Radio Resource Allocation Techniques for Efficient Spectrum Access

824

IEEE COMMUNICATION SURVEYS & TUTORIALS, VOL. 18, NO. 1, FIRST QUARTER 2016

Radio Resource Allocation Techniques for Efficient Spectrum Access in Cognitive Radio Networks Georgios I. Tsiropoulos, Octavia A. Dobre, Mohamed Hossam Ahmed, and Kareem E. Baddour

Abstract—This paper provides an overview of cognitive radio (CR) networks, with focus on the recent advances in resource allocation techniques and the CR networks architectural design. The contribution of this work is threefold. First, a systematic way to study the resource allocation problem is presented; various design approaches are introduced, such as signal-tointerference-and-noise ratio (SINR) or transmission power-based, and centralized or distributed methods. Second, CR optimization methods are presented, accompanied by a comprehensive study of the resource allocation problem formulations. Furthermore, quality of service criteria of the physical or/and the medium access control layers are investigated. Third, challenges in spectrum assignment are discussed, focusing on dynamic spectrum allocation, spectrum aggregation and frequency mobility. Such approaches constitute an emerging trend in efficient spectrum sharing and affect the performance of resource allocation techniques. The open issues for future research in this area are finally discussed, including adaptability-reconfigurability, dual accessibility, and energy efficiency. Index Terms—Cognitive radio, resource allocation, spectrum management, quality of service.

I. I NTRODUCTION

D

URING the last decade, significant research efforts have been undertaken to investigate and design efficient resource allocation schemes aiming at serving a continuously increasing number of users through the fixed spectrum range. As wireless devices become ubiquitous and offer new services with increased quality of service (QoS) characteristics and data rate demands, resource allocation becomes an even more difficult and challenging problem. The constantly growing request for advanced multimedia services, combined with the resource constraints of wireless networks, places increased stress on the fixed radio spectrum used by the current access technologies. The need for improved resource allocation techniques becomes even more essential as some spectrum usage measurements indicate that a significant amount of spectrum is used sporadically, leading to its underutilization [1]. Therefore, spectrum scarcity can be considered as a result of poor spectrum man-

Manuscript received February 17, 2014; revised July 23, 2014; accepted September 9, 2014. Date of publication October 16, 2014; date of current version January 27, 2016. G. I. Tsiropoulos is with National Technical University of Athens, 10682 Athens, Greece (e-mail: [email protected]). O. A. Dobre and M. H. Ahmed are with Memorial University of Newfoundland, St. John’s, NL A1B 3X9, Canada (e-mail: [email protected]; [email protected]). K. E. Baddour is with Communications Research Centre, Ottawa, ON K2H8S2, Canada (e-mail: [email protected]). Digital Object Identifier 10.1109/COMST.2014.2362796

agement policies rather than a physical shortage of frequencies [2]–[5]. Most of the existing spectrum allocation policies provide a fixed spectrum slice to each wireless access technology, so as to eliminate the interference among them [6], [7]. This static spectrum assignment prevents network users from efficiently utilizing the available spectrum since there are: 1) spectrum holes when no targeted devices exist in local area, 2) a poor utilization due to sparsely populated rural areas, and 3) blank periods where the communicating user does not transmit through the wireless channel. Thus, it is evident that a large portion of the assigned spectrum is used occasionally, leading to low levels of utilization for a significant amount of spectrum. Hence, dynamic spectrum access (DSA) techniques have been recently introduced to deal with the spectrum inefficiency problems [8], [9]. Several research groups are searching for improved resource allocation schemes, introducing at the same time a variety of technical terms, such as dynamic, cognitive radio (CR) or opportunistic spectrum access. DSA is a broad term, which encompasses various approaches to spectrum allocation, such as dynamic exclusive use, open sharing and hierarchical access models [10], [11]. The term CR was introduced by Mitola in 1998 as an intelligent radio technology able to adapt to the varying communication environment, which could be integrated into existing communication systems. Opportunistic spectrum access describes the overlay cognitive access method, where each cognitive user may transmit through available channels. The common idea in all cases is to realize a flexible spectrum allocation policy that dynamically allocates the available resources among competing users. Although DSA, CR and opportunistic spectrum access have different meanings, they usually are used as interchangeable terms. A detailed elaboration on these terms can be found in [11]. CR networks have been systematically investigated by researchers during the last decade, and a few surveys touching on the area of resource allocation in CR network have been published [9]–[14]. The specific research topics covered in these surveys are summarized in Table I. In [9], a wide range of topics related to the CR networks is presented, including spectrum sensing, management, mobility and sharing, which in turn affect the efficiency of resource allocation. In [10] and [11], one of the explored topics is spectrum opportunity identification, i.e., detection, tracking, and exploitation, with the latter connected to resource allocation. In [12], the effect of the CR

1553-877X © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

TSIROPOULOS et al.: RADIO RESOURCE ALLOCATION TECHNIQUES FOR EFFICIENT SPECTRUM ACCESS

825

TABLE I S URVEY PAPERS IN THE F IELD OF R ESOURCE A LLOCATION IN CR N ETWORKS

network architecture on resource allocation is considered. Additionally, spectrum sensing, detection, sharing and mobility are discussed; it is emphasized that an efficient spectrum sensing and detection procedure can enhance the resource allocation performance by selecting the most appropriate spectrum band for the CR users. In [13], the identification and allocation of the available spectrum to CR users are presented. In particular, the following topics related to spectrum allocation are discussed: 1) criteria for occupying the most appropriate portion of the spectrum, such as interference, power, fairness and delay, 2) employed approach, such as centralized or distributed, cluster-based and the use of a common control channel, 3) techniques employed to optimize the spectrum allocation, based on graph theory, linear programming, fuzzy logic, game theory or evolutionary algorithms. A recent survey paper, [14], investigates the resource allocation problem focusing on the employed criteria, such as interference, power, fairness, delay, and the approach or technique employed, i.e., centralized or distributed, graph- or game-theory based. When compared with the previously mentioned surveys, this work has the following contributions: 1) An extensive review of the resource allocation problem is presented according to different design techniques, such as signal-to-interference-and-noise ratio (SINR)versus transmission power-based, and centralized versus

distributed implementation. The main difference between our work and existing surveys is that we provide the basic mathematical formulation of each resource allocation problem. Thus, the paper can be characterized as a tutorial-oriented survey. 2) A comprehensive study of the most common CR optimization methods is presented in a tutorial fashion. Moreover, each method and its corresponding resource allocation problem are mathematically formulated and the QoS criteria employed in resource allocation techniques are investigated. 3) Research challenges in spectrum allocation are discussed. These include: a) distributed spectrum allocation techniques, which represent the recent trends in the design of efficient spectrum sharing techniques, b) mobility functions which enable the frequency handoff for SUs due to the changes in network parameters, and c) spectrum aggregation which constitutes a promising technique to increase the efficiency of existing spectrum allocation models. The rest of the paper is organized as follows. In Section II, the basic concepts and essential functionalities of the CR technology and resource allocation are presented. A broad classification of resource allocation techniques for CR networks and their requirements and parameters are provided in Sections III and IV, respectively. In Section V, the optimization methods employed for resource allocation are presented. Challenges in

826


spectrum allocation techniques are reviewed in Section VI. Discussions and important resource allocation open issues are presented in Section VII. Finally, Section VIII concludes the paper.

II. R ESOURCE A LLOCATION IN CR N ETWORKS A. CR Technology The aim of the CR technology is to enable flexible spectrum access by employing dynamic spectrum allocation techniques [11], [12]. Through the application of the CR technology, a number of users or networks adapt their operation and share access to the same spectral region in an opportunistic manner. Network users are divided into two main groups: licensed/primary (PU) and unlicensed/secondary users (SU). In the operating environments of interest, where heterogeneous wireless architectures are employed, asynchronous spectrum users with dissimilar QoS requirements of various services must coexist in a locally autonomous manner in a confined spectral bandwidth [12]. For this purpose, the CR technology exploits unused spectrum in the PUs frequency bands, which are referred to as spectrum holes or white spaces. SUs employ the spectrum holes in an opportunistic manner so as to transmit data under certain constraints, i.e., the interference of active SUs to PUs should be kept under a certain threshold. Thus, the CR networks aim to increase the efficient utilization of the spectrum range by exploiting the existence of underutilized radio frequency bands when others are crowded. To realize such an intelligent network, which should be able to adapt the spectrum allocation according to the changing radio frequency requirements and users’ needs subject to predefined QoS constraints, several intelligent techniques should be integrated by CR networks. The two lower layers of the communication protocol stack, i.e., the physical (PHY) and medium access control (MAC) layers, as well as the network layer, are modified to provide CR functionalities, such as spectrum sensing, opportunistic spectrum assignment and spectrum handoff [15], [16]. This paper considers mainly PHY and MAC.

B. Resource Allocation and Networking Functionalities When addressing the problem of resource allocation, we have to determine a set of parameters and aspects such as: a) the bandwidth allocated to each CR network user, b) users transmission rate and power, c) the mobility model of the CR network users and potential mobility support by the network, d) dynamic adaptation to the varying CR network conditions, and e) the level at which PUs are affected by SUs activity. To this end, the resource allocation problem can be divided into two main sub-problems: 1) the spectrum allocation and 2) determining the transmission parameters, such as transmission rate and power. These parameters will be further elaborated within this section. The other aspects, i.e., mobility support and dynamic adaptation are mainly related to the CR network architecture and MAC design specifications [17], [18]. These aspects will be studied in Section VI.

Spectrum Allocation: In a CR networking environment, efficient spectrum management techniques should be employed taking into consideration the QoS requirements of the supported services. The main challenges that a CR spectrum allocation technique should take into consideration are as follows: 1) Determining the available spectrum slots in the spectrum range under consideration. This functionality is commonly referred to as spectrum sensing. Thus, an efficient CR resource allocation scheme should sense the common medium within the available spectrum band and determine the unused portions of the spectrum [19], [20]. 2) Selecting the best available spectrum portions to serve an SU. This selection, which is called spectrum decision, is based on specific criteria, such as the interference introduced towards PUs and the satisfaction of SUs QoS criteria. 3) Sharing the spectrum, that is, to synchronize the access of the SU under consideration with the rest of communicating users within the CR network. This allocation is the output of a complex process where multiple network and QoS parameters are often taken into consideration. Moreover, certain policies, internal and possibly external, may also affect the spectrum allocation. 4) Releasing the occupied channel by the SU once a PU requests to communicate [21], [22]. Another strategy can also be considered, such as allocating less bandwidth to the SUs in case there is still available spectrum and the SUs QoS requirements are fulfilled. 5) Switching a communicating SU to another available frequency band when a PU starts transmitting in the initial channel of the SU. The term spectrum mobility is employed to describe this function. 6) Prioritizing the requirements of the PU which may need the spectrum temporarily assigned to an SU. Hence, if a PU requests a spectrum portion which is in use by an SU, the latter should vacate the channel in use and be handed off to another unused portion of the spectrum [12]. Alternatively, the SU can determine if it is possible to satisfy its own QoS requirements with fewer resources, without causing unwanted interference to the PUs. 7) Including learning capabilities based on past performance and experience so as to improve future performance. CR spectrum allocation techniques have to deal with the mutual interference from spectrally adjacent/overlapping transmissions, which may cause significant performance degradation [6]. As this interference is not managed by conventional radio resource allocation strategies, it is required to develop new technologies that enable users to exploit awareness of their local interference conditions to achieve efficient spectrum utilization while permitting reliable transmissions [6], [7]. Transmission Parameters: To efficiently serve an SU by satisfying its QoS requirements, the resource allocation technique has to determine a set of transmission parameters, such as transmission power and rate. Evidently, higher values of transmission power correspond to improved data rate, but on the other hand, the interference introduced to PUs is increased.


827

TABLE II D IFFERENT A PPROACHES FOR R ESOURCE A LLOCATION IN CR N ETWORKS

Thus, the resource allocation algorithm should consider the following challenges: 1) The PUs communication process should not be affected by the operation of SUs. As such, the CR network resource allocation algorithm should always consider the maximum interference that PUs can tolerate. 2) The CR network should satisfy the QoS requirements of SUs. Generally, the QoS requirements may be expressed in several ways, such as the minimum SINR value [23], minimum required data rate [24], or minimum value of SUs call blocking probability [25], [26]. III. C LASSIFICATION OF CR R ESOURCE A LLOCATION T ECHNIQUES A great variety of resource allocation techniques have been proposed in the literature, aiming at sharing the given spectrum among PUs and SUs, while not degrading the QoS of PUs and satisfying the QoS of SUs. The unique characteristic of CR networks, which differentiates them from ordinary wireless networks, is the coexistence of PUs and SUs. Thus, existing resource allocation techniques for wireless networks cannot be directly applied to CR networks. Moreover, the shared nature of the wireless channel requires new MAC functionalities, which should be incorporated into the resource allocation techniques. Such techniques, described in Section II-B, aim to achieve enhanced coordination among transmission attempts of CR network users. The CR network resource allocation techniques can be classified according to either the design concept that they implement or the parameter(s) employed in the decision-making

process [27]. A summary of the existing approaches in the literature is presented in Table II. A broad classification of the resource allocation schemes can be done according to the following design options: 1) Degree of Centralization for the Decision-Making Process: The resource allocation schemes can be distinguished into centralized or distributed. For the former, a central entity defines the resource allocation policy taking into account the actual status of the allocated resources, i.e., the way that network resources are already allocated to ongoing users, and the spatiotemporal utilization of the spectrum from the frequency sensing mechanism [28], [29]. The central entity decides whether to assign a particular channel to an SU or not, in a limited geographical region and for a given duration. On the other hand, distributed resource allocation schemes are performed by each CR network user independently [30], [31]. Thus, each SU decides based on the specific resource allocation scheme input parameters whether to transmit through an available wireless channel or not. In this case, it is necessary to have information exchange about some essential QoS metrics among neighboring users within a small area around the corresponding user. Although centralized resource allocation schemes can be more efficient, mainly due to the fact that they have access to the global information, they are more complex. Therefore, their implementation is difficult in real CR networks. Moreover, centralized resource allocation schemes necessitate a great amount of information exchange, which increases the CR network overhead. 2) The Amount of Available Information and Level of Cooperation Within the Network: According to this design feature, spectrum assignment schemes can be categorized as global or

828


Fig. 1. Classification of resource assignment schemes based on the information availability level.

DSA, local or direct access-based (DAB) schemes, and semilocal, as shown in Fig. 1. DSA schemes require information which may span the entire CR network [25], [32], [33]. Their objective is to determine a resource allocation scenario which enhances the overall performance of the CR network. However, such techniques need significant information exchange, demand high computational cost and cause increased latency. DSA schemes may be further classified based on the optimization method employed in the decision-making process into game theoretic, stochastic, graph theory, genetic and swarm intelligence schemes. These schemes are elaborated in [18]. In DAB schemes, the available information employed for the decision-making process is limited to the sender-receiver pair [34], [35]. Local resource allocation schemes are also found in the literature based on non-cooperative or selfish spectrum sharing approaches [36]. The aim of such schemes is to maximize an objective function which is much simpler than that for DSA schemes. The resource negotiation between both communicating parts is known as a sender-receiver handshake. Evidently, local resource allocation schemes are less complex and less efficient compared to the global ones. DAB schemes may be further categorized into contention- and coordinationbased schemes. In the first category, the SU sender and receiver exchange their sensing information through a simple handshake. Based on this information and the available resources, the pair negotiates to determine which channel to use. In the second category, the SU sender and receiver exchange their channel usage information with its neighbors. Thus, the overall system performance is improved on the basis that information exchange and users’ coordination are exploited. Apart from DSA and DAB, some semi-local resource allocation schemes have been introduced, which are based on the

availability of information for a small area around the network node [37], [38]. These techniques are also known as cooperative spectrum sharing schemes, since they exploit the interference measurements of each transmitting node within this small area, to ensure that the communication of one node does not affect the communication of the other nodes. Although the design features of the degree of centralization and the level of information availability seem to be similar, there is a slight difference between them. The former is associated with the area of the network that the resource allocation algorithm controls through its output, while the latter is related to the amount of information that the resource allocation decision is based on. Centralized systems usually have access to global information. On the other hand, distributed systems are typically based on local information. However, semi-local or even global information may be also available to distributed systems through information exchange. 3) Link Under Consideration: Resource allocation schemes may be distinguished upon their applicability to either uplink [39], [40] or downlink [41]–[43]. The main concern in the uplink operated schemes is the per user transmit power constraint since mobile terminals are battery oriented. Moreover, several additional design challenges such as the discrete nature of channel assignments (mainly in the orthogonal frequency division multiplexing (OFDM) systems), and the heterogeneity of channel conditions and service requirements [39] should be considered. On the other hand, resource allocation schemes in the downlink need information feedback from the mobile terminals to the base stations (BSs) to provide an efficient resource allocation policy [44]. In some cases, both links should be considered by an efficient resource allocation scheme, since an SU may not be admissible in one of the two links. The asymmetry between uplink and downlink traffic increases the complexity of the resource allocation problem. 4) The Way That Spectrum Is Shared Between PUs and SUs: To facilitate the access of SUs to licensed spectrum, a hierarchical access model is employed to distinguish PUs and SUs. The main limiting factor and requirement in CR networks is that the interference experienced by PUs when SUs access the licensed spectrum should be under a predefined level. Based on this design constraint, CR resource allocation schemes can be grouped into three categories according to the way the spectrum is shared between PUs and SUs [18]. The first approach, known as spectrum underlay access method, allows SUs to use the licensed spectrum even if PUs are active. In this case, SUs operate under the noise floor of PUs, as shown in Fig. 2(a); thus, they should comply with severe transmission power constraints. Spread spectrum techniques are employed to spread the signal of SUs over a wide spectrum range, so that the transmission of an SU is regarded as noise by PUs. In underlay access methods, SUs may communicate even in the worst case scenario, where all PUs are active all the time. In short range links, SUs may achieve significantly high transmission rates with extremely low transmission power [45]. Spectrum underlay schemes can be efficient since they utilize a broad spectrum range and thus, a large bandwidth; however, they have increased complexity, which makes implementation difficult in real CR networks [7], [9].


Fig. 2. Access methods of PUs and SUs in CR networks: (a) underlay access method, (b) overlay access method.

In contrast to spectrum underlay, overlay access methods permit SUs to access the wireless channel only when it is available, facilitating an opportunistic channel access as shown in Fig. 2(b). Thus, there is a need to identify the spectrum holes in time and space, to decide when and where an SU may transmit. Spectrum overlay access methods have been first introduced in [46] as a spectrum pooling strategy. In some studies, the term interweave access method is employed. This is similar to the overlay access method [47], since it requires sensing of the spectrum holes [48], [49]. However, as mentioned in [50], in interweave access methods, ongoing users are considered as PUs, while new incoming users are considered as SUs. Although the spectrum underlay access method may introduce increased interference to PUs when compared with the overlay one, this definitely depends on the transmission parameters, e.g., transmission power, position in frequency, as well as the performance of the spectrum sensing procedure. The interference in overlay access methods mainly stems from the non-orthogonality of the transmitted signals between SUs and PUs [51]–[53]. When the available access methods are studied from the perspective of the type of information they require, then overlay and underlay access methods focus on the interference caused by the SU users activity and the codebooks of PUs, respectively. On the other hand, the interweave access methods necessitate considerable information about the ongoing users activity which can be obtained from PUs sensing [50]. Generally, overlay access methods are simpler and easier to implement in practice. IV. CR R ESOURCE A LLOCATION R EQUIREMENTS AND N ETWORK PARAMETERS In opportunistic resource allocation techniques, PUs still have absolute priority over the SUs. However, the access of SUs to the common medium results in the degradation of PUs

829

performance. Thus, in such a CR communication environment, the prioritization of PUs over SUs is of utmost importance. The challenge to achieve such a policy is to maintain the SUs interference to PUs under a certain threshold, while satisfying the QoS requirements of both PUs and SUs [54]. In addition to interference-based schemes, CR resource allocation techniques which base their decisions on the sustainable rate of SUs have also been proposed [55]. In this design approach, the SUs transmission rate supported by the CR network should satisfy the QoS requirements. Moreover, the rate-based resource allocation scheme should take into consideration both the transmission rate requirement and the power control constraints. Alternative CR resource allocation schemes approach the problem by determining the fairness among users or some QoS-based utility function. The main difficulty in such schemes is to quantify the fairness level or utility of an SU. In most cases, both parameters are determined as a function of a countable network parameter, such as the SINR or the achievable transmission rate. Most studies of CR networks in the literature consider a common pool of available spectrum portions from different spectrum owners and wireless networking technologies within the spectral range under consideration, from which each SU may occupy the available spectral resources according to the specific criteria of the employed dynamic spectrum allocation technique [56]. Through this policy, unlicensed users acquire access to licensed bands. There are several possible available technologies that could be employed to realize this opportunistic spectrum allocation policy. Among them, OFDM has already prevailed, as it facilitates the dynamic allocation of non-occupied spectrum portions to SUs. However, the nonorthogonality of the transmitted signals affects the performance of CR networks resulting in mutual interference between SUs and PUs [51]–[53]. Several optimization schemes have been proposed in the literature, aiming at maximizing the overall transmission rate, while maintaining the mutual interference among communicating users under a predefined threshold and satisfying users QoS requirements [57]–[59]. In what follows, we study the requirements of resource allocation techniques and present the way that several network parameters are determined. Specifically, we examine the interference model of CR networks and determine the SINR and the transmission rate of SUs. Then, we present several utility functions that exist in the literature. These utility functions are based on network parameters, such as the interference that PUs experience due to the SUs activity, the SINR, the transmission rate or the transmission power1 of SUs. A. Interference Model and Determination of SINR The majority of proposed optimization schemes examine the interference caused by communicating users to a channel which is in use by either a PU or an SU [60]–[62]. According to the user category, different requirements should be fulfilled regarding the SINR. 1 Evidently, the transmission power of SUs has a direct impact on the data rate, the SINR and the interference introduced to PUs.

830


TABLE III N OTATION TABLE

Requirement #1 (R1): The aggregate interference caused by all opportunistic transmissions to each PU should be less than a predefined threshold. Let M and N be the total number of SUs and PUs, respectively, and Pic denote the transmission power of the ith SU in the cth subcarrier, where i = 1, . . . , M, c = 1, . . . , K ,

and K denotes the maximum number of available subcarriers.2 Note that the terms subcarriers and channels are used interchangeably in this survey and refer to the certain capacity for

2 A descriptive list of notations is provided in Table III.


Fig. 3.

831

Spatial distribution of PUs and SUs in a CR system model.

transmitting information, measured by its bandwidth in Hz. An opportunistic transmission may be between either two communicating SUs or an SU and a BS. Evidently, the transmission power is bounded such that 0 ≤ Pic ≤ Pmax , where Pmax refers c and G c to the maximum allowed transmitting power. Let G ik ji denote the channel gain from the transmitter serving the kth SU to the ith SU, i = k, k = 1, . . . , M, and the transmitter serving the ith SU to the jth PU, j = 1, . . . , N , respectively, on channel c. Moreover, Gicj is the channel gain from the jth PU to the ith SU. Fig. 3 illustrates the links between SUs, as well as between SUs and PUs. If I˙cj denotes the total interference from all opportunistic to the jth PU on channel c, then M c transmissions G ji Pic and R1 can be defined as I˙cj = i=1

The power spectral density c ( f ) of the cth channel of the ith SU is expressed as [51]

I˙cj ≤ I˙max

where B j represents the occupied bandwidth of the jth PU. M c ˙ Thus, the ACI component in I˙cj is i=1 I ji and the total interference from all opportunistictransmissions to the jth PU M M ˙c I ji . G cji Pic + i=1 on channel c is rewritten as I˙cj = i=1 Requirement #2 (R2): The QoS SINR of SUs should be maintained above a predefined threshold. The SINR for channel c at the ith SU can be determined by

(1)

where I˙max is the predefined threshold of the maximum acceptable interference on PUs. I˙cj is also referred to as co-channel interference (CCI), since it denotes the total interference of SUs transmitting on channel c to a PU operating on the same channel. Although CCI is the most important factor which should be taken into account in CR networks, interference generated by SUs operating on adjacent channels to the one that a PU transmits should be also considered. This is called adjacent channel interference (ACI) and is caused by sidelobe leakage of the neighboring channels employed by opportunistic transmissions into the bandwidth occupied by the PU [63]. ACI mainly depends on the power allocated to each neighboring channel operated by an SU and the spectral distance d ji between the ith SU and the jth PU band, i.e., between the center frequency of the channel occupied by the ith SU and that by the jth PU. Evidently, the relative position of the users and the respective channel gains also affect the ACI. Therefore, (1) should be reformulated so as to take into consideration both CCI and ACI. Let f c denote the bandwidth of channel c and Ts be the OFDM symbol duration. For simplicity, channels may be assumed to have equal bandwidth, i.e., f c = W, 1 ≤ c ≤ K .

c ( f ) = Pic Ts sin c2 (π f Ts ).

(2)

Given c ( f ), the interference introduced by the ith SU onto the jth PU is [51] I˙cji

2 = G cji

+B j /2 d ji

c ( f ) d f

d ji −B j /2

2 = G cji Pic Ts

d ji +B j /2

sin c2 (π f Ts ) d f,

(3)

d ji −B j /2

γic =

No + Iic +

G iic Pic M

k=1,k=i

c Pc Gik k

,

(4)

where No is the power of the additive white Gaussian noise, Iic is the total interference caused by all PU transmissions on M c c channel c to the ith SU and k=1,k =i Gik Pk represents the c total interference of SUs to the ith SU. Ii can be expressed as Iic =

N

Iicj ,

(5)

j=1

Iicj

denotes the interference introduced by the jth PU on where channel c. An exact formula to determine Iicj is given in [51], [52]. Note that the ACI from the PUs to the ith SU should also be taken into account, similar to the previous discussion. R2 can be formulated as γic ≥ γith ,

(6)

832


where γith is the threshold for the minimum SINR required for the ith SU to efficiently satisfy its QoS requirements, such as minimum guaranteed transmission rate and/or maximum bit error rate (BER). This depends on the specific coding, modulation and detection schemes being employed [19]. B. Determination of Attainable Data Rates The attainable data rate between an SU and the BS or another SU (in either downlink or uplink direction) constitutes a critical parameter which is employed by advanced resource allocation techniques in CR networks [64]–[66]. Generally, the data rate varies according to the SINR and the coding and/or modulation techniques employed. Thus, the attainable data rate is given as c 0, i f γic < γith c c (7) ri γi = f γi , i f γic ≥ γith , where the data rate is equal to zero when the SINR experienced by the SU is below the SINR threshold and the function f (·) depends on various factors, such as the coding and/or modulation techniques and the BER requirement [67]. Evidently, f (·) is a monotonically non-decreasing function of the SINR. For simplicity, we can define the attainable data rate as a function of the SINR experienced by SUs. Therefore, given γic , the maximum attainable rate ric of the ith SU on channel c can be determined by the Shannon’s capacity theorem [63] as ric = W log2 1 + γic . (8) In multicast CR networks, where multiple SUs can be served through a single transmission from the BS, SUs are grouped into several classes according to the service requested [22]. Evidently, the BS transmits at a given data rate for all SUs within the same multicast group. However, each SU has its own maximum data rate that can be supported, which mainly depends on the SINR experienced. To this end, the maximum data rate transmission of the BS should be less than or equal to the minimum attainable data rate of SUs within the multicast group. Otherwise, SUs whose attainable data rates are less than the one operated by the BS will not be able to decode the transmitted data. Let Ms denote the total number of SUs within the sth SU group, s = 1, . . . , S, where S represents the total number S of user classes supported by the CR network. Evidently, s=1 Ms = M. c denote the minimum SINR experienced by an SU Let γi,min within the sth SU group on channel c, expressed as c γi,min = min γic . i∈Ms

(9)

c on the Then, the minimum data rate transmission rs,min same channel c of the BS for the sth SU group is determined by c c rs,min . (10) = W log2 1 + γi,min c c The values of the network parameters γi,min and rs,min can be employed to satisfy Requirement 2.

C. User Utility Quantification The concept of a utility function is not new, since such functions have been widely employed in economics [68]. In our discussion, user utility is strictly related to the QoS level provided by the CR network and quantifies users’ satisfaction [69]. Thus, the higher the QoS of the network users, the more satisfied they are, resulting in higher utility values. User utility may be formulated as a function of various network parameters, depending on the design characteristics of the analysis performed and the employed QoS metric [70]. In CR networks, user utility may be defined as a function of the SINR experienced by the SUs or the transmission rate allocated to SUs by the CR network, which constitute the main QoS metrics of CR opportunistic resource allocation techniques. More sophisticated utility functions are based on the data that can be successfully transmitted per joule of energy consumed [71]. Apart from measuring the utility of SUs, several resource allocation schemes which take into consideration the utilities of both SUs and PUs have been recently proposed in the literature [72]–[78]. A descriptive summary of the utility functions employed in CR networks is shown in Table IV, and a more detailed discussion is provided in the sequel. 1) General Requirements of Utility Functions: Utility functions have been used in several studies for different wireless network technologies, e.g., CR networks [71], [72], NGN’s [79], IP-based networks [80] and peer-to-peer CR networks [81]. However, the main idea in almost every study is to define a function of several network parameters, such as SINR or data rate, with specific characteristics. c c , yi,max ] be the QoS metric for the ith SU Let yic ∈ [yi,min on channel c employed in the analysis, which may be either the SINR or the transmission rate of SUs. The upper bound c on yic readily maps to a QoS constraint when a further yi,max increase in yic cannot effectively increase the user utility. Morec corresponds to the minimum QoS constraint of the over, yi,min service offered by the CR network to SUs, below which the service cannot be offered. Then, the utility function U (yic ) for the ith SU should have the following properties [79]: Property 1 (P1): The utility function of the ith SU U (yic ) is a differentiable and continuous function of the QoS metric yic . Property 2 (P2): The utility U (yic ) is bounded in the closed set [0, 1]. Property 3 (P3): U (yic ) is an increasing function of yic , with ∂U (yic )/∂ yic > 0, c limc U (yic ) = 0, and limc

yic →yi,max

yi →yi,min

U (yic ) = 1.

Exact utility functions which fulfill the properties mentioned above can be obtained through field tests and user surveys [80]. A typical exponential utility function is given by r U yic = 1 − e a

c yic −yi,max

,

(11)

where ra denotes the constant absolute risk aversion coefficient for SUs, which is related to their behavior under uncertainty. Such utility functions are used due to their applicability and convenience under uncertainty conditions [80].


833

TABLE IV U TILITY F UNCTION C LASSIFICATION BASED ON THE Q O S M ETRIC U SED FOR I TS C ALCULATION

Note that risk aversion coefficients are used in behavioral economics to define the consumers and investors attitude towards risk [81]. 2) SINR-Based Utility Functions: In the literature, a great variety of utility functions which are specifically designed for QoS quantifications in terms of SINR have been proposed (see, e.g., [24], [82]–[86]). SINR-based utility functions are employed by CR resource allocation schemes to consider SUs satisfaction along with power control constraints. Let U (γic ) be the SINR-based utility function of the ith SU on channel c. In addition to the general requirements defined above for the utility functions, U (γic ) is selected to be concave over γic > 0. A typical function complying with such requirements is U γic = gi ln γic ,

(12)

where gi is a weighting factor for the ith SU, gi > 0 [82], [86]. The significance of this utility function is that it ensures the convexity of the resource allocation problem since it satisfies 2 c γic ∂ 2 U γic / ∂γic Ci γi = − = 1, ∂U γic /∂γic

(13)

where ∂U (γic )/∂γic and ∂ 2 U (γic )/(∂γic )2 denote the first and second derivative of U (γic ), respectively [88]. 3) Rate-Dependent Utility Functions: Although the transmit power and SINR experienced by SUs constitute critical parameters for the SUs QoS determination, another important QoS metric is the transmission rate of SUs, ric , i = 1, . . . , M. Accordingly, several CR resource allocation schemes which parameterize their utility as a function of the offered trans-

mission rate have been proposed in the literature [89]–[92]. A typical rate-dependent utility function which obeys the law of diminishing returns is [89] c ζ c ri /K (14) U ri = ζ , 1 + ric /K where K and ζ are essential parameters for determining the exact shape of the sigmoid utility function U (ric ) [68]. Evidently, U (ric ) is normalized within the interval [0,1). Moreover, U (ric ) = 1/2 when the transmission rate equals K . 4) Interference-Based Utility Functions and Energy Efficient Transmission: Some of the recent studies on resource allocation techniques for CR networks employ interference-based utility functions for PUs and SUs [72], [76]. The PUs utility functions quantify the QoS degradation of PUs to a lower but acceptable level, while the SUs utility functions measure the efficient use of SUs energy. The confining constraint on PUs is the interference that SUs generate, when they transmit data within the licensed frequency spectrum. Therefore, as long as the total interference from all opportunistic transmissions to the jth PU on channel c is less or equal to the maximum acceptable interference on PUs, that is I˙cj ≤ I˙max , then the PUs can meet their QoS requirements. Evidently, the PUs utility is severely penalized when their SINR is less than the one which corresponds to their minimum QoS requirements. On the other hand, the PUs utility is also slightly penalized when SINR is significantly higher than the target one. This is essential especially in wireless networks, since higher SINR values correspond to higher energy consumption by the mobile terminal and increased interference to all other wireless nodes operating at the same frequency bands.

834


Based on the above-mentioned remarks, the PUs utility function can be defined as

2

U j I˙cj = I˙max − μ1 I˙cj − I˙max u I˙cj − I˙max

I˙max − I˙cj max c ˙ ˙ −1 u I − I j , (15) − μ2 e where 1 ≤ j ≤ N and u(x) is the step function defined as 1, f or x ≥ 0 u(x) = (16) 0, f or x < 0, and μ1 and μ2 are weighting coefficients [93]. Although the main requirement for SUs in a CR network is to satisfy the interference condition, they also try to achieve the most energy efficient transmission. Thus, the SUs utility function should incorporate multiple parameters, i.e., SINR, transmission rate and transmission power [72], [73]. To define such an utility function, we first need to describe the energy efficiency function (EEF) f i (γic ), which is indicative of the successful transmission of a data frame through the wireless channel. Note that a data frame is retransmitted when it has at least one bit error, which increases the energy consumption and prolongs occupancy of the wireless channel. Generally, EEF is measured in bits/joule, since it measures the total number of bits which are successfully transmitted through the wireless channel per joule of energy consumed. Several EEFs have been proposed in the literature [72], [74], [75], [77], [78]; however, for simplicity, we will focus on two EEFs herein. Given γic , the first EEF is defined as [72]

c L f i γic = 1 − e−0.5γi , (17) where L is the total number of bits per packet. The second EEF is also based on γic and is given by [71] f i γic = ln K γic , (18) where K is a parameter determined by the target SINR of SUs. Given the EEF, the utility function can be expressed as c c c ric f i γic U γi , ri , Pi = , (19) Pic where ric /Pic is the ratio of the ith SU throughput to its transmission power [77], [78]. V. O PTIMIZATION M ETHODS FOR CR N ETWORK R ESOURCE A LLOCATION Several CR resource allocation techniques have been proposed in the literature, aiming mainly at maximizing the SU QoS levels and optimizing the overall network performance without degrading the QoS experienced by PUs below a certain level [94]–[98]. According to its specific design characteristics, each CR resource allocation technique considers different network performance criteria and user QoS parameters. Thus, based on the optimization criteria employed by each resource allocation technique, different optimization problems are ob-

tained. CR resource allocation techniques studied in the recent literature may be categorized as presented below. A. Opportunistic Allocation Based on Joint Channel Assignment and Power Control The main trend in current research is to address the dual optimization problem of channel assignment and power allocation so as to realize an opportunistic spectrum access that allows SUs to exploit the underutilized spectrum [56], [96]. By employing the channel assignment indicator αic , where 0, i f channel c is available αic = (20) 1, i f channel c is occupied, and the transmission power Pic of channel c for the ith SU, the joint channel assignment and power allocation problem can be formulated as a mixed-integer linear programming function, which aims to determine the maximum number of opportunistic SUs that can be served by the CR network subject to linear constraints. That is [61] M K

αic ,

(21-a)

αic ≤ 1,

∀ i = 1, 2, . . . , M,

(21-b)

αic ≤ 1,

∀ c = 1, 2, . . . , K ,

(21-c)

max c

αi ∈{0,1}

c=1 i=1

subject to K c=1 M i=1

G iic Pic −γith

M

k=1,k=i

c c G ik Pkc + αic −1 Gik Pmax + No

≥ γith No , ∀ i = 1, 2, . . . , M, I˙cj ≤ I˙max ,

(21-d) c = 1, 2, . . . , K ,

∀ c = 1, 2, . . . , K ,

j = 1, 2, . . . , N , (21-e)

0 ≤ Pic ≤ Pmax , ∀ i = 1, 2, . . . , M, c = 1, 2, . . . , K . (21-f) Particularly, constraints (21-b) and (21-c) imply that an SU may occupy only a single channel and a channel may be assigned to a single SU, respectively. For constraint (21-c), it is assumed that all SUs are served by the same BS; otherwise, SUs should be grouped according to the BS which serves them and this restriction should be applied to each group of SUs. Moreover, constraints (21-d) and (21-e) refer to restrictions given in (6) and (1), respectively. As proven in [60], the optimization problem in (21) is NP-hard; therefore, a unified analytical solution cannot be obtained. B. Consideration of PU Activity In the CR communicating environment there is a delay from the moment that a spectrum portion is available to the time that it is allocated and occupied by the potential SUs. This delay depends mainly on the performance of the spectrum sensing


process performed by the CR networks [20]. During this period, the available spectrum portion is idle, which diminishes the overall CR network performance. This effect is of particular concern when a spectrum band is frequently released and reallocated. To quantify this performance degradation, several loss functions have been proposed in the literature [97], [100]– [102]. Let L(Pic ) be the loss function on channel c used by the ith SU, which depends on the transmitted power Pic [22], [98]. L(Pic ) should satisfy the following two conditions: Condition #1 (C1): L(Pic ) > 0 for Pic > 0. Condition #2 (C2): L(Pic ) = 0 for Pic = 0. A typical formula for the loss function, which can express either the rate loss or the SINR degradation, is [102] L Pic = H Pic ,

(22)

where H is the normalized average cost per unit power; this is referred to as the linear loss function. In case where L(Pic ) is employed in the rate allocation problem, H denotes the normalized average cost measured in transmission rate units. Given the loss function L(Pic ), the joint channel assignment and power allocation problem given in Section V-A can be reformulated, with the constraint (21-c) rewritten as G iic Pic − γith

M k=1,k=i

c c Pkc + aic − 1 Gik Pmax + No G ik − L Pic ≥ γith No

(23)

and the above optimization conditions (C1) and (C2) included. C. Combined Rate Allocation and Power Control A slightly different class of CR allocation problems, which are extensively studied in the literature, involves the rate and transmitted power of SUs [56], [103], [104]. These optimization problems aim to maximize the aggregated transmission rate of SUs satisfying at the same time their QoS requirements and complying with restrictions imposed by PUs. Given γic and W , the maximum attainable rate ric of the ith SU on channel c can be obtained by (8). Hence, the combined rate allocation and power control can be formulated as K M c , (24-a) wi log2 1 + γi max W {γic } c=1 i=1 where wi ≥ 0 is the weighting factor of the ith SU and M is subjected to the constraint i=1 wi = 1. The parameter constraints of the maximization problem in (24-a) can be expressed as c c ≤ ric ≤ ri,max , ri,min

I˙cj ≤ I˙max ,

∀ c = 1, 2, . . . , K ,

0 ≤ Pic ≤ Pmax ,

(24-b) j = 1, 2, . . . , N ,

∀ i = 1, 2, . . . , M,

(24-c)

c = 1, 2, . . . , K . (24-d)

Apart from the previous constraints, additional restrictions may be imposed according to the specific design characteristics of

835

each CR resource allocation approach. For example, the authors in [105] provide a joint rate and power optimization problem subject to the violation probability that the SINR will fall under its threshold value. The loss function mentioned in the previous subsection can also be employed. In this case, L(Pic ) can be expressed in the data rate units and the optimization problem given by (24-a) can be rewritten as K M M K c c max . wi log2 1 + γi − L Pi W {γic ,Pic } c=1 i=1 c=1 i=1 (25) D. Fairness Criteria: Max-Min and Proportional Fairness Recent research studies have evolved the strict channel/rate allocation and power control optimization problem towards CR resource allocation techniques that incorporate fairness concerns [106]–[108]. Moreover, integrated resource allocation schemes have been recently introduced; these schemes try to ensure a balance between maximizing the aggregated SUs transmission rate and maintaining fairness in the resource allocation among such users [22], [109]. Let ri,min denote the minimum transmission rate requirement for the ith SU. Accordingly, the total aggregated minimum transmission M rate requirement of all SUs can be expressed as ri,min . If the total transmission rate that a CR Rmin = i=1 network can support in the available spectrum portion at a given time is less than or equal to Rmin , then some or all SUs are served at their minimum transmission rate requirements. However, if the network load is low, additional spectrum portions will be available; thus, the total transmission rate supported by the CR network may be higher than Rmin . For this reason, SUs would increase their transmission rates above their minimum requirement. The increase of SUs transmission rate should be realized in a fair manner, preventing the possibility of groups of SUs monopolizing the additional resources. Hence, the integrated rate-power allocation with fairness provision can be formulated as [110] K c ¯ , (26) f R max R¯ c

c=1

c ], c = 1, 2, . . . , K , and f ( R¯ c ) is where R¯ c = [r1c , r2c , . . . , r M the objective function which strikes a balance between rate allocation and fairness among SUs. The constraints of the optimization problem are the same as the ones described in Section V-C. Although several objective functions may be employed, two of them prevail in the literature, since they can be easily integrated into the resource allocation problem formulation. The first one is called max-min fairness and aims to maximize the minimum transmission rate within the group of communicating SUs (see, e.g., [60], [70], [111]). Therefore, the optimization problem described in (26) is rewritten as [111]

max mini ric .

(27)

836


The second objective function aims to achieve proportional fairness [112], [113] and is defined as f ( R¯ c ) =

M

ln ric .

(28)

i=1

Another fairness metric that is widely used in the literature is the Jain’s fairness index (JFI), defined as J F I = M c 2 M ( i=1 ri ) /M i=1 (ric )2 [114]. However, JFI is mainly employed to evaluate the performance of resource allocation schemes and is not typically used to determine the resource allocation decision. VI. C HALLENGES IN CR S PECTRUM A SSIGNMENT Since wireless networks parameters vary over time and space, spectrum assignment should be performed in a flexible and dynamic way. To implement such spectrum assignment features in CR networks, three design issues should be addressed: 1) The distributed spectrum assignment techniques where no central entity is employed. Thus, spectrum management is performed locally based on the information exchanged among neighboring users. 2) Frequency mobility where ongoing SUs are handed off to other available frequency bands when their channels are no longer available. 3) Frequency aggregation techniques, which aim to utilize idle and discontinuous spectrum fragments to accommodate channels. Thus, additional SUs may be served and spectrum can be used in a more effective way. More details on these aspects are presented in the following sub-sections. A. Distributed Spectrum Assignment Techniques Most of the proposed spectrum assignment techniques consider a central entity which is responsible for resource allocation among competing SUs [115]–[118]. Thus, if the construction of a centralized infrastructure is not preferable, distributed solutions should be applied, where SUs’ access over the available spectrum is based on local policies [92], [119]–[121]. To this end, distributed spectrum allocation techniques and coordination among multiple SUs constitute a challenging research field in CR networks. In such a resource allocation scheme, SUs may decide on the way that they access the available spectrum range either by themselves or by cooperating with their neighbors. In the latter case, SUs exchange information obtained from the sensing algorithms or from neighboring nodes, e.g., regarding channel assignments of other SUs. Therefore, each SU calculates a QoS metric and exchanges this information with its neighboring SUs. Based on the corresponding information that the SU under consideration receives from the surrounding SUs, it can determine the traffic load of each available channel, and afterwards occupy a channel according to the criterion employed for spectrum assignment, e.g., attainable data rate or interference introduced to PUs [122], [123]. If we consider the legacy distributed MAC protocols designed for ad-hoc networks, all SUs may access available spectrum applying a greedy policy. However, greedy design approaches deteriorate

CR network performance, particularly in cases when the CR network operates near congestion. Similar to most design approaches, there are advantages and disadvantages to distributed spectrum assignment techniques. Distributed techniques are more flexible in the sense that they can quickly adjust to potential network changes. Moreover, in most cases, the network change may affect only a local area; thus, a distributed resource allocation technique can address this change without performing a large scale adjustment of allocated resources. Furthermore, distributed design approaches are faster when compared to the centralized ones, since they do not demand information exchange from all the network nodes. Hence, distributed schemes require reduced signaling exchanges, which results in lower messaging overhead and reduced power consumption. Although distributed spectrum assignment techniques make faster resource allocation decisions, they do not provide an optimal solution since the outcome is based on local knowledge of the network state. Moreover, fairness concerns are satisfied locally for a group of neighboring SUs and not globally for the entire CR network. Another drawback of distributed schemes is the impact of false measurements by SUs on the spectrum assignment decision. Accordingly, inaccurate or misleading information about the metric information that SUs exchange may significantly affect the spectrum assignment algorithm outcome, deteriorating its performance. In what follows, we discuss the distributed CR resource allocation techniques by presenting works on this research field, as well as main challenges. Moreover, distributed resource allocation techniques integrating game theory features, as well as the way that information can be exchanged among CR network users and how this is employed for routing purposes, are discussed. 1) Distributed CR MAC: Several distributed resource allocation schemes for CR networks have been proposed in the literature, with different ways of addressing the resource allocation problem. In [124] and [125], the authors consider an integrated framework of spectrum sensing and frequency allocation which is modeled through a partially observable Markov decision process. The main concept in both studies is to limit the band that an SU node senses either by restricting the frequency range or by applying energy consumption limitations, which in turn lead indirectly to the same operating point. The main disadvantages of such distributed schemes are the challenging multichannel MAC problems that must be resolved [126]–[129]. Moreover, well known MAC problems, which are already solved for a single channel scenario, should be studied in the multichannel scenario. A typical example is the wellknown hidden terminal problem. In the single channel scenario, this problem is easily solved by exchanging request-to-send (RTS) and clear-to-receive (CTR) messages. This solution cannot be applied in the multichannel scenario where nodes operate in different channels; thus, they may not receive the RTS and CTR transmitted by the other network nodes. This problem becomes even more critical in the case of CR networks where the protection of PUs is of utmost importance. In [130], the authors apply the principles of mesh networks on a cognitive infrastructure towards the cognitive


Fig. 4.

837

A taxonomy of local information sources.

mesh network (COMNET). In COMNET, the network consists of mesh clients (MCs) and mesh routers (MRs), where each MR serves several MCs forming small clusters. In this distributed resource allocation technique, the resources are allocated by the MR to the MCs within the same cluster. SUs which are the MCs of the cluster operate in the ISM band. However, they sense the frequency spectrum of PUs and may employ a portion of the PUs band in the licensed spectrum only if it is idle. While MCs compete to exploit the blank periods of PUs, MRs communicate with each other through out-of-band communication on dedicated channels [131]. 2) Main Challenges: Distributed resource allocation techniques in CR networks deal with important challenges which in turn have a strong impact on the performance of these techniques. Such challenges are [132]: i) The time-varying characteristics of the network parameters and the adaptability of the resource allocation technique to this variable environment. The dynamic parameters of CR networks are the following: • activity of PUs, which is time-varying and affects the available spectrum that SUs may use; • users mobility; • time-varying characteristics of wireless channels and fading conditions; • bursty nature of emerging applications. Due to the time-varying nature of the network characteristics, SUs need to exchange information in a repeated way. They need to be able to learn the network parameters and dynamically adapt their operation and transmission strategy to the actual network conditions. ii) The amount of available information at the network nodes. The distributed techniques applied to SUs are sensitive to the amount of information available to the nodes. Moreover, this information is useful only if it can be conveyed in a timely manner. If the delay for transmitting this information over the network exceeds the timeline

constraint, then, it is useless and the distributed techniques cannot converge to an appropriate solution. Thus, the amount of information and delay in its transmission over the network are critical for the accuracy and efficiency of distributed resource allocation techniques. 3) Game Theoretic and Bargaining Spectrum Sharing Techniques: A large group of distributed resource allocation techniques are strictly related to bargaining games, game theoretic approaches and auction policies [133], [134]. More details are provided as follows. Bargaining Approach Based on Local Information: The information obtained by local sensing, i.e., observation of the local area or communication with neighboring nodes, is considered as local. In the IEEE 1900.4 standard [135], the information is categorized into terminal and network classes which correspond to local and global information, respectively. The information that can be obtained through the application which is running in the device itself, can also be considered as local information. As illustrated in Fig. 4, local information sources can be broadly categorized into two main groups according to the information related with: 1) The application that is used by the SU user. Examples of illustrative parameters are delay, packet loss and bandwidth, which are directly related to the application performance. 2) The device used by the SU. Some parameters such as the active links, e.g., block error rate, power allocated and SINR, and the channels employed by the SU, e.g., channel ID and frequency range, depend on the device employed. Local bargaining is explored in [36], where the authors introduce an approach to realize a dynamic and decentralized resource allocation policy. The key idea is to allow users to be self-organized into bargaining groups, wherein a spectrum assignment technique is employed to obtain a conflict-free

838


resource allocation assignment. The authors propose a local bargaining framework, which consists of two strategies. The first one is called one-to-one fairness bargaining, where each node/user of the bargaining group chooses to bargain with one node/user of the same group at a time. The second one is called one-buyer-multi-seller bargaining, where a node/user tries to purchase a set of channels from the other users in the same bargaining group. Moreover, the proposed resource allocation technique is combined with the feed poverty strategy to consider fairness issues and eliminate user starvation [36]. An important advantage of the proposed local-based bargaining approach is that it significantly reduces the number of computations required to adapt topology changes. Game Models for Spectrum Sharing: In static or one shot games, the system obtains a resource allocation policy in a non-repeated or once and for all fashion [136]. Due to users’ selfishness, static games lead to inefficient Nash equilibrium (NE) points. To include a dynamic aspect in the spectrum sharing technique and alleviate the impact of the users’ selfish behavior, repeated games are employed, which play the same static game multiple times [137]–[139]. Repeated techniques have been proposed also for the waterfilling algorithm to allocate the power and available spectrum among competing users in a cooperative framework [140]. However, it was observed that the iterative waterfilling algorithm may lead to inefficient resource allocation policies [140]. According to [136], a resource allocation solution is characterized as optimal if it is not possible to improve a certain parameter of the system without degrading the performance of some other parameters. A systematic study of a repeated spectrum sharing technique is proposed in [136], where the authors examine the efficiency, fairness and self-enforcing compatibility in a non-cooperative game framework. In self-enforcing or incentive compatible spectrum allocation there is no motivation for users to deviate from it [136]. The metric of each static game is the total throughput of active users and the overall payoff of the repeated game is the normalized summation of the payoff at each static game. In each round, the repeated technique remembers the output of the previous rounds. To obtain the solution in repeated games, the Folk theorem is employed, which ensures that any obtained NE dominates the min-max payoff [140]. Otherwise, i.e., if the payoff obtained by the min-max strategy is greater than the one obtained through repeated games, then the user/player is prone to deviate from the NE. The Folk theorem defines a discount factor, δ, where δ ∈ (0, 1) [136], [141], as the factor by which players discount future payoffs relative to payoffs in the current period. A discount factor close to one indicates that the player cares just as much about future periods as the present one, while a discount factor close to zero indicates that the player does not care about any future periods. A payoff of E in the current period is equivalent to a payoff of δ × E in the next period of the game. In the infinite horizon repeated game [136], the utility Ui of the ith SU/player is given by

Ui = (1 − δ)

∞ t=0

δ t Ri (t),

(29)

where Ri (t) denotes the utility of the ith SU/player at instant t. Instead of Ri (t), a utility function may also be used [137]. Evidently, the utility of each game round is employed in (29), which shows that in each round the past actions are used by the SU/player to decide on the current or future ones. Auction-Based Techniques: Auction-based techniques are used to realize a dynamic and distributed resource allocation policy in the cases where the resources are not governed by a central entity or controlled by a group of privileged users, i.e., PUs. However, in some cases auction-based techniques may also be applied to centralized infrastructures [142]–[144]. Generally, the solution is obtained through a repeated process on a negotiation basis. Two surveys of auction-based resource allocation schemes in CR networks and different auction types are provided in [145] and [146]. In [147], the authors propose an auction-based spectrum sharing technique where the resources are controlled by an entity called manager, which is either an independent organization or a government agency. Thus, SUs can purchase an amount of bandwidth from the manager to use a local and relatively short-term data service. The authors propose two auction-based resource allocation techniques employing either the users’ SINR or transmission power. In both cases, a utility function Ui (θi , γi ) is employed to quantify the user’s benefit from spectrum utilization, where θi is a user dependent parameter and γi is the user SINR. In each round of the repeated auction game, the SUs/players submit their bids, which represent their willingness to pay for the allocated resources. The cost Ci for the ith SU is proportional to the resources allocated, and is given either by [147] Ci = π s γi

(30)

or by Ci = π p pi G i

0

(31)

for the SINR or transmission power auction scenario, respectively. π s and π p are the price parameters for each scenario and G i 0 is the channel gain from the ith SU to the measurement point. Generally, the measurement point is an independent entity [147]; therefore, G i 0 is neither the path gain of SUs communication link nor the path gain of the link between a PU and the BS. A simpler cost function neglects the channel gain and (31) is rewritten as Ci = π p pi [148]. However, in other studies, no measurement entity is used and the cost is π p p¯ i j [149], where p¯ i j represents defined by Ci = j∈K , j=i

the amount of power that the ith SU buys from the jth SU. Given the utility Ui (θi , γi ) and the cost Ci , the surplus function Si (θi , γi ) = Ui (θi , γi ) − Ci should be maximized. Similar to (30) and (31), Ci has been defined in [150] through a pricingbased approach as Ci = sˆi qi + u i , where sˆi is the expected number of successfully transmitted packets by the ith SU per period, and qi and u i are the usage and flat price of the ith SU, respectively. The social optimal allocation can be obtained through the Vickrey-Clarke-Groves (VCG) auction by maximizing the total utility [151], [152]. However, VCG cannot be applied in this


case due to the required information from users, the interference constraints and the computational burden. The authors in [147] show that under some mild conditions, where less information exchange and less computational effort are required from the central entity, the proposed iterative auction techniques converge to the solution obtained by VCG. An VCG auction-based resource allocation scheme is also proposed in [153]. In this case, four different parties are taken into account in resource allocation: a) the spectrum regulator, b) the service provider, c) the PUs, and d) the SUs. In this four party co-winning strategy, PUs may suffer from the possible interference introduced by SUs. However, this is kept below a certain threshold. Moreover, PUs are compensated by the service provider for the QoS degradation that they may experience. 4) Essential Procedures: Cooperative distributed resource allocation techniques demand the exchange of information among CR network users to define their transmission strategy. Based on the information from nearby network users, they determine network parameters in the MAC or network layer, such as minimum-delay route and frequency channel that the user will employ. Two basic procedures that are essential in distributed resource allocation, namely, periodic information sharing procedure and route selection procedure, are discussed in the sequel. Periodic Information Sharing Procedure: A significant information exchange among SUs is necessary to coordinate the cooperation in spectrum sensing and sharing. Such information may be the location of SUs, the way that users are clustered into groups, the estimation of the location of PUs and their transmission power and which users transmit at each time epoch. The general concept of information sharing is split into six steps (A-F), which are presented as follows. In Step A, the information exchange among neighboring nodes is realized. In this step, the large amount of information that is typically required leads to increased overhead for SUs. To improve the sensing performance, distributed cooperative sensing has been proposed, which is based on evolutionary game theory (EGT) [154], [155]. EGT defines the way that SUs collaborate in the multiuser distributed CR networks and provides a convenient method to address the strategic uncertainty that players—in this case the SUs—encounter in a game. In Step B, SUs exploit the information exchange to learn how the neighboring users behave and determine user behavior patterns. In the EGT-based distributed cooperative sensing schemes, SUs/players learn during the strategic interactions and approach a robust equilibrium strategy known as evolutionary stable strategy [155], [156]. During the game, players may try different strategies, since they are uncertain about the actions of the other players, and they learn through interactions. Thus, the percentage of players that use a particular pure strategy may change at each play of the game. In the EGT games, replicator dynamics may be employed to address the uncertainty of players’ available actions and specify how many players change their strategy at each play of the game [157]. In Step C, the procedure determines the available resources and forms a temporary resource allocation matrix among competing SUs. This process is based on the information from Step A regarding the spectrum availability and the interference intro-

839

duced by the active PUs and SUs. Given the resource matrix, an action is taken in Step D and the information regarding the availability of the spectrum, the interference introduced to active users and the resource allocation matrix is updated accordingly. In Step E, this information is broadcast to the network. The procedure is repeated at the next time period (Step F). Route Selection: Based on the available information at the SU device, a set of feasible actions concerning the route of a given information packet can be determined. Given the channel availability and the interference introduced to active users, the channel condition for each link toward the destination can be estimated, forming the channel condition matrix. Then, the delay of each link to the next relay is estimated and compared with the delay requirement. The actions for which the delay estimation is greater than the delay deadline are dropped from the action set. The path with the minimum delay is selected and an RTS request is sent to the next relay. If a CTS response is received by the SU under consideration, then the packet is transmitted to the next relay. In some CR networks, SUs may keep an action information table for future transmissions. The efficiency of the route selection procedure depends on the accuracy of the channel sensing techniques [9], the grade of coordination in the RTS/CTS handshake [124], [158] and the way that delay vectors are computed for each available action [159].

B. Spectrum Mobility for CR Network Users Spectrum allocation techniques assign each SU with the best available channel according to the network parameters which are determined at the time that the SU requests to be served. However, network parameters may change during a communication session. Thus, an important function of the CR network resource allocation techniques is their ability to switch the channel of a communicating SU to another available frequency band, when the SU’s channel becomes unavailable. This is referred to as the spectrum mobility or spectrum handoff [15]. Spectrum mobility should be transparent to SUs; therefore, spectrum mobility management techniques in CR networks should ensure smooth and fast frequency transition. Moreover, spectrum mobility techniques should ensure a minimum QoS performance degradation for ongoing SU sessions during a spectrum handoff. To facilitate such a performance, the network protocols should be able to adapt their operation to the channel parameters. Thus, the network protocol parameters in a CR network protocol stack should change according to the operating frequency. Most spectrum handoff algorithms employed in CR networks mainly rely on the mobility-based handoff schemes developed for cellular networks [12], [16]. In general, a spectrum handoff for an SU may be triggered by the detection of a PU [25], [160] or it may be performed in a proactive way [1], [161]. In the latter case, SUs are allowed to employ more than one channel, which is not used by PUs [161]. Multiple SUs may share the same channels at the same time as long as their respective SINR is acceptable.

840


Spectrum mobility is also essential when channel conditions change either in time or space. Accordingly, spectrum handoff may be performed in the following cases [9], [12]: 1) Spectrum mobility due to PUs detection. As previously mentioned, SUs employ the licensed frequency band for data transmission such that both SUs and PUs coexist in the same frequency band. However, when a PU is detected that owns a channel already assigned to a communicating SU, then the application running on the SU should be handed off to another available frequency band. 2) Spectrum mobility due to time-varying channel conditions. Since wireless channel characteristics are timevarying, an ongoing SU may need to be handed off to another available channel. This decision may be triggered by a significant QoS degradation over time in the communication link assigned to the SU. The design of such spectrum allocation schemes is challenging and necessitates that the CR network provides an efficient channel sensing function. 3) Spectrum mobility in space domain. To support SUs mobility, CR networks should be able to support SUs handoff in the same way that handoff is realized in cellular networks. Thus, when an ongoing SU moves from one place to another, its communication is not terminated. In this case, the SU is handed off to another channel, which is served by another BS or node of the CR network. 4) Apart from the inter-cell handoff spectrum mobility, due to SUs mobility, intra-cell handoff, known also as vertical handoff, may be employed to “move” the ongoing SU between different networks. Vertical handoff constitutes the only solution for a network that operates near congestion, and cannot serve either additional SUs or existing SUs. The latter should be “moved” to another frequency range due to the PUs activity. SUs calls that are handed off to another network may release enough resources to the network in order to serve a new incoming call request. C. Carrier Aggregation Aware Spectrum Assignment Based on its QoS requirements, each user demands a certain bandwidth of the spectrum for transmission. The problem is to allocate the available spectrum in a way that satisfies the maximum number of communicating users. The majority of spectrum allocation schemes in the literature are based on continuous channel assignment algorithms which assume that each channel consists of a continuous spectrum portion. These schemes can result in low overall spectrum utilization as smaller spectrum fragments, which cannot accommodate a channel, are left idle. To address this problem, an SU should be able to access several discontinuous spectrum fragments simultaneously, which can accommodate a channel [162]–[164]. An SU may be able to transmit data through all the spectrum fragments at the same time. To realize such a policy, discontinuous OFDM (DOFDM) has been proposed as a transmission technique. As shown in the example presented in Fig. 5, DOFDM carriers span over three different spectrum fragments, which are separated by spectrum portions that serve PUs. Each of the three available

Fig. 5. Three available and discontinuous spectrum fragments form a channel to serve an SU employing DOFDM.

spectrum fragments cannot serve an SU by itself; however, by aggregating them into a single channel, they can accommodate the traffic of an SU. In this case, the data of an SU can be transmitted by all the fragments which form the channel. By applying this policy, even the smallest available segments of the spectrum may be used, leading to increased spectrum efficiency. Although spectrum aggregation is a promising technique for CR networks, there exist limitations which diminish its efficiency. For example, spectrum aggregation cannot span across large frequency bands due to transceivers limitations [165], [166]. As a result, some spectrum fragments may not be usable as they may extend beyond the transceivers maximum frequency span. VII. D ISCUSSION AND O PEN R ESEARCH I SSUES CR networks have been developed to address the problem of spectrum scarcity by increasing the utilization level of the available bandwidth. To realize the opportunistic spectrum access of CR networks, it is required to develop efficient resource allocation techniques, whose main characteristics and design have been investigated in the literature in recent years (see, e.g., [9], [12], [18], [27]). A detailed survey of common approaches and techniques used in resource allocation schemes for CR networks has been presented in this study. However, there are many design aspects and open issues that require investigation. In this section, the most essential open research issues in resource allocation for CR networks are discussed. A. Adaptability—Reconfigurability Generally, when a new SU requests to be served by a CR network, there may either be enough resources to provide a bandwidth to satisfy the SU demand or the request may occur under congestion, which means that there is not enough available spectrum to provide even the minimum bandwidth to the new CR call request. Evidently, in the former case the new CR call request is admitted since QoS requirements in terms of bandwidth demands are met. In the latter case, the existing spectrum assignment techniques in the literature cannot serve the additional call request. Thus, the new SU call request will be blocked. Moreover, the same problem exists when a PU is detected in a channel already assigned to an ongoing SU. In this case, the SU should vacate the channel and be handed off to another available one. If a CR network cannot detect any other available channel, the call is dropped.


Hopefully, SU call blocking and dropping may be avoided if bandwidth adaptation algorithms, also known as QoS degradation schemes, are applied to obtain an optimal solution in terms of the resource allocation serving as many SUs as possible. Consequently, the CR network provides the necessary QoS guarantees to ongoing SUs and gets the most out of the available spectrum. To this end, the terms adaptability and reconfigurability denote the capability of SUs to adjust their transmission parameters in real-time, without any need for hardware modifications. There are several network parameters that can be reconfigured in an ongoing SU session, as explained below [167]. 1) Bandwidth Adaptation: If multiple channels are employed to serve an SU or channels are not considered with their traditional meaning, namely that they have a specific central frequency and a predefined bandwidth range, then the bandwidth allocated to one or more ongoing SUs may be reallocated. In this case, smaller bandwidth portions, compared to the initial ones, are assigned to each SU taking into consideration their minimum QoS requirements. Accordingly, enough bandwidth range may be economized so as one or more additional SUs may be served by the CR network. Generally, rate diminishing techniques go along with rate restoration procedures, which are triggered when available spectrum is detected by the sensing function of the CR network, and are responsible to restore the transmission rate of SUs to their initial or even higher values. 2) Operating Frequency/Channel: Ongoing SUs may change the operating frequency to another available one, according to their location or the signal quality they experience. Thus, based on the output of the CR network sensing, the most suitable available operating frequency may be determined. Accordingly, the SU will be handed off to the most appropriate frequency. 3) Modulation Technique: The modulation technique employed by a CR network can be adapted to the service offered, the QoS requirements of SUs and the channel conditions. Evidently, loss-sensitive services necessitate modulation schemes that provide low BER, whereas delay sensitive services require appropriate modulation techniques, which focus on data rate rather than BER. 4) Transmission Power: Transmission power of SUs may vary within a predetermined range. Higher power values are undesirable since SUs battery is rapidly exhausted and greater interference is introduced towards ongoing SUs and PUs. For this purpose, power control has been introduced which enables the dynamic transmission power configuration within the allowable power range. Power control should also consider the actual channel condition and the QoS requirements of the ongoing SU. Through power control, the transmission power is reduced to the minimum acceptable value. Accordingly, the corresponding interference is decreased, which allows more users to share the available spectrum. 5) Access Technology: A CR network may be subscribed to various primary networks. Consequently, SUs should be able to connect to every access technology that each primary employs. When an SU cannot be served by the available resources of a certain primary network, SUs may be readjusted to employ available resources from other primary networks.

841

An aspect not covered in the literature is the additional processing time that the QoS degradation schemes require. Moreover, such schemes necessitate a considerable information exchange. Both issues are open for future research. B. Dual Access In the typical network model, a user may be either a PU or an SU served by a primary or CR network, respectively. A CR network may be subscribed to multiple primary networks, such as GPRS and UMTS. An SU may be served by an available channel belonging to one of these primary networks. Thus, if the resource allocation algorithm cannot find any available channel in a primary network, it requests that the other primary networks accommodate the SU call. However, this is not the case of PUs, which are served exclusively by only one primary network. For this reason, an open research topic is to study the option of enabling dual access to PUs. Particularly, PU devices may be equipped with both primary and CR network transceivers. Therefore, if the primary network is congested, PUs may act as SUs which will be served by the CR network. By the application of dual access policy, the performance of primary networks can be improved, the spectrum utilization level can be increased and call blocking probabilities of PUs can be significantly decreased. C. Energy Efficiency Energy efficiency is one of the issues of major concern in wireless networks. Thus, the objective of minimizing the energy consumption needs to be studied in CR networks, as well. Energy-aware resource allocation schemes enable battery life preservation. Several studies in the literature examine power control in spectrum assignment for CR networks, e.g., [93], [161]; however, they focus on minimizing the interference rather than minimizing the energy consumption. The energy consumption problem can be categorized into two main subproblems, as follows: 1) Energy Consumption Due to Communication: The energy consumption problem in data transmission depends on the frequency range under use and the distance between the communicating entities. Higher transmission power values correspond to increased power consumption, which in turn results in decreased battery lifetime. Several power control algorithms for CR networks have been proposed in the literature, which are integrated into spectrum assignment techniques, e.g., [167]– [169]. Such techniques attempt to reduce power consumption to enable users to remain connected for a longer time. Moreover, power-aware spectrum assignment techniques also consume energy; thus, their implementation should be simple and fast. 2) Energy Consumption Due to Spectrum Sensing: Spectrum sensing is an essential component which has a great impact on the overall performance of CR networks. However, it is one of the main sources of energy consumption and depends directly on the spectral range that is scanned. By reducing the spectral range under investigation, energy consumption can be decreased; however, the performance of spectrum assignment techniques will be adversely affected. Thus, there is a need

842


for novel resource allocation schemes capable of balancing the tradeoff between the spectrum sensing range energy consumption, and spectrum assignment performance. Some studies in this field reduce the spectral sensing range via a reinforcement learning-based approach [170], [171]. These schemes enable SUs to improve the performance of the sensing process by learning from their prior experiences. The need to develop energy efficient resource allocation schemes remains an open topic for research. The emerging energy efficient studies should take into consideration the necessity for guaranteeing QoS requirements under varying conditions. D. Proactive Spectrum Allocation In contrast to reactive spectrum allocation in which spectrum allocation is performed as a consequence of link failure or QoS requirements violation, proactive spectrum allocation policies are developed to avoid link failure occurrences [172]. Proactive policies are based on machine learning (ML) [173] and accurate models for users activity prediction [174]. In [175], ML in CR networks has been defined as the ability of the CR users to remember lessons learned from past experience and act quickly in the future. Thus, CR network users are able to adapt to a changing environment. Several ML learning algorithms are available for proactive spectrum allocation in CR networks, including hidden Markov models [176], neural networks [177], [178], genetic algorithms [179] and partially observable Markov decision processes [124]. According to [173], the remarkable benefits of the proactive spectrum allocation are the following: 1) The time and energy spent by CR network users to find an available channel can be decreased. 2) The number of handoffs and service interruptions can be reduced. 3) The interference introduced by SUs towards PUs can also be reduced. Note that proactive policies may be employed not only in resource allocation in CR networks [180], [181], but also in other CR network functionalities such as in spectrum sensing [182], [183] and routing [184], [185]. Proactive spectrum allocation in CR networks is certainly an issue that requires further investigation. VIII. C ONCLUSION Future wireless networks will likely rely on the CR technology, as this provides a solution to address the problem of spectrum underutilization. CR networks aim to exploit unused spectrum portions not only in free frequency bands, but also in the licensed ones, to serve unlicensed users in an opportunistic manner. CR networks have the ability to sense the wireless medium for ongoing transmissions, in order to determine spectrum opportunities to be employed by the CR network to accommodate SUs. The main design requirement of CR technology is to serve SUs without lowering the QoS experienced by PUs. Through the efficient operation of the CR

networks, both PUs and SUs can be served within the same spectrum. In this work, a review of resource allocation techniques for CR networks has been presented. Furthermore, the review was enriched by a tutorial coverage of resource allocation problems. To understand the concept of cognitivity and how a spectrum band can be used to accommodate additional users, a description of the CR technology has been first presented. Then, a classification of existing resource allocation techniques has been provided, which highlights how their operation differs according to their design concept. Various network and QoS criteriabased on network parameters and QoS metrics-employed in resource allocation techniques have also been discussed. Based on the criterion and design concept employed, CR spectrum allocation techniques and their respective optimization methods have been examined. Since spectrum mobility is a key function of an efficient resource allocation scheme in CR networks, the way that SUs are handed off to another available channel and why this is needed have been additionally presented. Finally, several research challenges and open issues have been discussed, which can be considered for further investigation. R EFERENCES [1] FCC, Notice of Proposed Rulemaking and Order, Washington, DC, USA, Tech. Rep. ET Docket 03-222, Dec. 2003. [2] Report of the Spectrum Efficiency Working Group, Federal Communications Commission Spectrum Policy Task Force, Washington, DC, USA, Nov. 2002. [3] K. N. Steadman, A. D. Rose, and T. T. N. Nguyen, “Dynamic spectrum sharing detectors,” in Proc. IEEE DySPAN, 2007, pp. 276–282. [4] M. Wellens and P. Mähönen, “Lessons learned from an extensive spectrum occupancy measurement campaign and a stochastic duty cycle model,” in Proc. IEEE TridentCom, 2009, pp. 1–9. [5] V. Valenta, R. Maršalek, G. Baudoin, M. Villegas, M. Suarez, and F. Robert, “Survey on spectrum utilization in Europe: Measurements, analysis and observations,” in Proc. CROWNCOM, 2010, pp. 1–5. [6] G. Alnwaimi, K. Arshad, and K. Moessner, “Dynamic spectrum allocation algorithm with interference management in co-existing networks,” IEEE Commun. Lett., vol. 15, no. 9, pp. 932–934, Sep. 2011. [7] N. Nie, C. Comaniciu, and P. Agrawal, “A game-theoretic approach to interference management in cognitive networks,” in Wireless Communications (The IMA Volumes of Mathematics and Its Applications), P. Agrawal, D. M. Andrews, P. J. Fleming, G. Yin, and L. Zhang, Eds. New York, NY, USA: Springer-Verlag, 2006. [8] M. M. Buddhikot, “Understanding dynamic spectrum access: Models, taxonomy and challenges,” in Proc. IEEE DySPAN, 2007, pp. 649–663. [9] F. Akyildiz, W.-Y. Lee, M. C. Vuran, and S. Mohanty, “NeXt generation/ dynamic spectrum access/cognitive radio wireless networks: A survey,” Comput. Netw., vol. 50, no. 13, pp. 2127–2159, Sep. 2006. [10] Q. Zhao and A. Swami, “A survey of dynamic spectrum access: Signal processing and networking perspectives,” in Proc. IEEE ICASSP, 2007, pp. 1349–1352. [11] Q. Zhao and B. M. Sadler, “A survey of dynamic spectrum access,” IEEE Signal Process. Mag., vol. 24, no. 3, pp. 79–89, May 2007. [12] F. Akyildiz, W.-Y. Lee, M. C. Vuran, and S. Mohanty, “A survey on spectrum management in cognitive radio networks,” IEEE Commun. Mag., vol. 46, no. 4, pp. 40–48, Apr. 2008. [13] J. Mitola, “Cognitive radio,” Ph.D. dissertation, KTH, Stockholm, Sweden, Dec. 1998, Licentiate proposal. [14] E. Z. Tragos, S. Zeadally, A. G. Fragkiadakis, and V. A. Siris, “Spectrum assignment in cognitive radio networks: A comprehensive survey,” IEEE Commun. Surveys Tuts., vol. 15, no. 3, pp. 1108–1135, 2013. [15] I. Christian, S. Moh, I. Chung, and J. Lee, “Spectrum mobility in cognitive radio networks,” IEEE Commun. Mag., vol. 50, no. 6, pp. 114–121, Jun. 2012. [16] W.-Y. Lee and I. F. Akyildiz, “Spectrum-aware mobility management in cognitive radio cellular networks,” IEEE Trans. Mobile Comput., vol. 11, no. 4, pp. 529–542, Apr. 2012.


[17] C. Cormio and K. R. Chowdhury, “A survey on MAC protocols for cognitive radio networks,” Ad Hoc Netw., vol. 7, no. 7, pp. 1315–1329, Sep. 2009. [18] A. De Domenico, E. C. Strinati, and M. Di Benedetto, “A survey on MAC strategies for cognitive radio networks,” IEEE Commun. Surveys Tuts., vol. 14, no. 1, pp. 21–44, 2012. [19] A. Al-Habashna, O. A. Dobre, R. Venkatesan, and D. C. Popescu, “Second-order cyclostationarity of mobile WiMAX and LTE OFDM signals and application to spectrum awareness in cognitive radio systems,” IEEE J. Sel. Topics Signal Process., vol. 6, no. 1, pp. 26–42, Feb. 2012. [20] W.-Y. Lee and I. F. Akyildiz, “A spectrum decision framework for cognitive radio networks,” IEEE Trans. Mobile Comput., vol. 10, no. 2, pp. 161–174, Feb. 2011. [21] R. Chen, J.-M. Park, and J. H. Reed, “Defense against primary user emulation attacks in cognitive radio networks,” IEEE J. Sel. Areas Commun., vol. 26, no. 1, pp. 25–37, Jan. 2008. [22] D. Ngo, C. Tellambura, and H. Nguyen, “Resource allocation for OFDMA-based cognitive radio multicast networks with primary user activity consideration,” IEEE Trans. Veh. Technol., vol. 59, no. 4, pp. 1668–1679, May 2010. [23] M. Cheung, V. Wong, and R. Schober, “SINR-based random access for cognitive radio: Distributed algorithm and coalitional game,” IEEE Trans. Wireless Commun., vol. 10, no. 11, pp. 3887–3897, Nov. 2011. [24] C. G. Yang, J. D. Li, and Z. Tian, “Optimal power control for cognitive radio networks under coupled interference constraints: A cooperative game-theoretic perspective,” IEEE Trans. Veh. Technol., vol. 59, no. 4, pp. 1696–1706, May 2010. [25] V. Tumuluru, P. Wang, D. Niyato, and W. Song, “Performance analysis of cognitive radio spectrum access with prioritized traffic,” IEEE Trans. Veh. Technol., vol. 61, no. 4, pp. 1895–1906, May 2012. [26] A. Alshamrani, X. S. Shen, and L.-L. Xie, “QoS provisioning for heterogeneous services in cooperative cognitive radio networks,” IEEE J. Sel. Areas Commun., vol. 29, no. 4, pp. 819–830, Apr. 2011. [27] Y. Xiao and F. Hu, Cognitive Radio Network. New York, NY, USA: Taylor & Francis, 2009. [28] S. Bayhan and F. Alagoz, “Scheduling in centralized cognitive radio networks for energy efficiency,” IEEE Trans. Veh. Technol., vol. 62, no. 2, pp. 582–595, Feb. 2013. [29] M. Hasegawa, H. Hirai, K. Nagano, H. Harada, and K. Aihara, “Optimization for centralized and decentralized cognitive radio networks,” Proc. IEEE, vol. 102, no. 4, pp. 574–584, Apr. 2014. [30] Y. Liang, L. Lai, and J. Halloran, “Distributed cognitive radio network management via algorithms in probabilistic graphical models,” IEEE J. Sel. Areas Commun., vol. 29, no. 2, pp. 338–347, Feb. 2011. [31] J.-A. Bazerque and G. B. Giannakis, “Distributed spectrum sensing for cognitive radio networks by exploiting sparsity,” IEEE Trans. Signal Process., vol. 58, no. 3, pp. 1847–1862, Mar. 2010. [32] J. Chunxiao, C. Yan, K. J. R. Liu, and R. Yong, “Renewal-theoretical dynamic spectrum access in cognitive radio network with unknown primary behavior,” IEEE J. Sel. Areas Commun., vol. 31, no. 3, pp. 406–416, Mar. 2013. [33] S. Min, X. Chunsheng, Z. Yanxiao, and C. Xiuzhen, “Dynamic spectrum access: From cognitive radio to network radio,” IEEE Wireless Commun., vol. 19, no. 1, pp. 23–29, Feb. 2012. [34] J. Jia, Q. Zhang, and X. Shen, “HC-MAC: A hardware-constrained cognitive MAC for efficient spectrum management,” IEEE J. Sel. Areas Commun., vol. 21, no. 1, pp. 106–117, Jan. 2008. [35] H. Su and X. Zhang, “Cross-layer based opportunistic MAC protocols for QoS provisionings over cognitive radio wireless networks,” IEEE J. Sel. Areas Commun., vol. 26, no. 1, pp. 118–129, Jan. 2008. [36] H. Zheng and L. Cao, “Device-centric spectrum management,” in Proc. IEEE DySPAN, 2005, pp. 56–65. [37] S. Chunhua, Z. Wei, and K. Ben, “Cluster-based cooperative spectrum sensing in cognitive radio systems,” in Proc. IEEE ICC, 2007, pp. 2511–2515. [38] M. Jia, Y. Wotao, L. Husheng, E. Hossain, and H. Zhu, “Collaborative spectrum sensing from sparse observations in cognitive radio networks,” IEEE J. Sel. Areas Commun., vol. 29, no. 2, pp. 327–337, Feb. 2011. [39] J. Huang, V. Subramanian, R. Berry, and R. Agrawal, “Joint scheduling and resource allocation in uplink OFDM systems for broadband wireless access networks,” IEEE J. Sel. Areas Commun., vol. 27, no. 2, pp. 226–234, Feb. 2009. [40] W. Wang, W. Wang, Q. Lu, and T. Peng, “An uplink resource allocation scheme for OFDM-based cognitive radio networks,” Int. J. Commun. Syst., vol. 22, no. 5, pp. 603–623, May 2009.

843

[41] P. Mitran, L. Le, C. Rosenberg, and A. Girard, “Resource allocation for downlink spectrum sharing in cognitive radio networks,” in Proc. IEEE VTC Fall, 2008, pp. 1–5. [42] J. Huang, V. Subramanian, R. Agrawal, and R. Berry, “Downlink scheduling and resource allocation for OFDM systems,” in Proc. CISS, 2006, pp. 1272–1279. [43] R. Agrawal, V. Subramanian, and R. Berry, “Joint scheduling and resource allocation in CDMA systems,” in Proc. WiOpt, Cambridge, U.K., 2004, pp. 24–26. [44] M. H. Ahmed, “Call admission control in wireless networks: A comprehensive survey,” IEEE Commun. Surveys Tuts., vol. 7, no. 1, pp. 50–69, 2005. [45] S. Srinivasa and S. A. Jafar, “Cognitive radios for dynamic spectrum access—The throughput potential of cognitive radio: A theoretical perspective,” IEEE Commun. Mag., vol. 45, no. 5, pp. 73–79, May 2007. [46] J. Mitola, “Cognitive radio for flexible mobile multimedia communications,” in Proc. IEEE Int. Workshop Mobile Multimedia Commun., 1999, pp. 3–10. [47] Y.-C. Liang, K.-C. Chen, G. Y. Li, and P. Mahonen, “Cognitive radio networking and communications: An overview,” IEEE Trans. Veh. Technol., vol. 60, no. 7, pp. 3386–3407, Sep. 2011. [48] A. Ghasemi and E. Sousa, “Collaborative spectrum sensing for opportunistic access in fading environments,” in Proc. IEEE DySPAN, 2005, pp. 131–136. [49] G. Ganesan and Y. Li, “Agility improvement through cooperative diversity in cognitive radio,” in Proc. IEEE GLOBECOM, 2005, pp. 2505–2509. [50] A. Goldsmith, S. A. Jafar, I. Maric, and S. Srinivasa, “Breaking spectrum gridlock with cognitive radios: An information-theoretic perspective,” Proc. IEEE, vol. 97, no. 5, pp. 894–914, May 2009. [51] T. Weiss, J. Hillenbrand, A. Krohn, and F. Jondral, “Mutual interference in OFDM-based spectrum pooling systems,” in Proc. IEEE VTC, 2004, pp. 1873–1877. [52] A. V. Oppenheim and R. W. Schafer, Digital Signal Processing. Upper Saddle River, NJ, USA: Prentice-Hall, 1975. [53] B. Farhang-Boroujeny and R. Kempter, “Multicarrier communication techniques for spectrum sensing and communication in cognitive radios,” IEEE Commun. Mag., vol. 46, no. 4, pp. 80–85, Apr. 2008. [54] Y. Xu, J. Lui, and D. Chiu, “On oligopoly spectrum allocation game in cognitive radio networks with capacity constraints,” Comput. Netw., vol. 54, no. 6, pp. 925–943, Jun. 2010. [55] A. Tajer, N. Prasad, and X. Wang, “Beamforming and rate allocation in MISO cognitive radio networks,” IEEE Trans. Signal Process., vol. 58, no. 1, pp. 362–377, Jan. 2010. [56] W. Wei, K. G. Shin, and W. Wenbo, “Joint spectrum allocation and power control for multihop cognitive radio networks,” IEEE Trans. Mobile Comput., vol. 10, no. 7, pp. 1042–1055, Jul. 2011. [57] Y. Wang, P. Ren, Q. Du, and C. Zhang, “Optimal resource allocation for spectrum sensing based cognitive radio networks with statistical QoS guarantees,” Mobile Netw. Appl., vol. 17, no. 6, pp. 711–720, Dec. 2012. [58] X. Chen and C. Yuen, “Efficient resource allocation in rateless coded MU-MIMO cognitive radio network with QoS provisioning and limited feedback,” IEEE Trans. Veh. Technol., vol. 62, no. 1, pp. 395–399, Jan. 2013. [59] I. Wajid, M. Pesavento, Y. C. Eldar, and D. Ciochina, “Robust downlink beamforming with partial channel state information for conventional and cognitive radio networks,” IEEE Trans. Signal Process., vol. 61, no. 14, pp. 3656–3670, Jul. 2013. [60] L. Le and E. Hossain, “Resource allocation for spectrum underlay in cognitive radio networks,” IEEE Trans. Wireless Commun., vol. 7, no. 12, pp. 5306–5315, Dec. 2008. [61] A. T. Hoang and Y. Liang, “Downlink channel assignment and power control for cognitive radio networks,” IEEE Trans. Wireless Commun., vol. 7, no. 8, pp. 3106–3117, Aug. 2008. [62] D. T. Ngo and T. Le-Ngoc, “Distributed resource allocation for cognitive radio networks with spectrum-sharing constraints,” IEEE Trans. Veh. Technol., vol. 60, no. 7, pp. 3436–3449, Sep. 2011. [63] T. S. Rappaport, Wireless Communications: Principles and Practice, 2nd ed. Upper Saddle River, NJ, USA: Prentice-Hall, 2002. [64] S. Ghalib and B. A. Ozgur, “Performance analysis of CSMA-based opportunistic medium access protocol in cognitive radio sensor networks,” Ad Hoc Netw., vol. 15, pp. 4–13, Apr. 2014. [65] N. Qiang and C. C. Zarakovitis, “Nash bargaining game theoretic scheduling for joint channel and power allocation in cognitive radio systems,” IEEE J. Sel. Areas Commun., vol. 30, no. 1, pp. 70–81, Jan. 2012.

844


[66] S. Wang, Z.-H. Zhou, M. Ge, and C. Wang, “Resource allocation for heterogeneous cognitive radio networks with imperfect spectrum sensing,” IEEE J. Sel. Areas Commun., vol. 31, no. 3, pp. 464–475, Mar. 2013. [67] A. T. Hoang, Y.-C. Liang, and M. H. Islam, “Power control and channel allocation in cognitive radio networks with primary users’ cooperation,” IEEE Trans. Mobile Comput., vol. 9, no. 3, pp. 348–360, Mar. 2010. [68] H. R. Varian, Intermediate Microeconomics: A Modern Approach. New York, NY, USA: Norton, 1987. [69] J. Hou, J. Yang, and S. Papavassiliou, “Intergration of pricing with call admission control to meet QoS requirements in cellular networks,” IEEE Trans. Parallel Distrib. Syst., vol. 13, no. 9, pp. 898–910, Sep. 2002. [70] C. Peng, H. Zheng, and B. Y. Zhao, “Utilization and fairness in spectrum assignment for opportunistic spectrum access,” Mobile Netw. Appl., vol. 11, no. 4, pp. 555–576, May 2006. [71] P. Zhou, W. Yuan, W. Liu, and W. Cheng, “Joint power and rate control in cognitive radio networks: A game-theoretical approach,” in Proc. IEEE ICC, 2008, pp. 3296–3301. [72] S. K. Jayaweera and T. Li, “Dynamic spectrum leasing in cognitive radio networks via primary-secondary user power control games,” IEEE Trans. Wireless Commun., vol. 8, no. 6, pp. 3300–3310, Jun. 2009. [73] T. Li and S. K. Jayaweera, “A novel primary secondary user power control game for cognitive radios,” in Proc. IEEE MILCOM, 2008, pp. 1–7. [74] E.-V. Belmega, S. Lasaulce, and M. Debbah, “A survey on energyefficient communications,” in Proc. IEEE PIMRC Workshops, 2010, pp. 289–294. [75] A. Akbari, M. A. Imran, and R. Tafazolli, “Maximising average energy efficiency for two-user AWGN broadcast channel,” in Proc. FutureNetw, 2011, pp. 1–8. [76] S. K. Jayaweera, G. Vazquez-Vilar, and C. Mosquera, “Dynamic spectrum leasing: A new paradigm for spectrum sharing in cognitive radio networks,” IEEE Trans. Veh. Technol., vol. 59, no. 5, pp. 2328–2339, Jun. 2010. [77] C. U. Saraydar, N. B. Mandayam, and D. J. Goodman, “Efficient power control via pricing in wireless data networks,” IEEE Trans. Commun., vol. 50, no. 2, pp. 291–303, Feb. 2002. [78] F. Meshkati, H. V. Poor, S. C. Schwartz, and N. B. Mandayam, “An energy-efficient approach to power control and receiver design in wireless data networks,” IEEE Trans. Commun., vol. 53, no. 11, pp. 1885–1894, Nov. 2005. [79] G. I. Tsiropoulos, D. G. Stratogiannis, H.-H. Chen, and P. G. Cottis, “Utility-based probabilistic call admission control for complete fairness in wireless networks,” Int. J. Commun. Syst., vol. 1, no. 1, pp. 1099– 1131, Jul. 2012. [80] P. Fishburn and A. Odlyzko, “Dynamic behavior of differential pricing and quality of service options for the internet,” Decis. Support Syst., vol. 28, no. 1/2, pp. 123–136, Mar. 2000. [81] K. Arrow, “The theory of risk aversion,” in Aspects of the Theory of Risk Bearing. Helsinki, Finland: Yrjo Jahnssonin Sattio, 1965, pp. 90–109, [Reprinted in: Essays in the Theory of Risk Bearing, Markham Publ. Co., Chicago, 1971]. [82] N. Gatsis, A. G. Marques, and G. B. Giannakis, “Utility-based power control for peer-to-peer cognitive radio networks with heterogeneous QoS constraints,” in Proc. IEEE ICASSP, 2008, pp. 2805–2808. [83] N. Gatsis, A. Marques, and G. Giannakis, “Power control for cooperative dynamic spectrum access networks with diverse QoS constraints,” IEEE Trans. Commun., vol. 58, no. 3, pp. 933–934, Mar. 2010. [84] E. Dall’ Anese, S.-J. Kim, G. B. Giannakis, and S. Pupolin, “Power control for cognitive radio networks under channel uncertainty,” IEEE Trans. Wireless Commun., vol. 10, no. 10, pp. 3541–3551, Oct. 2011. [85] F.-Y. Tsuo, W.-L. Lee, C.-Y. Wang, and H.-Y. Wei, “Power control game with SINR-pricing in variable-demand wireless data networks,” in Proc. IEEE VTC, 2010, pp. 1–5. [86] D. A. Schmidt, C. Shi, R. A. Berry, M. L. Honig, and W. Utschick, “Pricing algorithms for power control and beamformer design in interference networks,” IEEE Signal Process. Mag., vol. 26, no. 5, pp. 53–63, Sep. 2009. [87] Z. Liang and W. T. Chee, “Cognitive radio network duality and algorithms for utility maximization,” IEEE J. Sel. Areas Commun., vol. 31, no. 3, pp. 500–513, Mar. 2013. [88] S. Stanczak, M. Wiczanowski, and H. Boche, Resource Allocation in Wireless Networks: Theory and Algorithms. Berlin, Germany: Springer-Verlag, 2006. [89] O. Ileri, D. Samardzija, T. Sizer, and N. B. Mandayam, “Demand responsive pricing and competitive spectrum allocation via a spectrum server,” in Proc. IEEE DySPAN, 2005, pp. 194–202.

[90] D. Niyato and E. Hossain, “Spectrum trading in cognitive radio networks: A market-equilibrium-based approach,” IEEE Wireless Commun., vol. 15, no. 6, pp. 71–80, Dec. 2008. [91] C. Yang, J. Tao, W. Chonggang, and Z. Lei, “CRAC: Cognitive radio assisted cooperation for downlink transmissions in OFDMA-based cellular networks,” IEEE J. Sel. Areas Commun., vol. 30, no. 9, pp. 1614–1622, Oct. 2012. [92] Y.-W. Chan, F.-T. Chien, R. Y. Chang, M.-K. Chang, and Y.-C. Chung, “Spectrum sharing in multi-channel cooperative cognitive radio networks: A coalitional game approach,” Wireless Netw., vol. 19, no. 7, pp. 1553–1562, Jan. 2013. [93] L. Yu, C. Liu, and W. Hu, “Spectrum allocation algorithm in cognitive ad-hoc networks with high energy efficiency,” in Proc. ICGCS, 2010, pp. 349–354. [94] G. Mengyao and W. Shaowei, “Fast optimal resource allocation is possible for multiuser OFDM-based cognitive radio networks with heterogeneous services,” IEEE Trans. Wireless Commun., vol. 11, no. 4, pp. 1500–1509, Apr. 2012. [95] M. V. Nguyen, T. Q. Duong, C. S. Hong, S. Lee, and Y. Zhang, “Optimal and sub-optimal resource allocation in multi-hop cognitive radio networks with primary user outage constraint,” IET Netw., vol. 1, no. 2, pp. 47–57, Jun. 2012. [96] H. A. B. Salameh, “Throughput-oriented channel assignment for opportunistic spectrum access networks,” Math. Comput. Model., vol. 53, no. 11/12, pp. 2108–2118, Jun. 2011. [97] X. Kang, H. K. Garg, Y.-C. Liang, and R. Zhang, “Optimal power allocation for OFDM-based cognitive radio with new primary transmission protection criteria,” IEEE Trans. Wireless Commun., vol. 9, no. 6, pp. 2066–2075, Jun. 2010. [98] Z. Hasan, G. Bansal, E. Hossain, and V. K. Bhargava, “Energy-efficient power allocation in OFDM-based cognitive radio systems: A risk-return model,” IEEE Trans. Wireless Commun., vol. 8, no. 12, pp. 6078–6088, Dec. 2009. [99] E. Bedeer, O. A. Dobre, M. H. Ahmed, and K. Baddour, “A multiobjective optimization approach for optimal link adaptation of OFDM-based cognitive radio systems with imperfect spectrum sensing,” IEEE Trans. Wireless Commun., vol. 13, no. 4, pp. 2339–2351, Apr. 2014. [100] V. Vapnik, “An overview of statistical learning theory,” IEEE Trans. Neural Netw., vol. 10, no. 5, pp. 988–999, Sep. 1999. [101] Y. Tachwali, B. F. Lo, I. F. Akyildiz, and R. Agusti, “Multiuser resource allocation optimization using bandwidth-power product in cognitive radio networks,” IEEE J. Sel. Areas Commun., vol. 31, no. 3, pp. 451–463, Mar. 2013. [102] Z. Hasan, E. Hossain, C. Despins, and V. K. Bhargava, “Power allocation for cognitive radios based on primary user activity in an OFDM system,” in Proc. IEEE GLOBECOM, 2008, pp. 1–6. [103] E. Bedeer, O. Amin, O. A. Dobre, M. Ahmed, and K. Baddour, “Energyefficient power loading for OFDM-based cognitive radio systems with channel uncertainties,” IEEE Trans. Veh. Technol., vol. 64, no. 6, pp. 2672–2677, Jun. 2015. [104] L. Akter and B. Natarajan, “Distributed approach for power and rate allocation to secondary users in cognitive radio networks,” IEEE Trans. Veh. Technol., vol. 60, no. 4, pp. 1526–1538, May 2011. [105] D. I. Kim, L. B. Le, and E. Hossain, “Joint rate and power allocation for cognitive radios in dynamic spectrum access environment,” IEEE Trans. Wireless Commun., vol. 7, no. 12, pp. 5517–5527, Dec. 2008. [106] J. Tang, R. Hincapie, G. Xue, W. Zhang, and R. Bustamante, “Fair bandwidth allocation in wireless mesh networks with cognitive radios,” IEEE Trans. Veh. Technol., vol. 59, no. 3, pp. 1487–1496, Mar. 2010. [107] P. Thulasiramanz, J. Chen, and X. Shen, “Multipath routing and maxmin fair QoS provisioning under interference constraints in wireless multihop networks,” IEEE Trans. Parallel Distrib. Syst., vol. 22, no. 5, pp. 716–728, May 2011. [108] Z. Cai, S. Ji, J. He, L. Wei, and A. G. Bourgeois, “Distributed and asynchronous data collection in cognitive radio networks with fairness consideration,” IEEE Trans. Parallel Distrib. Syst., vol. 25, no. 8, pp. 2020–2029, Aug. 2014. [109] D. Xu, E. Jung, and X. Liu, “Efficient and fair bandwidth allocation in multichannel cognitive radio networks,” IEEE Trans. Mobile Comput., vol. 11, no. 8, pp. 1372–1385, Aug. 2012. [110] R. Fan, H. Jiang, Q. Guo, and Z. Zhang, “Joint optimal cooperative sensing and resource allocation in multichannel cognitive radio networks,” IEEE Trans. Veh. Technol., vol. 60, no. 2, pp. 722–729, Feb. 2011. [111] T. Zhang, B. Wang, and Z. Wu, “Spectrum assignment in infrastructure based cognitive radio networks,” in Proc. IEEE NAECON, 2009, pp. 69–74.


[112] F. P. Kelly, A. K. Maulloo, and D. Tan, “Rate control for communication networks: Shadow prices, proportional fairness and stability,” J. Oper. Res. Soc., vol. 49, no. 3, pp. 237–252, Mar. 1998. [113] J. Mo and J. Walrand, “Fair end-to-end window-based congestion control,” IEEE/ACM Trans. Netw., vol. 8, no. 5, pp. 556–567, Oct. 2000. [114] R. Jain, D. Chiu, and W. Hawe, “A quantitative measure of fairness and discrimination for resource allocation in shared systems,” Digit. Equip. Corp., Maynard, MA, USA, Tech. Rep. DEC-TR-301, 1984. [115] K. Hakim, S. K. Jayaweera, G. El-Howayek, and C. Mosquera, “Efficient dynamic spectrum sharing in cognitive radio networks: Centralized Dynamic Spectrum Leasing (C-DSL),” IEEE Trans. Wireless Commun., vol. 9, no. 9, pp. 2956–2967, Sep. 2010. [116] T. Jiang, H. Wang, and Y. Zhang, “Modeling channel allocation for multimedia transmission over infrastructure based cognitive radio networks,” IEEE Syst. J., vol. 5, no. 3, pp. 417–426, Sep. 2011. [117] D. N. Nguyen and M. Krunz, “Price-based joint beamforming and spectrum management in multi-antenna cognitive radio networks,” IEEE J. Sel. Areas Commun., vol. 30, no. 11, pp. 2295–2305, Dec. 2012. [118] R. E. Irwin, A. B. MacKenzie, and L. A. DaSilva, “Resource-minimized channel assignment for multi-transceiver cognitive radio networks,” IEEE J. Sel. Areas Commun., vol. 31, no. 3, pp. 442–450, Mar. 2013. [119] X. Li, T. Drive, and S. A. R. Zekavat, “Distributed channel assignment in cognitive radio networks,” in Proc. IWCMC, 2009, pp. 989–993. [120] A. Plummer and S. Biswas, “Distributed spectrum assignment for cognitive networks with heterogeneous spectrum opportunities,” Wireless Commun. Mobile Comput., vol. 11, no. 9, pp. 1239–1253, Sep. 2011. [121] L. Ding, T. Melodia, S. Batalama, J. Matyjas, and M. Medley, “Crosslayer routing and dynamic spectrum allocation in cognitive radio ad hoc networks,” IEEE Trans. Veh. Technol., vol. 59, no. 4, pp. 1969–1979, May 2010. [122] M. Shin, S. Lee, and Y.-A. Kim, “Distributed channel assignment for multi-radio wireless networks,” in Proc. IEEE MASS, 2006, pp. 417–426. [123] M. Gong, S. Midkiff, and S. Mao, “Design principles for distributed channel assignment in wireless ad hoc networks,” in Proc. IEEE ICC, 2005, pp. 3401–3406. [124] Q. Zhao, L. Tong, and A. Swami, “Decentralized cognitive MAC for opportunistic spectrum access in ad hoc networks: A POMDP framework,” IEEE J. Sel. Areas Commun., vol. 25, no. 3, pp. 589–600, Apr. 2007. [125] Y. Chen, Q. Zhao, and A. Swami, “Distributed spectrum sensing and access in cognitive radio networks with energy constraint,” IEEE Trans. Signal Process., vol. 57, no. 2, pp. 783–797, Feb. 2009. [126] A. Nasipuri, J. Zhuang, and S. R. Das, “A multichannel CSMA MAC protocol for multihop wireless networks,” in Proc. IEEE WCNC, 1999, pp. 1402–1406. [127] J. P. R. Draves and B. Zill, “Routing in multiradio multihop wireless mesh networks,” in Proc. ACM MobiCom, 2004, pp. 114–128. [128] J. So and N. H. Vaidya, “Multichannel MAC for ad hoc networks: Handling multichannel hidden terminals using a single transceiver,” in Proc. MobiHoc, 2004, pp. 222–233. [129] A. Raniwala and T. Chiueh, “Architecture and algorithms for an IEEE 802.11-based multichannel wireless mesh network,” in Proc. IEEE INFOCOM, 2005, pp. 2223–2234. [130] K. R. Chowdhury and I. F. Akyildiz, “Cognitive wireless mesh networks with dynamic spectrum access,” IEEE J. Sel. Areas Commun., vol. 26, no. 1, pp. 168–181, Jan. 2008. [131] F. Akyildiz, X. Wang, and W. Wang, “Wireless mesh networks: A survey,” Comput. Netw. J., vol. 47, no. 4, pp. 445–487, Mar. 2005. [132] H. Shiang and M. van der Schaar, “Distributed resource management in multi-hop cognitive radio networks for delay sensitive transmission,” IEEE Trans. Veh. Technol., vol. 58, no. 2, pp. 941–953, Feb. 2009. [133] Z. Ji and K. J. R. Liu, “Dynamic spectrum sharing: A game theoretical overview,” IEEE Commun. Mag., vol. 45, no. 5, pp. 88–94, May 2007. [134] B. Wang, Y. Wu, and K. J. R. Liu, “Game theory for cognitive radio networks: An overview,” Comput. Netw. J., vol. 54, no. 14, pp. 2537– 2561, Oct. 2010. [135] S. Buljore et al., “Architecture and enablers for optimized radio resource usage in heterogeneous wireless access networks: The IEEE 1900.4 Working Group,” IEEE Commun. Mag., vol. 47, no. 1, pp. 122–129, Jan. 2009. [136] R. Etkin, A. Parekh, and D. Tse, “Spectrum sharing for unlicensed bands,” IEEE J. Sel. Areas Commun., vol. 25, no. 3, pp. 517–528, Apr. 2007. [137] C. Li and J. Xie, “Repeated game-inspired spectrum sharing for clustering cognitive ad hoc networks,” Int. J. Distrib. Sensor Netw., vol. 2013, pp. 514 037-1–514 037-10, Jan. 2013.

845

[138] M. Le Treust and S. Lasaulce, “A repeated game formulation of energyefficient decentralized power control,” IEEE Trans. Wireless Commun., vol. 9, no. 9, pp. 2860–2869, Sep. 2010. [139] W. Yuan, H. Leung, W. Cheng, S. Cheng, and B. Chen, “Participation in repeated cooperative spectrum sensing: A game-theoretic perspective,” IEEE Trans. Wireless Commun., vol. 11, no. 3, pp. 1000–1011, Mar. 2012. [140] J. W. Friedman, “A non-cooperative equilibrium for supergames,” Rev. Econ. Stud., vol. 38, no. 1, pp. 1–12, Jan. 1971. [141] I. Obara, “Folk theorem with communication,” J. Econ. Theory, vol. 144, no. 1, pp. 120–134, Jan. 2009. [142] B. Eraslan, D. Gozupek, and F. Alagoz, “An auction theory based algorithm for throughput maximizing scheduling in centralized cognitive radio networks,” IEEE Commun. Lett., vol. 15, no. 7, pp. 734–736, Jul. 2011. [143] M. Dong, G. Sun, X. Wang, and Q. Zhang, “Combinatorial auction with time-frequency flexibility in cognitive radio networks,” in Proc. IEEE INFOCOM, 2012, pp. 2282–2290. [144] D. Niyato, E. Hossain, and Z. Han, “Dynamic spectrum access in IEEE 802.22-based cognitive wireless networks: A game theoretic model for competitive spectrum bidding and pricing,” IEEE Wireless Commun., vol. 16, no. 2, pp. 16–23, Apr. 2009. [145] Y. Zhang, D. Niyato, P. Wang, and E. Hossain, “Auction-based resource allocation in cognitive radio systems,” IEEE Commun. Mag., vol. 50, no. 11, pp. 108–120, Nov. 2012. [146] Y. Zhang, C. Lee, D. Niyato, and P. Wang, “Auction approaches for resource allocation in wireless systems: A survey,” IEEE Commun. Surveys Tuts., vol. 15, no. 3, pp. 1020–1041, 2013. [147] J. Huang, R. Berry, and M. L. Honig, “Auction-based spectrum sharing,” Mobile Netw. Appl., vol. 11, no. 3, pp. 405–418, Jun. 2006. [148] A. Wang et al., “Joint subcarrier and power allocation for physical layer security in cooperative OFDMA networks,” EURASIP J. Wireless Commun. Netw., vol. 2013, no. 1, pp. 193:1–193:10, Apr. 2013. [149] Y. Liu, M. Tao, and J. Huang, “An auction approach to distributed power allocation for multiuser cooperative networks,” IEEE Trans. Wireless Commun., vol. 12, no. 1, pp. 237–247, Jan. 2013. [150] L. Yang, H. Kim, J. Zhang, M. Chiang, and C. W. Tan, “Pricing-based decentralized spectrum access control in cognitive radio networks,” IEEE/ACM Trans. Netw., vol. 21, no. 2, pp. 522–535, Apr. 2013. [151] A. Mas-Colell, M. D. Whinston, and J. R. Green, Microeconomic Theory. London, U.K.: Oxford Univ. Press, 1995. [152] V. Krishna, Auction Theory. San Diego, CA, USA: Academic, 2002. [153] H.-B. Chang and K.-C. Chen, “Auction-based spectrum management of cognitive radio networks,” IEEE Trans. Veh. Technol., vol. 59, no. 4, pp. 1923–1935, May 2010. [154] B. Wang, K. J. R. Liu, and T. C. Clancy, “Evolutionary game framework for behavior dynamics in cooperative spectrum sensing,” in Proc. IEEE GLOBECOM, 2008, pp. 1–5. [155] B. Wang, K. J. R. Liu, and T. C. Clancy, “Evolutionary cooperative spectrum sensing game: How to collaborate?” IEEE Trans. Commun., vol. 58, no. 3, pp. 890–900, Mar. 2010. [156] J. M. Smith, Evolution and the Theory of Games. Cambridge, U.K.: Cambridge Univ. Press, 1982. [157] J. W. Weibull, Evolutionary Game Theory. Cambridge, MA, USA: MIT Press, 1995. [158] H. Wu et al., “Distributed channel assignment and routing in multiradio multi-channel multi-hop wireless networks,” IEEE J. Sel. Areas Commun., vol. 24, no. 11, pp. 1972–1983, Nov. 2006. [159] D. Bertsekas and R. Gallager, Data Networks. Upper Saddle River, NJ, USA: Prentice-Hall, 1987. [160] L.-C. Wang, C.-W. Wang, and C.-J. Chang, “Modeling and analysis for spectrum handoffs in cognitive radio networks,” IEEE Trans. Mobile Comput., vol. 11, no. 9, pp. 1499–1513, Sep. 2012. [161] S. Gao, L. Qian, and D. R. Vaman, “Distributed energy efficient spectrum access in wireless cognitive radio sensor networks,” in Proc. IEEE Wireless Commun. Netw. Conf., 2008, pp. 1442–1447. [162] A. M. Wyglinski, “Effects of bit allocation on non-contiguous multicarrier-based cognitive radio transceivers,” in Proc. IEEE VTC Fall, 2006, pp. 1–5. [163] S. Chen and A. M. Wyglinski, “Efficient spectrum utilization via crosslayer optimization in distributed cognitive radio networks,” Comput. Commun., vol. 32, no. 18, pp. 1931–1943, Dec. 2009. [164] D. Chen, Q. Zhang, and W. Jia, “Aggregation aware spectrum assignment in cognitive ad-hoc networks,” in Proc. Int. Conf. Cogn. Radio Oriented Wireless Netw. Commun., 2008, pp. 1–6. [165] Q. Zhang, J. Jia, and X. Shen, “Optimal spectrum sensing decision for hardware-constrained cognitive networks,” in Cognitive Wireless

846

[166] [167] [168] [169] [170]

[171] [172] [173] [174] [175] [176] [177] [178] [179] [180] [181] [182] [183] [184] [185]


Communication Networks. New York, NY, USA: Springer-Verlag, 2007. “A study of the provision of aggregation of frequency to provide wider bandwidth services,” Office Commun., London, U.K., Aug. 2006, QINETIQ/06/01773. X. Li, D. Wang, and J. McNair, “Residual energy aware channel assignment in cognitive radio sensor networks,” in Proc. IEEE Wireless Commun. Netw. Conf., 2011, pp. 398–403. P. Setoodeh and S. Haykin, “Robust transmit power control for cognitive radio,” Proc. IEEE, vol. 97, no. 5, pp. 915–939, May 2009. J. W. Mwangoka, K. B. Letaief, and Z. Cao, “Joint power control and spectrum allocation for cognitive radio networks via pricing,” Phys. Commun., vol. 2, no. 1/2, pp. 103–115, Mar.–Jun. 2009. T. Jiang, D. Grace, and Y. Liu, “Cognitive radio spectrum sharing schemes with reduced spectrum sensing requirements,” in Proc. IET Semin. Cogn. Radio Softw. Defined Radios—Technol. Tech., 2008, pp. 1–5. D. Grace, J. Chen, T. Jiang, and P. D. Mitchell, “Using cognitive radio to deliver ‘Green’ communications,” in Proc. CROWNCOM, 2009, pp. 1–6. I. F. Akyildiz, W.-Y. Lee, and K. R. Chowdhury, “CRAHNs: Cognitive radio ad hoc networks,” Ad Hoc Netw., vol. 7, no. 5, pp. 810–836, Jul. 2009. M. E. Monteiro, “Cognitive radio: Survey on communication protocols, spectrum decision issues, future research directions,” Wireless Netw. J., vol. 18, no. 2, pp. 147–164, Feb. 2012. L. Yang, L. Cao, and H. Zheng, “Proactive channel access in dynamic spectrum network,” in Proc. CROWNCOM, 2007, pp. 482–486. C. Clancy, J. Hecker, and E. Stuntebeck, “Applications of machine learning to cognitive radio networks,” IEEE Wireless Commun., vol. 14, no. 4, pp. 47–52, Aug. 2007. L. R. Rabiner, “A tutorial on hidden Markov models and selected applications in speech recognition,” Proc. IEEE, vol. 77, no. 2, pp. 257–285, Feb. 1989. S. Haykin, Neural Networks: A Comprehensive Foundation. Upper Saddle River, NJ, USA: Prentice-Hall, 1998. K. Tsagkaris, A. Katidiotis, and P. Demestichas, “Neural networkbased learning schemes for cognitive radio systems,” Comput. Commun., vol. 31, no. 14, pp. 3394–3404, Sep. 2008. D. Goldberg, Genetic Algorithms in Search, Optimization, Machine Learning. Reading, MA, USA: Addison-Wesley, 1989. M. Di Felice, K. Chowdhury, C. Wu, W. Meleis, and L. Bononi, “Learning-based spectrum selection in cognitive radio ad hoc networks,” in Proc. IEEE WWIC, 2010, pp. 133–145. A. Anandukumar, N. Michael, and A. Tang, “Opportunistic spectrum access with multiple users: Learning under competition,” in Proc. IEEE INFOCOM, 2010, pp. 1–9. G. Zhao, G. Y. Li, C. Yang, and J. Ma, “Proactive detection of spectrum holes in cognitive radio,” in Proc. IEEE ICC, 2009, pp. 1–5. G. Zhao, G. Y. Li, and C. Yang, “Proactive detection of spectrum opportunities in primary systems with power control,” IEEE Trans. Wireless Commun., vol. 8, no. 9, pp. 4815–4823, Sep. 2009. K.-C. Chen et al., “Routing for cognitive radio networks consisting of opportunistic links,” Wireless Commun. Mobile Comput., vol. 10, no. 4, pp. 451–466, Apr. 2010. I. Pefkianakis, S. H. Wong, and S. Lu, “SAMER: Spectrum aware mesh routing in cognitive radio networks,” in Proc. IEEE DySPAN, 2008, pp. 1–5.

Georgios I. Tsiropoulos received the Diploma (with first-class honors) in electrical and computer engineering, the M.Sc. degree in technoeconomics, and the Ph.D. degree from National Technical University of Athens (NTUA), Athens, Greece, in 2005, 2009, and 2010, respectively. From 2006 to 2009, he served as a Technical Consultant with the Ministry of Transport and Communications of the Hellenic Republic. Since 2006, he has been a Research Assistant with the Wireless and Satellite Communications Group, NTUA. Additionally, since 2011, he has been an ICT Consultant at the Union of Hellenic Chamber of Commerce, Athens. He has been also engaged in many professional activities. He has authored or coauthored over 25 papers in refereed professional journals and conferences. His research interests span the different areas of the wireless, satellite, mobile, and heterogeneous networks, including radio resource management, quality of service, and mobile ad hoc networks. Dr. Tsiropoulos is a member of the Technical Chamber of Greece, the IEEE, and other international organizations.

Octavia A. Dobre received the Dipl.Ing. and Ph.D. degrees in electrical engineering from Politehnica University of Bucharest (formerly Polytechnic Institute of Bucharest), Bucharest, Romania, in 1991 and 2000, respectively. In 2005, she joined Memorial University of Newfoundland, St. John’s, Canada, where she is currently an Associate Professor. Previously, she was with New Jersey Institute of Technology, Newark, NJ, USA, and Polytechnic Institute of Bucharest. In 2000, she was a recipient of a Royal Society Scholarship at the University of Westminster, London, U.K., and in 2001, she held a Fulbright Fellowship at Stevens Institute of Technology, Hoboken, NJ. Her research interests include cognitive radio systems, spectrum sensing techniques, blind signal recognition and parameter estimation techniques, transceiver optimization algorithms, dynamic spectrum access, cooperative wireless communications, network coding, resource allocation, and optical communications. She has authored or coauthored over 130 referred journal and conference papers in these areas. Dr. Dobre has been a Cochair for the IEEE GLOBECOM 2015 (Communications Theory Symposium), IEEE WCNC 2015 (PHY and Fundamentals Track), IEEE CrownCom 2015 (Emerging Applications for Cognitive Networks Track), IEEE VTC Fall 2014 (Spectrum Sensing and Cognitive Radio Track), IEEE ICC 2013 and IEEE GLOBECOM 2013 (Signal Processing for Communications Symposium), IEEE VTC Spring 2013 (Multiple Antenna Systems and Services Track), ICNC 2012 (Signal Processing for Communications Symposium), and IEEE CCECE 2009 (Signal and Multimedia Processing Track). She is currently serving as the Chair of the IEEE Communications Society’s Women in Communications Engineering Committee and the Cochair of the IEEE Measurement and Instrumentation Society’s Wireless and Telecommunications in Measurements Technical Committee. She is a Senior Editor of IEEE Communications Letters, an Editor of IEEE Communications Surveys and Tutorials and Elsevier Physical Communication, and a Guest Editor of the IEEE Communications Magazine. She has served as an Editor of IEEE Communications Letters, a Guest Editor of the IEEE Journal of Selected Topics in Signal Processing, and a Lead Guest Editor of the Elsevier Physical Communication “Cognitive Radio: The Road for its Second Decade” special issue. She is a Registered Professional Engineer in the province of Newfoundland, Canada.

Mohamed Hossam Ahmed received the B.Sc. and M.Sc. degrees in electronics and communications engineering from Ain Shams University, Cairo, Egypt, in 1990 and 1994, respectively, and the Ph.D. degree in electrical engineering from Carleton University, Ottawa, Canada, in 2001. From 2001 to 2003, he was a Senior Research Associate with Carleton University. In 2003, he joined the Faculty of Engineering and Applied Science, Memorial University of Newfoundland, St. John’s, Canada, where he is currently an Associate Professor. He has authored or coauthored over 100 papers in international first-class journals and conferences. His research interests include radio resource management in wireless networks, multihop relaying, cooperative communication, vehicular ad hoc networks, cognitive radio networks, and wireless sensor networks. His research is sponsored by NSERC, CFI, Bell Aliant, and other governmental and industrial agencies. Dr. Ahmed is a Senior Member of the IEEE. He serves as a Cochair of the Signal Processing Track in ISSPIT’14 and served as a Cochair of the Transmission Technologies Track in VTC’10-Fall and the Multimedia and Signal Processing Symposium in CCECE’09. He serves as an Editor of IEEE Communication Surveys and Tutorials and the EURASIP Journal on Wireless Communications and Networking (JWCN) and as an Associate Editor of the Wiley International Journal of Communication Systems and Wiley Wireless Communications and Mobile Computing. He served as a Guest Editor of a special issue on Fairness of Radio Resource Allocation, EURASIP JWCN in 2009 and a special issue on Radio Resource Management in Wireless Internet, Wiley Wireless Communications and Mobile Computing Journal in 2003. He was a recipient of the Ontario Graduate Scholarship for Science and Technology in 1997; the Ontario Graduate Scholarship in 1998, 1999, and 2000; and the Communication and Information Technology Ontario graduate award in 2000. He is a Registered Professional Engineer in the province of Newfoundland, Canada.


Kareem E. Baddour received the bachelor’s degree from Memorial University of Newfoundland, St. John’s, Canada, and the M.Sc. and Ph.D. degrees in electrical engineering from Queen’s University, Kingston, Canada. Since 2006, he has been a Research Scientist at the Communications Research Centre Canada, Ottawa, Canada. His research interests are in signal processing for wireless communications with a current focus on dynamic spectrum access networks. He has authored or coauthored numerous articles in these areas. Dr. Baddour has served on the technical program committees of major international communications conferences. He was a Cochair of the Cognitive Radio and Spectrum Sensing Track of IEEE VTC Fall 2012 and the First and Second Workshops on Cognitive Radio Advances, Applications and Future Emerging Technologies at ISWCS 2013 and 2014. He was a Guest Editor of the Elsevier PHYCOM “Cognitive Radio: The Road for its Second Decade” special issue. He was a corecipient of the Best Paper Award at the International Symposium on Wireless Communication Systems in 2010.

847