Efficient Radio Resource Management Algorithms in Opportunistic Cognitive Radio Networks Athina Bourdena1, Evangelos Pallis2, Georgios Kormentzas1 and George Mastorakis2 1Department

2Department

of Information and Communication Systems Engineering, University of the Aegean, Samos, Greece of Applied Informatics and Multimedia, Technological Educational Institute of Crete, Heraklion, Crete, Greece

ABSTRACT Cognitive radio (CR) paradigm was introduced, towards addressing challenges, related with radio spectrum scarcity and increased needs for wireless networking services provision. In this direction, CR networks exploit novel networking architectures, as well as dynamic radio spectrum access techniques and methods, alleviating problems, regarding limited wireless networking resources and their inefficient usage/exploitation. CR terminals exploit innovative mechanisms to identify un-used parts of radio spectrum, such as TV White Spaces (TVWS) in UHF/VHF bands, following an interference-free opportunistic manner. However, introduction of CR networks creates new challenges that are highly related to the fluctuation of TVWS, as they vary over time and location, as well as issues related to diverse Quality of Service (QoS) requirements. In this context, this paper proposes two radio resource management (RRM) algorithms, enabling for the opportunistic exploitation of TVWS in a centralized cognitive radio networking architecture. Efficient administration of radio spectrum resources is achieved, by exploiting a novel RRM framework, adopted in a spectrum broker, which is in charge to effectively orchestrate the available wireless networking resources. Efficient RRM algorithms performance, as a matter of maximum-possible spectrum broker benefit and radio spectrum utilization, as well as minimum-possible spectrum fragmentation is evaluated, by considering a fixed-price and an auction-based optimization approach. Experimental tests that were conducted under controlled simulation conditions, confirmed the validity of both RRM algorithms adopted in the proposed CR networking architecture, identifying fields for further research and experimentation.

KEY WORDS Opportunistic cognitive radio networks, TV White Spaces, radio resource management algorithms, spectrum broker, centralized networking architectures

*Correspondence A. Bourdena, Department of Information and Communication Systems Engineering, University of the Aegean, Samos, Greece E-mail: [email protected] † This paper was presented in part at the 1st ACM Workshop on High Performance Mobile Opportunistic Systems (HP-MOSys 2012), 15th ACM International Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems, in cooperation with R8 IEEE Computer Society C16, Paphos, Cyprus, 21-25 October 2012.

1. INTRODUCTION Demand for radio spectrum was dramatically raised during previous years due to the ongoing growth of mobile communication services provision, creating new challenges in wireless networks resources management and administration. Important studies and reports on radio spectrum exploitation [1] (e.g. Federal Communications Commission studies) have proved that the existing “command-and-control” framework/policy is insufficient to address current demands for wireless services provision. Furthermore, other reports [2] revealed that a great number of radio spectrum blocks/resources, like TV White Spaces (TVWS) [3] is under-utilized or idle for long time periods in certain geographical areas. TVWS include VHF/UHF channels that are either released by the digital switchover process (“Spectrum/Digital Dividend”), or are completely un-used primarily at specific locations, according to frequency allocation rules and wireless networks planning principles (“Interleaved Spectrum”) [3]. In addition, dynamic

spectrum access (DSA) paradigm was introduced, towards addressing vital radio spectrum utilization challenges [4], [5]. DSA is highly related with cognitive radio (CR) networks [6], [7], enabling for opportunistic exploitation of radio spectrum resources [4], [8], in CR networking architectures [9]. Towards this direction, CR networking architectures are categorized, either as centralized ones, if the decision of radio spectrum access is taken, by a central controller/module or distributed ones, in case that the decision is taken locally, by each individual frequency-agile device. Such devices have the ability to search for radio spectrum opportunities, discover potential un-exploited channels and adapt their operation parameters [10], [11]. Nevertheless, adoption of opportunistic CR networking architectures cannot be an easy process, particularly in licensed radio spectrum channels, where “command-and-control” policy prohibits

operation of secondary communication systems. Therefore, novel policies for radio spectrum administration and management are vital to be introduced, towards permitting the efficient operation of opportunistic CR networking architectures in unlicensed frequency bands. Along with the envisaged policies are the “Spectrum of Commons” (i.e. unlicensed policy) and the “Real-time Secondary Spectrum Market-RTSSM” (i.e. licensed policy) [12], [13], [14]. The former (i.e. “Spectrum of Commons”) represents the case, where coexistence with incumbent primary transmissions (e.g. DVB-T) is guaranteed, by avoiding possible interference, in comparison to a fixed radio spectrum allocation policy. There is no spectrum manager in such a case, to control over network resources allocation, similarly to the wireless ISM bands, where communication systems are required to fulfill a number of technical regulations, in order to guarantee excellent coexistence with other existing wireless networks. Nevertheless, even though this process promotes effectiveness through radio spectrum sharing, Quality of Service (QoS) is not possible to be assured, which is a vital issue especially for QoS-sensitive applications and services. On the other hand, RTSSM policy enables primary systems (i.e. licensed holders of radio spectrum) to deal rights of wireless networking resources with secondary systems, in order to establish a secondary market for opportunistic spectrum exploitation. In this case, primary systems run admission control algorithms, allowing secondary systems to access radio spectrum, only when QoS is sufficient. Leasing of secondary radio spectrum usage may also occur through an intermediary networking entity, such as a spectrum broker, using radio resource management (RRM) algorithms to determine channels, at which a secondary system is able to operate together with economic issues of transactions occurred. Secondary systems on the other hand, dynamically demand access to radio spectrum and are charged according to a spectrum utilization basis process. RTSSM policy is the most appropriate solution, for radio spectrum leasing and RRM, especially for applications that require QoS guaranteed sporadic/dynamic spectrum access, while it respects requirements of all involved participants (i.e. radio spectrum regulators, radio spectrum broker, primary and secondary systems). Spectrum regulators interest to efficiently exploit TVWS, while radio spectrum broker wishes to increase its revenue/profit, no matter

how many TVWS are required. Primary systems allow for the opportunistic operation of secondary systems, on condition that the latter do not cause possible interference. Also, secondary systems try to increase the opportunity of accessing radio spectrum bands, under a guaranteed QoS. By attempting to satisfy all participants, selfish behavior, conflicts or cooperation arise among them. Broad research work has been conducted based on economic related issues, towards addressing conflict or cooperation problems, based on game theory [15], contract theory [16], auction-based processes [17] and commodity pricing [18]. Current related research schemes include auction-based algorithms, adopted to address radio spectrum assignment challenges [19], elaborating on fairness, efficiency and valuation independence issues [20]. A vital issue for current research approaches is to assure economic properties, such as truthfulness [20], indicating that bids offered by secondary systems in order to request admission to wireless network resources, reveal their actual valuation. In this direction, this paper proposes a radio resource management framework and two optimization algorithms, enhancing the research approaches presented in previous authors’ published work at [12], [13], [21] and [22]. The proposed RRM algorithms enable for opportunistically lease of un-used TVWS, by secondary systems (i.e. cellular/wireless network providers), respecting several constrains and guaranteeing QoS related requirements, like restrictions associated with maximum transmission power thresholds and possible interference. Towards addressing such challenges, TVWS leasing methods are anticipated, functioning into a central networking unit, namely spectrum broker. This unit operates by optimally assigning radio spectrum resources to secondary systems in specific geographical locations, according to a combinatorial auction-based process. Spectrum broker is able to enhance its revenue, by minimizing spectrum fragmentation, in case that a fixed-price policy is adopted [23], or maximizing revenues and radio spectrum exploitation efficiency, according to an auction-based policy. The auctionbased algorithm, proposed in this paper, takes into account both frequency and time domains, exploiting the second revenue model (i.e. spectrum auctions), while research work regarding the fixed-price model is analytically presented in [12], [13]. In this context and according to this introductory section, next section elaborates on related work and research motivation, while section 3 discusses the

proposed radio resource management framework, based on a centralized CR networking architecture that opportunistically operates following RTSSM policy. Section 4 elaborates on radio spectrum leasing problem formulation and provides performance evaluation results based on simulation experiments of both RRM algorithms, in terms of spectrum broker benefit, radio spectrum utilization, spectrum fragmentation and TVWS access probability. Finally, section 5 concludes the paper, by highlighting challenges for future research approaches and further experimentation.

2. RELATED WORK AND RESEARCH MOTIVATION A vital challenge in dynamic spectrum access issues is to provide efficient radio spectrum utilization, satisfying most of opportunistic networking devices and increasing profit of spectrum seller (i.e. spectrum broker). The most optimal allocation of the available resources (e.g. radio spectrum opportunities) in a CR network, is possible to be achieved, by considering efficient co-operation between an RRM entity and a radio spectrum trading entity. The former is in charge to efficiently assign available wireless networking resources (e.g. TVWS), as a matter of maximum possible utilization of them and minimum spectrum fragmentation, by utilizing optimization approaches and methods [9]. Although excellent efforts [24], [25], [26], [27] have been proposed, towards addressing RRM challenges in CR networks, all research approaches mainly focus on resource management among primary and secondary systems, assuming an ideal spectrum sensing by secondary ones. However, an ideal spectrum sensing is impractical due to several issues such as, network connectivity, short sensing periods and hardware limitations [28]. Moreover, the ongoing developments regarding the provision of multiple wireless networking services require to support diverse QoS. Most existing research approaches elaborate on a single type of service provided by secondary systems. Moreover, RRM in a broker-based architecture that receives radio spectrum information from a geo-location database with heterogeneous services with guaranteed QoS, has been studied in [29], while the economic interaction among spectrum broker and secondary systems is discussed in the same research work.

In addition, the networking entity that is in charge to trade wireless resources, undertakes economic issues and transactions of radio spectrum leasing, by considering a price per unit of spectrum (e.g. cost per MHz). The aim of this networking entity (e.g. spectrum broker) is to increase its profit, during spectrum assignment process, while secondary systems (e.g. spectrum buyers) wish to maximize utility of frequency resources usage and their fulfillment, regarding QoS provisioning. Such objectives are usually in conflict among them and a stable solution, regarding radio spectrum assignment and pricing is required, in order both resources seller and buyer are fulfilled based to their requirements. Towards this direction, pricing schemes are considered as a vital issue that is directly related to spectrum assignment, keeping fairness among secondary systems and offering profit to radio spectrum seller (e.g. spectrum broker). Moreover, an interesting research field in CR networks is resource allocation and trading from a game theoretic point of view. More specifically in several cases, secondary systems that dynamically and intelligently adapt their transmission characteristics do not have incentives to collaborate among them and behave selfishly. A game theoretic perspective investigates this behavior and provides solutions to wireless networking resources sharing, which is described as a multi-objective optimization problem. The spectrum sharing issue is possible to be modeled as a game between primary and secondary systems with incentives to conflict or to cooperate. Based on this approach, game theory is capable to provide comprehensive and straightforward equilibrium criteria, in order to evaluate game optimality, under various game settings. Extensive and comprehensive research approaches have also been anticipated, according to economic interactions or decision-making in CR networks. Figure 1 presents the classification of solutions in radio spectrum allocation and trading problem based on game theory that attempts to study conflicts and cooperation among individual players, through mathematical tools, towards modeling dynamic radio spectrum sharing problem in CR networks [30]. This is an optimization approach of multi-players, while this concept is valid when actions of a number of players are common.

Figure 1 Economic-oriented Classification of Radio Spectrum Allocation Solutions

Radio spectrum management issues in economic perspective can usually be executed following either a) cooperative or b) non-cooperative game theory or c) price theory concept or d) market theory concept. In a cooperative game, all participants collaborate, by acting as a distinct entity, in order to maximize overall utility. For instance, a bargain game is usually exploited, towards formulating interaction between cooperative systems/players, considering that a system is possible to influence action of further players. Moreover, participants in non-cooperative games behave selfishly, while individual players decide separately. In a non-cooperative environment, players incentives usually reach to conflict. A non-cooperative game is usually exploited, in order to reach to an equilibrium solution, optimizing payoff for all players. Such an example is Nash equilibrium [31], where each player in the game is not possible to achieve an improved solution, by deviating the equilibrium. Moreover, price theory is exploited, in order to determine the value of an item, based on player incentives. In case of radio spectrum sharing, wellsuited pricing schemes are important to determine the price per spectrum unit, by exploiting economic models and increasing payoff for both primary and secondary players. Pricing is a vital issue, towards increasing revenue of service providers, as well as protecting from unnecessary competition and optimally allocating radio spectrum resources. Furthermore, market theory is an efficient method used for radio spectrum trading. More specifically, market theory can be divided/grouped into Monopoly, Oligopoly and Competitive Equilibrium [9], while cooperative and non-cooperative games exploit game

models, towards analyzing strategic interactions between players for wireless networking resources sharing, such as Stackelberg game, Bertrand game, Cournot game, Coalition game, game with Learning, Repeated game, Bargaining game and Modified Minority game [32]. Additionally, price theory can be performed, by either exploiting Bargain theory or Auction theory. Bargain theory is an alternative strategy, mostly used in cases that price of resources is fixed. In such a case, both buyer and seller come to an agreement, after a conflict regarding the value of the resource. On the other hand, Auction theory [20] has been widely used for determining optimal allocation of scarce resources that have an undetermined or variable price. Through an auction-based process, sellers aim to improve revenue, by assigning the available radio spectrum to buyers who increase their profit. Each player in Auction theory sets a bid that reflects the player’s value. A variety of auction-based processes can be exploited in radio spectrum allocation and trading challenges. The most widely exploited are single, double and combinatorial auction-based processes [9]. Bidders in a single auction set their bids and the auctioneer decides the winner of the process. The major types of single auctions include English auction (i.e. increasing-price auction), Vickrey auction (i.e. first-price sealed-bid and second-price sealed-bid auctions) and Dutch auction (i.e. decreasing-price auction). Also, another set of auction-based processes includes the double auction. In this case, several radio spectrum players trade among them, regarding heterogeneous or homogenous resources/items. Items can be auctioned at the same time (i.e. simultaneous

auction) or sequentially (i.e. sequential auction) [33]. The last set of an auction-based process includes combinatorial auction, where a single seller has multiple heterogeneous items for sale and bidders place their bids for parts of items on an "all-ornothing" basis, (i.e. a bidder will pay for items, only if it gets all of them). The most known combinatorial auction-based processes are the first-price auction [34] and Vickerey-Clark-Grooves (VCG) auction [35]. Although, radio spectrum sharing problem may have a number of possible solutions, the most of related approaches have drawbacks in the applicability. For instance, several approaches based on Nash equilibrium usually experience extreme competition between secondary and primary systems in noncooperative games. This results some times that the solution of the game is usually not achieved or is inefficient. However, auction-based processes are a solution for radio spectrum sharing problems that fulfill requirements of sellers, towards increasing their payoff, as well as buyers requirements to maintain their spectrum usage as possible at the lowest cost. Spectrum auction-based processes have extensively been exploited, in order to solve several types of wireless resources allocation problems. Authors in [36], developed an auction-based algorithm that allows systems to fairly contend for fading channels. For this scope they utilize mechanisms based on a second-price auction process. Also, authors in [37], propose a cognitive radio network that is interference tolerant, where single primary systems share radio spectrum together with secondary systems with limited interference. Two common multi-unit auctions are presented in this research work. The first one is a Vickrey auction, while the second one is a sequential process of a first-price auction, towards assigning wireless resources and utilizing power allocation strategies for an optimum performance. Furthermore, authors in [38] propose two auction-based mechanisms, in order to share radio spectrum between a group of systems, considering several restrictions related with interference temperature at measurement points, while [39] presents auction-based approaches (i.e. SNR and power auctions) to compute relay selection, as well as relay power allocation. The power auction achieves in this case an optimum resources allocation, by increasing total rate, while SNR auction is more flexible regarding efficiency and fairness. Moreover, research approach proposed in [33], analyzes radio spectrum allocation problem, under sequential and concurrent auctions, by considering a

number of bidders that compete to access multiple parts of radio spectrum. Furthermore, authors in [40], consider radio spectrum assignment, as a double auction with multiple primary and secondary systems, proposing an approach based on dynamic pricing, in order to optimize spectrum efficiency and maintain incentives of the systems. On the other hand, an integrated assigning, billing and pricing system is considered in [41] for networking architectures based on cognitive radio paradigm, while a joint power/channel assigning scheme is proposed in [42] that is used to optimize network performance. In a general context, during an auction-based process for leasing radio spectrum resources, bidders propose their bids to the auctioneer, including information related with the price of bidding, as well as the quantity per unit of spectrum. Auctioneer is then in charge to compute winning bidders and enable access to radio spectrum for them. Therefore, radio spectrum is traded at a specific price that is defined, according to the auction process. In such a case, secondary systems state their needs, towards obtaining admission to wireless networking resources, by sending their bids. An auction-based approach enables a number of secondary network operators to dynamically control the availability of radio spectrum, in contrast to research approaches based on fixed price markets, where networking systems are only permitted to inactively access resources, under a first-come-firstserved basis [43]. In this context and considering all the above mentioned state of the art research work, no related research approach has yet considered an auction-based process, by simultaneously assigning TVWS, during both time and frequency domains.

3. RADIO RESOURCE MANAGEMENT FRAMEWORK FOR OPPORTUNISTIC COGNITIVE RADIO NETWORKS This part elaborates on system design of a centralized topology, operating under the RTSSM policy, where radio resource administration and spectrum leasing is carried over a prototype RMM framework. An overview of spectrum broker reference architecture is presented in Figure 2. The centralized topology was adopted, as the most appropriate solution in this case, since QoS guarantee is crucial in the proposed system. Furthermore, such a centralized topology enables for radio spectrum trading, towards creating a spectrum market for secondary resources usage, by secondary systems. Spectrum broker entity in this network topology orchestrates the available portions of radio

spectrum that is assigned to each secondary system, by keeping QoS provisioning at an acceptable level. In this reference architecture, spectrum broker is an intermediary entity among the Geo-location database (i.e. TVWS information provider) and secondary systems that negotiate to access radio spectrum. The Geo-location database includes useful data, in order to create a pool of TVWS channels that are available in particular geographical locations. Furthermore, information, regarding digital terrestrial television protected areas, protection rules and propagation models, are exploited to calculate maximum thresholds of operation transmission power of secondary systems, towards avoiding possible interference with primary systems. Geo-location database is possible to dynamically store new data and continuously change several parameters, regarding the protection of possible interference caused to primary systems. In addition, TVWS spectral utilization efficiency is better than using a sensing alone detection approach. This is primarily due to the ability of Geo-location enabled TVWS devices to accurately determine protected service contours. Secondary systems and devices have in principle two methods to determine if a channel is occupied or not. The first choice is to use sensing techniques, where channels are detected to find incumbent signals at or above certain signal strength. The second choice is to

exploit a Geo-location database, where for a certain region attribution of channels to primary users is presented. In this respect, the proposed CR network adopts a hybrid approach, where local sensing information is combined with Geo-location database information to compute the TVWS spectrum pool, in order to obtain optimum channel occupancy and minimize false alarm rate (a drawback of autonomous sensing technique). Moreover, with a database, part of the complexity associated with sensing and maximum power computation is transferred to the core network, decreasing complexity and power demand of TVWS devices. The database has the ability to be dynamically updated and continuously adjust interference protection parameters in line with the evolution of incumbent standards, e.g. DVB-T2. In addition, TVWS spectral utilization efficiency is better than using sensing alone detection. This is primarily due to the ability of Geo-location enabled TVWS devices to accurately determine protected service contours. The CR networking architecture, presented in Figure 2 is centralized (i.e. broker-based) consisting of a spectrum broker that is in charge to manage TVWS usage and control economic transactions of resources exploitation, as well as several secondary systems, such as wireless network providers (e.g. cellular network operators) that request TVWS access.

Figure 2. CR networking architecture exploiting RTSSM policy

Spectrum broker entity consists of four sub-systems in the proposed CR networking architecture (i.e. TVWS occupancy repository, spectrum trading and policies

repository, a RRM module and trading module). More specifically, TVWS occupancy repository is exploited to collect relevant information stemming from Geo-

location database that includes all relevant data, concerning TVWS availability per geographical area, as well as the maximum acceptable transmission power of operation by secondary systems functioning without causing possible interference to primary systems. More specifically, TVWS occupancy repository generates frequency portfolios that contain required information, advertised to secondary systems before the submission of their bids during an auctionbased procedure. On the other hand, the RRM module, incorporated in spectrum broker is possible to match TVWS (i.e. spectrum supply) with spectrum requirements of secondary systems and allocate the available radio resources, according to specific QoS requirements. The adopted assigning mechanisms of TVWS inside this module, implement optimization algorithms, using relevant data from Geo-location database. This data is used to determine radio spectrum channels and transmission power, permitted by secondary systems to transmit, by not causing spectrum fragmentation, optimizing QoS and guaranteeing fairness in radio spectrum exploitation. Furthermore, trading module is in charge to calculate revenue/benefit of spectrum broker in the proposed CR networking architecture, by leasing TVWS to most economically significant secondary systems. The fourth spectrum broker entity (i.e. spectrum trading and policies repository) includes data regarding TVWS leasing process and the price per unit of spectrum that can be used during trading phase, by generating a price-portfolio. The proposed spectrum broker entity supports two alternative TVWS assigning mechanisms, by following either a fixed-price approach or an auction-based process. In case that a fixed-price policy is adopted, an optimization algorithm (e.g. Backtracking, Simulated Annealing, Genetic Algorithm) is exploited to obtain the best-matching solution, towards minimizing an

objective function, considering radio spectrum fragmentation and QoS priority level of secondary systems, when a number of cellular operators have to be served before other ones. Alternatively, during the auction-based process, spectrum broker collects a number of bids stemming from secondary systems, towards calculating a TVWS assigning output together with price of every spectrum part from price portfolio, maximizing spectrum broker revenue. This auctionbased process is then again repeated, as soon as available TVWS exist for leasing. Operation of the proposed CR networking architecture is following an approach based on three layers (Figure 3). Each layer denotes a vital procedure, regarding assignment of the available TVWS. The first layer includes local resource managers (i.e. LRM) at secondary users' side, the second layer incorporates spectrum managers (i.e. SM), while the third one exploits the proposed spectrum broker (i.e. SB). In more detail, LRM are in charge to assign TVWS within the area of secondary systems. LRM computes required resources needed, by considering radio link operation, as well as traffic load. According to bandwidth requirements provided by secondary systems through local resource managers, spectrum manager are then in charge to assign TVWS access to wireless network providers. Furthermore, each spectrum manager provides relevant data to the central spectrum broker entity of the proposed CR networking architecture, according to resources requests stemming from secondary systems. A negotiation request is also sent, when a secondary system desires additional resources, in comparison to its initial needs. Spectrum broker is responsible to perform TVWS assigning process, either by exploiting fixed-price approaches or auction-based processes, considering negotiations, as well as requests for bandwidth from secondary systems.

Figure 3. Layers of CR networking architecture operation

More specifically, spectrum broker of the proposed CR networking architecture initially advertises related information, regarding TVWS portions, which are available for leasing and relevant data, regarding limitations of maximum tolerable transmission power per channel. Such data is derived from the Geolocation database and is maintained inside the TVWS occupancy repository. Spectrum broker advertises both spectrum and price portfolios towards the available secondary systems, including information regarding transmission characteristics and price per unit of TVWS. Secondary systems (i.e. bidders) provide then their bids, indicating their interest related with specific units of radio spectrum and their offered price. Spectrum broker firstly gathers all relevant bids and stores such data to RRM module (Figure 2). This module then processes all bids and analyses transmission requirements of secondary systems and characteristics, regarding the available radio spectrum portions. Spectrum broker also maintains a list for every TVWS unit, including all bids from secondary systems per time period. This is an auction-portfolio created, towards deciding for the most important secondary system per time slot, when the auctionbased process is adopted. If two secondary systems submit their bids with the same requirements, the factor of time defines priority of the bid, in order to be on a higher position in auction-portfolio. This portfolio is examined by trading module inside the spectrum broker, considering the price per spectrum, which is defined in policies related with spectrum

auction. Finally, optimum results are achieved, by joining RRM together with trading module output, towards facilitating radio spectrum broker to lease portions of TVWS based on RTSSM policy. In such a case, spectrum broker considers to compute bestmatching solutions, based on optimization-based approaches. This is an NP-hard problem, requiring approximation algorithms, in order to solve the auction-based process. Even though optimized TVWS spectrum allocation policies have to be adopted from a business logic point of view, dynamic joint radio resource management (JRRM) techniques need also to be implemented at local resource manager level. As shown in Figure 3, each secondary system, via its spectrum manager module, allocates required spectrum to its LRM entities and latter serve all mobile end users, located in a specific geographical area. More specifically, secondary systems can be wireless networking solutions, such as LTE, WiMAX, WiFi, HSPA, GSM systems etc, while at the secondary user level (Figure 3), each LRM is deployed in one of the heterogeneous radio access technologies (RAT) base stations. For example, in Figure 3, the first spectrum manager entity depicted at the left hand side, could reside in an LTE system, which serves many geographical areas, by its corresponding eNodeBs. An LRM entity is deployed in each e-NodeB mainly dealing with interference management, admission control and scheduling issues. The second spectrum manager entity depicted at the right hand

side could represent a WiMAX system, following the same rationale with the pre-mentioned LTE system. LRM entities, apart from implementing admission control and scheduling algorithms, send also feedback to their inter-related spectrum managers. This context information enables LRM entities to provide optimal QoS/QoE services to mobile end users and the whole proposed CR networking architecture is possible to keep the network key performance indicators at acceptable levels. illustrates logical diagram of the proposed RRM framework and trading modules, based on a decision-making approach, where a “Process Data” function is initially taking place for producing all possible combinations, and therefore a set of “Possible Allocation Solutions”. Figure 4

Figure 4. Logical diagram of the proposed RRM framework and trading modules

As soon as all possible allocation solutions are established, the RRM modules calculate the optimum ones and create the spectrum portfolio that will be used by the broker during trading process. This spectrum portfolio is a first-stage filtering resulted from an iterative process namely as “IsValidSolution” in Figure 4, which examines if a possible allocation solution fulfills the secondary systems technical requirements. In this framework, the heuristic algorithm (i.e. Simulated Annealing (SA) [44])

creates the spectrum portfolio by obtaining solutions that respect secondary systems demand, power, priority QoS and bandwidth requirements, as well as TVWS transmission thresholds. In such a case, the possible allocation solution is registered in spectrum portfolio, otherwise it is discarded. To this extent, in the second-stage filtering, the selection of the bestmatching solution (i.e. Optimal Solution) is the result of an optimization process, targeting either to minimize spectrum fragmentation (i.e. fixed-price policy) or to maximize the profit (i.e. auction-based trading), whichever is appropriate. In both optimization approaches (i.e. fixed-price or auction-based), the second-stage filtering is performed by SA algorithm [44] that is applied as an optimization technique to solve the combinatorial problem. More specifically, SA algorithm substitutes step by step the present allocation solution, by another random solution. The result is selected, by exploiting a probability, which is related with the variance among related function values, as well as with a global parameter “T” (i.e. temperature) [44]. This probability is high, when “T” is increasing so that SA algorithm is not fixed at a certain local optimum. On the other hand, when this probability is decreasing, the probability of a local optima to exist is low. When “T” is zero, SA algorithm decreases to the greedy algorithm. This phase is repeated until the system reaches to a state, which is good enough. SA algorithm goes through O(logn) temperature stages. For each step, SA algorithm examines 0(n) solutions, as well as accepted variations. The calculation discards an alteration of the present tour in 0(1) time. If a modification is established, the mean path reversal includes 0(n) exchanges. As a result, run time Tn of SA algorithm has the complexity Tn=O((n2+n)logn)). Since the most of the phases take place at low “T” values, where the most variations are discarded, 0(nlogn) is not minor in comparison with 0(n2 logn).

4. RADIO SPECTRUM LEASING PROBLEM FORMULATION, PERFORMANCE EVALUATION AND EXPERIMENTAL RESULTS This section elaborates on the description of spectrum leasing problem, by formulating it under the fixedprice and auction-based approaches. Spectrum broker is responsible in both cases to efficiently trade TVWS and allocate them to various bidders (i.e. secondary systems) that compete to each other, in order to obtain

the available resources. According to initial measurements conducted in [45], [46] frequencies availability, denoted as BW, comprises of ten TV channels (i.e. 8MHz each), scattered among digital terrestrial television radio spectrum, while the aggregated available radio spectrum is 80 MHz. TVWS can be leased to participants during the allocation process on conditions that fulfill bandwidth and transmission power requirements. Such participants can be network operators, such as LTE, WiMax, UMTS, WiFi, Public Safety networks. In this research work, frequency/channel and time domains are taken into account, towards leasing radio spectrum, no matter what is the process that is followed. Error! Reference source not found. In case of fixed-price allocation process (i.e. fixedprice algorithm in Table 1), the spectrum broker reaches to the most optimal allocation solution, by minimising an objective function “C(A')” (i.e. equation 1), as a matter of allowable transmission power (P(i,f)), requested bandwidth (BW(i,f)), spectrum fragmentation (Frag(i,f)), when a secondary system “i” is assigned to a specific frequency “f” and/or secondary systems prioritisation (Pr(i)) (e.g. in case that a number of secondary systems must be served before other ones, due to higher QoS level priority). :

′

, ∈

,

,



1

∈

Spectrum fragmentation (or Fragmentation Score) was estimated by taking into account the number of unused spectrum-portions as well as the size of each individual fragment, according to (2) [47],

1

∑ ∑



2

where “n” is the number of the scattered fragments (i.e. number of unused spectrum portions), “fi“ is the bandwidth of the i-th fragment (e.g. in MHz), while “p” is a constant, which in our experiments was equal to “2” as proposed in [47]. In such a case, it is evident that when Fragmentation Score (Z) is equal to “0” there is only fragment and therefore the spectrum is considered as un-fragmented, while as Z increases towards “1”, the number of fragments also increases and the spectrum becomes more-and-more fragmented (many blocks of unexploited frequencies). On the other hand, auction-based process is best suited in cases that total requirement for radio

spectrum is considered more than the available radio sources (i.e. equation 3). 3

According to the auction-based process (i.e. auctionbased algorithm in Table 2), spectrum broker determines the optimal allocation solution, considering the maximization of its own income. To occur this, spectrum broker undertakes the trading mechanism that collects bids from secondary systems, in order to lease radio spectrum. It computes the assigning solution through this mechanism together with price per part of spectrum from price portfolio. This auction-based procedure is then repeated, until all resources are leased. Table 1: Fixed-Price Algorithm Pseudo-Code 1: Inputs: TVWSpool, Location(x,y), Powermax, DemandSS 2: Update TVWS repository from Geo-location database 3: Estimate the spectrum-unit price 4: Create and advertise price-portfolio 5: Receive secondary systems request R= {R1,…, RI}, where Ri = {xi, ti} 6: for all Requests do 7:

Sort Ri in descending order based on priority and update the price-portfolio

8: end for 9: Calculate the minimum fragmentation (Frag(i,f)) for all secondary system requests 10: Create initial solution S 11: for i = 1 to subset of variable length do 12: 13: 14: S 15:

Generate a new solution Si if (Objective_function(S) ≤ Objective_function(Si)) then save the new allocation solution Si to best found end if

16: end for 17: return Best Allocation Solution

Furthermore, when the auction-based algorithm is followed, radio spectrum sellers are denoted as N = {1,2,…,n}. N is 1 in the proposed CR networking architecture, according to the simulation scenario (i.e. Spectrum Broker, leasing the available TVWS S = {1,2,…,s} to I = {1,2,…,i} secondary systems). Each buyer “i” is able to purchase xi portions of radio spectrum for a specific time period ti, by reporting a price Pi(b) = {xi, ti} (i.e. Bid Price of m for specific portion of spectrum in a specific time), while

spectrum broker leases yn portions of radio spectrum for a specific time ti, by reporting a price Pn(s) = {yn, ti} (i.e. Asking Price of m for specific portion of spectrum in a specific time). Finally, xi,n is the quantity, which is leased by “i” secondary system from spectrum broker. The pair (i,s) in the pseudocode of Table 1 represents possible combinations of solutions, regarding “s” TVWS to “i” Secondary systems. In case that spectrum broker benefit has to be maximized, an optimization problem is formulated as follows, based on linear programming (i.e. equation 4): :

,

4

Table 2: Auction-Based Algorithm Pseudo-Code 1: Inputs: TVWSpool, Location(x,y), Powermax, DemandSS 2: Update TVWS repository from Geo-location database 3: Estimate the spectrum-unit price 4: Create and advertise price-portfolio 5: Receive secondary systems bids P(b) = {P1(b),…, PI(b)}, where Pi(b) = {xi, ti} 6: for all Bids do Sort Pi(b) in descending order based on price and

7:

create the auction-portfolio 8: end for 9: Calculate the highest valuation S[i,s] for all TVWS slots (i,s)  {1, 2,…, m} 10: set Soptimal = S[i,s] //Random solution for algorithm initiation 11: for slot =1 to m do //Iteration process in order to find the best solution

Moreover, the number of competitive secondary systems was up to fifty base stations, exploiting LTE technology, thus the same number of TVWS can be leased/allocated to more than one LTE base station, based on the re-used distance and on condition that no interference is caused. Both price models allow a player to reserve radio spectrum with temporarily exclusive rights for longer than the next time slot. The winners obtain these rights for the allocated spectrum until the next auction. The valuation of spectrum (i.e 1MHz) in both cases was resulted, considering a number of parameters [46], [48], such as benchmark price, price factor over year, population density, allocation area, degree of competition, incentives of operators in low/medium/high density areas and traffic conditions (i.e. low/medium/high). Performance of the proposed research approach was evaluated, considering a quantitative and qualitative comparison among both algorithms, in order to estimate the average values of spectrum broker benefit/utility, spectrum fragmentation, spectrum utilization and probability of accessing TVWS, according to RTSSM policy. More specifically, spectrum broker benefit was estimated based on the total income by secondary systems during four spectrum leasing time periods. Figure 5 presents the average benefit/income of spectrum broker, under various numbers of bidders (i.e. LTE secondary systems) for both algorithms during all time periods (i.e. four experimental tests were conducted). It is observed that the proposed algorithm based on the auction process performs better providing higher revenue compared to the fixed-price algorithm.

12: if (S[i,s]) ≤ (S[i+1, s+1]) // Check if the current solution is better or not to the neighbor solution 13:

then save the new allocation solution (S[i+1, s+1]) to the best found

14: end if 15: end for 16: return Best Solution

Towards verifying the validity of algorithms (i.e. fixed-price approach and auction-based process), a set of experimental test was designed and conducted. A simulation test-bed, based to the overall design specifications was set-up, comprising four timeauctions/allocation per hour, (each one of 15-minutes long intervals). The available TVWS based on real data [45] are ten, while the number of frequency-time slots is forty. They are computed, by multiplying available TVWS and time-auctions per hour.

Figure 5. Spectrum Broker Average Benefit

Moreover, spectrum utilization was estimated as the percentage of the exploited bandwidth over the totally available TVWS, while spectrum fragmentation was estimated by considering the unused parts of radio spectrum and the size for each individual fragment. In cases that spectrum fragmentation is equal to “0”, radio spectrum is considered as un-fragmented after the allocation process, while as the value increases towards to “1”, the number of fragments left after the assigning process, are increased and spectrum becomes more-and-more fragmented (i.e. many blocks of unexploited frequencies exist after TVWS allocation process). It has to be noted here that towards avoiding possible interference among LTE secondary systems that are assigned with consecutive channels, frequency parts of 1MHz are left un-used as guard intervals. As the number of secondary systems is getting higher, the number of guard intervals also increases, resulting to a higher radio spectrum fragmentation, after frequency allocation process. Figure 6 and Figure 7 present evaluation results, regarding average spectrum utilization and spectrum fragmentation respectively. As spectrum utilization is increasing, resulting to an increased number of secondary systems that exploit TVWS, fragmentation is also increasing, getting worst. In case of the fixedprice algorithm, the most optimum solution is provided, based on minimizing spectrum fragmentation, resulting to lower levels of utilization. On the other hand, auction-based algorithm creates a more fragmented radio spectrum after the allocation process, when the number of secondary systems served, is higher in comparison to the fixed-price algorithm.

Finally, probability of accessing TVWS is a metric that defines the possibility of a secondary system permitted to operate, exploiting radio spectrum resources. The auction-based algorithm provides a higher probability of using TVWS, in comparison to the fixed-price algorithm, as depicted in Figure 8. This implies that bidders are encouraged to participate in the auction-based process, increasing the possibility to access the available TVWS.

Figure 7. Average Spectrum Fragmentation

Figure 8. Average Probability of accessing TVWS

5. CONCLUSIONS

Figure 6. Average Spectrum Utilization

This paper proposes two RRM algorithms that were designed, developed and incorporated in a centralized CR networking architecture, towards enabling for efficient TVWS exploitation according to RTSSM policy. In this direction, a detailed description of an infrastructure-based CR network is presented, where

dynamic TVWS allocation among secondary systems is coordinated by a spectrum broker entity. This entity administrates economic issues of transactions, related with TVWS leasing, by utilizing either a fixed-price or an auction-based RRM algorithm. For efficient system performance, as a matter of maximumpossible radio resource exploitation and trading revenue, the paper elaborates on the study and development of a prototype RRM framework at the spectrum broker side, which is based on optimization methods to obtain best-matching solutions. Towards evaluating system performance, several sets of experiments were designed and conducted in a controlled conditions environment (i.e. simulation tests), where a number of LTE secondary systems are accessing the available TVWS. The experimental results verified the validity of both proposed RRM algorithms, including a qualitative and quantitative comparison among them. The provided results include the evaluation of spectrum broker benefit/utility, spectrum fragmentation, spectrum utilization and probability of accessing TVWS, according to RTSSM policy. Fields for future research and experimentation include the enhancement of the proposed research approach, by adopting energy-efficient schemes and algorithms, incorporated in spectrum broker entity, optimizing TVWS allocation decisions regarding energy conservation issues.

6. REFERENCES [1] First Report and Order, “Federal Commission Std., 2010. Available:

Communication

[7] A. Wyglynski, M. Nekovee, and T. Hou. Cognitive Radio Communications and Networks: Principles and Practice, Academic Press, 2009. [8] K.-C. Chen and R. Prasad, “Cognitive Radio Networks”, Hoboken, NJ: Wiley, 2009. [9] E. Hossain, D. Niyato, and Z. Han, “Dynamic spectrum access and management in cognitive radio networks”, first ed. Cambridge University Press, 2009. [10] Mitola, III, “Software radio architecture: A mathematical perspective”, IEEE J. Sel. Areas Commun., vol. 17, no. 4, pp. 514–538, Apr. 1999. [11] S.-Y. Tu, K.-C. Chen, and R. Prasad, “Spectrum sensing of OFDMA systems for cognitive radio networks”, IEEE Trans. Veh. Technol., vol. 58, no. 7, pp. 3410–3425, Sep. 2009. [12] A. Bourdena, E. Pallis, G. Kormentzas, C. Skianis, and G. Mastorakis. “Real-Time TVWS Trading Based on a Centralised CR Network Architecture”, in Proc. IEEE Globecom2011, IEEE International Workshop on Recent Advances in Cognitive Communications and Networking, Texas, Houston, USA, pp. 964-969, 2011. [13] A. Bourdena, E. Pallis, G. Kormentzas, and G. Mastorakis, “A prototype cognitive radio architecture for TVWS exploitation under the real time secondary spectrum market policy”, Physical Communications, Special Issue on “Cognitive Radio for LTE Advanced & Beyond”, Physical Communications, Elsevier (to appear). [14] I. F. Akyildiz, W.-Y. Lee, M. C. Vuran, and S. Mohanty, “A Survey on Spectrum Management in Cognitive Radio Networks”, IEEE Communications Magazine, 2008. [15] W. Saad, “Coalitional Game Theory for Distributed Cooperation in Next Generation Wireless Networks”, Ph.D. dissertation, University of Oslo, 2010.

http://fjallfoss.fcc.gov/edocs_public/attachmatch/FCC-10196A1_Rcd.pdf

[16] L. Gao, X. Wang, Y. Xu, and Q. Zhang, “Spectrum trading in cognitive radio networks: A contract-theoretic modeling approach”, IEEE Journal on Selected Areas in Communications, vol. 29, no. 4, pp. 843–855, 2011.

[2] F. Force. Report of the spectrum efficiency working group, Washington DC, 2002. Available: http://transition.fcc.gov/sptf/files/SEWGFinalReport_1.pdf

[17] J. Huang, R. Berry, and M. Honig, “Auction-based spectrum sharing”, ACM Journal on Mobile Networks and Applications, vol. 11, no. 3, pp. 405–418, 2006.

[3] Unlicensed Operation in the TV Broadcast Bands, Final Rules, Available: http://edocket.access.gpo.gov/2009/pdf/E93279.pdf.

[18] L. Duan, J. Huang, and B. Shou, “Investment and Pricing with Spectrum Uncertainty: A Cognitive Operator’s Perspective”, IEEE Transactions on Mobile Computing, vol. 10, no. 11, pp. 1590–1604, 2011.

[4] I. Akyildiz, W. Lee, M. Vuran, and S. Mohanty, “NeXt generation/ dynamic spectrum access/cognitive radio wireless networks: A survey”, Computer Networks, vol. 50, no. 13, pp. 2127–2159, Sep. 2006. [5] 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. [6] J. Mitola III and G. Q. Maguire, “Cognitive Radio: Making Software Defined Radio More Personal”, IEEE Pers. Commun., vol. 6, no. 4, Aug. 1999.

[19] S. Sodagari, A. Attar, and S. Bilen, “On a truthful mechanism for expiring spectrum sharing in cognitive radio networks”, IEEE Journal on Selected Areas in Communications, vol. 29, no. 4, pp. 856–865, 2011. [20] V. Krishna, “Auction theory”, Academic press, 2009. [21] A. Bourdena, E. Pallis, G. Kormentzas, and G. Mastorakis, “A centralised broker-based CR network architecture for TVWS exploitation under the RTSSM policy”, in Proc. 2nd IEEE Workshop on Convergence among Heterogeneous Wireless Systems in Future Internet (CONWIRE 2012), IEEE ICC2012, Ottawa, Canada, pp. 7228-7232, 2012.

[22] A. Bourdena, G. Mastorakis, E. Pallis, C. X. Mavromoustakis, G. Kormentzas, E. Karditsis, “A Radio Resource Management Framework for Opportunistic TVWS Access”, Submitted in 1st ACM Workshop on High Performance Mobile Opportunistic Systems (HP-MOSys 2012), 15th ACM International Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems, in cooperation with R8 IEEE Computer Society C16, Paphos, Cyprus, 21-25 October 2012. [23] The Economics of Spectrum management: A Review, Australian Communication and Media Authority (ACMA), 2007. [24] L. B. 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. [25] Y. Wu and D. H. K. Tsang, “Distributed power allocation algorithm for spectrum sharing cognitive radio networks with QoS guarantee”, in Proc. IEEE INFOCOM’09, Rio de Janeiro, Brazil, Apr. 2009.

[34] Y. Shoham, K. Leyton-Brown, “Multi-agent Systems: Algorithmic, Game-Theoretic, and Logical Foundations”, Cambridge University Press, 2009. [35] Kiho Yoon, “The participatory Vickrey–Clarke–Groves mechanism”, Journal of Mathematical Economics, Elsevier, Volume 44, Issues 3–4, February 2008, Pages 324–336 [36] J. Sun, E. Modiano, and L. Zheng, “Wireless channel allocation using an auction algorithm”, IEEE J. Sel. Areas Commun., vol. 24, no. 5, pp. 1085– 1096, May 2006. [37] Tianyu Wang, Lingyang Song, Zhu Han, Xiang Cheng and Bingli Jiao, “Power Allocation using Vickrey Auction and Sequential First-Price Auction Games for Physical Layer Security in Cognitive Relay Networks”, ICC, page 1683-1687. IEEE, 2012. [38] J. Huang, R. Berry, and M. L. Honig, “Auction-based spectrum sharing”, ACM Mobile Netw. Appl. J., vol. 11, no. 3, pp. 405–418, Jun. 2006.

[26] F. Wang, M. Krunz, and S. Cui, “Price-based spectrum management in cognitive radio networks”, IEEE J. Sel. Topics Signal Proc., vol. 2, no. 1, pp. 74–87, Feb. 2008.

[39] J. Huang, Z. Han, M. Chiang, and H. V. Poor, “Auctionbased Resource Allocation for Cooperative Communications”, IEEE J. Sel. Areas Commun., vol. 26, no. 7, pp. 1226–1237, Sep. 2008.

[27] F. F. Digham, “Joint power and channel allocation for cognitive radios”, in Proc. IEEE WCNC’08, Las Vegas, USA, Apr. 2008.

[40] Z. Ji, K.J. Ray Liu, “Belief-Assisted Pricing for Dynamic Spectrum Allocation in Wireless Networks with Selfish Users”. In Proc. of IEEE SECON, 2006.

[28] K. Liu, Q. Zhao, and B. Krishnamachari, “Dynamic multichannel access with imperfct channel state detection”, IEEE Trans. Signal Proc., vol. 58, no. 5, pp. 2795–2808, May 2010.

[41] C. Kloeck, H. Jaekel, and F. K. Jondral, “Dynamic and Local Combined Pricing, Allocation and Billing System with Cognitive Radios”, in Proc. IEEE DySPAN ’05, pp. 73-81, 2005.

[29] A. Bourdena, E. Pallis, G. Kormentzas, C. Skianis, G. Mastorakis, “QoS provisioning and policy management in a broker-based CR network architecture”, in Proc. IEEE Globecom2012, IEEE International Workshop on Recent Advances in Cognitive Communications and Networking, Anaheim, San Francisco, USA, 3-7 December, 2012.

[42] F. Wang, M. K. Shuguang Cui, “Spectrum Sharing in Cognitive Radio Networks”, in Proc. IEEE INFOCOM ’08, pp. 1885-1893, 2008.

[30] E. Altman, T. Boulogne, R. El-Azouzi, T. Jime ́nez, and L. Wynter, “A survey on networking games in telecommunications”, Computers and Operations Research, vol. 33, no. 2, pp. 286–311, 2006.

[44] Steven S. Skiena, “The Algorithm Design Manual”, Second Edition, Springer, ISBN: 978-1-84800-069-8

[31] D. Niyato and E. Hossain, “Competitive pricing for spectrum sharing in cognitive radio networks: dynamic game, inefficiency of Nash equilibrium, and collusion”, IEEE Journal on Selected Areas in Communications, vol. 26, no. 1, pp. 192– 202, 2008.

[43] C. Kloeck, H. Jaekel, F. K. Jondral, “Auction Sequence as a New Resource Allocation Mechanism”, VTC05 Fal, Dallas, 2005.

[45] http://www.ictcogeu.eu/pdf/COGEU_D4%201(ICT_248460).pdf [46] The European 7th Framework Programme project: COGEU (ICT-248560) - Deliverable 6.1 - Dynamic Radio Resource Management algorithms for an efficient use of TVWS, 2010

[32] Petri Mähönen, Marina Petrova, “Minority game for cognitive radios: Cooperating without cooperation”, Physical Communication, Elsevier, Volume 1, Issue 2, June 2008, Pages 94–102.

[47] Measuring Fragmentation of Two-Dimensional Resources Applied to Advance Reservation Grid Scheduling, Julius Gehr, Jorge Schneider, 9th IEEE/ACM International Symposium on Cluster Computing and the Grid

[33] S. Sengupta and M. Chatterjee, “Sequential and concurrent auction mechanisms for dynamic spectrum access”, in Proceedings of International Conference on Cognitive Radio Oriented Wireless Networks and Communications (CrownCom) 2007, July–August 2007.

[48] The European 7th Framework Programme project: COGEU (ICT-248560) - Deliverable 6.4 - System level evaluation platform and simulation results, 2012.