IEEE SYSTEMS JOURNAL, VOL. 5, NO. 3, SEPTEMBER 2011 417 ...folk.uio.no/yanzhang/IEEESystemSep2011.pdf · media system performance over cognitive radio ... radio network (CCRN) consisting

IEEE SYSTEMS JOURNAL, VOL. 5, NO. 3, SEPTEMBER 2011 417

Modeling Channel Allocation for MultimediaTransmission Over Infrastructure Based

Cognitive Radio NetworksTigang Jiang, Honggang Wang, Member, IEEE, and Yan Zhang, Senior Member, IEEE

Abstract—Multimedia applications have being widely supportedby cognitive radio networks in which high bandwidth can be pro-vided by fully utilizing radio spectrum. However, the researchcommunities lack a theoretical modeling and analysis for multi-media system performance over cognitive radio networks. In thispaper, a novel Markov-Chain based model of centralized cognitiveradio network (CCRN) consisting of APs and user equipments(UEs) equipped with several antennas (one antenna for channelsensing, the others for data communication) is proposed. In theproposed model, AP will scan the idle primary channels (PCs)that are not used by primary users (PUs), and then transmit datathrough selected idle channels. Each selected channel in CCRNwill be divided into many sub-time slots and some of them areused as the signalling control and the other slots are used only fordata transmissions. In addition, three types of communicationsare considered and an appropriate number of time slots arepre-assigned to ensure QoS of multiple types of multimedia traffic.We used a multi-dimension Markov chain to model time slotallocation and proposed a general channel allocation model formultimedia traffic over CCRN. We validated our model throughboth theoretical analysis and simulations.

Index Terms—Channel allocation, cognitive radio, MarkovChain, multimedia.

I. INTRODUCTION

F EW research engages in studying multimedia transmis-sion over cognitive radio networks. In [1], the authors ba-

sically proposed a content-aware multimedia delivery schemeand formulate the spectrum allocation problem as an auctiongame. In [2], the authors studied a dynamic channel selection tosatisfy the rate requirements and delay deadlines of heteroge-neous multimedia users through dynamic channel selection. In

Manuscript received November 14, 2010; revised March 14, 2011; acceptedMarch 14, 2011. Date of publication June 27, 2011; date of current version Au-gust 24, 2011. This work was supported by the National Natural Science Foun-dation of China under Grants 61001083 and 60802024, the 863 Project of Chinaunder Grant 2009AA011801, the National ST Major Project of China underGrants 2008ZX03006-001 and 2008ZX03003-005, the UMass President’s 2010Science & Technology Initiatives award and the Joseph P. Healey Endowmentaward.T. Jiang is with the School of Communication and Information Engineering,

University of Electronic Science and Technology of China (UESTC), Chengdu,China (e-mail: [email protected]).H. Wang is with the Department of Electrical and Computer Engineering,

University of Massachusetts-Dartmouth, North Dartmouth, MA 02747-2300USA (e-mail: [email protected]).Y. Zhang is with Simula Research Laboratory, 1325 Lysaker, Norway

(e-mail: [email protected]).Color versions of one or more of the figures in this paper are available online

at http://ieeexplore.ieee.org.Digital Object Identifier 10.1109/JSYST.2011.2159640

[3], a joint source-channel coding approach is proposed for dis-tributed multimedia transmission over second users. The foun-tain code is studied because of its efficiency in error correc-tion capability over lossy wireless channel. The study in [3]also shows that the trade-offs between link reliability, spectralefficiency and coding overheads for multimedia transmission.In [4], the authors specially developed a cross-layer optimiza-tion approach to optimize the video quality. The quality opti-mization is formed as a mixed integer nonlinear programming(MINLP) problem. A distributed algorithm is also developedto solve the MINLP problem. In [5], the authors presented agame-theoretic solution for wireless resource management fordelay-sensitive multimedia applications. The authors in [7] pro-posed an approach that jointly optimizes multimedia intra re-freshing rate and lower layer access strategy for high qualitymultimedia transmission. The optimization is formed as a par-tially observableMarkov decision process. All these approachesabove basically consider both multimedia features and channelselection for cognitive radio networks. In [8], the video trans-mission system is formulated as a switching control dynamicMarkovian game. The control scheme based on the game canimprove video transmission PSNR quality. In [9], the authorsstudied voice service over cognitive radio networks and an an-alytical model is proposed to estimate the capacity for the sec-ondary users. In [10], the authors proposed an open cognitivearchitecture that can support efficient multi-service operations.In [11], the voice capacity of the CR system based on the theoryof effective bandwidth (EB) is derived. The authors show thatthe simulation results match the analytical results well. A calladmission control policy for QoS provisioning in CR networksis also developed. The authors in [13]–[15] studied the principleof resource allocations and multimedia scheduling services inwireless networks. An analysis of channel allocation scheme forwireless cellular networks is given in [16]. In [17], a spectrumsharing scheme across multiple service providers via cognitiveradio nodes is proposed. The authors in [18], [19] have studiedthe multimedia transmission over cognitive radio networks in across-layer manner. The works shows the significance of jointdesign of multimedia coding and channel allocation of cogni-tive radio networks. In [20], the authors proposed an end-to-endsystem to optimize the user-perceived video quality at the re-ceiver end under the constraint of packet delay bound in CCRN.Some recent related studies on cognitive radio network havebeen conducted in [21], [22].However, in the CCRN, each communication node is usu-

ally equipped with multiple antennas and these antennas are dy-

1932-8184/$26.00 © 2011 IEEE

418 IEEE SYSTEMS JOURNAL, VOL. 5, NO. 3, SEPTEMBER 2011

Fig. 1. Infrastructure based cognitive radio networks.

TABLE ISYMBOLS AND NOTATIONS

namically set to sending/receiving/idle status. The number ofworking antennas may be dynamically changed according tothe arrival of PUs. Most of existing works ignore the practicaland real configurations. To be best of our knowledge, there areno existing research works that model an infrastructure basedcognitive radio networks for multimedia transmission. A the-oretical analysis model for this network is strongly requiredto evaluate multimedia system performance in CCRN. In thispaper, we studied an infrastructure based cognitive radio net-works in which both access point (AP) and primary user basestation (PBS) co-exist in the networks and the AP is acted asthe center of slot allocation and data relay. Fig. 1 shows theproposed infrastructure which incudes three types of communi-cations: between UEs within CCRN, from UEs in CCRN to thenodes outside CCRN, and from nodes outside CCRN to the UEsin CCRN. As described in Fig. 1, the PBS manages all the PCsthat can be freely used by the nodes of primary user equipment(PUE), while some idle PCs can be used by other second usersin the CCRN. The CCRN consists of access points (AP) thatconnects user equipment (UEs). AP communicates with other

TABLE IITIME SLOT DEFINITION

nodes outside CCRN via either wire or wireless connections.All APs and UEs are equipped with multiple antennas.In the infrastructure based conative radio networks, free chan-

nels of a PU are scanned and reused by the CCRN. AP decideswhat channels could be re-used and maintains the resource allo-cation and data relay. For example, we can assume that TDMAbased protocol is applied. Each frame last 10ms consisting of 40OFDM signals, and each signal (signal slot) last 250 s. Eachchannel reused by AP will be divided into multiple time slotsand be reallocated to SUs. Two types of frames will be alsoclassified due to the needs of signalling and data exchange be-tween the AP and UEs. The signaling from AP to UE includesthe synchronizing, cognition information broadcast and alloca-tion table broadcasting. The signalling from UE to AP includesupper layer application QoS requirements which will be sent byUE to AP via a fixed slot. In the paper, we focus on the time slotsallocation and assume the QoS requirement information always

JIANG et al.: MODELING CHANNEL ALLOCATION FOR MULTIMEDIA TRANSMISSION OVER INFRASTRUCTURE BASED COGNITIVE RADIO NETWORKS 419

TABLE IIICALL TYPE DEFINITION

can be sent to AP successfully. There is an interrupt from phys-ical layer of UE every 10 ms. The major symbols we used in thepaper are listed in Table I.

II. BASIC DEFINITION AND ASSUMPTIONS

In the proposed infrastructure based Cognitive radio net-works, we assume that there are PCs (primary channels)that are used by the PUs (Primary Users). In many scenarios,

PCs are used by PUs while ( PCscan be used by the SUs. The maximum number of channels theSUs can use is . is the number of antennas each UE isequipped. When is satisfied, only channelscan be used. If , then all the antennas should be turnedoff and the communication of the CCRN will be interrupted.Each channel will be divided into sub channels and ofthem should be used as signalling such as time synchronizationand time slot allocation. The other ( ) time slotsare only used for data transmissions. If the number of channelsthe CCRN used is , then antennas are working and

antennas should be turned off. If there are totaltypes of multimedia traffic in CCRN, the least number of timeslots should be guaranteed for each traffic in order to assurethe QoS. For example, some multimedia traffic require at leastsub slots for transmissions, then the system should allocate( ) time slots. On the other hand, any type of traffic

require at most sub slots for data transmissions. For eachnode equipped with multi-antennas, number of slots shouldbe allocated to the user, and may be larger than . However,it must follow . This is because theworking antennas will be in receiving/sending status at a sametime due to the limitation of hardware.

A. Time Slot Definition

Multiple slot allocation patterns can applied for differentcommunication modes. For the communication among users inthe CCRN, the number of time slots for the uplink allocated tothe sender is equal to the number of time slots for the downlinkallocated to the receiver. Because the communication betweena user within CCRN and a node outside of CCRN involvesboth wireless and wired connections, we should consider thedifference for the time slot allocation specifically. In addition,both AP and UE need to exchange signalling informationfor the uplink and downlink. The detail of multiple time slotdefinition is shown in Table II.

B. Call Type Classification

According to the different communication modes inTable II, we can classify the calls into three types as shownin Table III. Among the call types described in Table III,for a given type- , type- ortype- call, the call will be allocated ( )slots for transmission if it is accepted by the system. Then thecall on communication is denoted as type- ,type- or type- . If somechannels CCRN used are reoccupied by the arriving PUs, theAP will try to re-allocate the slots to the ongoing users toavoid the call termination. In this scenario, some calls may bedropped due to the unavailability of time slots. If only a singledirectional link (uplink or downlink) slots allocation is consid-ered and there are enough number of slots to be allocated, totalslots on a single direction link will be allocated for the SU.Because the number of working antenna is , the allocatedtotal slots are , . The function can bedefined as follows:

if

if(1)

In (1), represents the total time slots allocated to a SU.The time slots allocated to the SU per channel (or per antenna)is , which is expressed in the following:

C. States Definition and Basic Formulas

Given a fixed , the state of the SUs can be modeled asfollows:

(2)

Where is the number of calls of type –as described above. The elements in are indexed byfour integer , and . We reordered the index of


and , and only use one integer to index the ele-ments. The total number of elements for a fixed is

andthe number of total states is . For acertain , the element indexed by ( ) will be indexedby , . After we map the fourelements index to , can be described in(3):

(3)

or

The state of includes a series of sub state blocks. Spaceis represented as , where

is the number of PUs. The total state space can be shown as. All elements in but not belonging to

are zero. we defined two functions. one function is totranslate four elements index to one integer index as shown inthe following:

(4)

The other is to convert one integer index to four elements index:

(5)

The sub index of these four elements can be expressed in (6):

(6)

Therefore, the elements in the state , can be noted as , ,or separately, where . If ,then is . The elements for state can be noted as ,or .For the communication within CCRN, the time slots must

be allocated to the sender and the receiver simultaneously sothat the reduced number of available slots is the twice of thatallocated to the sender. Therefore, and can be expressedin (7)–(8):

(7)

(8)

III. CHANNEL ALLOCATION MODELINGFOR MULTIMEDIA TRAFFIC

For multimedia traffic over cognitive radio networks, thereare eight possible events which lead to the state change inspace, which are listed as follows.1) Accepting a type- call.2) Leaving of a type- call ( ).3) Accepting a type- call.4) Leaving of a type- call ( ).5) Accepting type- call.6) Leaving of a type- call ( )7) Accepting a PU.8) Leaving of a PU.These events above can change the state from to . For

any events, if the destination state and the corresponding tran-sition rate can be determined, multiple Markov chain can besolved.

A. Accepting of a Type- ,or Type-Call

In this case, and represent the low and high num-bers of slots allocated to the call, respectively. Then

and . In addition, the transitionrate equals to (here ) if the destination states exist.The destination states can be analyzed in the following:1) IfThe call will be allocated slots to thesender and the receiver per channel (or per antenna), andbecomes an ongoing call of type- ,where is the number of slots allocated to thesender and receiver simultaneously. The number oftime slots on these channels allocated to a pairof UEs within CCRN is . For the new state ,

, each element of is assame as that of .

2) IfThe call will be allocated ( ) slots per channel,and become an ongoing call of type- .Then for , . Each elementof is also as the same as that of .

3) If and , there will be notenough idle time slots that can be utilized. A call willbe allocated slots per channel and be-come an ongoing call of type- . There-fore, the system should try to release time slots from othertype- calls ( ) and al-locate them to new calls. The time slot allocation has twosteps: 1) releasing time slots; 2) allocating released slotsand other idle slots to the new call. After the releasing,becomes a temporary state . Let function

return the number of slots, which can be re-allocated perchannel from the calls indexed from 1 to in the setblock.

(9)

where function return 1 if else return 0. If


, the call is a type- call. The uplink andthe downlink should release equal number of time slots.There must exist that satisfiesandAlso, there must exist non-negative integers and in thefollowing,

where is the number of de-allocation candidates oftype- , and is the numberof calls of type- which neednot be de-allocated. Therefore, the accepted call indexposition of can be located according to the element of.

B. Leaving of a Type- , orType- Call

The transition rate is

though the call using total slots, but the uplink and thedown link use slots respectively.The transition state: if type- call leaves, then

slots will be released. AP should re-allocate thoseslots per channel to the other ongoing calls as possible as it

can. Firstly, AP should find who should be upgraded. Secondly,AP reallocates slots to them. After the call leaves, the systemcame into a temporary state . locate the and setblocks in and respectively, in the according blocks,

The other elements between and are the same. Afterthe channel re-allocation, will move from to .

C. Accepting of a Type- orType- Call in the Up Link of UE

The transition rate is .Destination state: for only the uplink slots should be allo-

cated, the time slot allocation strategy is as the same as 1) and, .

D. Leaving of a Type- orType- Call

Transition rate is .Destination state: as the situation of B), if a

type- call leaves, then slots willbe released. AP should re-allocate those time slots to the otherongoing calls as possible as it can. The situation is as similar as2, only the release slots number is instead of .

E. Accepting a Type- or Type-Call in the Down Link of UE

The transition rate is .

Destination state: the situation is the same as ).

F. Leaving of a Type- orType- Call

Transition rate is .Destination state: The situation is the same as ).

G. Accepting a PU

Translation rate isDestination state analysis is as follows.1) If , the PU will be rejected.2) If and , the PU will be accepted.CCRNmay handoff the working channel if the PUwants tore-claim the channel CCRN used. We assume the handofftime is very short and the communication in CCRN are notaffected. Then for , , the other each elementof and that of is the same.

3) If and , the PUwill be accepted and theCCRN must release one channel. The slots in CCRN willbe re allocated. If some SUs’ basic time slot requirementcan not be satisfied, then the SUs will be dropped. , theavailable slots number is wherea) if , then all communication in CCRN will bedropped.

b) if , then AP start channel re-allocationWe assume AP firstly reallocates time slots, then releasechannel. The reallocation has two steps. Firstly, AP triesto allocate slots to the ongoing users with basic slots re-quirement, the state will move to a temporary state .Secondly, if some free slots are available after the first step,then try to allocate the available slot to the users for betterservice, state will transit to state .

H. Leaving of a PU

The transition rate is . Destination state is determinedbelow: Once a PU leaves the system, the scan antenna of theAP can find the event, then AP will decide whether occupies thereleased SC or not. If , this means all antennas areworking, AP will not use the released SC all element of andthat of are the same. If , AP will start a idle antennawith the channel of the released SC, , at thattime, additional number of slots can be re-allocated. In thisscenario, AP starts an idle antenna and sends a control commandto all the UEs in the CCRN. The new slot allocation table createdby AP also will be send to the UEs through downlink headerslot. The AP will try to allocate those slots to the UEs forfull channel utilization. Once all the transition states

and the corresponding transition rates are obtained,each state probability can be obtained through queueing theory.

IV. NETWORK PERFORMANCE METRICS

In our study, we evaluated the proposed instructress basedcognitive networks performance using three metrics: blockingprobability, dropping probability and throughput. The blockingof PUs only occurs when an arrival SU find that there are noavailable channels in the system. A simple BCC (blockingcall cleared) queuing model can be used to describe it. Anarriving SU is blocked only if the AP can not find the enough


available slots to satisfy the basic slot requirement of the SUtype . The blocking probability of SUs oftype-(k,m,n,0) is expressed in the following:

(10)

Where is the available slots of state according for-mula (8).The blocking probability of PUs is expressed as follows:

(11)

The forced termination probability is equal to the ratio of thenumber of terminated SU connections to the sum of terminatedand complete SU connections. For a state , the numberof working antennas is and the total ongoing calls is

. A channel must be released when a PU arrives. Fora state where there are working antennas, the new numberof working antennas is . Let integer numberfollow .

represents the number of ongoing calls that will be reserved.The total number of dropped calls is .If the number of working antenna number decreases, only the

calls indexed from 1 to are preserved and the calls indexedby to will be dropped in the new state .So we have the dropping probabilities of the three types of SUs:

(12)

where , 2 or 3 means the dropping probability of threetypes of SUs. is the number of states in the sub statespace of the state .

represents the throughput that is defined as the averagenumber of service completions for SUs per second, which canbe written in the following:

(13)

where is the number of working antennas in state .is the number of states of and is the total states

set of .In this paper, we will conduct simulations to verify our ana-

lytical model described before. The simulation is independent ofthe theoretical analysis. The behavior of type-calls is as similar as that of type- calls becausethey only use single direction slots. Therefore, we only con-sider one of the two types. In our case study, the parametersare set as follows: , , ; There are twotypes of calls in the system: 1) type- call, ortype- , , ; 2) type-call or type- , , ; to 6 calls/second, , , ,

Fig. 2. Example of the state.

Fig. 3. Example of the state translation.

if only one slot used for transmission. The transi-tion diagram can be expressed as in Fig. 2. In Fig. 2, the non-cir-cled elements all are zeros. Therefore, we only select mean-ingful elements in the transition diagram, and rewrite a concisestate with six elements in Fig. 2 because those zero-elementshave no effects on our theoretical analysis. Therefore, the statestransition diagram can be expressed in Fig. 3. Once all the tran-sition states and the corresponding transitions are obtained, theprobability of each state can be obtained using the queueingtheory.We validated the simulation results with theoretical analysis.

The simulation platform is NS2. The parameters of the simu-lation and the theoretical analysis are same. Both the numberof PCs and antennas are 2; The signaling slot number isset to zero and the slot number for data transmission is 2.Two call type- and type- aredefined, in which is 1 and is 2; the arrival rate of PUsranges from 1 to 6 calls/s. The arrival rate of the two typesof SUs is 2 calls/s and is 4 calls/s. The service rate ofPUs is 2 calls/second and the service rate of SUs is 1. Theflow directions of SUs are explained in Fig. 4.The simulation time is 10 000 seconds. The simulation and

theory results are shown from Figs. 5–7. From these figures, wecan see that the simulation result perfectly matches the theoret-ical results.Table IV shows the number of calls handled by the simulation

system when is set to 6 calls/s. For example, for the first


Fig. 4. Flow direction of SUs.

Fig. 5. Blocking probability of PUs.

Fig. 6. Blocking probability of SUs.

and second rows, there are total 59 907 arrived PUs during thefirst 10 000 s, and 28 038 of them are accepted while the rest of31 869 are blocked due to lack of PCs.When the simulation timeis expired, there are still 2 PUs in communications. The numberof successful serviced PUs is 28 036. For the third and fourthrows, there are total 20 021 SUs for arrived multimedia stream1, in which 4450 of them are accepted and the other 15 571SUs are blocked. There are no remaining SUs of stream 1 whenthe simulation time exceeds 10 000 s. Therefore, the number

Fig. 7. Dropping probability of SUs.

TABLE IVNUMBER OF CALLS HANDED BY SIMULATION SYSTEM

successful serviced SUs are 1073. The simulation scenarios ofmultimedia stream 2 are similar with stream 1.In Fig. 5, we can see with the increasing of arrival rate of PUs,

the blocking probability of PUs increase, it is clearly for the ser-vice resource is limited, and simulation and theoretical resultsperfectly match each other, they also match the classic queueingtheory of BCC (blocked calls clear) model and Cooper’s result(see [8, Fig. A.1, p. 316], with ), the curves in Fig. 5 isalmost the same as [8].Fig. 6 shows some interesting results. With the increasing

rate of PUs, the blocking probability of stream 1 will deceasefirstly and then increase later. This is reasonable because thereare only two PCs and each PC is divided into two slots. APshould allocate the stream 1 to double direction slots, half foruplink (allocate to the source) and half for downlink. With theincreasing of , fewer antennas are working and the blockingprobability of stream 2 will increase. Because the arrival rate ofsteam 2 is the twice of that of steam 1, there are less stream 2calls in the system and the stream 1 have larger probability tobe accepted. Because is large, the acceptance of all stream 1and stream 2 will be affected.Fig. 7 demonstrated that the dropping probability of two mul-

timedia streams increase with the growth of the arrival rate ofPU. Larger represents that the UEs have higher probabilityto turn off its working antennas. Under large some ongoingSUs must be dropped and therefore the throughput is reduced.In Fig. 7, we observed that the dropping probability of stream 1is larger than that of stream 2. This is because that the 1st stream


Fig. 8. Throughput of SUs.

TABLE VTRAFFIC STREAMS IN THE SYSTEM

requires two-way (uplink and the downlink) slot allocations be-tween two nodes within the CCRN. The 2nd stream only needsone direction link (uplink) slots allocation. If a PU arrives andwants to recycle the channel used by SUs, stream 1 will havelarger probability to be dropped. In Fig. 8, the throughput ofstream 2 is larger than that of stream 1, for both the blockingprobability and dropping probability of stream 2 are smaller (seeFigs. 6 and 7) than those of stream 1. The calls of stream 2 havelarger accepted probability and less dropped probability by theservice system. Therefore, the radio of successful serviced callsof stream 2 is larger than that of stream 1 and the throughput ofstream 2 is larger than that of stream 1.From the simulation and analysis results described from

Figs. 5–8, the correctness of the proposed modeling is verified.If there are many types of stream in the system, or the numbersof channels, slots and antennas are large, the multiple Markovchain is hard to be expressed due to page limit in this paper.Next, we studied a more complex scenario. We assume thereare total nine types of traffic in terms of different communica-tion range and time slots requirements in the system, which isshown in Table V.

Fig. 9. Blocking probability of PUs.

Fig. 10. Blocking probability of SUs.

Let , , , to 6 calls/s,. All the arrival rates of nine types of SUs are equal to

2 calls/s, and the service rate of each SU is assumed to 1 call/sif the SU only uses one slot for transmission. In our simulation,the traffic intensity of PUs are from 0.5 to 3, and the trafficintensity of each group of SUs are two if only one slot used. Thenumber of data antenna of each UE is set to three (not includedthe scanning antenna). The total number of PCs is four, and eachchannel can be divided into five slots for CCRN communication.The simulation time lasts 5000 s.From Fig. 9, we can observe that the blocking probability of

PUs will become larger with the increasing traffic intensity ofPUs. Our simulation curve perfect matches the classic blockedcall clear modal. Fig. 9 also shows that the behavior of PUs isnot impacted by SUs.Fig. 10 shows the performances of SUs with different types

and slot requirements. In Fig. 10, we can observe that multi-media steamswithin the CCRN (first, fourth, and seventh streamin the figure) have higher blocking probability than the that ofmultimedia steams between the nodes in CCRN and outside


Fig. 11. Dropping probability of SUs.

Fig. 12. Throughput of SUs.

CCRN. This is because that each communication with CCRNneeds double time slots including both the uplink and the down-link and the other communications only need slot allocation ineither uplink or downlink. Therefore, they have lower blockingprobability (streams 2, 3, 5, 6, 8, and 9).We can observe that the blocking probability for streams 1,

4, and 7 increases with the increasing the least required numberof the time slots However, for the communications between thenodes in CCRN and nodes outside CCRN, the blocking proba-bility also decrease if the number of the least slots requirementsdecrease. Thus, stream 8 and 9 have higher blocking probabilitythan that of streams 2, 3, 5, and 6.In Fig. 11, the dropping probability for all SUs increases with

the increase of arrival of PUs. For streams 8 and 9, they have lessdropping rate because they have larger minimum /maximumslots requirements. Fig. 11 also shows that the dropping prob-ability is sensitive to the calls ability of release slots. If the in-terval between its minimum and maximum slots requirement islarger, the call has less dropping rate. For example, streams 2, 3,

5, and 6 have larger dropping rate than streams 8 and 9. It indi-cated that the communications within CCRN have larger drop-ping rates. This is because those calls communication withinCCRN should release double number of slots including uplinkand downlink. For example, stream 1 (stream 4 or stream 7) haslarger dropping rate than that of streams 2 and 3 (streams 5 and6 or streams 8 and 9).In Fig. 12, the throughput of each stream decreases with the

increase of the arrival rate of PUs. Obviously, more PUs in thesystem, less channels can be used by CCRN. The number ofworking antennas of CCRN reduces, the slots in CCRN willdecrease and less SUs can be successfully served, which re-duces the throughput of each stream (and of all the steams ofSUs). Streams 1, 4, and 7 have less throughput than that of othersteams because the SUs of the three streams are hardly to be ac-cepted in the CCRN due to double directional slot allocation.

V. CONCLUSIONS

In the paper, we proposed a novel resource allocationmodel for an infrastructure based cognitive radio networks.In the CCRN, each communication node is equipped withmultiple antennas and these antennas are dynamically set tosending/receiving/idle status. The number of working antennasis dynamically changed according to the arrival of PUs. Theyare in sending/receiving state at the same time due to the hard-ware limits. Three types of calls with different slot requirementsare studied. We proposed to use a multi-dimensional Markovchain to model the channel allocation for multimedia streamingover CCRN. The behavior of multimedia traffic over cognitiveradio networks was studied by both theoretical analysis andsimulation. The match between the simulation and theoreticalanalysis proved that our proposed model is accurate.

REFERENCES[1] Y. Chen, Y. Wu, B. Wang, and K. J. R. Liu, “Spectrum auction games

for multimedia streaming over cognitive radio networks,” IEEE Trans.Commun., vol. 58, no. 8, pp. 2381–2390, Aug. 2010.

[2] H.-P. Shiang and M. van der Schaar, “Queuing-based dynamic channelselection for heterogeneous multimedia applications over cognitiveradio networks,” IEEE Trans. Multimedia, vol. 10, no. 5, pp. 896–909,Aug. 2008.

[3] H. Kushwaha, Y. Xing, R. Chandramouli, and H. Heffes, “Reliablemultimedia transmission over cognitive radio networks using fountaincodes,” Proc. IEEE, vol. 96, no. 1, pp. 155–165, Jan. 2008.

[4] D. Hu, S. Mao, Y. T. Hou, and J. H. Reed, “Scalable video multicastin cognitive radio networks,” IEEE J. Sel. Areas Commun., vol. 28, no.3, pp. 334–344, Apr. 2010.

[5] M. van der Schaar and F. Fu, “Spectrum access games and strategiclearning in cognitive radio networks for delay-critical applications,”Proc. IEEE, vol. 97, no. 4, pp. 720–740, Apr. 2009.

[6] H. P. Schang andM. van der Schaar, “Queuing-based dynamic channelselection for heterogeneous multimedia applications over cognitiveradio networks,” IEEE Trans. Multimedia, vol. 10, no. 5, pp. 896–909,Aug. 2008.

[7] S. Ali and R. Yu, “Cross-layer QoS provisioning for multimedia trans-missions in cognitive radio networks,” in Proc. IEEE WCNC, 2009.

[8] R. B. Cooper, Introduction to Queueing Theory, 2nd ed. Amsterdam,The Netherlands: North Holland, 1981.

[9] H. Mansour, J. W. Huang, and V. Krishnamurthy, “Multi-user scalablevideo transmission control in cognitive radio networks as a Markoviandynamic game,” in Proc 48th IEEE Conf. Decision and Control, 2009Held Jointly with the 2009 28th Chinese Control Conf., CDC/CCC2009, Dec. 15–18, 2009, pp. 4735–4740.

[10] P. Wang, D. Niyato, and H. Jiang, “Voice service support over cogni-tive radio networks,” in IEEE Int. Conf. Communications, Jun. 14–18,2009, pp. 1–5.


[11] P. Jung, A. Viessmann, C. Spiegel, A. Burnic, Z. Bai, G. H. Bruck,K. Statnikov, A. Waadt, S. Wang, X. Popon, R. R. Velilla, Saarnisaari,M. Alles, T. Brack, F. Kienle, F. Berens, S. Rotolo, F. M. Scalise, andN. Wehn, “Cognitive radio prototyping,” in 3rd Int. Conf. CognitiveRadio Oriented Wireless Networks and Communications, CrownCom2008, May 15–17, 2008, pp. 1–6.

[12] S. Gunawardena and W. Zhuang, “Voice capacity of cognitive radionetworks,” in IEEE Int. Conf. Communications (ICC), May 23–27,2010, pp. 1–5.

[13] L. Zhou, X. Wang, W. Tu, G. Mutean, and B. Geller, “Distributedscheduling scheme for video streaming over multi-channel multi-radiomulti-hop wireless networks,” IEEE J. Sel. Areas Commun., vol. 28,no. 3, pp. 409–419, Apr. 2010.

[14] L. Zhou, N. Xiong, L. Shu, A. Vasilakos, and S.-S. Yeo, “Context-aware multimedia service in heterogeneous networks,” IEEE Intell.Syst., vol. 25, no. 2, pp. 40–47, Mar./Apr. 2010.

[15] L. Zhou and Y. Zhang, “Distributed media-service scheme for P2P-based vehicular networks,” IEEE Trans. Veh. Technol., vol. 60, no. 2,pp. 692–703, Feb. 2011.

[16] W. Du, W. Jia, G. Wang, and W. Lu, “Analysis of channel alloca-tion scheme for wireless cellular networks,” Int. J. Ad Hoc UbiquitousComput., vol. 4, no. 3/4, pp. 201–209, 2009.

[17] X. Li and S. A. Zekavat, “Spectrum sharing across multiple serviceproviders via cognitive radio nodes,” IET Commun., vol. 4, no. 5, pp.551–561, Mar. 2010.

[18] D. Hu and S. Mao, “Streaming scalable videos over multi-hop cogni-tive radio networks,” IEEE Trans. Wireless Commun., vol. 9, no. 11,pp. 3501–3511, Nov. 2010.

[19] D. Hu, S. Mao, Y. T. Hou, and J. H. Reed, “Fine grained scalabilityvideo multicast in cognitive radio networks,” IEEE J. Sel. AreasCommun., Special Issue on Wireless Video Transmission, vol. 28, no.3, pp. 334–344, Apr. 2010.

[20] H. Luo, S. Ci, D. Wu, and H. Tang, “Cross-layer design for real-timevideo transmission in cognitive wireless networks,” in Proc. IEEE IN-FOCOM-2010 Workshop on Cognitive Wireless Communications andNetworking, Mar. 2010.

[21] S. Xie, Y. Liu, Y. Zhang, and R. Yu, “A parallel cooperative spectrumsensing in cognitive radio networks,” IEEE Trans. Veh. Technol., vol.59, no. 8, pp. 4079–4092, Oct. 2010.

[22] J. Xiang, Y. Zhang, and T. Skeie, “Medium access control protocolsin cognitive radio networks,”Wireless Commun. Mobile Comput., vol.10, no. 1, pp. 31–49, 2010.

Tigang Jiang received the B.Eng degree in materialengineering from Xi’an Jiaotong University, China,in 1998, the M.Eng. degree in computer engineeringfrom Southwest Jiaotong University, China, in 2001,and the Ph.D. degree in communication engineeringfrom the Southwest Jiaotong University in 2005.Currently, he is an Associate Professor in

the School of Communication and InformationEngineering, University of Electronic ScienceTechnology of China (UESTC), Chengdu. He haspublished more than 30 journal and conference

papers. His research interests include parallel and distributed computing,wireless networks, cognitive radio networks and pattern recognition.

Honggang Wang (M’06) received the Ph.D. degreein computer engineering from the University of Ne-braska-Lincoln in 2009.He is an Assistant Professor in the Electrical and

Computer Engineering Department, University ofMassachusetts-Dartmouth. His research interestsinclude networking, wireless sensor networks, mul-timedia communication, network and informationsecurity. He serves as an Associate Editor of Wiley’sSecurity and Communication Networks (SCN).Dr. Wang is the TPC member for IEEE Globecom

2010–2011, IEEE WiMob 2008, 2009 and a Co-chair of IWCMC 2010 Multi-media over Wireless Symposium (http://iwcmc.com/).

Yan Zhang (SM’10) received the Ph.D. degree fromNanyang Technological University, Singapore.Since August 2006, he is working with Simula

Research Laboratory, Lysaker, Norway, where heis currently Senior Research Scientist. He is alsoan Associate Professor (part-time) at the Universityof Oslo, Norway. He is a regional editor, associateeditor, on the editorial board, or guest editor of anumber of international journals. He is currentlyserving the Book Series Editor for the book serieson Wireless Networks and Mobile Communications

(Auerbach Publications, CRC Press, Taylor & Francis Group). His researchinterests include resource, mobility, spectrum, energy, and data management inwireless communications and networking.Dr. Zhang serves as organizing committee chairs for many international

conferences, including AINA 2011, WICON 2010, IWCMC 2010/2009,BODYNETS 2010, BROADNETS 2009, ACM MobiHoc 2008, IEEE ISM2007, and CHINACOM 2009/2008.

Documents

IEEE SYSTEMS JOURNAL, VOL. 5, NO. 3, SEPTEMBER 2011 417 ...folk.uio.no/yanzhang/IEEESystemSep2011.pdf · media system performance over cognitive radio ... radio network (CCRN) consisting