Embed Size (px)
Citation preview
1
PCF-Based LTE Wi-Fi Aggregation forCoordinating and Offloading the Cellular Traffic to
D2D NetworkBushra Ismaiel,Student Member IEEE, Mehran Abolhasan, Senior Member, IEEE, Wei Ni, Senior Member, IEEE,David Smith, Member, IEEE, Daniel Franklin, Member, IEEE, Eryk Dutkiewicz, Member, IEEE, Marwan Krunz,
Fellow, IEEE and Abbas Jamalipour, Fellow, IEEE
Abstract—Device-to-Device (D2D) communication is a promis-ing technology towards 5G networks. D2D communication canoffload traffic using licensed/unlicensed band by establishinga direct communication between two users without traversingthe base station (BS) or core network. However, one of themajor challenges of D2D communication is resource allocationand guaranteeing Quality-of-Service (QoS). In this paper, weestablish an optimal queuing scheduling and resource allocationproblem for three-tier heterogeneous network based on LTEWi-Fi aggregation (LWA), to offload voice/multimedia trafficfrom licensed band to unlicensed band using Scalable MACProtocol (SC-MP) under various static delay constraints. Theaccess mechanism used for Wi-Fi in SC-MP is Point Coordi-nation Function (PCF), which further offloads the multimediatraffic using D2D communication in unlicensed band. Resourceallocation and optimal joint queuing scheduling problems areformulated with diverse QoS guarantee between licensed andunlicensed band to minimize the bandwidth of licensed band.Furthermore, an iterative algorithm is proposed to express thenon-convex problem as series of sub-problems based on BlockCoordinate Descent (BCD) and difference of two convex functions(D.C) program. We have simulated the proposed scheme usingtwo scenarios; voice traffic using licensed band and voice trafficusing both licensed and unlicensed band whereas, multimediatraffic uses unlicensed band for both scenarios. The simulationresults show that both the schemes perform better than theexisting scheme and scenario 2 outperforms scenario 1.
Index Terms—Device-to-Device communication, Effective ca-pacity, Wireless local area network (WLAN), Long term evolution(LTE) network, Resource allocation, Quality-of-Service (QoS)
I. INTRODUCTION
The rapid growth in demand for wireless devices has accel-erated the development of next-generation wireless networks,
Manuscript was submitted on March 11, 2018; revised on July 18, 2018;accepted on Sep 22, 2018. (Corresponding Author: Bushra Ismaiel.)
B. Ismaiel, M. Abolhasan, D. Franklin and Eryk Dutkiewicz areassociated with the School of Electrical and Data Engineering,University of Technology Sydney, Ultimo, NSW 2007, Australia. (e-mail: [email protected]; [email protected];[email protected]; [email protected])
W. Ni and D. Smith are the Senior Research Scientist and associatedwith Data61, CSIRO, Clayton, VIC 3169, Australia. (e-mail: [email protected]; [email protected]).
Marwan Krunz is the Professor in the Department of Electrical andComputer Engineering at the University of Arizona (UA), ECE Bldg.,Rm. 365, 1230 East Speedway Blvd., Department of Electrical and Com-puter Engineering University of Arizona, Tucson, AZ 85721. (e-mail:[email protected]).
A.Jamalipour is a Professor and is associated with School of Electricaland Information Engineering, University of Sydney, Camperdown NSW 2006,Australia. (e-mail: [email protected]).
namely, 5G. In 5G, it is proposed to use a wider rangeof frequencies and spectrum to solve the capacity problem.In a dense heterogeneous network environment, the licensedspectrum is easily congested due to limited resources. Thishas motivated the operators to exploit the potential use of theavailable unlicensed spectrum. Recently, several new schemeshave been proposed to overcome the coverage and congestionproblems which include; deployment of small cells (Femto,Pico), cognitive radio, LTE unlicensed (LTE-U), LTE li-censed assisted access (LTE-LAA) and LTE Wi-Fi aggregation(LWA).
Joint deployment of cellular networks and wireless Localarea networks (WLAN) is more attractive to the operators asit utilizes both licensed and unlicensed band. There are severaladvantages of joint deployment such as; it is cost-effective,WLAN works with less complex protocol as compared withthe cellular network (LTE, LTE-A), better spatial reuse and itenabling access to local services utilizing unlicensed band [1].Currently, LTE-U/LTE-LAA has been the center of attractionnow adays but the real challenge for LTE-U/LTE-LAA isthe contention problem with the Wi-Fi users, as the greaterpreference is given to LTE-U/LTE-LAA users, which willenable them dominate the spectrum [2], [3]. LWA does notrequire another protocol like CSAT/LBT, but rather utilizesthe same existing Wi-Fi protocol and it transmits LTE traf-fic through Wi-Fi access point (AP) associated with LWAbase stations [2]. LWA has several advantages that include;improved quality-of-service(QoS), cost effective, and increasesystem capacity. Furthermore, joint operation between WLANand LTE is simple in LWA as LTE operates using LTE protocoland Wi-Fi operates using Wi-Fi traditional protocol [4].
D2D communication is another technology that can in-crease the capacity and coverage of the wireless networksby enabling immediate communication between the users.D2D communication allows two proximity users to performdirect data transmissions between each other with little or noinvolvement of a base station (BS). D2D technology, has alsobeen proposed as an effective solution for efficiently utilizingthe radio resources, hence increasing the spectral efficiency.Furthermore, it eases the core network data processing loadsby performing traffic offloading [5]. D2D communication willbe an integral part of 5G technology because of its particularfeatures; increase in spectral efficiency, capacity gain, andlower end-to-end delay [6], [7]. Resource allocation is the key
Copyright c© 2018 IEEE. Personal use of this material is permitted. However, permission to use thismaterial for any other purposes must be obtained from the IEEE by sending a request to [email protected].
2
aspect in D2D communication, and most of the work has beendone using cellular network [8].
Nowadays, there are high demands for video and multi-media content over cellular networks. Downloading similarvideos can waste the spectrum resource. However, D2D com-munication can be significant in caching videos and sharingcontent [9]. D2D users can act cooperatively, form clusters andfacilitate content dissemination [10], [11]. Service providerscan exploit D2D communication to additionally offload trafficin a congested domain particularly utilizing unlicensed band.
The novelty of the paper is to investigate an optimalqueue scheduling and resource allocation problem of diversetraffic in three-tier heterogeneous network with Wi-Fi accessmechanism based on IEEE 802.11 PCF and further offloadingmultimedia files through D2D communication. There is anongoing work [12], that proposes an optimal solution foroffloading traffic using LWA. Our proposed scheme builds onand further extends the approach proposed in [7]. The diversetraffic; voice and multimedia, will be offloaded from licensedband to unlicensed band in three-tier network using IEEE802.11 PCF as the access mechanism for Wi-Fi and furthermultimedia traffic is offloaded using D2D communication. Ourkey contributions can be summarized as:
• A three-tier network based on LWA technology is pro-posed to offload diverse traffic from licensed to un-licensed band with modification in resource allocationscheme based on IEEE 802.11 PCF to access Wi-Fichannels and further offload multimedia files throughD2D communication. The optimal joint queue schedulingand resource allocation problem of three-tier network isformulated to minimize the bandwidth of licensed bandby guaranteeing the QoS.
• A closed-form expression of effective capacity is derivedusing 6 state semi-Markovian model.
• An iterative algorithm is proposed to convexify the prob-lem using generic block coordinated descent (BCD) anddifference of convex functions (D.C) program.
• Simulation is performed using two scenarios of the pro-posed schemes; in the first scenario the voice traffic useslicensed band whereas multimedia traffic uses unlicensedband and in the second scenario half of the voice trafficgoes through licensed band and other half of the trafficgoes to unlicensed band along with the multimedia traffic.Our result proved that both scenarios perform better thanSMS and scenario 2 outperforms scenario 1.
The paper is organized as follows. In Section II, foundationand related works are discussed and surveyed. In SectionIII, the system model is presented. In Section IV, the closedform expression for effective capacity of three-tier network isderived. In Section V, the optimal joint queue scheduling andresource allocation problems with the QoS guarantee betweenlicensed and unlicensed band are formulated to minimizethe usage of licensed bandwidth of three-tier heterogeneousnetwork. An iterative algorithm is proposed based on BCDand DC programs to solve the problem. In Section VI,the simulation results are shown, and finally, conclusion ispresented in Section VII.
II. RELATED WORK
The performance characteristics of PCF has been exten-sively studied in [13]–[15]. The performance of video trans-mission using PCF mechanism is discussed in [16], [17].
In [18], a cooperative ARQ MAC protocol is proposed thatis compatible to the legacy IEEE 802.11 DCF. It coordinatesthe transmissions among a set of relay nodes which act ashelpers in a bidirectional communication. The maximizationof the effective capacity with QoS and power constraints usingDCF is proposed in [19].
The aggregation of LTE and Wi-Fi has recently drawnextensive attention. In [20], LWA analytical model was pro-posed using Markovian model. In [21], convex optimizationtechnique was used to optimize the power of licensed bandand time duration of the unlicensed band to maximize the totalalgorithmic utility of users. In [22], a linear program techniquewas used to optimize the licensed bandwidth allocation andrate allocation in the unlicensed band to maximize the overallthroughput. In [23], a cross-system learning approach is pro-posed to optimize the power, cell range expansion bias, sub-band selection and traffic scheduling and the delay-toleranttraffic is steered to unlicensed band. The access mechanismused in Wi-Fi in [21]–[23] is DCF which is contrary to ourwork as we will be using PCF as an access mechanism andfurther offloading the traffic using D2D communication, whichis very useful for the dense environment as assigning QoS.
Some related work have been focused on the QoS of thequeues transmitted through a single interface. In [24], a semi-Markovian model was proposed for Wi-Fi based on distributedcoordinated function (DCF) access channel mechanism. Itprovides for an asymptotically tight linkage between sourcecharacteristics, system resources, and QoS. In [25], a uni-fied framework for LWA was proposed based on QoS classindicator, but there was no analytical model. A non-trivialeffort would be required to extend these works to multipleheterogeneous interfaces, as proposed in this paper. In [26],Karush-Kuhn-Tucker technique was used to maximize theeffective capacity of the mobile video traffic. In [27], [28],the effective capacity is focused on the modelling of ratefluctuations at the physical layer. In [12], author proposes tooffload traffic from licensed band to unlicensed band usingLWA technology under delay constraints and also deriveeffective capacity using semi-Markovian model. Our workdiffers from [12] as we will be introducing a three tier networkto offload voice/multimedia traffic from licensed to unlicensedband with modification in resource allocation scheme basedon IEEE 802.11 PCF to access Wi-Fi channels and furtheroffload multimedia files through D2D communication. We canconclude that our proposed scheme is scalable and performbetter with no contention problem in it.
To the best of our insight, no analytical work has beenpublished until now for offloading traffic from LTE to Wi-Fi based on IEEE 802.11 PCF mechanism and further of-floading multimedia traffic through D2D communication. Thispaper presents an analytical model for three-tier heterogeneousnetwork using Scalable MAC protocol for Wi-Fi network,to evaluate an optimal joint queue scheduling and resource
3
allocation problem under various static delay constraints. Thesimulation results illustrate that the proposed scheme hassignificant performance gain as compared to the state of art.
III. SYSTEM MODEL
Tier 3D2D
LTE BS
Wi-Fi AP 2Tier 1
Tier 2
Wi-Fi AP 1
Fig. 1: Proposed Three-Tier Heterogeneous Network
Our proposed architecture builds on the LTE-Wi-Fi ag-gregation (LWA) technology with modification in resourceallocation scheme based on IEEE 802.11 PCF to accessWLAN channel and D2D communication integration. Asthe architecture is proposed for 5G, it is also assumed thatSoftware Defined Network (SDN) is deployed at the core endso that the decisions made are structured, rapid and congestioncontrolled. SDN and routing protocol is out of the scope ofour paper but for further understanding [29] and [30] can bereferred. The Fig. 1, shows the architecture of proposed three-tier network.
We consider a macro LTE base station (BS) with an LTE airinterface in the licensed band overlayed by a Wi-Fi base station(BS) with Wi-Fi air interface in the unlicensed band. We havealso considered another Wi-Fi node operating in an unlicensedband within the coverage of three-tier network. Two frequencybands are considered; band a is the licensed bandwidth andband b is the unlicensed bandwidth. The total bandwidth isgiven as Bn where n = a, b. Both voice and multimedia trafficis taken into account and licensed and unlicensed bandwidthare further sub-divided. Let Bav and Bam be the licensedbandwidth for voice and multimedia and Bbv and Bbm bethe unlicensed bandwidth for voice and multimedia. There aretotal K number of users, where γk,n represent the signal-to-interference-noise ratio (SINR) of a user k in band n andk = 1,. . . ,K. The packets are delivered to the users eitherusing licensed LTE air interface or using unlicensed Wi-Fi airinterface.
The Wi-Fi users operating in the unlicensed band useScalable MAC Protocol (SC-MP) based on point coordinationfunction (PCF) to access the channel [31]. PCF consist ofcontention free period (CFP) and a contention period (CP)and works on a polling based mechanism. PCF has pointcoordinator (PC), which is located at a Wi-Fi Access Point(AP). During CFP the point coordinator (PC) creates a pollinglist and polls the node accordingly. PC ensures that the intervalbetween two polling stations is not more than PCF inter-frame
spacing (PIFS) [32]. The polling scheme used is the bestsignal-to-noise ratio (SINR) scheme. The advantage of bestSINR polling scheme is that it will give better throughputwith less delay. The PC creates a polling list based on bestSINR polling scheme and poll the user accordingly. PC has tosense if the channel is idle for PCF inter-frame spacing (PIFS)period and then a beacon frame is transmitted at the start ofCFP. PC polls the first user from the polling list and if the userhas data to transmit, a time slot (TS) is allocated to that user.If the user does not have any data to transmit a null frameis transmitted. Similarly, all the users are polled accordinglyfrom the polling list and if the cycle ends before all users arepolled then in the next cycle PC will resume polling wherethe polling list ended.
We further apply D2D communication for cachingvideo/multimedia files using CP in PCF [31]. When a useris polled by WLAN BS (WBS) and requests a video, WBSchecks if any of the neighbouring user has downloaded therequested video. If the requested video is already downloadedby the neighbouring user, WBS will check the distance be-tween the two users and if distance is less than the thresholdvalue it will allow D2D communication between the two usersby assigning a free channel for a specific time period. If thevideo is not downloaded, WBS assigns a time slot to the userto download the video. Once the video is downloaded, it willbe stored in the cache memory of WBS with the locationinformation of the user. If downloaded video is not requestedby any other user for a certain time, WBS deletes the videofrom the cache memory as it will save the storage space [31].
We assume Rayleigh block flat-fading channels in bothlicensed and unlicensed bands. The packets arriving at thescheduler for the user k can be scheduled to proceed usingeither licensed LTE air interface or unlicensed Wi-Fi airinterface. Two transmit queues are formed for the two airinterface as shown in the Fig. 2. A binary variable yn,k isdefined to select the band for the packets. The bandwidthallocated to a user with binary variable yn,k is αn,k. Thechannel remains unchanged during a time frame T, but canvary independently across different time frames.
The QoS exponent θk is the QoS packets intended for theuser k. The larger the value of θk, the more stringent is theQoS and the smaller the value of θk the more loose is the QoS.If both air interfaces are selected, we propose to decoupleθk between the two interfaces. QoS exponent of user k inband n is denoted as θn,k. We propose to precisely designthe binary variable yn,k, bandwidth αn,k and QoS θn,k suchthat the maximum number of packets can be delivered withoutcompromising θk for all the users.
We also propose that each user n requires a minimumdata rate of Rk, a delay bound of Dk
th and the maximumprobability threshold of the delay bound being violated isPkth. The effective capacity Cn,k(θn,k), can be defined to themaximum consistent arrival rate at the input of the transmitqueue for user k in band n, as given by [27], [33], [34];
Cn,k(θn,k) = − limt→∞
1
θn,ktlog(E{e−θn,kSn,k(t)}) (1)
4
Licensed band
Controller
Voice
Multimedia
Unlicensed band
Multimedia
Multimedia
Voice
Voice
Fig. 2: Queuing Model Of Three-Tier Network
where Sn,k(t) is the number of bits successfully delivered toWi-Fi user k during (0,t] and E{.} denotes expectation.
According to the effective capacity theory [26], the delay-bound violation probability of the transmit queue in eachindividual band can be approximated as;
Pr{D > Dkth} ≈ e−θn,kCn,k(θn,k)D
nth ,∀k ∈ K,∀n ∈ N (2)
Overall, the delay-bound violation probability needs to satisfy[26];∑n∈N
yn,ke−θn,kCn,k(θn,k)D
kthCn,k(θn,k)∑
n∈Nyn,kCn,k(θn,k)
≤ P kth,∀k ∈ K, (3)
where∑n∈N
yn,kCn,k(θn,k) is the total number of packets
delivered to user and∑n∈N
yn,ke−θn,kCn,k(θn,k)D
kthCn,k(θm,n)
accounts for the total number of packets delivered to user kbefore the delay bound through the two bands.
IV. EFFECTIVE CAPACITY OF LTE AND SCALABLE MACPROTOCOL (SC-MP)
In this section we will discuss the effective capacity ofSC-MP operating in unlicensed band and LTE operating inlicensed band, while preserving the QoS.
A. Effective Capacity Of SC-MP in Unlicensed Band
A semi-Markovian model is used to derive the effectivecapacity of SC-MP of K number of users with total 3K stateswith each user having 3 states. In Fig. 3, a semi-Markovianmodel is shown with 2 users having 6 states in total.
Two types of traffic is considered in the framework for theallocation of a time slot; voice and video/multimedia.Moreover, D2D communication is also applied tovideo/multimedia traffic for offloading the traffic. Wehave assumed that duration of voice traffic in a time slotis random and the voice duration can vary for every userdepending on the number of packets in the buffer. Similarly,for multimedia traffic either the user is allocated a time slot totransmit or the user will be allowed for D2D communication.The probability that a user has data in the buffer is P,
P(r1|v2)
P(r1|m2)P(r2|r1)
r1
v1 m1
r2
v2 m2
P(r1|r2)
Fig. 3: SC-MP Semi-Markovian Model
and 1-P is the probability that there is no data in thebuffer. The probability P can either be a voice traffic or avideo/multimedia traffic.
P = PV or P = PM (4)
where PV is the probability of voice traffic and PM is theprobability of video/multimedia traffic. Two types of channelcondition are used in modelling the SC-MP; J and E. E is thechannel condition that should be satisfied for the allocation ofthe time slot for voice and video/multimedia traffic. Whereas,J is the channel condition that should be satisfied for D2Dcommunication between two users. D2D communication willbe allowed if the video requested by the user is alreadydownloaded by the neighbouring user/users and is in the cachememory of the AP and satisfies the channel condition J.The probabilities are defined as, PND= Probability that videorequested is not downloaded by any neighbouring users, PS=Probability that channel condition E is not satisfied and PD=Probability that channel condition J is not satisfied.
5
Each user has 3 unique states: rk = polling state of a Wi-Fi user, vk = time slot allocation for a Wi-Fi user havingvoice traffic and satisfying the channel condition E, mk = atime slot is allocated for video/multimedia traffic to a Wi-Fiuser after satisfying the channel condition E or allow D2Dcommunication to a Wi-Fi user if the neighbour has alreadydownloaded the video/multimedia file and satisfy the channelcondition J, where k = 1, . . ., K.
The transition probability from polling state r1 to v1 foruser1 is given as
PA1 = P (v1|r1) = PV 1(1− PS1) (5)
Similarly, the transition probability for user2 is given as
PA2 = P (v2|r2) = PV 2(1− PS2) (6)
where PV 1 and PV 2 are the probabilities of voice traffic foruser1 and user2. 1-PS1 and 1-PS2 are the probabilities thatsatisfies the channel condition E for time slot allocation foruser1 and user2 for voice and video/multimedia traffic.The transition probability from r1 to m1 is given as
PF1 = P (m1|r1) = PM1(1− PS1)
+(PM1)(PS1)[1− (PND1
+((1− PND1)PD1)Nn)]
(7)
Similarly, for user2 the transition probability is given as
PF2 = P (m2|r2) = PM2(1− PS2)
+(PM2)(PS2)[1− (PND2
+((1− PND2)PD2)Nn)]
(8)
where PM1 and PM2 are the probabilities of video/multimediatraffic for user1 and user2. 1-PS1 and 1-PS2 are the prob-abilities that satisfies the channel condition E for time slotallocation for user1 and user2, Nn is the total number ofneighbours that downloaded the video, PND1 and PND2 arethe probabilities that video requested by user1 and user2 isnot downloaded by any neighbouring users and PD1 and PD2
are the probabilities that channel condition J is not satisfiedfor D2D communication for user1 and user2. The transitionprobability for polling state r1 to polling state r2 is given as
PE1 = 1− PA1 − PF1 (9)
PE2 = 1− PA2 − PF2 (10)
The transition probability for all other states are 1. The statetransition probabilities generate the transition matrix Q, usingsemi-Markovian model for 2 users and is given as
Q =
0 PA1 PF1 PE1 0 00 0 0 1 0 00 0 0 1 0 0PE2 0 0 0 PA2 PF2
1 0 0 0 0 01 0 0 0 0 0
(11)
The duration of both states r1 and r2 is Tr1 and Tr2 andis constant for both users. The duration for v1 and v2 israndom and denoted by Tv1 and Tv2 , which depend on thenumber of packets in the buffer. The duration for m1 and
m2 is random and is denoted by Tm1 and Tm2, it can eithertransmit video/multimedia traffic through time slot or canallow D2D communication between users depending on thechannel condition and neighbouring downloaded traffic.
The moment generating functions (MGF) of Tr1, Tv1,Tm1, Tr2, Tv2, and Tm2 are M1(t)=etTr1, M2(t)=ΣetTv1PA1,M3(t)=ΣetTm1PF1, M4(t)=etTr2, M5(t)=ΣetTv2PA2, andM6(t)=ΣetTm2PF2.
We define two auxiliary variables, s and u, with referenceto [24], [35]. We now create a diagonal matrix Γ(s,u). Thediagonal elements of Γ(s,u) are the MGFs of the six state semi-Markovian model and is given in (12). The matrix H(s,u) isa non-negative irreducible matrix, as it cannot be arranged toupper triangular matrix, e.g, by using the Gaussian-NewtonMethod [36]. The spectral radius of H(s,u), denoted byφ(s,u)=ρ(H(s,u)) is a simple eigenvalue of H(s,u) and ρ(.)represents the spectra radius.We can write for each permissible pair of s and uH(s,u)=Γ(s,u)Q and is given in (13).
With reference to [35], the effective capacity C = u(s)s , when
φ(s,u(s)) = 1 and θ = -s.As the result, the effective capacity C can be evaluated by
solving φ(-θ,-θC)=1 for θ>0 [24], [35]. Meanwhile, φ(-θ,-θC)is the eigenvalue of (H(-θ,-θC)), we have (14), where I is theidentity matrix and |.| is the determinant of the matrix. Thedeterminant is given in (15) where,
X = M2(−Rθ + θC)PA1PE2 +M3(−Rθ + θC)PF1PE2+
M5(−Rθ + θC)PE1PA2 +M6(−Rθ + θC)PE1PF2
Y = PA1PA2M2(−Rθ + θC)M5(−Rθ + θC) + PA1PF2
M2(−Rθ + θC)M6(−Rθ + θC) + PA2PF1
M3(−Rθ + θC)M5(−Rθ + θC) + PF2PF1
M3(−Rθ + θC)M6(−Rθ + θC)
We substitute the λ = φ(-θ,-θC) = 1 in (15).Substituting the value of X and Y and solving (15) through
the Matlab, the closed form expression of effective capacityfor voice and multimedia of 2 users of SC-MP in unlicensedband is given as;
Cbv(αbv, θb) =z
θb+ αbv log(1 + γb) (16)
Cbm(αbm, θb) =z
θb+ αbm log(1 + γb) (17)
where γb is signal-to-interference-noise ratio (SINR) of a userin unlicensed band and z is a constant value and varies with thevalue of probabilities defined, that is; PV , PM , PD and PNDfor every user. We can be calculate the value of z by insertingthe values of PV , PM , PD and PND in equation (15).
B. Effective Capacity of LTE in Licensed Band
The effective capacity of LTE in the licensed band of userk is given by [12], [37];
Ca,k(αa,k, θa,k) = − 1
θa,kTlog(Eγ{e−θa,kαa,kT log2(1+γa,k)})
(18)
6
Γ(s, u) =
M1(−u) 0 0 0 0 0
0 M2(Rs− u) 0 0 0 00 0 M3(Rs− u) 0 0 00 0 0 M4(−u) 0 00 0 0 0 M5(Rs− u) 00 0 0 0 0 M6(Rs− u)
(12)
H(s, u) =
0 M1(−u)PA1 M1(−u)PF1 M1(−u)PE1 0 00 0 0 M2(Rs− u) 0 00 0 0 M3(Rs− u) 0 0
M4(−u)PE2 0 0 0 M4(−u)PA2 M4(−u)PF2
M5(Rs− u) 0 0 0 0 0M6(Rs− u) 0 0 0 0 0
(13)
|H(-θ,-θC)-φ(-θ,-θC)I| =∣∣∣∣∣∣∣∣∣∣∣∣
−φ(−θ,−θC) M1(−u)PA1 M1(−u)PF1 M1(−u)PE1 0 00 −φ(−θ,−θC) 0 M2(Rs− u) 0 00 0 −φ(−θ,−θC) M3(Rs− u) 0 0
M4(−u)PE2 0 0 −φ(−θ,−θC) M4(−u)PA2 M4(−u)PF2
M5(Rs− u) 0 0 0 −φ(−θ,−θC) 0M6(Rs− u) 0 0 0 0 −φ(−θ,−θC)
∣∣∣∣∣∣∣∣∣∣∣∣(14)
= φ(−θ,−θC)6 − φ(−θ,−θC)4PE1PE2M1(θC)M4(θC)− φ(−θ,−θC)3M1(θC)M4(θC)X − φ(−θ,−θC)2
M1(θC)M4(θC)Y(15)
where γa,k is the signal-to-interference-noise ratio (SINR) ofuser k in licensed band.
The closed form expression of effective capacity in (18) forvoice and multimedia can be written as:
Cav,k(αav,k, θa,k)
= − 1
θa,kTlog(Eγ{e−θa,kαav,kT log2(1+γa,k)}) (19)
Cam,k(αam,k, θa,k)
= − 1
θa,kTlog(Eγ{e−θa,kαam,kT log2(1+γa,k)}) (20)
where Cav,k(αav,k, θa,k) is the effective capacity of voice inlicensed band and Cam,k(αam,k, θa,k) is the effective capacityin of multimedia in licensed band.
V. MINIMIZING THE BANDWIDTH OF LICENSED BAND
The important objective of this section is to minimize therequirement of licensed bandwidth in three-tier heterogeneousnetwork while guaranteeing the QoS for all users. For currentproblem formulation, we have assumed that voice traffic usesboth licensed and unlicensed band, whereas multimedia trafficgoes through the unlicensed band to minimize the usage oflicensed band. In next Section VI, we compare this scenariowith another scenario where voice traffic goes to licensedband and multimedia traffic goes through unlicensed band andobserve their performance.
According [12], we can also formulate the problem in ourscheme as
P1 : minimizeαn,k,θn,k,yn,k
∑k∈K
yn,kαn,k (21a)
s.t,
∑n∈N
yn,ke−θn,kCn,k(θn,k)D
kthCn,k(θn,k)∑
n∈Nyn,kCn,k(θn,k)
≤ P kth,∀k ∈ K
(21b)∑k∈K
yb,kαb,k ≤ Bb (21c)
∑n∈N
yn,kCn,k(αn,k, θn,k) ≥ Rk,∀k ∈ K (21d)
yn,k = 0, 1,∀k ∈ K,∀n ∈ N (21e)
αn,k ≥ 0, θn,k ≥ 0,∀k ∈ N, ∀n ∈ N (21f)
where (21b) describes to meet the requirement of QoS of userk, (21c) states the total unlicensed bandwidth that should notexceed Bb, (21d) represents the minimum data rate of the userk, , (21e), and (21f) are the generic restraints to specify theproblem.It can be observed that from P1, e−θn,kCn,kD
kth and yn,kCn,k
are coupled and the effective capacity in (21d) is not jointconvex on αn,k and θn,k. As yn,k being a binary variable andaccording to [12], it is said that
yn,kCn,k(αn,k, θn,k) = Cn,k(yn,kαn,k, θn,k) (22)
7
By using the inequality [38];
0 ≤ αn,k ≤ yn,kχ (23)
where χ is a predefined constant. If both the licensed andunlicensed bands are selected to transmit the packets (21d)can be rewritten as∑
n∈Ne−θn,kCn,k(θn,k)D
kthCn,k ≤ P kth
∑n∈N
Cn,k (24)
Using Chebyshev’s sum inequality [39], (21d) can be writtenas
1
|N |∑n∈N
e−θn,kCn,k(θn,k)Dkth
∑n∈N
Cn,k ≤ P kth∑n∈N
Cn,k (25)
where |N | stands for cardinality. If a packet is transmitted touser k in band n the (21b) can be rewritten as ;
e−θn,kCn,k(θn,k)Dkth ≤ P kth (26)
Combining (25) and (26), (21b) can be rewritten as∑n∈N
e−θn,kCn,k(θn,k)Dkth − 1 + yn,k ≤ P kth
∑n∈N
yn,k (27)
The (21e), can be relaxed as the intersection of the followingregion [38];
0 ≤ yn,k ≤ 1,∀k ∈ K,∀n ∈ N (28)
∑n∈N
∑k∈K
(yn,k − (yn,k)2) (29)
We state two variables for simplification; ωn,k = αn,kθn,k and
βn,k =1
θn,k, where αn,k = ωn,kβn,k. The effective capacity
of SC-MP and licensed band in (16), (17), and (19) can berewritten in terms of auxiliary variables as
Cav,k(ωav,k, βa,k) = −βa,kT
log(Eγ{e−ωav,kTlog2(1+γa)})(30)
Cbv(αbv, βb) = zβb + ωbvβblog2(1 + γb) (31)
Cbm(αbm, βb) = zβb + ωbmβblog2(1 + γb) (32)
Using (30), (31) and (32), (27) can be rewritten as
e
log(Eγ{e−ωav,kTlog2(1+γa)})T
Dkth
+
e−(2z+ωbvβblog2(1+γb)+ωbmβblog2(1+γb)Dkth − 2+∑
n∈Nyn,k ≤ P kth
∑n∈N
yn,k,∀k ∈ K
(33)
As an outcome and according to [12], P1 can be simplifiedto be a continuous problem and can be written as
P2 : minimize{ωn,k},{βn,k},{yn,k}
∑k∈K
ωa,kβa,k (34a)
e
log(Eγ{e−ωav,kTlog2(1+γa)})T
Dkth
+
e−(2z+ωbvβblog2(1+γb)+ωbmβblog2(1+γb)Dkth − 2+∑
n∈Nyn,k ≤ P kth
∑n∈N
yn,k,∀k ∈ K
(34b)
∑k∈K
(ωbv,k + ωbm,k)βb,k ≤ Bb (34c)
s.t,∑n∈N
Cn,k(ωn,k, βn,k) ≥ Rk,∀k ∈ K (34d)
∑n∈N
yn,k ≥ 1,∀k ∈ K (34e)
∑n∈N
∑k∈K
(yn,k − (yn,k)2) (34f)
ωn,k ≥ 0, βn,k ≥ 0, 0 ≤ yn,k ≤ 1,∀n ∈ N, ∀k ∈ K (34g)
0 ≤ ωa,nβa,n ≤ yn,kχ,∀k ∈ K, ∀n ∈ N (34h)
As (34a), (34c), and (34d) are affine, therefore P2 is linearon {βn,k}.
Using difference of convex (DC) programming, P2 can bereformulated using P3
P3: minimize{ωn,k},{yn,k}
∑k∈K
ωa,kβa,k + λ∑n∈N
∑k∈K
yn,k−
λ∑n∈N
∑k∈K
(yn,k)2,(35)
s.t(34b)− (34d), (34f)− (34h), (36)
where λ is a large penalty factor.Let;
f1(ωn,k, yn,k) =∑k∈K
ωa,kβa,k + λ∑n∈N
∑k∈K
yn,k (37)
f2(yn,k) = λ∑n∈N
∑k∈K
(yn,k)2 (38)
f1(ωn,k, yn,k)− f2(yn,k) (39)
where (39) is the difference of two convex functions. Hence,P3 is a DC program in {ωn,k}, {yn,k}.
According to [40], P3 is equivalent to P2 for large value ofλ.
An algorithm is summarized based on a block coordinateddescent (BCD) framework [41]. In the algorithm we initialnumber of users K, minimum data rate for each user Rk,signal-to-noise ratio of user γn,k, requirement for QoS, andtotal unlicensed bandwidth Bb. The algorithm consist of 2loops. In the inner loop, given {ωn,k} and {yn,k}, P2 islinear programming on {βn,k}, and can be solved efficientlyusing interior-point method [42] and has a complexity ofO((2XY )3(3Y + X + XY )). In the outer loop, given{βn,k}, P2 is a DC program in {ωn,k, yn,k} and has acomplexity of O((XY )3). The total complexity of algorithm1 is O((XY )6(3Y +X +XY )) based on [38].
8
VI. SIMULATION
In this section, we evaluate the performance of the proposedalgorithm through simulation. There is one LTE Base station(LTE-BS) over-layed by WLAN BS (WBS). In our proposedscheme the WLAN BS contend the channel with the existingWi-Fi systems before it can poll the users. We have alsoconsidered another Wi-Fi node operating in an unlicensed bandwithin the coverage of proposed three-tier network. The chan-nel model used for licensed and unlicensed bands are the ITU-UMi Models. We have assumed 10 users uniformly distributedand all have the same minimum data rate requirement and QoSrequirement. We assumed the time frame length of LTE T =1 ms.
We have used both voice and multimedia traffic, for voicewe used persistent scheduling and for video we use adaptivescheduling. We have taken two different scenarios for ourproposed scheme. In scenario 1, the voice traffic is assigned tothe licensed band and the multimedia traffic is assigned to theunlicensed band. In scenario 2, half of the voice traffic goes onthe licensed band and other half will go through the unlicensedband along with the multimedia traffic. The proposed scenariosare also compared with the existing algorithm, Static mappingscheme (SMS), to evaluate the performance.
Static Mapping Scheme (SMS): In SMS cellular usersare first ordered in descending order in terms of SINR ofthe unlicensed band. The LWA BS sequentially assigns theunlicensed bandwidth and the QoS exponent to the orderedusers by solving the following equations [12];
e−θb,kCb,kDkth = P kth (40)
Cb,k(αb,k, θb,k) =z
θb,k+ αb,klog(1 + ηb,k) (41)
LWA BS will keep on assigning the unlicensed bandwidthuntil the total allocated bandwidth reaches the total unlicensedbandwidth or all the QoS users are satisfied. If the unlicensedband is insufficient to satisfy the QoS of the users, the restof the traffic goes to the licensed band. The remaining usersare ordered in descending SINR of the licensed band. TheLWA BS sequentially allocate the licensed bandwidth and theQoS exponent to the ordered users by solving the followingequations,
e−θa,kCa,kDkth = P kth (42)
Ca,k(αa,k, θa,k) = − 1
θa,kTlog(Eγ{e−θa,kαa,kT log2(1+γ)})
(43)
A static mapping table is established where a design parameterη (0 ≤ η ≤ 1) is decided according to the QoS Class Indicator(QCI) or the types of traffics. In the simulations, we assignedη = 0.7.In Fig. 4, we investigate the bandwidth of licensed band with
the delay requirements, where PV =0.3, PM=0.7, PD=0.6 andPND=0.4. There are in total 10 users within the coveragearea of LTE and Wi-Fi. The figure shows that as the delaybound increases the required bandwidth of the licensed banddecreases. It can also be observed that the our proposed
0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Delay bound [s]
0
0.5
1
1.5
2
2.5
3
3.5
Lic
ense
d B
and
wid
th (
Hz)
106
Proposed Scenario 1- PS-1SMS[7]Proposed Scenario 2- PS-2
PS-1
PS-2
[7]
SMS
Fig. 4: Licensed Bandwidth vs Delay Bound: PV =0.3,PM=0.7, PND=0.4, and PD=0.6
scenarios performs better in terms of delay as compared tothe existing scheme in [12] and SMS scheme. Also, proposedscenario 2 can perform much better than scenario 1, when thevoice traffic is split between the licensed and unlicensed band.
0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Delay bound [s]
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5L
icen
sed
Ban
dw
idth
(H
z)106
Proposed Scenario 1-PS-1[7]SMSProposed Scenario 2 -PS-2
SMS
[7]
PS-1
PS-2
Fig. 5: Licensed Bandwidth vs Delay Bound: PV =0.6,PM=0.4, PND=0.4, and PD=0.4
In Fig. 5, the values of probabilities are changed to furtherobserve the behaviour of licensed bandwidth with the delayrequirements. The Fig. 5 is plotted for PV =0.4, PM=0.6,PD=0.4 and PND=0.4. It can be observed that by increasingthe probability of voice users will increase the requirementof licensed bandwidth as well. Also, our proposed schemeperforms better than the existing scheme in [12] and the SMSscheme.
In Fig. 6 and 7, the required licensed bandwidth is investi-gated with the number of transmitters by changing the value ofprobabilities. The required licensed bandwidth increases withthe increase of transmitters. In Fig. 7, the voice probabilityis increased and the requirement for licensed bandwidth alsoincreases. Also, it can be observed that our proposed schemecan reduce the licensed bandwidth as compared to the existingscheme in [12] and SMS. Scenario 2 can provide 50% less
9
2 3 4 5 6 7 8 9 10
Number of Transmitters
0
0.5
1
1.5
2
2.5
Lic
ense
d B
and
wid
th (
Hz)
106
Proposed Scenario 1- PS-1SMS[7]Proposed Scenario 2- PS-2
[7]
PS-1
PS-2
SMS
Fig. 6: Licensed Bandwidth vs Number of Transmitters:PV =0.3, PM=0.7, PND=0.4, and PD=0.6
2 3 4 5 6 7 8 9 10
Number Of Transmitters
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
Lic
ense
d B
and
wid
th
106
Proposed Scenario 1- PS-1SMS[7]Proposed Scenario 2- PS-2
PS-1
PS-2
SMS
[7]
Fig. 7: Licensed Bandwidth vs Number of Transmitters:PV =0.6, PM=0.4, PND=0.4, and PD=0.4
requirement of licensed band as than scenario 1.In Fig. 8 and 9, we investigate the delay bound with the
number of transmitters by changing the probabilities. Allusers have the same delay threshold Dn
th = 0.05 s. Thedelay increases as the transmitter increases but our proposedschemes can reduce the delay as compared to the existingscheme in [12] and the SMS scheme. Scenario 2 can furtherreduce the delay as compared with scenario 1.
VII. CONCLUSION
In the proposed paper, we investigate an optimal queuescheduling and resource allocation problem under variousstatistical delay constraints of three-tier network. The three-tier network is based on LWA technology with modificationin resource allocation scheme based on IEEE 802.11 PCFto access WLAN channels. Multimedia files are further of-floaded using D2D communication. A closed-form expressionof effective capacity for the unlicensed band is derived usingsemi-Markovian process. Secondly, we formulate the optimaljoint queue scheduling and resource allocation problem withthe QoS guarantee between licensed and unlicensed bandto minimize the bandwidth of licensed band. An iterative
2 3 4 5 6 7 8 9 10
Number of Transmitters
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Del
ay B
ou
nd
[s]
Proposed Scenario 1- PS-1SMS[7]Proposed Scenario 2- PS-2
SMS
[7] PS-1
PS-2
Fig. 8: Delay Bound vs Number of Transmitters: PV =0.3,PM=0.7, PND=0.4, and PD=0.6
2 3 4 5 6 7 8 9
Number of Transmitters
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Del
ay b
ou
nd
[s]
Proposed Scenario 1- PS-1SMS[7]Proposed Scenario 2- PS-2
SMS
[7] PS-1
PS-2
Fig. 9: Delay Bound vs Number of Transmitters: PV =0.6,PM=0.4, PND=0.4, and PD=0.4
algorithm is proposed to convexify the problem as a seriesof block coordinated descent (BCD) and difference of convexfunctions (D.C) program. Further, the simulation was carriedout with our proposed scheme with two scenarios, in thefirst scenario all the voice traffic uses the licensed band andmultimedia traffic uses the unlicensed band, and the secondscenario split the voice traffic to licensed and unlicensed bandwhereas all the multimedia traffic goes to unlicensed band.The simulation results showed that our proposed scheme canprovide improved performance than the existing scheme in[12] and SMS scheme. Also scenario 2 can perform 50% betterthan scenario 1.
REFERENCES
[1] K. Doppler, M. Rinne, C. Wijting, C. B. Ribeiro, and K. Hugl, “Device-to-device communication as an underlay to LTE-advanced networks,”IEEE Communications Magazine, vol. 47, no. 12, 2009.
[2] 3GPP, “Collaboration on LTE - WLAN Integration,” http://www.3gpp.org/news-events/3gpp-news/1771-{WLAN LTE}, 2016, [Online;2016].
[3] L. Li, J. P. Seymour, L. J. Cimini, and C.-C. Shen, “Coexistence of wi-fiand laa networks with adaptive energy detection,” IEEE Transactions onVehicular Technology, 2017.
[4] Q. Cui, Y. Gu, W. Ni, X. Zhang, X. Tao, P. Zhang, and R. P. Liu,“Preserving reliability of heterogeneous ultra-dense distributed networks
10
in unlicensed spectrum,” IEEE Communications Magazine, vol. 56,no. 6, pp. 72–78, 2018.
[5] S. Andreev, A. Pyattaev, K. Johnsson, O. Galinina, and Y. Koucheryavy,“Cellular traffic offloading onto network-assisted device-to-device con-nections,” IEEE Communications Magazine, vol. 52, no. 4, pp. 20–31,2014.
[6] J. Liu, N. Kato, J. Ma, and N. Kadowaki, “Device-to-device commu-nication in lte-advanced networks: A survey,” IEEE CommunicationsSurveys & Tutorials, vol. 17, no. 4, pp. 1923–1940, 2015.
[7] J. Liu, N. Kato, H. Ujikawa, and K. Suzuki, “Device-to-device commu-nication for mobile multimedia in emerging 5g networks,” ACM Trans-actions on Multimedia Computing, Communications, and Applications(TOMM), vol. 12, no. 5s, p. 75, 2016.
[8] J. Dai, J. Liu, Y. Shi, S. Zhang, and J. Ma, “Analytical modelingof resource allocation in d2d overlaying multihop multichannel uplinkcellular networks,” IEEE Transactions on Vehicular Technology, vol. 66,no. 8, pp. 6633–6644, 2017.
[9] N. Golrezaei, P. Mansourifard, A. F. Molisch, and A. G. Dimakis, “Base-station assisted device-to-device communications for high-throughputwireless video networks,” IEEE Transactions on Wireless Communica-tions, vol. 13, no. 7, pp. 3665–3676, 2014.
[10] A. Antonopoulos, E. Kartsakli, and C. Verikoukis, “Game theoretic d2dcontent dissemination in 4g cellular networks,” IEEE CommunicationsMagazine, vol. 52, no. 6, pp. 125–132, 2014.
[11] E. Datsika, A. Antonopoulos, N. Zorba, and C. Verikoukis, “Cross-network performance analysis of network coding aided cooperativeoutband d2d communications,” IEEE Transactions on Wireless Com-munications, vol. 16, no. 5, pp. 3176–3188, 2017.
[12] Y. Gu, Q. Cui, W. Ni, P. Zhang, and W. Zhuang, “Optimal Schedulingacross Heterogeneous Air Interfaces of LTE/WiFi Aggregation,” IEEEInternational Conference on Communications (ICC), 2018.
[13] B. P. Crow, I. Widjaja, J. G. Kim, and P. Sakai, “Investigation of theieee 802.11 medium access control (mac) sublayer functions,” in IN-FOCOM’97. Sixteenth Annual Joint Conference of the IEEE Computerand Communications Societies. Driving the Information Revolution.,Proceedings IEEE, vol. 1. IEEE, 1997, pp. 126–133.
[14] M. A. Visser and M. El Zarki, “Voice and data transmission over an802.11 wireless network,” in Personal, Indoor and Mobile Radio Com-munications, 1995. PIMRC’95. Wireless: Merging onto the InformationSuperhighway., Sixth IEEE International Symposium on, vol. 2. IEEE,1995, pp. 648–652.
[15] J. Wu and G. Huang, “Simulation study based on qos schemes for ieee802.11,” in Advanced Computer Theory and Engineering (ICACTE),2010 3rd International Conference on, vol. 6. IEEE, 2010, pp. V6–534.
[16] S. Sharma, K. Kim, C. Oh, and A. Ahmad, “Performance evaluationof ieee 802.11 mac protocol for multimedia services,” Proceedings ofAPCC/ICCS, pp. 181–185, 1998.
[17] T. Suzuki and S. Tasaka, “Performance evaluation of integrated videoand data transmission with the ieee 802.11 standard mac protocol,” inGlobal Telecommunications Conference, 1999. GLOBECOM’99, vol. 1.IEEE, 1999, pp. 580–586.
[18] A. Antonopoulos, C. Verikoukis, C. Skianis, and O. B. Akan, “Energyefficient network coding-based mac for cooperative arq wireless net-works,” Ad Hoc Networks, vol. 11, no. 1, pp. 190–200, 2013.
[19] J. Li, Y. Ding, Q. Ye, N. Zhang, and W. Zhuang, “On effective capacityand effective energy efficiency in relay-assisted wireless networks,”IEEE Transactions on Vehicular Technology, 2017.
[20] O. Galinina, A. Pyattaev, S. Andreev, M. Dohler, and Y. Koucheryavy,“5G multi-RAT LTE-WiFi ultra-dense small cells: Performance dy-namics, architecture, and trends,” IEEE Journal on Selected Areas inCommunications, vol. 33, no. 6, pp. 1224–1240, 2015.
[21] F. Liu, E. Bala, E. Erkip, M. C. Beluri, and R. Yang, “Small-cell trafficbalancing over licensed and unlicensed bands,” IEEE Transactions onVehicular Technology, vol. 64, no. 12, pp. 5850–5865, 2015.
[22] A. R. Elsherif, W.-P. Chen, A. Ito, and Z. Ding, “Resource allocation andinter-cell interference management for dual-access small cells,” IEEEJournal on Selected Areas in Communications, vol. 33, no. 6, pp. 1082–1096, 2015.
[23] M. Bennis, M. Simsek, A. Czylwik, W. Saad, S. Valentin, and M. Deb-bah, “When cellular meets WiFi in wireless small cell networks,” IEEEcommunications magazine, vol. 51, no. 6, pp. 44–50, 2013.
[24] E. Kafetzakis, K. Kontovasilis, and I. Stavrakakis, “A novel effectivecapacity-based framework for providing statistical QoS guarantees inIEEE 802.11 WLANs,” Computer Communications, vol. 35, no. 2, pp.249–262, 2012.
[25] Q. Cui, Y. Shi, X. Tao, P. Zhang, R. P. Liu, N. Chen, J. Hamalainen, andA. Dowhuszko, “A unified protocol stack solution for LTE and WLAN
in future mobile converged networks,” IEEE Wireless Communications,vol. 21, no. 6, pp. 24–33, 2014.
[26] A. A. Khalek, C. Caramanis, and R. W. Heath, “Delay-constrained videotransmission: Quality-driven resource allocation and scheduling,” IEEEJournal of Selected Topics in Signal Processing, vol. 9, no. 1, pp. 60–75,2015.
[27] D. Wu and R. Negi, “Effective capacity: a wireless link model forsupport of quality of service,” IEEE Transactions on wireless commu-nications, vol. 2, no. 4, pp. 630–643, 2003.
[28] X. Zhang, J. Tang, H.-H. Chen, S. Ci, and M. Guizani, “Cross-layer-based modeling for quality of service guarantees in mobile wirelessnetworks,” IEEE Communications Magazine, vol. 44, no. 1, pp. 100–106, 2006.
[29] M. Abolhasan, J. Lipman, W. Ni, and B. Hagelstein, “Software-definedwireless networking: centralized, distributed, or hybrid?” IEEE Network,vol. 29, no. 4, pp. 32–38, 2015.
[30] R. Masoudi and A. Ghaffari, “Software defined networks: A survey,”Journal of Network and Computer Applications, vol. 67, pp. 1–25, 2016.
[31] B. Ismaiel, M. Abolhasan, D. Smith, W. Ni, and D. Franklin, “ScalableMAC protocol for D2D communication for future 5G networks,” in 14thIEEE Annual Consumer Communications & Networking Conference(CCNC),. IEEE, 2017, pp. 542–547.
[32] A. Kopsel, J.-P. Ebert, and A. Wolisz, “A performance comparison ofpoint and distributed coordination function of an IEEE 802.11 WLAN inthe presence of real-time requirements,” in 7th International Workshopon Mobile Multimedia Communication (MoMuC2000), 2000.
[33] L. Musavian and Q. Ni, “Effective capacity maximization with statisticaldelay and effective energy efficiency requirements,” IEEE Transactionson Wireless Communications, vol. 14, no. 7, pp. 3824–3835, 2015.
[34] A. H. Anwar, K. G. Seddik, T. ElBatt, and A. H. Zahran, “Effectivecapacity of delay-constrained cognitive radio links exploiting primaryfeedback,” IEEE Transactions on Vehicular Technology, vol. 65, no. 9,pp. 7334–7348, 2016.
[35] K. Kontovasilis and N. Mitrou, “Effective bandwidths for a class of nonMarkovian fluid sources,” ACM SIGCOMM Computer CommunicationReview, vol. 27, no. 4, pp. 263–274, 1997.
[36] Q. Cui, Y. Gu, W. Ni, and R. P. Liu, “Effective capacity of licensed-assisted access in unlicensed spectrum: From theory to applications,”IEEE Journal on Selected Areas in Communications, 2017.
[37] F. Jin, R. Zhang, and L. Hanzo, “Resource allocation under delay-guarantee constraints for heterogeneous visible-light and RF femtocell,”IEEE Transactions on Wireless Communications, vol. 14, no. 2, pp.1020–1034, 2015.
[38] E. Che, H. D. Tuan, and H. H. Nguyen, “Joint optimization of coop-erative beamforming and relay assignment in multi-user wireless relaynetworks,” IEEE Transactions on Wireless Communications, vol. 13,no. 10, pp. 5481–5495, 2014.
[39] Inequalities. By GH Hardy, JE Littlewood, G. Polya.. University Press,1952.
[40] B. Khamidehi, A. Rahmati, and M. Sabbaghian, “Joint sub-channelassignment and power allocation in heterogeneous networks: An efficientoptimization method,” IEEE Communications Letters, vol. 20, no. 12,pp. 2490–2493, 2016.
[41] N. Tekbiyik, T. Girici, E. Uysal-Biyikoglu, and K. Leblebicioglu,“Proportional fair resource allocation on an energy harvesting downlink,”IEEE Transactions on Wireless Communications, vol. 12, no. 4, pp.1699–1711, 2013.
[42] S. Boyd and L. Vandenberghe, Convex optimization. Cambridgeuniversity press, 2004.
Bushra Ismaiel -is currently working towards herPhD degree with the School of Electrical andData Engineering, University Of Technology Syd-ney, Australia. She completed her B.E from Univer-sity Of Texas at Dallas, USA and M.E from Univer-sity of South Carolina, USA. Her current researchinterests are in D2D communication, IoT, Wireless,Applications of IoT and 5G. She has worked inWireless industry for 5-6 years as a Team Lead, RFEngineer and Junior Project Manager
11
A/Prof Mehran Abolhasan -completed his B.E inComputer Engineering and PhD in Telecommunica-tions on 1999 and 2003 respectively at the Universityof Wollongong. He is currently an Associate Profes-sor and Deputy Head of School at School of Electri-cal and Data Engineering at UTS. A/Prof. Abolhasanhas authored over 120 international publications andhas won over 3 million dollars in research fund-ing. His Current research Interests are in SoftwareDefined Networking, IoT, Wireless Mesh, WirelessBody Area Networks, Cooperative Networks, 5G
Networks and Beyond and Sensor networks. He is currently a Senior Memberof IEEE.
Wei Ni -received the B.E. and Ph.D. degrees in Elec-tronic Engineering from Fudan University, Shanghai,China, in 2000 and 2005, respectively. Currentlyhe is a Team Leader, Senior Scientist, and ProjectLeader at CSIRO, Sydney, Australia. He also holdshonorary positions at the University of New SouthWales (UNSW), Macquarie University (MQ) andthe University of Technology Sydney (UTS). Hisresearch interests include optimization, game theory,graph theory, as well as their applications to networkand security. Dr Ni serves as Vice Chair of IEEE
NSW VTS Chapter since 2018, served as Editor for Hindawi Journal ofEngineering from 2012 – 2015, secretary of IEEE NSW VTS Chapter from2015-2018, Track Chair for VTC-Spring 2017, Track Co-chair for IEEE VTC-Spring 2016, and Publication Chair for BodyNet 2015. He also served asStudent Travel Grant Chair for WPMC 2014, a Program Committee Memberof CHINACOM 2014, a TPC member of IEEE ICC14, ICCC15, EICE14, andWCNC10.
David Smith -(S’01-M’04) received the B.E. de-gree in electrical engineering from the Universityof New South Wales, Sydney, NSW, Australia, in1997, and the M.E. (research) and Ph.D. degrees intelecommunications engineering from the Universityof Technology, Sydney, NSW, Australia, in 2001and 2004, respectively. Since 2004, he was withNational Information and Communications Technol-ogy Australia (NICTA, incorporated into Data61of CSIRO in 2016), and the Australian NationalUniversity (ANU), Canberra, ACT, Australia, where
he is currently a Senior Research Scientist with Data61 CSIRO, and anAdjunct Fellow with ANU. His current research interests include wirelessbody area networks, game theory for distributed networks, disaster tolerantnetworks, 5G networks, IoT, antenna design, distributed optimization for smartgrid and privacy for networks. He has published over 100 technical refereedpapers. Dr. Smith was the recipient of four conference Best Paper Awards.
Daniel Franklin - completed his PhD in Telecom-munications Engineering, “Enhancements to Chan-nel Models, DMT Modulation and Coding for Chan-nels Subject to Impulsive Noise”, at the Universityof Wollongong (2007) and also holds a Bachelor ofEngineering (electrical) - Honours I, University ofWollongong (1999). He has been a member of theIEEE since 1998. He is currently a Senior Lecturerin the Faculty of Engineering and Information Tech-nology at the University of Technology Sydney. Hiscurrent research and commercial interests include
multimedia and signal processing, wireless communication systems, positronemission tomography, radiotherapy, radiation dosimetry and pharmacokineticmodelling.
Eryk Dutkiewicz - received his B.E. degree inElectrical and Electronic Engineering from the Uni-versity of Adelaide in 1988, his M.Sc. degree in Ap-plied Mathematics from the University of Adelaidein 1992 and his PhD in Telecommunications fromthe University of Wollongong in 1996. His industryexperience includes management of the Wireless Re-search Laboratory at Motorola in early 2000. He iscurrently the Head of School of Electrical and DataEngineering at the University of Technology Sydney,Australia. He also holds a professorial appointment
at Hokkaido University in Japan. His current research interests cover 5G andIoT networks.
Marwan M. Krunz - is the Kenneth VonBehrenEndowed Professor in the ECE Department at theUniversity of Arizona. He is also an affiliated facultymember of the University of Technology Sydney(UTS). He directs the Broadband Wireless Accessand Applications Center, a multi-university industry-focused NSF center that includes affiliates fromindustry and government labs. He previously servedas the UA site director for Connection One, an NSFIUCRC that focuses on wireless communicationcircuits and systems. In 2010, Dr. Krunz was a
Visiting Chair of Excellence at the University of Carlos III de Madrid. Hepreviously held visiting research positions at UTS, INRIA-Sophia Antipolis,HP Labs, University of Paris VI, University of Paris V, University of Jordan,and US West Advanced Technologies. Dr. Krunzs research interests lie in theareas of wireless communications and networking, with emphasis on resourcemanagement, adaptive protocols, and security issues. He has published morethan 270 journal articles and peer-reviewed conference papers, and is a co-inventor on several US patents. He is an IEEE Fellow, an Arizona EngineeringFaculty Fellow (2011-2014), and an IEEE Communications Society Distin-guished Lecturer (2013 and 2014). He was the recipient of the 2012 IEEETCCC Outstanding Service Award. He received the NSF CAREER award in1998. He currently serves as the Editor-in-Chief for the IEEE Transactionson Mobile Computing. He previously served on the editorial boards for theIEEE Transactions on Cognitive Communications and Networks, IEEE/ACMTransactions on Networking, IEEE TMC, IEEE Transactions on Network andService Management, Computer Communications Journal, and IEEE Com-munications Interactive Magazine. See http://www2.engr.arizona.edu/ krunz/for more details.
Abbas Jamalipour -(S’86-M’91-SM’00-F’07) isthe Professor of Ubiquitous Mobile Networking atthe University of Sydney, Australia, and holds a PhDin Electrical Engineering from Nagoya University,Japan. He is a Fellow of the Institute of Electrical,Information, and Communication Engineers (IEICE)and the Institution of Engineers Australia, an ACMProfessional Member, and an IEEE DistinguishedLecturer. He has authored six technical books, elevenbook chapters, over 450 technical papers, and fivepatents, all in the area of wireless communications.
He was the Editor-in-Chief IEEE Wireless Communications (2006-08), VicePresident-Conferences (2012-13) and a member of Board of Governors of theIEEE Communications Society, and has been an editor for several journals. Hewas a General Chair or Technical Program Chair for a number of conferences,including IEEE ICC, GLOBECOM, WCNC and PIMRC. Dr. Jamalipour is anelected member of the Board of Governors (2014-16 and 2017-19), ExecutiveVice-President, Chair of Fellow Evaluation Committee, and the Editor-in-Chief of the Mobile World, IEEE Vehicular Technology Society. He is therecipient of a number of prestigious awards such as the 2016 IEEE ComSocDistinguished Technical Achievement Award in Communications Switchingand Routing, 2010 IEEE ComSoc Harold Sobol Award, and the 2006 IEEEComSoc Best Tutorial Paper Award.
