Anaytica Modeing of Wi-Fi and LTE-LAA Coexistence: Throughput and Impact of Energy Detection Threshod Morteza Mehrnoush, Vanin Sathya, Sumit Roy, and Monisha Ghosh University of Washington, Seatte, WA-99 University of Chicago, Iinois- Emai: {mortezam, sroy}@uw.edu, {vanin, monisha}@uchicago.edu. Abstract With both sma-ce LTE and Wi-Fi networks avaiabe as aternatives for depoyment in unicensed bands (notaby GHz, the investigation into their coexistence is a topic of active interest, primariy driven by industry groups. Notaby, GPP has standardized LTE Licensed Assisted Access (LTE-LAA that seeks to make LTE more co-existence friendy with Wi-Fi by incorporating simiar sensing and back-off features. Nonetheess, the resuts presented by industry groups offer itte consensus on important issues ike respective network parameter settings that promote fair access as required by GPP. Answers to such key system depoyment aspects, in turn, require credibe anaytica modes, on which there has been itte progress to date. Accordingy, in this work, we deveop a new framework for estimating the throughput of Wi-Fi and LTE-LAA in coexistence scenarios via suitabe modifications to the ceebrated Bianchi [] mode. The impact of energy detection (ED threshod on Wi-Fi and LTE-LAA coexistence is expored as a byproduct of this mode and corroborated via a Nationa Instrument (NI experimenta testbed that vaidates the resuts for LTE-LAA access priority cass and. Index Terms Wi-Fi, LTE-LAA, GHz Unicensed band Coexistence. I. INTRODUCTION The increasing penetration of high-end handhed devices using high bandwidth appications (e.g mutimedia streaming has ed to an exponentia increase in mobie data traffic and a consequent bandwidth crunch. Operators have had to resort to provisioning high bandwidth end-user access via sma ce LTE or. Wi-Fi networks to achieve desired peruser throughput. However, in hot-spot (very high demand scenarios, dense depoyment of such sma ces inevitaby eads to the need for time-sharing of the unicensed spectrum between LTE and Wi-Fi, for exampe, when two operators respectivey depoy overapping Wi-Fi and sma-ce LTE networks. An immediate frequency band of interest for such coexistence operation is the GHz UNII bands in US where a significant swath of additiona unicensed spectrum was earmarked by the FCC in []. Wi-Fi networks have been architected for operating in unicensed spectrum via a time-sharing mechanism among other Wi-Fi nodes based on the Distributed Coordination Function (DCF, and with non-wi-fi networks via energy detection (ED and dynamic frequency seection (DFS []. On the This work was supported by the Nationa Science Foundation (NSF under grant. other hand, there are two specifications for unicensed LTE operation with a view to coexistence: LTE Licensed Assisted Access (LTE-LAA and LTE Unicensed (LTE-U. LTE-LAA has been deveoped by GPP and integrates a Listen-Before- Tak (LBT mechanism [], [] - simiar to CSMA for Wi-Fi - to enabe spectrum sharing wordwide in markets where it is mandated. LTE-U empoys an (adaptive duty-cyce based approach - denoted as Carrier Sense Adaptive Transmission (CSAT - to adapt the ON and OFF durations for LTE channe access []. Specificay, GPP has sought to achieve a notion of fair coexistence [], [] whereby LAA design shoud target fair coexistence with existing Wi-Fi networks to not impact Wi-Fi services more than an additiona Wi-Fi network on the same carrier, with respect to throughput and atency. On the other hand, LTE-U is proposed for regions where LBT is not required and is promoted by the LTE-U forum []. As currenty specified, both LTE-LAA and LTE-U utiize carrier aggregation between a icensed and an (additiona unicensed carrier for enhanced data throughput on the downink (DL, and a upink traffic is transmitted on the icensed carrier. In this work, we ony focus on Wi-Fi/LTE-LAA coexistence and defer Wi-Fi/LTE-U coexistence for future work. Despite significant efforts ed by industry, there does not exist as yet a credibe anaytica mode for investigating the coexistence mechanism proposed by GPP. Further, as discussed in the next section, many of the industry resuts based on simuations or experiments remain independenty unverifiabe (as can be expected - using proprietary toos or interna aboratory resources and ack the commony accepted anaytica basis to create the necessary transparency. Consequenty, resuts on this topic appear to be divided into two camps - one (LTE caiming that fair coexistence is feasibe under the rues as proposed whereas the other (Wi-Fi suggesting significant negative impact and unfairness due to the presence of LTE. This has created an impasse with dueing positions based on incompatibe resuts and no pathway as yet emerging for crafting a methodoogy that buids confidence on both sides. In this context, we beieve that our approach incorporating a mix of fundamenta modeing backed by carefu, transparent experiments wi make a significant contribution towards practica depoyment of LTE-LAA LBT oad based equipment (LBE coexisting with Wi-Fi [], [] for the indoor scenario in Fig.. The specific nove contributions of this work incude:
Users For A Users For B Operator A Operator B Fig. : Wi-Fi AP (Operator A and LTE-LAA enb (Operator B coexistence network scenario. A new anaytica mode for throughput of LTE-LAA and Wi-Fi for simpe coexistence scenarios consistent with GPP, assuming saturation; Modeing the impact of energy detection (ED threshod and exporing its impact on the throughput of Wi-Fi and LTE-LAA; Vaidating anaytica resuts via experimenta resuts using the Nationa Instruments (NI Labview patform. This paper is organized as foows. Section II discusses the reated research in industry and academia. Section III contains a sef-contained description of the media access contro (MAC protocos of LTE-LAA and Wi-Fi. In Section IV, a new mode for coexistence of Wi-Fi and LTE-LAA is deveoped and used to estimate throughput. Section V investigates the impact of ED threshod on coexistence throughput. Section VI provides detaied numerica and experimenta resuts of Wi-Fi / LTE- LAA coexistence and expores the impact of ED threshod. Section VII concudes the paper. II. RELATED WORK Interest in LTE/Wi-Fi co-existence has been driven by the finaization of GPP Reease [], that inspired significant industry-driven exporation of this topic. One of the eary works to expore GHz LTE - Wi-Fi coexistence [] from a radio resource management perspective shows that Wi-Fi can be severey impacted by LTE transmissions impying that achieving some measure of fair coexistence of LTE and Wi- Fi needs to be carefuy managed, via some possibe mechanisms for joint depoyments. Nokia Research [9] proposed a bandwidth sharing mechanism whereby the impact on Wi- Fi network throughput can be controed by restricting LTE activity. In the situation where a arge number of Wi-Fi users try to access the network, users may spend a ong time in back-off (i.e. medium is ide; if LTE coud expoit these sient times the Wi-Fi performance woud not necessariy degrade but instead, the bandwidth utiization efficiency coud increase. In [], a performance evauation of LTE and Wi-Fi coexistence was conducted via simuation which shows that whie LTE system performance is sighty affected, Wi-Fi is significanty impacted by LTE transmissions. Wi-Fi channe access is most often bocked by LTE interference, causing the Wi-Fi nodes to stay in the isten mode more than 9% of the time. In [], the coexistence of Wi-Fi and LTE-LAA with LBT mechanism was investigated through an experimenta test set-up that particuary expored the impact of the reevant carrier sensing threshods specified by GPP. The proposed Listen Before Tak (LBT method in LTE-LAA empoys a somewhat different back-off technique from standard DCF in Wi-Fi, and the consequence of this asymmetry is worthy of carefu investigation. A second issue is the recommended LBT sensing threshod of - dbm, since LTE interference weaker than - dbm has been shown to be harmfu. Their investigation reveaed that LBT as recommended does not by itsef guarantee successfu coexistence with Wi-Fi, and fairness is impacted by mutipe factors - the contention parameters for channe access, as we as the sensing threshod and transmission duration. [] aso expored this topic through experiments and expored the reative issue of bandwidth asymmetry between LTE and Wi-Fi. Their resuts indicate that even sma bandwidth of LTE-LAA has a arge impact on Wi-Fi performance, however, the impact is dependent on where the LTE-LAA bandwidth is ocated reative to the Wi- Fi channe. The negative impact on Wi-Fi is reduced if the sma bandwidth of LTE resides on the guard band or center frequency of Wi-Fi. On the other hand, Quacomm [] investigated the coexistence of Wi-Fi with LTE-LAA and LTE-U through simuation and showed that significant throughput gain can be achieved by aggregating LTE across icensed and unicensed spectrum; further (and importanty, this throughput improvement does not come at the expense of degraded Wi-Fi performance and both technoogies can fairy share the unicensed spectrum. Ericsson in [] expored aspects of LTE-LAA system downink (DL operation such as dynamic frequency seection (DFS, physica channe design, and radio resource management (RRM. An enhanced LBT approach was proposed for improving coexistence of LTE-LAA and Wi-Fi and resuts from a system-eve simuation for GPP evauation scenarios showed that fair coexistence can be achieved in both indoor and outdoor scenarios. In summary, the industry driven research can be divided into two camps: one that predicts argey negative consequences for Wi-Fi with the proposed LTE-LAA coexistence mechanisms, and others who caim that fair coexistence is feasibe (with necessary tweaks or enhancements. This has ed to an impasse; progress on this important probem requires more carefu and transparent approach such as ours, that expore the very basis of the coexistence assumptions inherent in GPP prescriptions and provide some mode-based anaysis of the probem. We next summarize some of the pertinent academic ETSI Options A and B each specify a back-off procedure that is different from Wi-Fi.
iterature that aso ooked at this probem - either with intent to deveop anaytica modes to predict network throughput and/or idea-driven soutions for fair channe access. In [], the authors focused on various design aspects of LBT schemes for LTE-LAA as a means of providing equa opportunity channe access in the presence of Wi-Fi. In [], authors proposed an LBT enhancement agorithm with contention window size adaptation for LTE-LAA in order to achieve not ony fair channe access but aso Quaity of Service (QoS fairness. In [], authors designed the MAC protoco for LTE-LAA system for fair and friendy coexistence - the LTE transmission time is optimized for maximizing the overa normaized channe rate contributed by both LTE-LAA and Wi-Fi, whie protecting Wi-Fi. The prior art on anaytica modes for coexistence throughput of Wi-Fi and LTE-LAA [], [9] is the cosest in spirit to this contribution, as they seek to adapt the Bianchi mode for the coexistence scenario. In [9], the channe access and success probabiity are evauated for the coexistence of LTE- LAA and Wi-Fi; however, they have not considered the LTE- LAA LBT impementation defined in [], [] which foows the exponentia back-off. Simiary, in [], the coexistence of LTE-LAA and Wi-Fi is evauated by computing the network throughput for fixed contention window size (again not conforma with the GPP standard. In [], the throughput performance of LTE-LAA and Wi-Fi is cacuated for both fixed and exponentia back-off; however, the packet transmission for LTE-LAA is exacty simiar to Wi-Fi where the packet can be transmitted with 9µs resoution whie in reaity, LTE-LAA foows the.ms transmission boundaries of LTE in GPP. Aso, the resuts from their Markov mode for coexistence throughput coud not be vaidated with our approach, notaby the NI Labview experimenta resuts. In summary, a correct evauation of coexistence of Wi-Fi and LTE-LAA LBT as proposed by GPP standard [], [] is sti outstanding. Finay, whie much of the coexistence investigations are driven by the imperative for fair sharing as defined by GPP, we do not expore that issue in any depth in this work. This is mosty because this very important aspect deserves a much deeper investigation that can be accompished here and is eft for separate future work. Nonetheess, we quote a few reevant important prior art that has contributed to our understanding of the overa probem. In some important work [], [] derive the proportiona fair rate aocation for Wi-Fi/LTE- LAA (as we as Wi-Fi/LTE-U coexistence. Aso in [], the fairness in the coexistence of Wi-Fi/LTE-LAA LBT based on the GPP criteria is investigated through a custom-buit eventbased system simuator. Their resuts suggest that LBT (and correct choice of LBT parameters is essentia to proportiona fairness. III. COEXISTENCE OF LTE-LAA AND WI-FI: MAC PROTOCOL MECHANISMS In this section, the MAC protoco of Wi-Fi and LTE-LAA is presented. Whie LTE-LAA uses LBT to mimic Wi-Fi DCF, Proportiona fairness is we-known to achieve airtime fairness in rateheterogeneous Wi-Fi networks. Parameter TABLE I: Gossary of this paper. Definition W Wi-Fi minimum contention window m Wi-Fi maximum retransmission stage W LTE-LAA minimum contention window m LTE-LAA maximum retransmission stage PhyH preambe with Physica header MACH MAC header ACK Acknowedgment ength σ Wi-Fi sot time DIFS distributed interframe space SIFS short interframe space N B Wi-Fi packet data portion size Psize Wi-Fi data portion duration CCA Cear Channe Assessment ecca extended CCA T s LTE-LAA time sot for back-off T d LTE-LAA differ time e retry imit after reaching to m n w Number of Wi-Fi APs in coexistence n Number of LTE-LAA enbs in coexistence N Number of Wi-Fi APs in Wi-Fi ony system Wi-Fi Coision Probabiity τ w a Wi-Fi station Transmission Probabiity LTE-LAA Coision Probabiity τ an LTE-LAA station Transmission Probabiity T f LTE-LAA differ time T D transmission opportunity (TXOP D LT E Deay for next transmission r MCS data rate r w Wi-Fi data rate r LTE-LAA data rate δ propagation deay P trw Wi-Fi transmission probabiity P sw Wi-Fi successfu transmission probabiity P tr LTE-LAA transmission probabiity P s LTE-LAA successfu transmission probabiity T sw Wi-Fi successfu transmission duration T cw Wi-Fi coision transmission duration T s LTE-LAA successfu transmission duration T c LTE-LAA coision transmission duration T cc couped coision duration of LTE-LAA and Wi-Fi T E Tota average time T put w Wi-Fi throughput T put LTE-LAA throughput r(n samped received signa ɛ average energy M number of sampes in energy detector η energy detector threshod P d detection probabiity of energy detector detection probabiity of Wi-Fi P d detection probabiity of LTE-LAA δ propagation deay there are some key differences which are highighted, as these have a significant bearing on the respective channe access. A. Wi-Fi DCF The Wi-Fi MAC distributed coordination function (DCF empoys CSMA/CA [] as iustrated in Fig. and expained in the foowing. Each node attempting transmission must first ensure that the medium has been ide for a duration of DCF Interframe Spacing (DIFS using the ED and CS mechanism. When either of ED and CS is true, the Cear the abiity of Wi-Fi to detect the externa interference the abiity of Wi-Fi to detect and decode an incoming Wi-Fi signa preambe
Backoff Resume Backoff Busy Packet DIFS ACK SIFS Time sot DIFS L-STF L-LTF L-SIG (Sym (Sym (Sym Frame Contro Duratio n ID Address,, Seq. Contro Address Data Frame Body FCS Preambe ( OFDM Symbo MAC header ( Bytes N B Bytes Fig. : Wi-Fi CSMA/CA contention and frame transmission. The Wi-Fi frame structure with Preambe, MAC header, and data portion. Channe Assessment (CCA is indicated as busy. If the channe is ide and the station is not transmitting immediatey after a successfu transmission, the station transmits. Otherwise, if the channe is sensed busy (either immediatey or during the DIFS or the station is transmitting after a successfu transmission, the station persists with monitoring the channe unti it is measured ide for a DIFS, then seects a random back-off duration (counted in units of sot time and counts down. Specificay, a station seects a back-off counter uniformy at random in the range of [, i W ] where the vaue of i (the back-off stage is initiaized to and W is the minimum contention window chosen initiay. Each faied transmission due to packet coision resuts in incrementing the back-off stage by (binary exponentia back-off or BEB and the node counts down from the seected back-off vaue; i.e. the node decrements the counter every σµs corresponding to a backoff sot as ong as no other transmissions are detected. If during the countdown a transmission is detected, the counting is paused, and nodes continue to monitor the busy channe unti it goes ide; thereafter the medium must remain ide for a further DIFS period before the back-off countdown is resumed. Once the counter hits zero, the node transmits a packet. Any node that did not compete its countdown to zero in the current round, carries over the back-off vaue and resumes countdown in the next round. Once a transmission has been competed successfuy, the vaue of i is reset to. The maximum vaue of back-off stage i is m and it stays in m stage for one more unsuccessfu transmission. If the ast transmission was unsuccessfu, the node drops the packet and resets the backoff stage to i =. If a unicast transmission is successfu, the intended receiver wi transmit an Acknowedgment frame (ACK after a Short Interframe Spacing (SIFS duration post successfu reception; the ACK frame structure is shown in Fig. which consists of preambe and MAC header. A coision event occurs if and ony if two nodes seect the same back-off counter vaue at the end of a DIFS period. Bytes Bytes MAC header L-STF L-LTF L-SIG Frame Duration ID Address (Sym (Sym (Sym Contro FCS Preambe ( OFDM Symbo B. LTE-LAA LBT Bytes Bytes Bytes Fig. : Wi-Fi ACK frame structure LTE-LAA foows LBT approach for coexistence with Wi- Fi [] which is simiar in intent to CSMA/CA with the foowing key differences as iustrated in Fig. : (a LTE-LAA performs a CCA check using energy detect (CCA-ED where it observes the channe for the defer period (T d. The T d depends on the access priority cass number in Tabe II. There is no CS in LTE-LAA ike Wi-Fi for performing preambe detection. If sensed ide and the current transmission is not immediatey after a successfu transmission, the LTE-LAA node starts transmission; if sensed busy, it reverts to extended CCA (ecca whereby it senses and defers unti the channe is ide for T d, and then performs the exponentia back-off simiar to DCF (seects a back-off counter and decrements the back-off counter every sot time T s = 9 µs. (b As iustrated in Tabe II, LTE-LAA identifies channe access priority casses with different minimum and maximum contention window size. (c Whenever a coision happens, the back-off number is seected randomy from doubed contention window size for retransmission (i.e., [, i W ], where i is the retransmission stage for seecting the contention window size. When i exceeds the maximum retransmission stage m, it stays at maximum window size for e times (e is the retry imit after reaching to m where the e is seected from the set of vaues {,,..., }; then, i resets to. (d When an LTE-LAA enb gets access to the channe, it is
TABLE II: LTE-LAA LBT parameters per cass. Access Priority Cass # T d W m TXOP µs ms µs ms µs ms or ms 9 µs ms or ms aowed to transmit packets for a TXOP duration of up to ms when known a-priori that there is no coexistence node, otherwise up to ms. (e The minimum resoution of data transmission in LTE- LAA is one subframe (i.e., ms and LTE-LAA transmits the subframe per. ms sot boundaries; (f After the maximum transmission time, if data is avaiabe at the LTE-LAA buffer, it shoud perform the ecca for accessing the channe. In LTE, a Resource Bock (RB is the smaest unit of radio resource which can be aocated to a UE, equa to khz bandwidth over transmission time interva (TTI equa to one subframe. Each RB of khz bandwidth contains sub-carriers, each with OFDM symbos, equaing resource eements (REs. Depending upon the moduation and coding schemes (QPSK, -QAM, -QAM, each symbo or resource eement in the RB carries, or bits per symbo, respectivey. In an LTE-LAA system with MHz bandwidth, there wi be RBs avaiabe. Each subframe consists of OFDM symbos as indicated in Fig., of which to are physica downink contro channe (PDCCH symbos and the rest are physica downink shared channe (PDSCH data. As aready mentioned, LTE- LAA enb transmits the subframe in LTE sots resoution, i.e. enb transmits at the start of. ms LTE sots for (at east one subframe ( LTE sots. If the enb acquires the channe before the start of (next LTE sot, it may need to transmit a reservation signa to reserve the channe. After the transmission period, the receiver (or receivers transmits the ACK through the icensed band if the symbos are successfuy decoded. IV. COEXISTENCE ANALYSIS USING A MODIFIED BIANCHI MODEL In this section, we first modify the -D Markov mode in [] for anayzing the performance of the Wi-Fi DCF protoco. Then we propose a new Markov mode for LTE-LAA LBT system. Finay, the anaytica modeing for the interaction of Wi-Fi and LTE-LAA coexistence to cacuate the throughput of each system when a nodes are saturated is investigated. We assume co-channe coexistence of Wi-Fi and LTE-LAA where both systems transmit on a MHz channe. A. Anayzing Wi-Fi DCF using a Markov Mode The -dimensiona Markov chain mode [] for Wi-Fi DCF is shown in Fig. for the saturated nodes. Let {s(t = j, b(t = k} denote the possibe states in the Markov chain, where s(t is the retransmission stage and b(t the back-off counter. The Markov chain and correspondingy one step transition probabiity in [] is modified based on the expanation in section III-A. In [], when the back-off stage reaches to maximum vaue (i.e. m, it stays in m forever. However, in Wi-Fi when the maximum vaue is reached, the back-off stage stays at m for one more attempt; then it resets to zero in case of an unsuccessfu transmission. Considering the stationary distribution for the Markov mode as b j,k = im t P {s(t = j, b(t = k}, j (, m +, k (, W i, the modified one step transition probabiity of the Markov chain is, P {j, k j, k + } =, k (, W i j (, m + P {, k j, } = Pw W, k (, W j (, m + P {j, k j, } = Pw W i, k (, W i j (, m + P {, k m +, } = Pw W, k (, W m ( where is the coision probabiity of Wi-Fi nodes, W is the minimum contention window size of Wi-Fi, W i = i W is the contention window size at the retransmission stage i, and i = m is the maximum retransmission stage (i.e., i = j for j m and i = m for j > m. The first equation in ( represents the transition probabiity of back-off decrement; the second equation represents the transition probabiity after successfu transmission and seecting a random back-off at stage for contending for the next transmission; the third equation represents the transition probabiity after unsuccessfu transmission in which the contention window size (W i is doubed; the ast equation represents the transition probabiity after unsuccessfu transmission in (m + th stage in which the next random back-off vaues shoud be seected from the minimum contention window size. To simpify the cacuation, we introduce the foowing variabes for use in (: b j, = b j,, < j m + b j, = P j wb,, j m + b, = b m+, + ( j=m+ j= b j,, ( the ast equation impies that, j=m+ j= b j, = ( P m+ w b,. ( In each retransmission stage, the back-off transition probabiity is b j,k = W i k W i b j,, j m +, k W i. ( We can derive b, by the normaization condition, i.e., m+ j= b, = W i k= b j,k =, W ( ( ( m+ ( + m (Pw m+ Pw m+ ( +. m+ Pw (
CCA Td PDCCH PDSCH T D LTE Subframe 9 D LTE ACK Td Busy Td ecca Ts Contro OFDM Symbos Data OFDM Symbos Fig. : LTE-LAA LBT contention with CCA/eCCA and LTE subframe structure Pw Pw W W,, m, m+, Pw W, W W, m, Pw W,W -,W - m,w m - ( Hence, (the probabiity ( ( that a node transmits in a time sot W ' W ' W ' is cacuated using ( and ( as,,,,w' - W ' m'+, W ' m+, W ' m'+, m+,w m - Fig. : Markov chain mode for the Wi-Fi DCF with binary exponentia back-off m+ τ w = b j,,,,w' - j= = ( ( ( W m+ ( + m (Pw m+ Pw m+ ( ( ( Pw m+ m', m',. + m',w' m'- m'+,w' m'- P {, k j, } = P W Retry imit B. Anayzing LTE-LAA using a Markov Mode (e j=m +e In LTE-LAA, a nodes in an access priority cass use the m'+e m'+e LBT mechanism, for channe, contention. The two-dimensiona m'+e,w' m'- j= Markov chain mode of LTE-LAA LBT is iustrated in Fig. for the saturated nodes in which when the retransmission stage reaches m it stays in maximum contention window size for e times, then resets to the zero stage. Simiary, denoting {s(t = j, b(t = k} as the possibe states in the Markov chain, where ( s(t is the retransmission stage and b(t the back-off counter, the one step transition probabiity is as, P {j, k j, k + } =, k (, W i j (, m + e W P {j, k j, } = W i P {, k m + e, } = W, k (, W j (, m + e, k (, W i j (, m + e, k (, W m where is the coision probabiity of LTE-LAA nodes, W is the minimum contention window size of LTE-LAA, W i = i W is the contention window size at retransmission stage i, and i = m is the maximum retransmission stage (i.e., i = j for j m and i = m for j > m. The first equation in ( represents the transition probabiity of back-off decrement; the second equation represents the transition probabiity after successfu transmission and seecting a random backoff at stage for the contending for the next transmission; the third equation represents the transition probabiity after unsuccessfu transmission in which the contention window size (W i is doubed; the ast equation represents the transition probabiity after unsuccessfu transmission in (m +e th stage in which the next random back-off vaues reset to the minimum contention window size (W. To simpify the cacuation, we introduce some formuas derived from Fig. and ( which reates a of the states (b j,k to the b, : b j, = b j,, < j m + e b j, = P j b,, j m + e b, = b m +e, + (, ( j=m +e j= b j, the ast equation is rewritten as, b j, = ( ( P m +e + b,. (9 For each retransmission stage, the back-off transition probabiity is b j,k = W i k W i b j,, j m +e, k W i (
m+, m+, m+,w m- ( ( ( W ' W ', W ', m', m'+, m'+e, W ' W ' ( W ',, m', m'+, m'+e,,w'-,w'- m',w'm'- m'+,w'm'- m'+e,w'm'- Retry imit (e Fig. : Markon chain mode for the LTE-LAA LBT with binary exponentia back-off. The b, can be derived by imposing the normaization condition as m +e j= W i k= b j,k = m j= W j k= b j,k + m +e j=m + W m k= b j,k =, ( where b, is derived as in (. Hence, the probabiity that a node transmits in a time sot is cacuated as, τ = W ( m +e j= b j, = ( ( ( m + + P ( ( P m +e + m m + P m +e + P m +e +. + ( where simiary is couped to both Wi-Fi and LTE-LAA via τ w and τ. In order to compute the,, τ w, and τ for the coexistence of Wi-Fi and LTE-LAA, we jointy sove (, (, (, and (. The transmission probabiity of Wi-Fi is the probabiity that at east one of the n w stations transmit a packet during a time sot: P trw = ( τ w nw, ( and simiary the transmission probabiity of LTE-LAA is: P tr = ( τ n. ( The successfu transmission of a Wi-Fi node is the event that exacty one of the n w stations makes a transmission attempt given that at east one of the Wi-Fi APs transmit: P sw = n wτ w ( τ w nw P trw, ( Simiary the successfu transmission probabiity of LTE-LAA is cacuated as: P s = n τ ( τ n P tr. (9 To compute average throughput, we need the average time durations for a successfu transmission and a coision event, respectivey, given by: T sw =MACH + PhyH + Psize + SIFS + δ+ ACK + DIFS + δ T cw =MACH + PhyH + Psize + DIFS + δ, ( where the vaues of the parameters are isted wherever it is required for the cacuation (the parameters presented in terms of # of bits are converted to time based on the channe data rate provided in the numerica resuts section. The average time duration of successfu transmission event and coision event for LTE-LAA are C. Throughput of Wi-Fi and LTE-LAA in Coexistence We assume there are n w Wi-Fi APs and n LTE-LAA enbs co-channe and co-ocated networks, each with fu buffer. To be consistent with GPP, we consider ony downink (DL transmission, impying that the contention is between ony the APs and enbs. The coision probabiity of a Wi-Fi Aith at east one of the other remaining (n w Wi-Fi or n LTE- LAA stations is given by = ( τ w nw ( τ n, ( where is now couped to both Wi-Fi and LTE-LAA nodes via τ w and τ. Simiary, the coision probabiity for an LTE- LAA enb with at east one of the other remaining stations is = ( τ n ( τ w nw, ( T s = T D + D LT E T c = T D + D LT E, ( where the T D is the TXOP of LTE-LAA - which coud be up to ms for access priority cass and [] as expained in section III-B. D LT E is the deay for the next transmission which is one LTE sot (. ms. After transmission for TXOP, the transmitter waits for the ACK and then resumes contention for the channe for the next transmission. If an LTE enb wins channe contention before start of next LTE sot, the LTE-LAA transmits a reservation signa to reserve the channe unti the end of the current LTE sot to start packet transmission. The throughput of Wi-Fi is cacuated as: T put w = P trwp sw ( P tr P size T E r w, ( where P trw P sw ( P tr is the probabiity that Wi-Fi transmits a packet successfuy in one Wi-Fi sot time, and r w is the Wi-
b, = W ( ( m + + m + P m P m +e + + P m +e. ( + Fi physica ayer data rate. T E is the tota average time of a possibe events which is cacuated as, T E = ( P trw ( P tr σ + P trw P sw ( P tr T sw + P tr P s ( P trw T s + P trw ( P sw ( P tr T cw + P tr ( P s ( P trw T c + (P trw P sw P tr P s + P trw P sw P tr ( P s + P trw ( P sw P tr P s + P trw ( P sw P tr ( P s T cc, ( where T cc is the couped coision time interva which is the average time of the coision between Wi-Fi APs and LTE- LAA enbs, determined by the arger vaue between T cw and T c. Simiary the throughput of the LTE-LAA is cacuated as, T put = P trp s ( P trw T D T E r, ( where T D is the fraction of the TXOP in which the data is transmitted, i.e. PDCCH symbo in a subframe with OFDM symbos is considered, and r is the LTE-LAA data rate. V. IMPACT OF ENERGY DETECT (ED THRESHOLD ON WI-FI AND LTE-LAA COEXISTENCE The goa in this section is to investigate the effect of changing the ED threshod on the throughput performance of Wi-Fi and LTE-LAA in a coexistence network. In contending for accessing the channe, Wi-Fi performs preambe based CS for detecting other Wi-Fi stations and ED detecting for the externa interference. LTE-LAA uses the CCA-ED detection for detecting the same and other types of network as we as the externa interference through ED (CCA-ED. The ED threshod in generic Wi-Fi system is dbm and in LTE-LAA is dbm. The accuracy of preambe detection for ow threshod is very high, but cross-network ED based detection is imperfect at ow threshod vaues. Varying ED threshods of Wi-Fi and LTE-LAA networks thus eads to hidden node probems (whereby additiona packet drops occur due to the resuting cross-network interference that impact the respective networks differenty and ead to divergent network throughputs. A. Probabiity of Detection in Energy Detector Considering a samping rate of ns in a MHz Wi-Fi channe, the received signa in the presence of interference (H and no interference (H is: the preambe based CS is perfect because of its ow carrier sensing threshod of dbm H : (W/O Interference r(n = w(n H : (W Interference r(n = x s (n h(n + w(n, ( where r(n is the received signa, w(n is the AWGN noise, x s (n is the moduated interference signa, and h(n is the channe impuse response (assume normaized channe, i.e. n h(n =. The test statistic for the energy detector is: ɛ = M r(i, ( M i= where M is the ength of received sampe sequence for test statistics. For the DIFS duration where the Wi-Fi stations sense the channe for µs, the vaues of M is. The probabiity of detection is cacuated as []: ( η (σ P d = P (ɛ > η = Q n + σx M (σ x + σn, ( where η is the ED threshod, σ x is the signa power, and σ n is the power of noise. Given ED threshod and number of sampes, the detection probabiity can be cacuated. B. Modifying Anaytica Mode to Capture the ED Threshod In order to modify the throughput mode by incorporating ED threshod for cross network detection, we introduce as the cross network ED detection probabiity of Wi-Fi AP and P d as the cross network ED detection probabiity of LTE-LAA enb. To incorporate the ED detection probabiity of Wi-Fi, we first begin by rewriting the Wi-Fi coision probabiity ( as = ( ( τ n ( τ w nw + ( τ w nw, ( where ( ( τ n is the probabiity that at east one of the LTE-LAA enbs transmit, i.e is the probabiity that the LTE- LAA is active (or ON. To capture the imperfect detection of Wi-Fi nodes (, which resuts in additiona physica ayer coision and consequent packet oss, we mutipy the ( ( τ n term in eq ( by the as, = [( ( τ n ] ( τ w nw + ( τ w nw. (9 Simiary the coision probabiity of LTE-LAA based on eq ( can be recacuated considering the LTE-LAA detection probabiity (P d as: = [( ( τ w nw P d ] ( τ n + ( τ n. ( Using these modified equations, the anaytica throughput in the previous section can be re-cacuated as a function of the detection probabiity.
9 Ce A NI Labview Feet Ce B Tx Loopback Fig. : The Wi-Fi ony and LTE-LAA/Wi-Fi coexistence experimenta setup. Rx TABLE III: NI abview experimenta setup parameters. Parameters NI Experiment Transmission Power dbm Operating Channe Operating Frequency. GHz RF Transmission Loopback LTE-LAA Transmission Channe PDSCH, PDCCH Data Traffic Fu buffer TABLE IV: NI experiment compared with the theoretica derivation for three cases for Wi-Fi ony and Wi-Fi/LTE-LAA coexistence. VI. EXPERIMENTAL AND NUMERICAL RESULTS In this section, we first present experimenta resuts (using the Nationa Instruments (NI Labview patform for vaidation of our anaytica mode and compare with the numerica resuts obtained using the anaytica derivations. Further, we extend the numerica resuts to capture the effect of different parameters in different LTE-LAA access priority casses and expore the effect of ED threshod on throughput. A. Coexistence Experiment and Comparison with Anaytica Derivation The NI. Labview Appication Framework provides functiona eements of the Physica (PHY as we as the Medium Access Contro (MAC ayer. This Labview code incudes a receiver (RX and transmitter (TX impementation and functiona eements for channe state handing, sot timing management, and back-off procedure handing. The MAC is impemented on the Fied Programmabe Gate Array (FPGA and tighty integrated with the PHY to achieve requirements for interframe spacing (such as SIFS and DIFS, as we as sot timing management to aow frame exchange sequences, such as DATA ACK and basic DCF for CSMA/CA protoco. Our experiment uses two NI USRP 9R software defined radios (SDRs as iustrated in Fig.. Each of these can be configured to be either a Wi-Fi AP or an LTE enb. The hardware requirements consist of two computers which have at east GB RAM (Instaed Memory, -bit operating system, x-based processor, Inte(R Core i, CPU cock.ghz. The setup aso contains antennas covering the GHz ISM band. For experiments on the coexistence of Wi-Fi and LTE- LAA, one SDR is configured to be an LTE-LAA enb and the other to be a Wi-Fi AP (i.e., Ce A is LTE-LAA and Ce B is Wi-Fi in Fig., both transmitting on the same channe which is Channe. The Wi-Fi transmission foows the.a standard with MHz bandwidth. The LTE-LAA uses RBs to cover the tota MHz bandwidth and OFDM symbo in a subframe is assigned as the PDCCH symbo. The SDR transmission characteristics aong with other experimenta parameters under study are summarized in Tabe III. Three different experiments are performed to compare the throughput performance of the experiment with the theoretica System NI Experiment Theoretica Modeing Wi-Fi rate r w = 9 Mbps and LTE-LAA rate r =. Mbps Case (Wi-Fi Aggregate.. Case (Wi-Fi & LAA. and.9. and. Case (Wi-Fi & LAA. and..9 and. Wi-Fi rate r w = Mbps and LTE-LAA rate r =. Mbps Case (Wi-Fi Aggregate.. Case (Wi-Fi & LAA. and.. and. Case (Wi-Fi & LAA.9 and.9. and. Wi-Fi rate r w = Mbps and LTE-LAA rate r =. Mbps Case (Wi-Fi Aggregate.. Case (Wi-Fi & LAA. and.. and. Case (Wi-Fi & LAA. and.. and. throughput resuts cacuated from the anaytica modeing as foows: Case (Wi-Fi ony: This is the baseine for comparison. Both SDRs are configured to be Wi-Fi APs on the same channe, with cient per AP. The experiment is performed under the video data transmission priority access cass with W =, m =, and packet ength of N B = bytes. Case (Coexistence, LTE-LAA Cass : One of the SDRs is configured to be a Wi-Fi AP and the other as an LTE-LAA enb, cient per station. The LTE-LAA enb uses access priority cass [] with W =, m =, and TXOP = ms and the Wi-Fi AP uses data transmission with W =, m =, and packet ength of N B = bytes. Case (Coexistence, LTE-LAA Cass : One of the SDRs is configured to be a Wi-Fi AP and the other as a LTE-LAA enb, cient per station. The LTE-LAA enb uses access priority cass [] with W =, m =, and TXOP = ms and the Wi-Fi AP uses video transmission with channe access parameters of W =, m =, and packet ength of N B = bytes. We perform the experiment with three different data rates for Wi-Fi: 9 Mbps (BPSK, coding rate =., Mbps (QPSK, code rate =., and Mbps (QAM, code rate =.. LTE-LAA uses a RBs which coud give three data rates of. Mbps (QPSK, code rate =.,. Mbps (QPSK, code rate =., and. Mbps (QAM, code rate =.. In the NI Labview impementation, the retransmission stage (i resets to zero after i exceeds m, so we consider e = in the anaytica mode to compare with the NI Labview
TABLE V: Wi-Fi and LTE-LAA parameters. Parameter vaue PhyH µs MACH ( bytes/r w µs r Mbps ACK ( bytes/r µs δ. µs σ 9 µs e DIFS µs SIFS µs N B bytes Psize N B /r w µs D LT E. ms Throughput (Mbps n =n w 9 Wi-Fi Ony - Per User Wi-Fi Ony - Tota Coex - Wi-Fi per User Coex - LTE-LAA per User Coex - Tota experimenta resuts. The resuts are shown in Tabe IV (the vaues are in Mbps, where the theoretica throughput of the Wi-Fi ony network is from eq. ( in [] using the τ w ( cacuated in this paper and the theoretica throughput of the coexistence network is from ( and (. As can be seen in Tabe IV, the trend of aggregate experimenta throughputs in the Wi-Fi ony network (Case are simiar to the theoretica resuts for different data rates. In Case and, the measured throughput for coexistence of Wi-Fi and LTE-LAA show a simiar trend to theoretica resuts. The tota theoretica throughput in Case for the coexistence scenario is. Mbps (for r w = 9Mbps and r =.Mbps which is smaer than the aggregate theoretica throughput of. Mbps in Wi-Fi ony. In Case, the tota theoretica coexistence throughput is. Mbps (for r w = 9Mbps and r =.Mbps which is again smaer than the theoretica Wi-Fi ony throughput but by a ower margin as compared with Case. The reasons behind the smaer throughput of the coexistence network compared to the Wi-Fi ony network is discussed next. B. Numerica Resuts from Anaytica Derivation We evauate the coexistence of Wi-Fi and LTE-LAA through numerica resuts for the derivations in Sec. IV, using the parameters isted in Tabe V. For LTE-LAA LBT the cass and parameters from Tabe II are used. The per user throughput of each network is cacuated by dividing the tota throughput of the network over the number of nodes of that network (e.g. Wi-Fi per user throughput in coexistence network is T put w /n w. In Fig. we show the resuts for the Wi-Fi ony system with N APs and the coexistence system with n w = n = N/, W = W =, m = m =, LTE-LAA TXOP = ms, r w = 9 Mbps, and r w =. Mbps. The other parameters are isted in Tabe V. The number of Wi-Fi APs N in the Wi-Fi ony network is equay divided among the number of Wi-Fi APs and LTE-LAA enbs in the coexistence network, i.e. n w = n = N/. We observe that the tota throughput of Wi-Fi and LTE-LAA coexistence is ower than the tota throughput of the Wi-Fi ony network. In part, this is due to LTE-LAA channe access in a sotted fashion, impying that after accessing the channe it transmits a reservation signa to keep the channe unti the start of the next sot - this wastes some channe opportunity for data transmission and N Fig. : Throughput of the coexistence system (considering priority cass for LTE-LAA compared with the Wi-Fi Ony, W = W =, m = m =. Bottom x-axis (bue is the number of Wi-Fi APs in Wi-Fi ony network and top x-axis (red is the number of Wi-Fi APs and LTE-LAA enbs in coexistence network. hence the coexistence throughput decreases. Further, since the transmission duration of Wi-Fi is smaer than the LTE-LAA TXOP, the per user throughput of Wi-Fi in coexistence is ower than the LTE-LAA. In Fig. 9, the Wi-Fi ony system with N APs is compared against the coexistence system with n w = n = N/, W = W =, m = m =, LTE-LAA TXOP = ms, r w = 9 Mbps, and r w =. Mbps; other parameters are isted in Tabe V. The tota throughput of Wi-Fi and LTE-LAA is ower than the tota throughput of Wi-Fi ony network. In addition to the wasted channe opportunity for transmission due to the reservation signa, in this figure the transmission time of LTE-LAA is more than Wi-Fi which causes the LTE- LAA to transmit for a onger duration per access and have a higher per-user throughput than the Wi-Fi ony per user throughput. The arger difference between coexistence and Wi- Fi ony network throughput curves compared with Fig. arises because the LTE-LAA gets a higher opportunity for access and has a ower data rate compared with Wi-Fi. In Fig., the throughput performance of Wi-Fi APs with W =, m =, r w = 9 Mbps, and LTE-LAA enb with W =, m =, TXOP = ms, and r w =. Mbps is shown. The other parameters are isted in Tabe V. The Wi- Fi per user throughput in coexistence network for N is higher than the Wi-Fi ony per user throughput, which indicates a fair coexistence based on the GPP definition []. The coexistence network compared with the Wi-Fi ony network achieves a smaer tota throughput for fewer number of stations (N < but b higher throughput at number of stations N. This is due to the smaer maximum retransmission stage of Wi-Fi (i.e., m in which the Wi-Fi stations access the channe more frequenty, as we as the
n =n w 9 Wi-Fi Ony - Per User Wi-Fi Ony - Tota. Throughput (Mbps Coex - Wi-Fi per User Coex - LTE-LAA per User Coex - Tota N Fig. 9: Throughput of the coexistence system (considering priority cass for LTE-LAA compared with the Wi-Fi Ony considering W = W =, m = m =. Throughput (Mbps n =n w 9 Wi-Fi Ony - Per User Wi-Fi Ony - Tota Coex - Wi-Fi per User Coex - LTE-LAA per User Coex - Tota N Fig. : Throughput of the coexistence system with W =, m = and LTE-LAA with W =, m =. fewer number of Wi-Fi stations in coexistence network, i.e. n w = N, which causes fewer coision. This impies that in a specific network setup, the tota throughput of coexistence network coud be higher or ower than Wi-Fi ony network depending on the number of stations. The setup in Fig. exacty foows that in Fig., but the tota number of stations is fixed at, i.e., N = n w +n =, whie the number of LTE-LAA enbs and Wi-Fi APs in the coexistence network are varying. The goa is to investigate the effect of the asymmetric number of Wi-Fi and LTE-LAA stations in coexistence network. The maximum throughput of the coexistence system occurs at n w =, n = 9 which indicates that a arger number of LTE-LAA enbs deivers a Throughput (Mbps.... Wi-Fi Ony - Per User Wi-Fi Ony - Tota Coex - Wi-Fi Per User Coex - LTE-LAA Per User Coex - Tota (,9 (, (, (, (9, (,9 (, (, (, (9, (n w,n Fig. : Throughput of the coexistence system with W =, m = and LTE-LAA with W =, m =. higher portion of tota throughput in this scenario. Moreover, the throughput of coexistence network for any combination of Wi-Fi and LTE-LAA is higher than the Wi-Fi ony system. By increasing the number of Wi-Fi and decreasing the number of LTE-LAA nodes, the throughput of both systems are seen to decrease. For the tota range of the number of stations, the Wi- Fi per user throughput in coexistence network is higher than the Wi-Fi per user throughput in Wi-Fi ony system, which iustrates that the GPP fairness is achieved regardess of the number of stations for this (per user throughput metric. The throughput performance of the coexistence system with the parameters W = W =, m = m =, LTE-LAA TXOP = ms, and higher data rates for Wi-Fi and LTE- LAA, r w = Mbps and r w =. Mbps, is iustrated in Fig.. The other parameters are isted in Tabe V. The Wi- Fi per user throughput in coexistence is ower than the LTE- LAA per user throughput because the Wi-Fi frame airtime is smaer at higher Wi-Fi data rates whie the LTE-LAA airtime (TXOP is fixed. Smaer airtime of Wi-Fi, considering the same channe access parameters as LTE-LAA, eads to higher utiization of the channe by LTE-LAA. This aso resuts in ower Wi-Fi per user throughput in coexistence compared with per user throughput of Wi-Fi ony network, because in Wi-Fi ony network the airtime and channe access parameters are the same for a users whie in coexistence scenario, the LTE- LAA has a higher airtime. In Fig., the effect of changing the e (retry imit after reaching to m the in LTE-LAA is investigated for different scenarios; the curves for cass foow the parameters of Fig. and cass foow the Fig. 9. For cass where W = W = and m = m =, the tota throughput as we as the per user throughput of Wi-Fi and LTE-LAA increases noticeaby with retry imit, but much ess for cass with W = W = and m = m =. C. Effect of ED threshod on throughput In Fig., numerica resuts foowing the derivations in Sec. V iustrate the effect of varying ED threshod on the
Throughput (Mbps n =n w 9 Wi-Fi Ony - Per User Wi-Fi Ony - Tota Coex - Wi-Fi per User Coex - LTE-LAA per User Coex - Tota N Fig. : Throughput of the coexistence system with W =, m =, W =, m = and the higher data rates for Wi-Fi and LTE-LAA r w = Mbps and r =. Mbps. Throughput (Mbps Wi-Fi per User, cass=, e = LTE-LAA per User, cass=, e = Tota Coex, cass=, e = Wi-Fi per User, cass=, e = LTE-LAA per User, cass=, e = Tota Coex, cass=, e = Wi-Fi per User, cass=, e = LTE-LAA per User, cass=, e = Tota Coex, cass=, e = Wi-Fi per User, cass=, e = LTE-LAA per User, cass=, e = Tota Coex, cass=, e = 9 n =n w Fig. : Throughput versus different vaues of e in LTE-LAA. throughput of Wi-Fi in coexistence network with W = W =, m = m =, LTE-LAA TXOP = ms, and Wi-Fi packet ength bytes. Tabe VI shows the ED probabiity cacuated using ( for different ED threshods during the DIFS period with sampes, and other parameters as isted in Tabe V. By increasing the detection probabiity of Wi-Fi ( whie keeping the detection probabiity of LTE-LAA (P d equa to, the Wi-Fi throughput decreases; simiary by increasing the detection probabiity of LTE-LAA (keeping the detection probabiity of Wi-Fi equa to the Wi-Fi throughput increases. By decreasing the ED threshod in Wi-Fi, the Wi-Fi nodes detect LTE-LAA stations at ower transmit power and thus defer their transmission (i.e., enter back-off eading to ower Wi-Fi throughput. On the other hand, by decreasing the LTE-LAA ED threshod, the LTE-LAA detects more Wi-Fi stations with ower signa power and defers the transmission, TABLE VI: Energy Detector parameters and P d. Parameter Vaue signa to noise ratio (SNR db P n -9 dbm ED threshod - dbm - dbm - dbm ED detection probabiity (, P d... Throughput (Mbps....... =.,P d =. =.,P d =. =.,P d =. =.,P d =. =.,P d =. 9 n w =n Fig. : Throughput performance of Wi-Fi through changing the detection probabiity of Wi-Fi and LTE-LAA. so Wi-Fi stations have more opportunity for transmission which increases the throughput. Fig. shows the effect of changing the detection probabiity of Wi-Fi and LTE-LAA on the throughput of LTE-LAA in coexistence simiar to Fig.. By increasing the detection probabiity of Wi-Fi, the LTE-LAA throughput increases, and by increasing the detection probabiity of LTE-LAA, the LTE- LAA throughput decreases, athough impact of changing ED threshod on LTE-LAA throughput is proportionay smaer compared with Wi-Fi. A simiar concusion can be drawn from Fig.. These resuts are in ine with those presented by NI in []. VII. CONCLUSION In this work, we first presented a new mode for anayzing the throughput performance for Wi-Fi and LTE- LAA coexistence. We then modified the mode to incorporate ED sensing threshod to evauate the impact of threshod choices on throughput performance. We demonstrated that to achieve the maximum throughput in a coexistence scenario, the ED sensing threshod shoud be optimized. To vaidate the proposed mode, we aso set up a ab experiment with NI Labview and compared the experimenta throughput with the numerica resuts and showed very good correspondence between experiment and anaysis. The throughput performance of a Wi-Fi and LTE-LAA in coexistence system depends on the channe access parameters, TXOP of LTE-LAA, and data rates of Wi-Fi and LTE-LAA. By changing these parameters, the Wi-Fi or LTE-LAA achieves higher per user throughput in coexistence network compared with the per user throughput
Throughput (Mbps =.,P d =. =.,P d =. =.,P d =. =.,P d =. =.,P d =. 9 n w =n Fig. : Throughput performance of LTE-LAA through changing the detection probabiity of Wi-Fi and LTE-LAA. in Wi-Fi ony network. Finay, we note that these resuts aso form the bedrock of a thorough and objective ook at coexistence fairness in future, where we expect to show how the CSMA/CA and/or LBT parameters must be tuned to achieve access fairness. ACKNOWLEDGMENT This work was supported by the Nationa Science Foundation (NSF under grant. The authors woud ike to thank Thomas Henderson for the hepfu discussions and suggestions to make this work stronger. REFERENCES [] G. Bianchi, Performance Anaysis of the IEEE. Distributed Coordination Function, IEEE Journa of Seected Area Communications, vo., no., pp., Sep.. [] FCC, Revision of Part of the Commission s Rues to Permit Unicensed Nationa Information Infrastructure (U-NII Devices in the GHz Band, in Federa Communications Commission, Feb. [] ETSI, Broadband radio access networks (BRAN; GHz high performance RLAN; harmonized EN covering the essentia requirements of artice. of the R&TTE directive, EN 9, V.., June. [] GPP, rd Generation Partnership Project; Technica Specification Group Radio Access Network; Study on Licensed-Assisted Access to Unicensed Spectrum;(Reease, GPP TR.9, V.., June. [] ETSI, LTE; Evoved Universa Terrestria Radio Access (E-UTRA; Physica ayer procedures, ETSI TS, V.., Reease, May. [] http://www.teuforum.org. in LTE-U Forum. [] R. Kwan, R. Pazhyannur, J. Seymour, V. Chandrasekhar, S. R. Saunders, D. Bevan, H. Osman, J. Bradford, J. Robson, and K. Konstantinou, Fair co-existence of Licensed Assisted Access LTE (LAA-LTE and Wi-Fi in unicensed spectrum, in th Computer Science and Eectronic Engineering Conference (CEEC, Sept, pp.. [] F. M. Abinader, E. P. L. Ameida, F. S. Chaves, A. M. Cavacante, R. D. Vieira, R. C. D. Paiva, A. M. Sobrinho, S. Choudhury, E. Tuomaaa, K. Dopper, and V. A. Sousa, Enabing the coexistence of LTE and Wi-Fi in unicensed bands, IEEE Communications Magazine, vo., no., pp., Nov. [9] T. Nihtia, V. Tykhomyrov, O. Aanen, M. A. Uusitao, A. Sorri, M. Moisio, S. Iraji, R. Ratasuk, and N. Mangavedhe, System performance of LTE and IEEE. coexisting on a shared frequency band, in IEEE Wireess Communications and Networking Conference (WCNC, Apri, pp.. [] A. M. Cavacante, E. Ameida, R. D. Vieira, S. Choudhury, E. Tuomaaa, K. Dopper, F. Chaves, R. C. D. Paiva, and F. Abinader, Performance Evauation of LTE and Wi-Fi Coexistence in Unicensed Bands, in IEEE th Vehicuar Technoogy Conference (VTC Spring, June, pp.. [] N. Jinda and D. Bresin, LTE and Wi-Fi in Unicensed Spectrum: A Coexistence Study, in Googe white paper, Feb. [] Y. Jian, C. F. Shih, B. Krishnaswamy, and R. Sivakumar, Coexistence of Wi-Fi and LAA-LTE: Experimenta evauation, anaysis and insights, in IEEE Internationa Conference on Communication Workshop (ICCW, June, pp.. [] Q. T. Inc., Quacomm Research LTE in Unicensed Spectrum: Harmonious Coexistence with Wi-Fi, in Technica Report, June. [] A. Mukherjee, J. F. Cheng, S. Faahati, L. Faconetti, A. Furuskar, B. Godana, D. H. Kang, H. Koorapaty, D. Larsson, and Y. Yang, System architecture and coexistence evauation of icensed-assisted access LTE with IEEE., in IEEE Internationa Conference on Communication Workshop (ICCW, June, pp.. [] Kini, Ananth V and Canonne-Veasquez, Loïc and Hosseinian, Mohsen and Rudof, Marian and Stern-Berkowitz, Janet, Wi-fi-aa coexistence: Design and evauation of isten before tak for aa, in Annua Conference on Information Science and Systems (CISS,, pp.. [] T. Tao, F. Han, and Y. Liu, Enhanced LBT agorithm for LTE-LAA in unicensed band, in IEEE th Annua Internationa Symposium on Persona, Indoor, and Mobie Radio Communications (PIMRC,, pp. 9 9. [] S. Han, Y.-C. Liang, Q. Chen, and B.-H. Soong, Licensed-assisted access for LTE in unicensed spectrum: A MAC protoco design, IEEE Journa on Seected Areas in Communications, vo., no., pp.,. [] Y. Song, K. W. Sung, and Y. Han, Coexistence of Wi-Fi and Ceuar With Listen-Before-Tak in Unicensed Spectrum, IEEE Communications Letters, vo., no., pp., Jan. [9] C. Chen, R. Ratasuk, and A. Ghosh, Downink Performance Anaysis of LTE and WiFi Coexistence in Unicensed Bands with a Simpe Listen- Before-Tak Scheme, in IEEE st Vehicuar Technoogy Conference (VTC Spring, May, pp.. [] Y. Gao, X. Chu, and J. Zhang, Performance Anaysis of LAA and WiFi Coexistence in Unicensed Spectrum Based on Markov Chain, in IEEE Goba Communications Conference (GLOBECOM, Dec, pp.. [] C. Cano and D. J. Leith, Unicensed LTE/WiFi coexistence: Is LBT inherenty fairer than CSAT? in IEEE Internationa Conference on Communications (ICC, May, pp.. [] C. Cano, D. J. Leith, A. Garcia-Saavedra, and P. Serrano, Coexistence of WiFi and LTE in unicensed bands: A proportiona fair aocation scheme, in arxiv, May. [] Wireess LAN Medium Access Contro (MAC and Physica Layer (PHY specifications, IEEE Std.-,. [] D. Cabric, A. Tkachenko, and R. W. Brodersen, Experimenta study of spectrum sensing based on energy detection and network cooperation, in Proceedings of the First Internationa Workshop on Technoogy and Poicy for Accessing Spectrum, Aug. [] Nationa Instruments, Rea-time LTE/Wi-Fi Coexistence Testbed, in http://www.ni.com/white-paper//en/, Feb.