Coexistence Wi-Fi MAC Design for Mitigating...

1

Coexistence Wi-Fi MAC Design for MitigatingInterference Caused by Collocated Bluetooth

Alex Chia-Chun Hsu, Member, IEEE, David S. L. Wei, Senior Member,IEEE,and C.-C. Jay Kuo, Fellow, IEEE

�

Abstract—A non-collaborative coexistence mechanism for wireless-fidelity (Wi-Fi) and bluetooth (BT) systems based on dynamic packetfragmentation is proposed in this work. The basic idea is to adapt thepacket length of Wi-Fi in the MAC layer such that the fragmented packethas a better chance to survive the interference from the nearby BTdevices. We first develop an analytical model that specifies the informa-tion required by the Wi-Fi MAC layer to decide the best fragmentationstrategy. Then, this model is extended to analyze the throughput andtransmission delay of the Wi-Fi device. The analytical model is validatedby computer simulation. Furthermore, it is demonstrated by simulationresults that the proposed coexistence mechanism improves the perfor-mance of Wi-Fi in throughput and transmission delay significantly whilerelatively smaller performance improvement is observed for BT.

Index Terms—Fragmentation, Wi-Fi, Bluetooth, non-collaborative coex-istence mechanism, medium access control, unlicensed spectrum.

1 INTRODUCTION

The ISM unlicensed (UL) band is presently populatedby various wireless devices [2], [3], [4], [5]. Most ofthese devices are used for wireless local area networking(WLAN) with the wireless-fidelity (Wi-Fi) technology [2]or wireless personal area networking (WPAN) with theBluetooth (BT) technology [3]. Since WLAN and WPANare complementary rather than competing technologies,it is likely that Wi-Fi and Bluetooth devices will operateconcurrently in close proximity. Since both devices usethe same frequency band and their radio coverages over-lap with each other, the interference between them can bevery severe [6], [7], [8], [9], [10], [11], [12], [13], [14], [15],[16], [17], [18]. To provide a ubiquitous communicationenvironment through the shared frequency band, weshould not only enable devices to access the UL bandefficiently but also develop a coexistence mechanism thatdetects and mitigates the interference between differenttypes of wireless devices.

• Alex C.-C. Hsu is with MediaTek Inc., HsinChu, Taiwan, R.O.C. Email:[email protected].

• David S. L. Wei is with the Computer and Information Science Depart-ment, Fordham University, New York, NY. E-mail: [email protected].

• C.-C. Jay Kuo are with the Ming Hsieh Department of Electrical Engi-neering, University of Southern California, Los Angeles, CA 90089-2564,USA. E-mail: [email protected].

• A part of this work was presented at the Military CommunicationConference (MILCOM) 2006, Washington DC, October 2006 [1].

Due to the importance of coexistence, IEEE has createda coexistence task group, called the IEEE 802.15 TG2[19], to study and reduce the interference impact onthe throughput of coexisting wireless devices. The taskgroup has defined two classes of coexistence mechanism,namely, collaborative and non-collaborative. Althoughthe classification scheme is mainly developed for BT,the principle can be applied to other coexisting sce-narios. The collaborative mechanism works only whenthe information exchange is possible among coexistingheterogeneous networks [12], [13], [14], [19], [20], andthus is feasible only when the two systems are installedon the same device and controlled by a centralizedcontroller, e.g. a common driver. On the other hand, anon-collaborative mechanism is free from such a require-ment and it is likely to be employed in most of thepractical scenarios [1], [15], [21]. Under non-collaborativecoexistence, each device simply takes its own maneuverto reduce the interference.

A non-collaborative coexistence solution for Wi-Fi isthe main focus of this paper. The proposed mechanismdynamically adjusts the fragmentation level based on thecurrent level of the packet error rate (PER). Thus, it iscalled the dynamic fragmentation (DF) scheme, which istriggered only when it is needed. Our mechanism hastwo versions. The version for mobile networks is namedDF-I while the version for static networks is named DF-II. We first develop an analytical model to evaluate thePER of Wi-Fi and BT devices under interference, andthen show that the model can be employed to effectivelydetermine the right time (in terms of PER) for furtherfragmentation or move back to the previous state of non-fragmentation.

Furthermore, with the proposed DF scheme, 802.11MAC can distinguish whether the failed transmissionis due to interference caused by a BT device or due tocollision caused by a Wi-Fi station so that different actionscan be taken to achieve higher throughput. It will beshown by simulation results that our DF mechanismcould reduce the interference between Wi-Fi and BT andsignificantly improve the performance of Wi-Fi in boththroughput and transmission delay, although only slightperformance improvement is observed for BT.

The rest of the paper is organized as follows. Previous

Digital Object Indentifier 10.1109/TC.2013.224 0018-9340/13/$31.00 © 2013 IEEE

IEEE TRANSACTIONS ON COMPUTERSThis article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

2

work on coexistence is reviewed in Section 2. The dy-namic fragmentation (DF) scheme is described in Section3, where the analytical model for interference is given inSection 3.2, transmission time and transition thresholdare analyzed in Section 3.3, throughput and transmissiondelay are studied in Sections 3.4 and 3.5, respectively.Simulation results are given in Section 4 to validate ouranalytical model and to show the performance improve-ment on throughput and transmission delay. Concludingremarks are drawn in Section 5.

2 REVIEW OF PREVIOUS WORK

Coexistence issue of different types of wireless networksystems has already been rigorously studied [5], [6], [12],[16], [22], [23], [24], [25], [26], [27]. Some works focus oninvestigating the effects of interference between differenttypes of wireless systems and most of those works areperformed at the physical layer, while other works focuson devising techniques to mitigate interference betweenthe different systems. Our work focuses on the issue ofmitigating interference between Wi-Fi and BT.

With the rapid frequency hopping mechanism and abroader operating frequency band, BT is more capableof avoiding interference caused by Wi-Fi devices. The BTperformance in the presence of Wi-Fi has been studiedin [8], [10], [13], [14], [15], [21], [28], [29]. There havebeen works on non-collaborative mechanisms devel-oped to further strengthen the interference mitigationcapability of BT, including AFH (Adaptive FrequencyHopping) [15], [19], [30], BIAS (Bluetooth InterferenceAware Scheduling) [31], D-OLA (Data-OverLap Avoid-ance) [32], DAFH (Dynamic Adaptive Frequency Hop-ping) [33], and DCT (Dual Channel Transmission) [10].These mechanisms control the hopset to avoid over-lapping in frequency. The basic idea is to distinguishgood channels from bad ones and then let the hoppingsequence visit good channels more frequently than badones.

In contrast, Wi-Fi is more vulnerable to interferencecaused by BT devices due to its longer data packet andlack of frequency agility. Thus, it is crucial to developsome coexistence mechanisms for Wi-Fi. Unfortunately,not much such research work has been done so far.A scheduling mechanism, known as V-OLA (Voice-OverLap Avoidance) [32], has been proposed to avoidthe interference from the BT voice traffic by squeezingWi-Fi transmission into the idle period between con-secutive BT voice packets. However, it works when BTdevices have voice traffic only and, otherwise, requirescollocated BT devices to run D-OLA [32]. Consequently,it is not a pure non-collaborative solution. More recently,packet fragmentation has been proposed as an effectiveway to mitigate interference. An analytical model ispresented in [34] for finding the optimal fragmentationwith respect to PER by solving involved differentialequations. However, though the developed model isquite helpful in investigating the effects of interference

between Wi-Fi and BT, it may not be a useful tool tomitigate interference between the two systems due tothe burdensome complexity of the model, i.e. by the timewhen optimal fragmentation is obtained, the interferencestatus may have already been changed. Therefore, in-stead of finding the optimal packet length in arduouseffort as what [34] does, we aim to provide immediateperformance boost rather than gradual performance ad-justment. In this paper, we will show that to find theright timing for fragmentation is more effective thanto costly determine the optimal fragmentation. To thebest of our knowledge, so far we have not yet seenany other pure non-collaborative interference mitigationmechanism developed for MAC protocol of Wi-Fi.

Interference can be modeled in different ways [35][36] [37] [38], and, in general, can be categorized intotwo groups, namely physical model and protocol model. Inthe physical model [35] [36], SNRij denotes the signal-to-noise ratio at node nj for transmissions received fromnode ni. The total noise at nj consists of the ambientnoise plus the interference due to other ongoing trans-missions in the network. A transmission from node ni

to node nj is successful if SNRij ≥ SNRt, where SNRt

is a threshold signal-to-noise ratio. In the protocol model[36], there are V number of nodes in a wireless network,where nodes are denoted by ni, 1 ≤ i ≤ V , and dij repre-sents the distance between nodes ni and nj . Each node,ni, is equipped with a radio with communication rangeRi and a larger interference range R′

i. In this model, ifthere is a single wireless channel, a transmission fromni to nj is successful if both of the following conditionshold: (i) dij ≤ Ri, and (ii) any other node nk, such thatdkj ≤ R′

k, is not transmitting. It has been indicated in[38] that caution should be exercised before interpretingresults based on different interference models. Due tothe nature of this work, we took the approach of protocolmodel.

3 PROPOSED DYNAMIC FRAGMENTATION(DF) MECHANISM

3.1 Description of DF MechanismThere are two fundamental issues that Wi-Fi has tocope with to mitigate interference in a coexistence en-vironment. First, a Wi-Fi station cannot determine if apacket loss is due to collision or interference. The impo-tence of the PHY layer resolution on such an incidentlimits the capability of MAC to improve the coexistingperformance. As a result, Wi-Fi’s collision avoidancemechanism (which is CSMA/CA) would treat all packetloss incidents in the same way, i.e. double the backoffwindow and retransmit the packet. This leads to thesecond issue: CSMA/CA is not efficient in dealing withinterference. Basically, CSMA/CA is designed to solvethe traffic congestion problem among Wi-Fi stations.With a longer backoff window, the traffic is expected toaverage out over time and thus lower the collision rate.However, the interference rate (caused by BT) will not


3

be lowered by simply increasing the backoff time dueto the fact that there is no backoff window mechanismon the BT side. Failing to lower the interference rateby increasing the backoff time, CSMA/CA is ineffectiveand simply introduces unnecessary overhead. We wouldlike to devise a new mechanism to address the problemof packet loss due to interference, which motivates ourwork of dynamic fragmentation (DF) mechanism.

The basic idea is to develop an algorithm that adjuststhe Wi-Fi packet length using the existing fragmentationfunction of 802.11 [2] and the latest PER informationto reduce the interference rate caused by BT. In otherwords, legacy 802.11 MAC will be enhanced by theproposed DF algorithm so that it can handle the interfer-ence problem at run time. It is worthwhile to emphasizethat the DF algorithm aims at reducing the interferencerate but not the collision rate. The task of collision ratereduction is still on the shoulder of CSMA/CA.

As depicted in Fig. 1, there are two states in theproposed DF mechanism; namely, states 1 and 2. The en-tire communication payload is transmitted in one piecewithout fragmentation in state 1. On the other hand,the entire communication payload of a single packet isdivided into η fragments, which are transmitted sequen-tially, in state 2. The system collects the packet error rate(PER) information periodically in a fixed time interval.For state transition, we compare PER and threshold p.If the current system is in state 1 and the latest PER ishigher than p, the system transits from state 1 to state 2.If the current system is in state 2 and the latest PER islower than p, then the system transits from state 2 backto state 1. For all other situations, the system remains inits original state without state transition. Determining aproper threshold value, p, is critical in the proposed DFalgorithm. This can be achieved by comparing the frag-mentation cost and the throughput gain, which will beelaborated later. Although the proposed DF mechanismemploys two states only, to be generic, our analyticalmodel is developed to model the transition from the stateof n fragments to the state of ηn fragments such that itcan be applied to other scenarios with more complicatedfragmentation schemes as well.

Before discussing the selection of p, we explain howfragmentation works below. Fig. 2(a) shows the basicpacket transmission of legacy 802.11 without fragmen-tation. It waits for a DIFS, keeps listening until theend of backoff window BW (or contention window)and then sends data in one piece. Finally, an ACKcompletes the transmission. Fig. 2(b) shows the packettransmission with fragmentation. In this example, thepayload is divided into two fragments, i.e., DATA 1and DATA 2. The first fragment is sent with the fullcontention mechanism. Then, after SIFS, the secondfragment is sent without contention. Fragments areACKed separately. Fig. 2(c) shows what happens whensome fragment suffers from transmission failure. Eachof the failed fragments has to be retransmitted with thefull contention mechanism. The bottom line is that the

Fig. 1. Flowchart of Dynamic Fragmentation Mechanism

prior fragment has to be successfully received beforeany attempt of the next fragment. For transmissions inthe sequel, if it is not a retransmission, the contentionmechanism could be saved.

Fig. 2. Successful transmission of (a) legacy 802.11with no fragmentation and (b) two fragments with noretransmission; and retransmission of (c) DF-I for mobileWLAN and (d) DF-II for static WLAN.

3.2 Interference Rate AnalysisSince the performance affected by interference is of ourmain concern, following [32], we assume that there isno hidden node problem and the effect of RTS/CTS isignored to simplify the analysis. Nevertheless, accordingto our study, the outcome remains unaffected even withthe intervention of RTS/CTS. To conduct the analysis,we need to develop an interference model first. Interfer-ence between Wi-Fi and BT occurs when transmissions


4

of the two systems overlap both in frequency and time.Whenever the residual signal strength of one is higherthan the SINR threshold of the other, it results in a packetloss. Depending on the spatial relation, transmissionpower, the carrier sensing threshold and channel condi-tions, there are three scenarios of transmission failure: BTpacket loss, Wi-Fi packet loss, or both. Since the nature ofcollision due to BT/Wi-Fi interference is different fromthat due to the contention of multiple Wi-Fi stations,we refer the packet loss caused by interference betweenBT and Wi-Fi as an interference incident and use collisionexclusively for the packet loss caused by the contentionof Wi-Fi stations throughout the rest of this paper.

Normally, a Wi-Fi packet has a length equal to thatof several BT time slots. Let TW be the time durationof Wi-Fi packet transmission1, TB be the BT time slotand TBA represent the active time within each BT timeslot. Furthermore, we use t to denote the time intervalbetween the beginning of Wi-Fi packet and the beginningof the first overlapped BT time slot as shown in Fig. 3.

Fig. 3. Illustration of the Wi-Fi packet transmission andthe BT time slot.

According to [32], the number of BT time slots, N , thatoverlap with a Wi-Fi packet could be calculated by

N =

{�TW

TB�, if t ≤ �TW

TB�TB − TW ,

�TW

TB�+ 1, otherwise.

(1)

If the packet length in Eq. (1) is a random variable, wecan replace TW and N by E[TW ] and E[N ], respectively.The probability that a BT device hops on the frequenciesthat would interfere Wi-Fi transmission is denoted by Pf .Since there is an inactive period in each BT time slot, theutilization rate of a BT time slot is denoted by σ. If thereare multiple collocated WPANs, the number of WPANsis nBT . For the ith piconet πi, we define the traffic load(or piconet activity) Gi to be the probability of a packetin a time slot.

For a single WPAN and a single WLAN, we canexpress the PER of a Wi-Fi station as

PERWi−Fi = 1−(1−PfGσ)N ≈ NPfGσ, where σ =TBA

TB.

(2)

1. TW contains both DATA and ACK, and if any portion of theduration is interfered, it will cause retransmission.

For multiple collocated WPANs, the PER becomes

PERWi−Fi = 1−nBT∏i=1

(1− PfGiσ)N ≈

nBT∑i=1

NPfGiσ. (3)

The PER of WPAN under the interference from Wi-Fiwill not be needed in our model and is thus omitted.

3.3 Transmission Time and Threshold AnalysisFollowing [34], for a successful transmission with nofragmentation as shown in Fig. 2(a), the total time ofa complete transmission can be written as

DIFS +BW + Th + TDATA + SIFS + TACK , (4)

where DIFS and SIFS are two kinds of inter-frameintervals, BW is the time of the backoff window (orcontention window) and Th, TDATA and TACK are thetransmission times for the header, DATA and ACK of apacket, respectively. BW is a random number times theWi-Fi time slot.

If the packet is divided into n fragments as shown inFig. 2(b), the total transmission time with no retrans-mission can be expressed as

DIFS +BW + Th + TDATA

n + SIFS + TACK

+(n− 1)(SIFS + Th + TDATA

n + SIFS + TACK),(5)

which can be further simplified to

DIFS − SIFS +BW + n(SIFS + Th + TDATA

n+SIFS + TACK).

(6)

Note that Toh = Th + TACK + 2 × SIFS is a fixedoverhead for each fragment. It is easy to check thatEq. (4) is a special case of Eq. (5) when n = 1. Thus,the total transmission time, Ts, for n fragments withoutany retransmission is equal to

Ts = DIFS − SIFS +BW + n(Toh +TDATA

n). (7)

If there is a fragment loss, a retransmission would takeplace such as the scenario in Fig. 2(c). Then, the time fora single retransmission is

Tr = DIFS − SIFS +BW + Toh +TDATA

n. (8)

Finally, we could determine the total transmission timeof a packet, which is divided into n fragments andsuffers a total R retransmissions. The total transmissiontime would be the successful transmission time of nfragments plus R times the single retransmission time,i.e.

Tn,R = DIFS − SIFS +BW + n(TDATA

n + Toh)+R(DIFS − SIFS +BW + Toh + TDATA

n )= (R+ 1)(DIFS − SIFS) +

∑1+R BWi

+(n+R)(Toh + TDATA

n )

(9)

After the above analysis, we would like to deter-mine the threshold for state transition. In the proposedDF scheme, the decision on state transition depends


5

on whether the transition could reduce the expectedtransmission time. If the total transmission time canbe reduced by a new state, the state transition willbe conducted. From the state of n fragments to thenext state, the number of fragments changes from n ton′(= ηn) and the retransmission number will changeaccordingly. We use R and R′ to denote the numbersof retransmission before and after the state transition,respectively. Then, the condition for a transition to occurcan be expressed as

E[(R+ 1)(DIFS − SIFS) +∑1+R

i=1 BWi+(n+R)(TDATA

n + Toh)] >

E[(R′ + 1)(DIFS − SIFS) +∑1+R′

i=1 BWi+(n′ +R′)(TDATA

n′ + Toh)]

(10)

After rearranging the terms, the inequality becomes

(E[R]− E[R′])(DIFS − SIFS)+

(E[R]− E[R′]− (η − 1)n)Toh + (E[R]− E[R′]η )TDATA

n

+E[∑R

i=1 BWi]− E[∑R′

i=1 BWi] > 0.

(11)

To calculate the threshold, we need to find all theexpected values in Eq. (11). First, we have

E[R] = E[n∑

i=1

Ri], (12)

where R is the total number of retransmissions and E[Ri]is the expected number of retransmission for each frag-ment, and n times the value would be the expected valueof the total number of retransmission. It is assumed thatRi is geometrically distributed, whose probability is inform of fRi(k) = pk(1−p), where k is the retransmissioncount. Then, E[R] can be expressed as

E[R] = n× E[Ri] =np

1− p. (13)

Next, we want to find E[R′]. For terms in Eqs. (2)and (3), Pf , G and σ will remain the same after statetransition, only N , which is a function of n, will change.Based on Eqs. (2) and (3), we can obtain the following:

PERWi−Fi

PER′Wi−Fi

=p

p′=

N

N ′ ,

and then

p′ =N ′

Np. (14)

Therefore,

E[R′] =n′p′

1− p′=

n′N ′N p

1− N ′N p

=n′p

NN ′ − p

=ηnp

κ− p, (15)

where κ = NN ′ and η = n′

n .To find E[

∑Ri=1 BWi] and E[

∑R′

i=1 BWi], we have tocalculate a few parameters. Let CWmin and CWmax

be the minimum and maximum sizes of the backoffwindow (or contention window), and constants a and

b are defined by CWmin = 2a − 1 and CWmax = 2b − 1.For a fragment enters its kth retransmission, we have

BW (k) ∈ [0, 1, 2, ..., 2k+a − 1]× Tslot

so that the expected backoff window can be expressedas

E[BW (k)] =1

2(2k+a − 1)× Tslot. (16)

For any fragment, E[BW (k)] is the average backoff ofthe kth retransmission. Since the expected value of totalretransmissions of a packet is E[R], the expected valueof total backoff of a fragment is

∑E[Ri]k=1 E[BW (k)]. For a

packet divided into n equal length fragments, the totalbackoff of all fragments is

E[R∑i=1

BWi] = n×E[Ri]∑k=1

E[BW (k)].

Since the backoff window is doubled only up to theupper bound CWmax, we consider the following twocases.Case A: E[Ri] ≤ b − a and the backoff window isnot greater than the maximum contention window size.Then, we have

E[R∑i=1


E[BW (k)]

= n×E[Ri]∑k=1

1

2(2k+a − 1)× Tslot

=1

2(2a+1(2E[Ri] − 1)− E[Ri])× nTslot.

Case B: E[Ri] > b − a and the backoff window reachesthe maximum contention window size. Then, we get

E[R∑i=1


E[BW (k)]

= n× {b−a∑k=1

1

2(2a+k − 1) +

E[Ri]∑k=b−a+1

1

2(2b − 1)} × Tslot

= 12 (2

a+1(2b−a − 1)− 2b(b− a) + (2b − 1)E[Ri])× nTslot

Without loss of generality, we consider the case withE[Ri] ≤ b − a and E[R′

i] ≤ b − a. Then, the followingexpressions can be derived:

E[R∑i=1

BWi] =

E[Ri]∑k=1

1

2(2k+a − 1)× nTslot

=1

2(2a+1(2E[Ri] − 1)− E[Ri])× nTslot, (17)

and

E[R′∑i=1

BWi] =

E[R′i]∑

k=1

1

2(2i+a − 1)× ηnTslot

=1

2(2a+1(2E[R′

i] − 1)− E[R′i])× ηnTslot. (18)


6

Since all terms in Eq. (11) are now available, we couldplug them in to calculate the threshold. That is, we have

F (p) =(E[R]− E[R′])(DIFS − SIFS)

+ (E[R]− E[R′]− (η − 1)n)Toh

+ (E[R]− E[R′]η

)TDATA

n

+ E[R∑i=1

BWi]− E[R′∑i=1

BWi]. (19)

We can solve F (p) = 0 with F (p−)F (p+) < 0 for p.

3.4 Throughput AnalysisAs defined in [39], the throughput is the fraction ofthe total transmission time dedicated to the payloadtransmission. Mathematically, we have

throughput =TDATA

E[(R + 1)(DIFS − SIFS)+

1+R∑i=1

BWi + (n + R)(TDATA

n+ Toh)]

(20)According to Eq. (20), to improve the throughput, one

can either cut down overhead or reduce the numberof retransmissions. For fragmentation, there is a trade-off between retransmission and overhead. The more atransmission is fragmented, the more overhead it willincur. However, the fewer and cheaper retransmission isexpected. When the gain of reducing the retransmissionnumber cannot balance off the loss of the increased over-head, fragmentation is not a good solution. This explainsthe necessity of a careful analysis of the fragmentationcost so as to guarantee the performance improvement.

We investigate the overhead cost and find an oppor-tunity for further performance improvement below. Thetotal overhead cost can be broken down into four parts:the header, the inter-frame space, the ACK message andthe backoff window. The first three parts are fixed whilethe last one varies with PER. Moreover, the overhead ofthe backoff window grows exponentially with PER.

For a normal Wi-Fi transmission in a static environ-ment (without mobility), there should be no collision inthe middle of a transmission session. Collision happensonly when two Wi-Fi stations randomly choose the samebackoff window and they start their transmission atthe same time. In other words, collision either happensfrom the beginning of a transmission or it does nothappen at all. This scenario is quite different from that ofinterference. Since BT has no carrier sensing mechanism,interference could happen at any time during a Wi-Fi transmission. Since Wi-Fi performs carrier sensing,interference is unlikely to happen at the beginning ofa transmission.

Without DF, CSMA/CA alone cannot exploit thisinteresting feature. With DF, we can associate somesuitable interpretation with the transmission status of acertain fragment. For example, if the first fragment in a

sequence of multiple fragments is lost, it is most likelya collision. On the other hand, if subsequent fragmentsencounter a transmission failure, it is quite likely due tointerference. In CSMA/CA with DF, if a station enters afragmentation state but does not observe a reduction inPER or any transmission failure in the second fragmentand beyond, it is likely that the high PER is caused bycollision only. Thus, we should switch back to the non-fragmentation state since, under such a scenario, frag-mentation will fail to reduce PER but just decreases thethroughput by introducing some unnecessary overhead.

Except for the first fragment, retransmission of subse-quent fragments is most likely caused by interferenceso that the backoff window assigned by the collisionavoidance mechanism for those subsequent fragmentswill simply introduce unnecessary overhead but has noimpact on reducing the interference probability. Conse-quently, we should simply bypass them as shown inFig. 2(d). Under this new procedure, when a sender en-counters a transmission failure for the second fragmentor its subsequent ones, it will retransmit the fragmentimmediately after the ACK timeout without waiting forthe backoff window. This revised scheme is called DF-II.The original scheme, called DF-I, can be used for mobilenetworks while DF-II can improve the performance ofstatic networks.

For DF-II, since collision may incur the retransmis-sion of the first fragment, we should keep the backoffwindow for the retransmission of the first fragment.Let r be the retransmission count of the first fragment.Then, R − r is the retransmission count for subsequentfragments which are free from the backoff procedure. Inanalogy with Eq. (9), we can derive the total transmissiontime for successful transmission of n fragments plus Rretransmissions for DF-II as

Tn,R,II =(DIFS − SIFS +BW + n(TDATA

n+ Toh))

+ r(DIFS − SIFS +BW +TDATA

n+ Toh)

+ (R− r)(DIFS − SIFS +TDATA

n+ Toh)

=(R+ 1)(DIFS − SIFS) +1+r∑i=1

BWi

+ (n+R)(TDATA

n+ Toh) (21)

The threshold decision function for DF-II can be derivedby modifying Eq. (19) as

F (p) =(E[R]− E[R′])(DIFS − SIFS)

+ (E[R]− E[R′]− (η − 1)n)Toh

+ (E[R]− E[R′]η

)1

nTDATA

+ E[r∑

i=1

BWi]− E[r′∑i=1

BWi] (22)

By removing unnecessary backoff windows, we can


7

lower the price of fragmentation so that the thresholdof entering the fragmentation state is lowered.

With the system parameters given in Table 1, we canstudy the relationship between throughput and PERunder different schemes and then determine thresholdp accordingly. Three cases are examined in Fig. 4. Theyare the legacy 802.11 without fragmentation, DF-I andDF-II with fixed fragmentation (η = 2). The dashed linesin Fig. 4 are analytical throughput values as functions ofdifferent PER levels while the solid lines are the valuesobtained by computer simulation. We see a close matchbetween analytical and simulated results. Furthermore,the throughput performance favors no fragmentationin lower PER values but fragmentation in higher PERvalues, and fragmented DF-II outperforms fragmentedDF-I. These are consistent with our above analysis.

Fig. 4. Throughput as a function of PER for threeschemes.

802.11 Wi-Fi ParametersParameter Assigned ValueSlot time 20 μs

DIFS 50 μsSIFS 10 μs

PHY header 192 bitsMAC header 224 bits

Payload 12000 bitsACK 112 bits

CWmin 32CWmax 1024

BlueTooth ParametersTB 625 μsTBA 366 μs

TABLE 1Parameters for Wi-Fi and BT systems.

The legacy 802.11 intersects with fragmented DF-I andDF-II at PER equal to 0.4 and 0.3, respectively. It meansthat we should choose these values as threshold values.For example, if PER is less than 0.3, DF-II should stay instate 1 (no fragmentation). Otherwise, DF-II should be

in state 2. It is worthwhile to mention that the thresholdvalues are sensitive to the fixed transmission overhead,Toh, which accounts for the majority of the fragmentationcost, and the ratio of overlapped BT time slots in one Wi-Fi fragment of the current and the previous state, whichis κ in Eq. (15).

3.5 Delay Analysis

When a Wi-Fi packet is sent without retransmission norfragmentation, the transmission delay for such perfecttransmission comes from requirements of the standardsuch as DIFS, the contention window, SIFS, time forthe ACK message and headers. However, besides thesefactors, there is another type of delay caused by interfer-ence, i.e. time due to retransmissions. Here, transmissiondelay is defined as the difference between the actualtransmission time and the transmission time for the non-fragmented payload. It is basically the extra time causedby retransmission, the fragmentation overhead and thespacing time set by the standard. Mathematically, wehave

delay = n× Toh

+ (E[R] + 1)× (DIFS − SIFS + Toh +TDATA

n)

+

E[R]∑i=1

BWi. (23)

Since fragmentation decreases the retransmissionpenalty, the proposed DF mechanism can decrease thedelay caused by retransmissions. DF makes the statetransition, i.e. performs fragmentation, only when theexpected transmission time is less than that of the non-fragmentation case, thereby reducing the transmissiondelay. In other words, as far as the delay is concerned, DFin general outperforms scheduling-based algorithms, e.g.[32], in terms of packet transmission delay, which will beverified by computer simulation in the next section.

4 COMPUTER SIMULATION

4.1 Simulation Environment Setup

Our computer simulation environment consists of oneWLAN network and several piconets in proximity. A Wi-Fi device and a BT device are separated in less than 3meters since such a distance results in the most severeinterference effect [17] [18]. This scenario is common inoffices, households, airports, etc. For example, a PDA isconnected to a laptop via Wi-Fi (11 Mb/s), and a nearbypiconet, e.g. cellphone/headset, is communicating overthe BT link. The device separation distance is intention-ally assigned to capture the specific interference scenario.That is, if a BT device operates at Wi-Fi frequency band,and if some Wi-Fi stations are also active, then bothtransmissions would fail. However, if BT transmits out-side the Wi-Fi frequency band, concurrent transmissionis possible. In other words, Pf is exactly 22/79 in our


8

simulation. To reduce the simulation complexity, thefollowing assumptions have been made without lossof generality: (i) the propagation delay of WLAN andWPAN is neglected due to the short operational distance;(ii) WLAN is of low mobility, which is the most commonscenario of Wi-Fi; and (iii) piconets may be mobile, yetits mobility would have little impact on the interferencepattern due to the wide coverage of Wi-Fi.

The arrival rate of Wi-Fi packets is exponentially dis-tributed. The Wi-Fi packet length is fixed in simulation2.The Wi-Fi and BT parameters used in simulation aresummarized in Table 1. The header part in each Wi-Fi fragment consists of the PHY header (or called thepreamble) and the MAC header. With these parameters,we can calculate the best η, which is equal to two in oursimulation, with respect to the assigned packet length. Ahigher η value, i.e. with more number of fragments, willnot decrease κ of Eq. (15), but just increase the overhead.

Both SCO and ACL traffic scenarios are simulated forBT. For the SCO link, the most popular HV3-type linkis used and a packet is generated every six time slotsin both directions. A BT slave can support up to threeSCO links from the same master or two SCO links ifthe links are originated from different masters. An SCOpacket needs no ACK nor retransmission. If a BT slot isnot reserved by SCO, the master could establish the ACLlink on per slot basis. Each ACL link packet needs to beACKed in the next time slot. In the ACL simulation, theDH1-type link is used such that one data packet occupiesone BT time slot. The packet arrival rate of ACL is alsoexponentially distributed.

With different BT traffics, we consider three scenarios:• Wi-Fi runs on legacy 802.11 without any fragmen-

tation;• Wi-Fi runs on dynamic fragmentation DF-I;• Wi-Fi runs on dynamic fragmentation DF-II.

Each plotted value shown in the figures is the averageresults of at least 50 runs.

4.2 Simulation Results and DiscussionTo verify our analysis, we first see that the state tran-sition threshold obtained from the analytical model isclose to that obtained by simulation as shown in Fig. 4.Based on Eqs. (19) and (22), the threshold values for DF-I and DF-II are 0.38 and 0.31, respectively. They are alsoconfirmed in Fig. 4. Furthermore, the relation betweenthe PER of Wi-Fi and the BT traffic load is shown inFig. 5 which verifies Eq. (2). It is clear that simulationcurves are consistent with our analytical prediction.These results demonstrate the accuracy of our analyticalmodel, which can capture the interference phenomenonbetween Wi-Fi and BT well.

Fig. 6 shows the throughput improvement of Wi-Fiwith DF in the log scale. In our calculation, the Wi-Fi

2. For a variable payload length, parametermax Fragment(Payload) length can be used to control the fragmentlength.

Fig. 5. The Wi-Fi PER as a function of the BT traffic load.

Fig. 6. Throughput improvement for Wi-Fi with DF-I andDF-II.

throughput is defined in Eq. (20), which is the ratio oftime dedicated to payload to the total transmission time,including the time spent on retransmissions. When thePER of Wi-Fi is equal to 0.5 and 0.6, the throughputimprovement of DF-I is equal to 15% and 30% respec-tively. For DF-II, the throughput improvement becomes28% and 56%, respectively3. One noticeable trend is thatthe throughput improvement grows exponentially withPER. On the other hand, one might think that we mighthave negative throughput improvement for low PERsince the gain in reducing retransmissions caused byinterference could be lower than the induced overhead.

3. One may argue that the PER of a Wi-Fi is not supposed to be ashigh as 0.3. This is true if the collision is only caused by other Wi-Fidevices. However, if there are BT devices in a close proximity, it is quitepossible to have PER of up to 0.6. As an example, a common scenarioof collocated Wi-Fi and BT in a very close proximity is a smartphoneuser uses BT earphone to have a Skype chat through Wi-Fi.


9

However, since our algorithm is dynamically adjustedbased on the PER level, no negative improvement wouldactually happen. That is, when PER is below the thresh-old, there will be no fragmentation.

Fig. 7 shows the Wi-Fi throughput as a function ofdifferent Wi-Fi traffic loads when there are two SCO linksin presence. Since our mechanism is neither collaborativenor scheduling, our mechanism will not have an extrabenefit from the recursive nature of SCO traffic. TwoSCO links present a substantial traffic load on the BTside. Thus, the DF mechanism outperforms legacy 802.11even with a low Wi-Fi traffic load.

Fig. 7. Throughput of the Wi-Fi network, which coexistswith two SCO links on the BT piconet.

Fig. 8. Throughput of the Wi-Fi network in the presence ofACL link on the BT piconet, where the background Wi-Fitraffic is given by τWi−Fi = 0.3.

Simulation results of the Wi-Fi throughput as a func-tion of the BT ACL traffic load with two background Wi-

Fi traffic loads are shown in Figs. 8 and 9. We see a sig-nificant improvement on the Wi-Fi throughput in Fig. 9and the improvement increases exponentially with theBT traffic load. The crossing points between DF-I/DF-II and legacy 802.11 represent appropriate thresholds.This threshold is a function of the background Wi-Fitraffic load. Intuitively, a higher background traffic loadwill trigger the state transition earlier. Fig. 8 gives theresult of the same simulation setup except for a lowerbackground Wi-Fi traffic load. We see a higher threshold,which is consistent with our intuition.

Fig. 9. Throughput of the Wi-Fi network in the presence ofACL link on the BT piconet, where the background Wi-Fitraffic is given by τWi−Fi = 0.7.

Fig. 10. Throughput of the BT piconet with ACL trafficτBT = 0.7, which coexists with Wi-Fi.

Fig. 10 shows a slight improvement on the BT through-put. The curves of DF-I,II and legacy 802.11 are closeto each other when the Wi-Fi traffic load is low. We


10

see more visible performance improvement when theWi-Fi traffic load is sufficiently high. Even though theimprovement is not impressive at the BT side, ourscheme has no negative impact on the BT performanceat any event. Note that the proposed DF scheme cansignificantly decrease the cost of Wi-Fi retransmission(only the fraction of the entire packet), but cannotprevent any Wi-Fi transmission failure. For the samereason, it can not prevent any BT transmission failure.The small improvement at the BT side comes from lessinterference due to shorter retransmission forced by DF.Thus, the improvement is more substantial when theretransmission time possesses a significant portion of thetotal transmission time, which is always the case whenthe traffic load is sufficiently high.

With two BT piconets that adopt ACL traffic andcoexist with a Wi-Fi network, we show simulation resultsof the Wi-Fi throughput with respect to the combined BTtraffic load in Fig. 11. We see that Figs. 9 and 11 are verysimilar. This implies that, regardless of the number ofcoexisting BT piconets, the degree of interference sensedby Wi-Fi is the cumulative contributions of all coexistingpiconets, which corroborates Eq. (3).

Fig. 11. Throughput of the Wi-Fi network in the presenceof ACL links on multiple BT piconets, where the back-ground Wi-Fi traffic is given by τWi−Fi = 0.7.

Finally, Fig. 12 shows the average packet transmissiondelay of Wi-Fi as a function of the BT traffic load. Inthe simulation, we first calculate the total transmissiontime for files of fixed length, e.g. 1000 packets. Then,we subtract the time needed to transmit the payloadfrom the total transmission time to determine the totaldelay. Then, the total delay is divided by the numberof packets to get the average delay of a single packet.The improvement on packet transmission delay comesprimarily from reduced retransmissions caused by inter-ference. It is demonstrated by simulation results that DFcan improve the transmission delay significantly.

Fig. 12. The average Wi-Fi delay as a function of the BTtraffic load, where the background Wi-Fi traffic is given byτWi−Fi = 0.5.

5 CONCLUSION AND FUTURE WORK

A non-collaborative mechanism, called Dynamic Frag-mentation (DF), to improve the Wi-Fi performance inthe presence of BT interference was proposed in thiswork. We first developed an analytical model to char-acterize the interference between Wi-Fi and BT. Then,we proposed DF to reinforce the coexistence ability ofWi-Fi networks. With DF, a Wi-Fi station can performfragmentation dynamically to reduce interference andincrease throughput. We also investigated the scenarioof static networks and proposed an enhanced solutioncalled DF-II to improve the performance furthermore. Inaddition to the throughput improvement, DF helps Wi-Fi differentiate between interference and collision. Thus,if transmission failures are mostly caused by collision atthe first fragment, DF can recognize the difference andswiftly switch back to the non-fragmentation state toavoid unnecessary overhead. The fragmentation mech-anism has already been described in the legacy 802.11standard. Therefore, our proposed DF-I and DF-II solu-tions can be easily implemented. The derived analyticalresults were validated by simulation results. With thePER level higher than 0.6, we could get more than 56%improvement on throughput for static networks, and30% throughput improvement for mobile networks. Theimprovements grow exponentially with PER. Substantialimprovement on the delay has been observed as well.Due to the non-collaborative nature of our coexistencescheme, there is only slight throughtput improvementon the BT side. However, a non-collaborative coexistencescheme is preferable to a collaborative one.

The current DF mechanism divides the packet intofragments with equal length. It would be interesting todevelop an extended model using more advanced frag-mentation scheme as future work. Our analytical model


11

can provide clues in the design of such fragmentationmechanism. Besides, in addition to devise coexistenceschemes, efforts on regulation, standards, etc. are alsoneeded to enhance the coexistence of various kindsof wireless networks. Though there have been variouskinds of wireless networks, the standardized MACsfor the networks using 2.4 GHz ISM unlicensed bandcan be generally categorized into two types, namelyCSMA/CA (e.g. Wi-Fi or Zigbee) and frequency hopping(e.g. BT). Our work provides important insights into thecoexistence between CSMA/CA networks and frequencyhopping networks, which offers feasible solutions whensetting the standard for future wireless technologies forthe networks using ISM unlicensed band.

REFERENCES

[1] A. C.-C. Hsu, D. S. L. Wei, and C.-C. J. Kuo, “Coexistence mech-anism using dynamic fragmentation for interference mitigationbetween wi-fi and bluetooth,” in MILCOM’06, Washington D.C.,October 2006.

[2] Wireless LAN medium access control (MAC) and physical layer (PHY)specification, ANSI/IEEE Std. 802.11, 1999.

[3] Wireless medium access control (MAC) and physical layer (PHY)specifications for wireless personal area networks (WPANs), IEEE Std.802.15.1, 2002.

[4] I. C. Society, “802.15.4: Wireless Medium Access Control(MAC) and Physical Layer (PHY) Specifications for Low- RateWireless Personal Area Networks (LR-WPANs),” in Availableat: http://standards.ieee.org/getieee802/download/ 802.15.4-2003.pdf.,2003.

[5] T. Kondo, H. Fujita, M. Yoshida, and T. Saito, “Technology forWIFI/bluetooth and WiMax Coexistence,” Fujitsu Scientific andTechnical Journal, vol. 46, no. 1, pp. 72–78, 2010.

[6] K. L. Mills, “Networking for Pervasive Computing,” in NISTSpecial Publication 500-259, July 2005.

[7] L. Cabral and R. Lins, “An Analysis of the QoS in the Trans-mission in 802.11g Networks in the Presence of Bluetooth Inter-ference,” in Fifth International Conference on Wireless and MobileCommunications. ACM, 2009, pp. 76–81.

[8] I. Ashraf, K. Voulgaris, A. Gkelias, M. Dohler, and A. Aghvami,“Impact of Interfering Bluetooth Piconets on a Collocated p -Persistent CSMA-Based WLAN,” IEEE Transactions on VehicularTechnology, vol. 58, no. 9, pp. 4962–4975, 2009.

[9] R. J. Punnoose, R. Tseng, and D. Stancil, “Experimental Results forInterference between Bluetooth and IEEE 802.11b DSSS Systems,”in IEEE VTC Fall 2001.

[10] J. Li and X. Liu, “A Frequency Diversity Technique for Interfer-ence Mitigation in Coexisting Bluetooth and WLAN,” in IEEE ICC2007.

[11] A. Conti, D. Dardari, G. Pasolini, and O. Andrisano, “Bluetoothand IEEE 802.11b coexistence: analytical performance evaluationin fading channels ,” IEEE Journal on Selected Areas in Communi-cations, vol. 21, no. 2, pp. 259 – 269, February 2003.

[12] L. Sydanheimo, M. Keskilammi, and M. Kivikoski, “Performanceissues on the wireless 2.4 GHz ISM band in a multisystemenvironment,” IEEE Transactions on Consumer Electronics, vol. 48,no. 3, p. 643658, 2002.

[13] J. Liang, “Proposal for Collaborative Bluetooth and 802.11b MACMechanisms for Enhancing Coexistence,” in IEEE 802.15-01/080r0,January 2001.

[14] S. Shellhammer, “Collaborative Coexistence Mechanism: TDMAof Bluetooth and 802.11,” in IEEE 802.15-01/025r0, January 2001.

[15] B. Treister, H. Gan, K. Chen, H. Chen, A. Batra, and O.Eliezer,“Adaptive Frequency Hopping: A Non-Collaborative CoexistenceMechanism,” in IEEE 802.15-01/252r0, p. May 2001.

[16] R. Nawaz and S. Sun, “Bluetooth Interference Mitigation in802.11g,” in ICC 2008, 2008.

[17] J. Lansford, A. Stephens, and R. Nevo, “Wi-Fi (802.11b) andBluetooth: enabling coexistence,” IEEE Network, vol. 15, no. 5, pp.20–27, Sep 2001.

[18] N. Golmie, R. E. V. Dyck, A. Soltanian, A. Tonnerre, and O. Rebala,“Interference evaluation of Bluetooth and IEEE 802.11b systems,”ACM Wireless Networks, vol. 9, pp. 202–211, 2003.

[19] IEEE 802.15 WPAN task group 2 (TG2). [Online].Available: http://www.ieee802.org/15/pub/TG2-Coexistence-Mechanisms.html

[20] S. Shellhammer, “IEEE 802.15.2 Overview,” in IEEE P802.15-01/313r1, July 2001.

[21] B. Treister, K. Chen, H. Chen, and A. Batra, “Adaptive FrequencyHopping Clause 14.3,” in IEEE P802.15-TG2-366r1, July 2001.

[22] J. Kim, S. Park, S. H. Rhee, Y.-H. Choi, Y.-U. Chung, and H. Y.Hwang, “Coexistence of WiFi and WiMAX Systems Based on PS-Request Protocols,” Sensors, vol. 11, no. 10, pp. 9700–9716, 2011.

[23] S. Y. Shin, J. S. Kang, and H. S. Park, “Packet Error Rate Analysisof ZigBee under Interferences of Multiple Bluetooth Piconets,” inIEEE 69th Vehicular Technology Conference, 2009 (VTC Spring 2009).,pp. 1–5.

[24] C.-J. M. Liang, N. B. Priyantha, J. Liu, and A. Terzis, “Survivingwi-fi interference in low power ZigBee networks,” in Proceedingsof the 8th ACM Conference on Embedded Networked Sensor Systems,2010.

[25] F. Penna, C. Pastrone, M. Spirito, and R. Garello, “Measurement-Based Analysis of Spectrum Sensing in Adaptive WSNs under Wi-Fi and Bluetooth Interference ,” in IEEE 69th Vehicular TechnologyConference, 2009 (VTC Spring 2009), 2009.

[26] G. Tamilselvan and A. Shanmugam, “Interference Cancellationand Performance Enhancement in Coexistent HeterogeneousWireless Packet Networks using Packet Transmission Time Shar-ing,” International Journal of Computer and Electrical Engineering,vol. 2, no. 1, pp. 92–98, 2010.

[27] A. B. Bagula, A. Mehta, and M. Zennaro, “Experimental Evalua-tion of Interference Mitigation on The 2.4 GHz ISM band UsingChannel Hopping,” in IFIP Wireless Days, 2011.

[28] I. Howitt, “Bluetooth performance in the presence of 802.11bWLAN,” in IEEE Transactions on Vehicular Technology, vol. 51, no. 6,November 2002, pp. 1640–1651.

[29] N. Golmie, N. Chevrollier, and O. Rebala, “Bluetooth and WLANcoexistence: challenges and solutions,” IEEE Wireless Communica-tions Magazine, Dec 2003.

[30] S.-H. Lee and Y.-H. Lee, “Adaptive Frequency Hopping forBluetooth Robust to WLAN Interference,” IEEE CommunicationsLetters, vol. 13, no. 9, pp. 628–630, 2009.

[31] N. Golmie, “Bluetooth dynamic scheduling and interference mit-igation,” ACM Mobile Networks, vol. 9, no. 1, Feb 2004.

[32] C. Chiasserini and R. Rao, “Coexistence mechanisms for inter-ference mitigation in the 2.4-GHz ISM Band,” IEEE Transactionson Wireless Communications, vol. 2, no. 5, pp. 964–975, September2003.

[33] P. Popovski, H. Yomo, and R. Prasad, “Dynamic adaptive fre-quency hopping for mutually interfering wireless personal areanetworks,” IEEE Transactions on Mobile Computing, vol. 5, no. 8,pp. 991–1003, August 2006.

[34] I. Howitt and F. Awad, “Optimizing IEEE 802.11b packet fragmen-tation in collocated Bluetooth interference,” in IEEE Transactionson Communications, vol. 53, no. 6, June 2005, pp. 936–938.

[35] K. Jain, J. Padhye, V. Padmanabhan, and L. Qiu, “Impact of Inter-ference on Multi-Hop Wireless Network Performance,” WirelessNetworks, vol. 11, no. 4, pp. 471–487, 2005.

[36] P. Gupta and P. Kumar, “The capacity of wireless networks,” IEEETransactions on Information Theory, vol. 46, no. 2, pp. 388–404, 2000.

[37] M. Alicherry, R. Bhatia, and L. Li, “Joint channel assignmentand routing for throughput optimization in multiradio wirelessmesh,” IEEE Journal on Selected Areas in Communications, vol. 24,no. 11, pp. 960–1971, 2006.

[38] A. Iyer, C. Rosenberg, and A. Karnik, “What is the right modelfor wireless channel interference?” in Proceedings of the 3rd interna-tional conference on Quality of service in heterogeneous wired/wirelessnetworks. ACM Press New York, NY, USA, 2006.

[39] G. Bianchi, “Performance Analysis of the IEEE 802.11 DistributedCoordination Function,” IEEE Journal on Selected Areas in Commu-nications,, vol. 18, no. 3, pp. 535–547, 2000.


12

PLACEPHOTOHERE

Alex C.-C. Hsu received his B.S. degree fromNational Tsing-Hua University, Hsinchu, Tai-wan, in 1997, M.S. degree in Electrical andComputer Engineering from Purdue University,West Lafayette, Indiana, in 2002, and Ph.D.degrees in Electrical Engineering from Univer-sity of Southern California, Los Angeles, CA,in 2007. Dr. Hsu is currently working as Sr.engineer at MediaTek, Taiwan. His researchinterests are in wireless protocols (LTE/LTE-A/ISM), multi-carrier/RAT aggregation, Machine-

to-Machine communication, Device-to-Device communication, diversedata application, Multimedia Broadcast Multicast Services, and cognitiveradio.

PLACEPHOTOHERE

David S.L. Wei (SM’07) received his Ph.D. de-gree in Computer and Information Science fromthe University of Pennsylvania in 1991. He iscurrently a Professor of Computer and Informa-tion Science Department at Fordham University.From May 1993 to August 1997 he was on theFaculty of Computer Science and Engineeringat the University of Aizu, Japan (as an Asso-ciate Professor and then a Professor). Dr. Weihas authored and co-authored more than 90technical papers in the areas of distributed and

parallel processing, wireless networks and mobile computing, opticalnetworks, peer-to-peer communications, and cognitive radio networksin various archival journals and conference proceedings. He served onthe program committee and was a session chair for several reputedinternational conferences. He was a lead guest editor of IEEE Journalon Selected Areas in Communications for the special issue on MobileComputing and Networking, and was a guest editor of IEEE Journalon Selected Areas in Communications for the special issue on Peer-to-Peer Communications and Applications. He was the chair of IntelligentTransportation Forum of Globecom 2010, the general chair of IntelligentTransportation Workshop of ICC 2011, and the chair of Cloud SecurityForum and Intelligent Transportation Forum of Globecom 2011. He iscurrently a lead guest editor of IEEE Journal on Selected Areas inCommunications for the special issue on Networking Challenges inCloud Computing Systems and Applications, a lead guest editor ofIEEE Transactions on Cloud Computing for the special issue on CloudSecurity, and an Associate Editor of Journal of Circuits, Systems andComputers. Currently, Dr. Wei focuses his research efforts on cloudcomputing, wireless sensor networks, and cognitive radio networks.

PLACEPHOTOHERE

C.C. Jay Kuo (F’99) received the B.S. degree inelectrical engineering from the National TaiwanUniversity, Taipei, Taiwan, in 1980, and the M.S.and Ph.D. degrees in electrical engineering fromthe Massachusetts Institute of Technology, Cam-bridge, in 1985 and 1987, respectively. He isthe Director of the Media Communications Lab,University of Southern California, Los Angeles,where he is also a Professor of electrical en-gineering, computer science, and mathematicswith Ming Hsieh Department of Electrical Engi-

neering. He is the coauthor of about 200 journal papers, 850 conferencepapers, and 10 books. His research interests include digital image/videoanalysis and modeling, multimedia data compression, communicationand networking. Dr. Kuo is a Fellow of the American Association forthe Advancement of Science and the International Society for OpticalEngineers.


Coexistence Wi-Fi MAC Design for Mitigating...

Documents

Transcript of Coexistence Wi-Fi MAC Design for Mitigating...