance of PUs (A1.2) and to provide dedicated channelaccess to SUs (A2.1) in distributed (C1.2) SU networks. Thepurpose is to maximize the PUsโ and SUsโ utility functions๐๐and๐๐ in the presence of a PU communication node pair and๐พ SUs. The functionalities are modeled and solved using thegame theoretic approach and the Stackelberg equilibrium ina nonintracooperative (C2.2) mode and nonintercooperative(C3.2) mode, respectively. In this game theoretic approach,PUs and SUs are rational in nature, in which the PUs andSUs attempt to achieve their respective equilibrium point.The PU selects suitable SUs that transmit PU packets as relayusing their respective transmission power, while the SUs inreturn achieve a portion of channel access time set by the PUto transmit their own packets. The PU divides a unit time
slot into two sub-slots for primary transmission (PU-PU,PU-SU, and SU-PU) and secondary transmission (SU-SU),respectively. The length of the primary transmission subslotis ๐ผ, while the secondary transmission subslot is 1โ๐ผ. Highervalue of ๐ผ indicates that PUs are willing to lease its spectrumto SUs in order to maximize its packet transmission whileallocating the remaining time to SUs for their own packettransmission. Thus, PU must determine the optimal value of๐ผ (F1.2) thatmaximize its own and SUsโ transmission rate.ThePU utility function is defined as
๐๐ = ๐ผ๐ ๐ (๐ผ) , (20)
where ๐ ๐(๐ผ) is the achievable transmission rate through SUrelay nodes (F1.3) and it is dependent on transmitter power๐๐ก, channel gain ๐บ, and noise variance ๐
2. The SUsโ utilityfunction is defined as
๐๐ = ๐๐ (๐ ๐) ๐ก๐ โ1
2๐ผ๐๐, (21)
where ๐๐, ๐ ๐, and ๐ก๐ are the revenue, achievable transmissionrate, and allocation time of SU ๐, and ๐๐ is the transmissionpower used by SU ๐ to relay the PUsโ packets to PUdestination and therefore it is considered as a cost by SU๐. Therefore, the utility function of SU ๐ is the differencebetween its revenue in terms of achievable transmission rate(F2.3) and the energy cost that SU ๐ must bear to relay thePUsโ packets. It has been shown that PUsโ utility functionincreases with the increment of the ๐ผ value. Furthermore,as the distance between PUs and SUs decreases, it increasestheir utility functions significantly because of higher channelgain.
Zhang et al. [33] propose two PU F(1) and one SUF(2) functionalities, namely, determination of PUsโ and SUsโchannel access time (F1.2), relay selection (F1.3), and SUsโpacket transmission (F2.3) in order to enhance the networkperformance of PUs (A1.2) and to provide dedicated channelaccess to SUs (A2.1) in distributed (C1.2) SU networks. Thepurpose is to maximize the PUsโ and SUsโ utility functions๐๐ and ๐๐ in order to enhance their transmission rate in thepresence of a PU communication node pair and ๐พ SUs. Thefunctionalities are modeled and solved using game theoryand the Nash equilibrium in a non-intracooperative (C2.2)mode and inter-cooperative (C3.1) mode, respectively. Inthis game, the PU selects the suitable SUs as relay nodesto transmit PUsโ packets using their respective transmissionpower and in return, the SUs receive a portion of channelaccess time set by the PU to transmit their own packets. ThePU divides a unit time slot into three subslots for primarytransmission (PU-PU and PU-SU), relayed transmission(SU-PU), and secondary transmission (SU-SU), respectively.The length of the primary transmission subslot is 1 โ ๐ผ, therelay nodes transmission subslot is ๐ผ๐ฝ, and the secondarytransmission subslot is ๐ผ(1 โ ๐ฝ). Higher value of ๐ผ indicatesthat PUs are willing to lease its white spaces to SUs whilehigher value of ๐ฝ encourages SUs to collaborate more andrelay PU packets. Thus, the PU must determine optimal
-
16 International Journal of Distributed Sensor Networks
values of ๐ผ and ๐ฝ (F1.2) that maximize its own and SUsโtransmission rate. The PUsโ utility function is defined as
๐๐ = ๐ ๐๐ ๐ โ ๐ ๐๐ + ๐ผ๐๐๐๐, (22)
where ๐ ๐๐ ๐ and ๐ ๐๐ are the achievable transmission ratethrough SU relay nodes (F1.3) and PU-PU direct transmis-sion. These rates are dependent on transmission power ๐channel gain ๐บ and noise power๐ whereas ๐๐ is the cost perunit of transmission power consumed by PU source node totransmit its packets to SUs and PU destination node.The SUsโutility function is defined as
๐๐ = ๐ผ (1 โ ๐ฝ) log2 (1 +๐๐ ๐บ๐
๐) โ ๐ผ๐๐ ๐๐ , (23)
where ๐๐ is the cost per unit transmission power consumed bySU relay node ๐ to transmit PU source nodeโs packets to PUdestination node.Therefore, the utility function of SU ๐ is thedifference between its revenue in terms of the achievable rate(F2.3) and the energy cost that SU ๐ must bear to relay thePUsโ packets. It has been shown that, as the distance increasesbetween the PU and SUs, their utility function increases untila certain limit which then decreases.
Zhu et al. [34] propose two SU F(2) functions, namely,collaborative selection (F2.1) and determination of SUโs chan-nel access time (F2.2) in order to provide dedicated channelaccess to SUs (A2.1) in distributed (C1.2) SU networks. Thereare two types ofmarkets, namely, primarymarket (comprisedof SU service providers and PUs) and secondary market(comprised of SU service providers and SU hosts). The func-tionalities are modeled and solved using a hierarchical gametheoretic framework comprised of upper- and lower-levelgames and in a non-intracooperative (C2.2) mode and non-intercooperative (C3.2) mode, respectively. The purpose is tomaximize the SUsโ service provider and SU network utilityfunctions, ๐๐,๐(๐ก) and ๐๐ ,๐(๐ก), respectively. The hierarchicalgame theoretic framework is as follows.
(i) Secondary market allows SU hosts to purchase whitespaces from SU service providers on a short-termbasis (e.g., minutes), and it is a lower-level gamemodeled by evolutionary game. Each SU serviceprovider ๐ offers white spaces, which are representedby bandwidth ๐๐ and price ๐๐. Note that higher price๐๐ for a particular bandwidth ๐๐ reduces demandlevels, and so it improves network performance.Subsequently, each SU host competes and selects a SUservice provider. Hence, the secondary market imple-ments collaborator selection (F2.1). Each SU aims tomaximize its individual utility function defined as
๐๐ ,๐ (๐ก) = ๐ผ โ ๐๐ (๐ก)
๐๐, (24)
where ๐ผ is a constant based on network performancerequirement, in order to maximize its network per-formance satisfaction.The number of SUs that chooseservice provider ๐ is represented by ๐๐(๐ก).
(ii) Primary market allows SU service providers to pur-chase white spaces from PUs (or spectrum brokers)on a long-term basis (e.g., weeks or months), and itis a upper-level game modeled by differential game.Each SU service provider ๐ purchases some amountof white spaces ๐๐(๐ก) from PUs based on the selectionof SU service providers ๐ฅ๐(๐ก) in order to maximizeprofits. Hence, it implements the determination ofSUโs channel access time (F2.2). Note that higheramount of the purchased white spaces improvesnetwork performance and so it attracts more SUs;however, it reduces monetary revenues. Each SUservice provider ๐ adjusts the amount of white spaces๐๐(๐ก), and maximizes its profit defined as
๐๐,๐ (๐ก) = ๐๐ โ ๐๐ (๐ก) โ ๐ฝ๐ โ ๐2
๐(๐ก) , (25)
where ๐๐ โ ๐๐(๐ก) represents the monetary revenue,๐ฝ๐ โ ๐2
๐(๐ก) represents the cost paid to the PUs, and ๐ฝ๐
is a constant weight. Note that, with ๐2๐(๐ก), it causes
the cost to increase rapidly, and so it prevents aSU service provider ๐ from being too aggressive. AtNash equilibrium, each SU service provider obtainsmaximized profit. In differential game, the SU serviceproviders make decision simultaneously; however,some providers may make decision first, and theyare called the leaders. In this case, a Stackelbergdifferential game can be applied to achieve Stack-elberg equilibrium. In Stackelberg game, the leaderproviders make decisions first, followed by followerproviders. So, the leader providers can achieve higherpay-off, and the follower providers make decisionbased on the optimal strategies made by the leaderproviders. The spectrum leasing scheme has beenshown to increase SU service providersโ profits.