Performance Analysis of Non-Orthogonal Multicast in Two-tier Heterogeneous Networks
aa r X i v : . [ ee ss . SP ] J a n Performance Analysis of Non-Orthogonal Multicastin Two-tier Heterogeneous Networks
Yong Zhang ∗ , Bin Yang ∗ , Xiaohu Ge ∗ , Yonghui Li †∗ School of Electronic Information and Communications, Huazhong University of Science and Technology, Wuhan, China † School of Electrical and Information Engineering, University of Sydney, Sydney, AustraliaCorresponding author: Xiaohu Ge. Email: [email protected]
Abstract —With the explosive growth of mobile services,non-orthogonal broadcast/multicast transmissions can effectivelyimproves spectrum efficiency. Nonorthogonal multiple access(NOMA) represents a paradigm shift from conventional orthog-onal multiple-access (OMA) concepts and has been recognizedas one of the key enabling technologies for fifth-generation (5G)mobile networks. In this paper, a two-tier heterogeneous networkis studied, in which the wireless signal power is partitionedby the NOMA scheme. Moreover, the coverage probability, theaverage rate and the average QoE are derived to evaluate net-work performance. Simulation results show that compared withthe OMA method, non-orthogonal broadcast/multicast methodimprove both the average user rate and QoE in the two-tierheterogeneous network.
I. I
NTRODUCTION
With the explosive growth of mobile data traffic, especiallyvideo services, currently, cellular networks are facing hugechallenges to provide higher spectrum efficiency for mobileusers (MUs) [1]–[4]. However, in many cases, the MU’srequirements are roughly the same, e.g., requesting for hotresources. In this case, broadcast/multicast becomes a solutionto achieve higher network efficiency and improve quality ofexperience (QoE). Multicasting enables the same content to betransmitted to all users or a specific group of users [5]. Due tothe growth in data traffic and the number of connected devices,traditional orthogonal multicast cannot meet the requirementof 5G multicast services at low frequencies. Nonorthogonalmultiple access (NOMA) technology can achieve spectralefficiency improvement through superposition on the powerdomain [6], [7]. Many studies are dedicated to NOMA’sperformance analysis [8] and energy efficiency in cellularnetworks [9], [10]. Compared with the traditional water-fillingpower allocation strategy, NOMA scheme allocates morepower to users with poor channel conditions, resulting in abetter compromise between system throughput and user fair-ness [11]. However, in the practical multicast scenario, MUshave different ability to receive the same broadcast/multicastdata. Therefore, we consider a two-layer model which intro-duces NOMA into the network by dividing the user’s servicerequirements into two layers, i.e., the primary layer (PL) andthe secondary layer (SL). In each layer, MUs can provide thebest service as they can.Multicasting was studied in wireless networks [12], hetero-geneous networks (HetNets) [8], and device-to-device (D2D) communications. Amrico et al. [13] considered scalableMBMS video streams, with one basic layer to encode thebasic quality and consecutive enhancement layers for higherquality. In this work, only the most important stream (baselayer) is sent to all users in the cell. While less importantstreams (enhancement layers) are transmitted with less poweror coding protection, only user conditions with better channelscan receive additional information to improve video quality.At the same time, since the power domain non-orthogonaltransmission [14] enables multiple users to multiplex in thepower domain, it is necessary to decode their required datafrom the superimposed signals through continuous interfer-ence cancellation (SIC). SIC technology can significantlyincrease spectrum efficiency, reduce transmission delays, andsupport large-scale connectivity. SIC reduces the interferencepower by decoding and cancelling the interference signal. Anew SIC receiver was developed in [15], which decodes thesignal according to the downlink signal power and subtractsthe decoded signal from the received multi-user signal. How-ever, these studies have focused on cancelling interferenceby NOMA schemes. How to achieve the broadcast/multicastcommunications by NOMA scheme is surprisingly rare inthe open literature. Utilizing the NOMA scheme in wirelesssignal power partitions, a two-tier heterogeneous networkwith NOMA scheme is proposed in this paper. The maincontributions of this paper are summarized as follows:1) Considering the wireless signal power partition by theNOMA scheme, a two-tier heterogeneous multicast net-work is proposed to provide different QoEs for MUsbased on requirements and channel conditions.2) Based on the interference cancellation scheme, the cov-erage probability, the average rate and the average QoEare derived for the two-tier heterogeneous multicastnetwork with NOMA scheme.3) Compared with the orthogonal multiple access scheme,simulation results indicate the average user rate andthe QoE are improved in the two-tier heterogeneousmulticast network with NOMA scheme.The remainder of this paper is structured as follows. SectionII describes the system model. The coverage probability, theaverage rate and the average QoE of the two-layer hetero-geneous multicast network with NOMA scheme are derivedin Section III. The simulation results and discussions are
BSSBS MU SL PL P o w e r SBS SL PL P o w e r MBS
Fig. 1. System model and NOMA schemes for multicast communications intwo-tier heterogeneous network. presented in Section IV. Finally, conclusions are drawn inSection V. II. SYSTEM MODELA two-tier heterogeneous network is considered in thiswork, where the first tier is consist of low-band macrocellbase stations (MBSs) and the second tier consists of smallcell base stations (SBSs). According to [16], independentlyhomogeneous Poisson point process (PPP) can be used tomodel the locations of MBSs and small cells denoted as Φ M with density λ M and Φ S with density λ S , respectively.When multiple MUs request the same resource, e.g., pop-ular video resources or news, the base station (BS) performsmulticast transmission to improve transmission efficiency. Asmulticast MUs’ channel conditions are different, in order toprovide better services, the information is divided into twoparts in the power domain to provide different users withdifferent service, i.e., the primary layer (PL) which providesthe basic service and the secondary layer (SL) which aims toimprove the QoE of MUs. The received power from a smallcell BS and a macrocell BS are denoted as P rs and P rm ,respectively. Besides, assume that P rm > P rs . In this context,the MU has a large probability to connect with the macrocellBS even deploying a amount of SBSs. In order to reduce theload of the MBSs, cell extension technique is adopted witha bias factor b ( b > ). When (1) max { P rs } > max { P rm } ,MU would connect the SBS; (2) max { P rm } > b max { P rs } ,MU would connect the MBS; (3) max { P rs } < max { P rm }
To characterize shadowing effect in urban areas whichis a unique scenario in our analysis, both non-line-of-sight (NLOS) and line-of-sight (LOS) transmissions are incorpo-rated. Specifically, given the distance between a MU and a BS,saying d, the path-loss model can be described as follows: L U ( d ) = C U,i d − α U,i , wp P Ui ( d ) , (1) where U ∈ { S, M } , S and M means SBS andMBS,respectively. i ∈ { L, N } , L and M represents line-ofsight or non-line-of-sight, respectively. α U,L , α U,N are thepath loss exponents for BS LOS transmission and NLOStransmission respectively, C U,L and C U,N are the path lossfor BS LOS and NLOS transmission at the reference distance, P UL ( d ) is the probability that a link having length d is LOS,and P UN ( d ) = 1 − P UL ( d ) is the probability of the NLOSone. Regarding the mathematical form of P UL ( d ) , Bai[17]formulated P UL ( d ) = e − β U d , where β U is a parameter de-termined by the density and the average size of the blockages. B. Small scale fading
We describe h i as the fading of the link between the i-th BSand MU. Assume that each link is subjected to Nakagami-mdistribution. Then, H i = | h i | follows the normalized Gammadistribution. And N SL , N SN , N ML and N MN are the fadingparameters for the LOS link and the NLOS link in the SBSand MBS, respectively . C. Interference cancellation
The interference a MU received affects decoding perfor-mance in the future 5G networks. Therefore, in order toimprove the decoding capability, interference cancellationshall be applied. Assume that the useful signal is divided intotwo parts p = α p P r and p = (1 − α p ) P r by NOMA schemesfor multicast communications. p , p and α p present the PLsignal, SL signal and power allocation ratio, respectively. X , X , ..., indicate the interference signal, and without loss ofgenerality, we assume that X > X > ... > X k > ... First, the MU tries to decode signal p directly. If signal p can’t be decoded, the interference with the highest power willbe decoded at the receiver. Then subtract this interferenceand verify whether the user can decode signal p again.Due to reduce the interference cancellation complexity andlatency, we assume that only one interference cancellationis performed. After decoding signal p , signal p would bedecoded by subtracting the decoded signals.III. P ERFORMANCE ANALYSIS
A. The coverage probability of the primary signal
Assume that the transmission power of the SBS be P ts , andthe transmission power of the MBS be P tm , m = P tm /P ts ( m > ). As we assume that only the strongest interferences performed cancellation, p can be successfully decoded aslong as one of the following events is successful: p I Ω j + p + σ S ≥ T p I Ω j + p + σ S < T !| {z } A ∩ X (1) I Ω j + p + σ S ≥ T !| {z } B ∩ p I Ω j + p + σ S ≥ T !| {z } C . (2) In order to get the coverage probability of the primarysignal, three cases should be considered: (1) max { P rs } > max { P rm } (2) max { P rm } > b max { P rs } (3) max { P rs } < max { P rm } < b max { P rs } . So that coverage probability canbe expressed as P P ( α P , T PL ) = X i =1 P P,i ( α P , T PL ) , (3) where P P,i is the coverage probability of the primary signalin i-th case.
1) when max { P rs } > max { P rm } : In this case, the useris connected to the SBS and the strongest interference signalis definitely smaller than the useful signal so that interferencecancellation does not need to be applied in this case ,becauseif the useful signal cannot be successfully decoded, theinterference signal can not be decoded successfully. We definethe probability of coverage in this case as P P, ( α P , T PL ) = P P, ( P s > P m ) = P S,PL ( α P , T PL ) , (4) where P S,PL ( α P , T PL ) = P ( SINR PL > T PL )= P ( α p H S, L S ( d )(1 − α p ) H S, L S ( d ) + I S + I M + σ S > T PL )= P ( H S, > T PL α p − (1 − α p ) T PL · I S + I M + σ S L S ( d ) )= X s P S,s ( T PL α p − (1 − α p ) T PL ) ,s ∈ { L, N } , (5) and I S = X X s ∈ Φ S,s \ B C S,s H S,s d − α S,s
S,s + C S, ¯ s H S, ¯ s d − α S, ¯ s S, ¯ s I M = m X X s ∈ Φ M,i C M,i H M,i d − α M,i
M,i + C M, ¯ i H M, ¯ i d − α M, ¯ i M, ¯ i , (6) where s ∈ { L, N } , i ∈ { L, N } . Similar with [17], we canget the analytical expression as follows P S,s ( T ) = N s X n =1 ( − n +1 (cid:18) N s n (cid:19) Z ∞ e − nηsxαsTσ SCS,s e − Q S,n ( T,x ) − V S,n ( T,x ) − Q M,n ( T,x ) − V M,n ( T,x ) f S,s ( x ) dx, (7) Q S ,n ( T, x )=2 πλ S Z ∞ x F ( N S,s , nη
S,s x α S,s TN S,s t α S,s ) p S,s ( t ) tdt,Q M,n ( T, x ) = 2 πmλ M Z ∞ ϕ M ( x ) F ( N M,L , nϕ M ( x ) η S,s
TmN
M,L t α M,L ) p M,s ( t ) tdt ; ϕ M ( x ) = ( mC M,L C S,s x α S,s ) /α M,L ,V S,n ( T, x ) = 2 πλ S Z ∞ γ S ( x ) F ( N S, ¯ s , nγ S ( x ) η S,s TN S, ¯ s t α S, ¯ s ) p S, ¯ s ( t ) tdt ; γ S ( x ) = ( C S, ¯ s C S,s x α S,s ) /α S, ¯ s ,V M,n ( T, x )=2 πmλ M Z ∞ ξ M ( x ) F ( N M,N , nξ M ( x ) η S,s TN M,N t α M,N ) p M, ¯ s ( t ) tdt ; ξ M ( x ) = ( mC M,N C S,s x α S,s ) /α M,N , (8) where f S,s ( x ) =2 πλ S exp[ − Λ sS ([0 , x ])] exp[ − Λ LM ([0 , ϕ M ( x )])] · exp[ − Λ NM ([0 , ξ M ( x ))] , (9) Λ sBS ([0 , x ] = 2 πλ BS Z x rp BS,s ( r ) dr, BS ∈ { S, M } , s ∈ { L, N } . (10)
2) when max { P rm } > b max { P rs } : The user is connectedto the MBS and the strongest interference signal is definitelysmaller than the useful signal in this case so that interferencecancellation doesn’t need to be applied in this case. We definethe probability of coverage as P P, ( α P , T PL ) = P P, ( P m > bP s ) = P M,PL ( α P , T PL ) . (11) Similar with Eq. (5) P M,PL ( α P , T PL ) = X s P M,s ( T PL α p − (1 − α p ) T PL ) , s ∈ { L, N } . (12) P M,s ( T PL α p − (1 − α p ) T PL ) could be solved the same as Eq. (7)
3) when max { P rs } < max { P rm } < b max { P rs } : Inthis case, the user is connected to the SBS in the cellextension area, in which the maximum interference is greaterthan the desired signal. Therefore, cell cancellation shouldbe adopted to improve coverage performance. We definecoverage probability in this case as P P, . If interferencecancellation is performed only once, the event of successfullydecoding NOMA primary signal can be expressed as the unionof the following two events. Because event 0 is exclusive withevent 1, therefore we can get the expression as follows P P, ( α P , T PL ) = P P, ( α P , T PL ) + P P, ( α P , T PL ) . (13) From Eq. (2), it is found that event 1 consists of the eventA, B and C. Although the events A, B and C are relatedto each other which results in the difficulty to calculate P P, ( α P , T P L ) , we can get the approximation in some prac-tical scenarios. Through some practical simulation we foundthat there is a high probability: SIN R B > SIN R C , that is,event C ⊂ B . Therefore, P P, ( α P , T P L ) can be expressed as P P, ( α P , T PL ) = P ( ABC ) ≈ P ( AC ) = P ( C ) − P ( AC )= P ( C ) − P ( A ) = P ( A ) − P ( C ) . (14) irst, we can get the expression of P P, ( α P , T P L ) : P P, ( α P , T PL ) = P P, ( P s < P m < bP s )= P P, ( P m < b ∗ P s ) − P P, ( P m < P s )= P P, ( P s > P m/b ) − P P, ( P s > P m ) . (15) Similar to Eq. (4), its easy to calculate P P, ( P s > P m/b ) , P P, ( P s > P m ) . Second, calculating P ( A ) : P ( A ) = 1 − P F, ( α P , T P L ) . Finally, calculating P ( C ) .In order to obtain the expression of P ( C ) , we assume thatthe connected link is s ∈ { L, N } and the greatest interferencelink is i ∈ { L, N } . In this case, the connected one is SBSand the greatest interference is MBS. P C ( α P , T PL ) = P ( α p H S, L S ( d )(1 − α p ) H S, L S ( d ) + I S + I M + σ S > T PL )= P ( H S, > T PL α p − (1 − α p ) T PL · I S + I M + σ S L S ( d ) )= X s ∈{ L,N } ,i ∈{ L,N } P s,i ( T PL α p − (1 − α p ) T PL ) , (16) where I S = X X s ∈ Φ S,s \ B C S,s H S,s d − α S,s
S,s + C S, ¯ s H S, ¯ s d − α S, ¯ s S, ¯ s I M = m X Xs ∈ Φ M \ X (1) C M,i H M,i d − α M,i
M,i + C M, ¯ i H M, ¯ i d − α M, ¯ i M, ¯ i , (17) Referring to [17], we can obtain: P s,i ( T ) = N S,s X n =1 ( − n +1 (cid:18) N S,s n (cid:19) Z ∞ Z ( mCM,iCS,s x αS,sS,s ) /αM,i ( mCM,ibCS,s x αS,sS,s ) /αM,i · e − nηsxαsTσ SCS,s − Q S,n ( T,x ) − V S,n ( T,x ) · e − Q M,n ( T,x ) − V M,n ( T,x ) f ( x, R ) dRdx, (18) Q S ,n ( T, x )=2 πλ S Z ∞ x F ( N S,s , nη
S,s x α S,s TN S,s t α S,s ) p S,s ( t ) tdt, (19) Q M ,n ( T, x )=2 πλ M Z ∞ R F ( N M,s , nmC
M,i η S,s x α S,s TC S,s N M,i t α M,i ) · p M,i ( t ) tdt, (20) V S,n ( T, x )=2 πλ S Z ∞ ( CS,sCS,s x αS,s ) /αS,s F ( N S,s , nC
S,s η S,s x α S,s TC S,s N S,s t α S,s ) p S,s ( t ) tdt, (21) V M,n ( T, x )=2 πλ M Z ∞ ( CM,iCM,i R αM,i )1 /αM,i F ( N M,s , nmC
M,i η S,s x α S,s TC S,s N M,i t α M,i ) p M,i ( t ) tdt, (22) f ( x, R ) = exp[ − Λ sS ([0 , ( C S,s C S,s x α S,s ) /α S,s ])]exp[ − Λ iM ([0 , ( C M,i C M,i R α M,i ) /α M,i ])] f M,i ( R ) f S,s ( x ) , (23) f S,s ( x ) = 2 πλ s p s ( x ) x · exp( − πλ s Z x p s ( t ) tdt ) ,f M,i ( R ) = 2 πλ m p m ( R ) R · exp( − πλ m Z R p m ( t ) tdt ) . (24) B. The coverage probability of the primary signal and thesecond signal1) If the maximum interference has been successfully can-celed when decoding the primary layer signal: the SINR ofthe second layer signal can be expressed as follows:
SINR = (1 − α p ) H S L S ( d ) P X i ∈ Φ S \ B H S,i L S ( d i ) + m P X i ∈ Φ M H M,i L S ( d i ) + σ S . (25) In this case, successful decoding of both signals can berepresented as the following events: p I Ω j + p + σ S < T PL !| {z } A ∩ X (1) I Ω j + p + σ S ≥ T PL !| {z } B ∩ p I Ω j + p + σ S ≥ T PL !| {z } C ∩ p I Ω j + σ S ≥ T SL !| {z } D . (26) As already mentioned above, in general, C ⊂ B , mean-while p is bigger than p because α p is bigger than 0.5and T P L is smaller than T SL , therefore SIN R B /T P L >SIN R D /T SL , that is D ⊂ B . And we found that the value of SIN R A is smallest. Also when α p ≤ T PL (1+ T SL ) T SL + T PL (1+ T SL ) , wecan get SIN R C /T P L > SIN R D /T SL and if else, we canget SIN R C /T P L < SIN R D /T SL Therefore, the successfulprobability can be expressed as: P PSL = P ( ABCD ) = P ( CD )= ( P ( A ) − P ( C ) , α p ≤ T PL (1+ T SL ) T SL + T PL (1+ T SL ) P ( A ) − P ( D ) , α p > T PL (1+ T SL ) T SL + T PL (1+ T SL ) . (27) Similar to the solution of P ( C ) P ( D ) = P SL ( α P , T PL ) = P SL ( SINR > T SL )= P SL ( H S, > T SL (1 − α p ) · I S,L + I S,N + I M,L + I M,N + σ S L S ( d ) )= P SL,L ( T PL − α p ) + P SL,N ( T PL − α p ) . (28)
2) If the maximum interference need not decode whendecoding the first layer signal: the SINR of the second layersignal can be expressed as:
SIN R = p I Ω0 j + σ S In this case,successful decoding of both signals can be represented as thefollowing events: p I Ω j + p + σ S ≥ T PL !| {z } M ∩ p I Ω j + σ S ≥ T SL !| {z } N . (29) imilar the successful probability can be expressed as: P ( MN ) = ( P ( M ) , α p ≤ T PL (1+ T SL ) T SL + T PL (1+ T SL ) P ( N ) , α p > T PL (1+ T SL ) T SL + T PL (1+ T SL ) , (30) the solution of P ( M ) , P ( N ) is similar to Eq. (5) in case 1 C. The average rate of users
Suppose that we can successfully decode the NOMA pri-mary layer with threshold T P L and the NOMA second layerwith threshold T SL The average rate of users can be expressedas: R ave = Z ∞ T PL P ( SINR PL = T ) log(1 + T ) dT + Z ∞ T SL P ( SINR SL = T ) log(1 + T ) dT. (31) D. The average quality of experience
The most widely used is the ”mean Opinion Score” (MOS)proposed by the International Telecommunications Union(ITU) to evaluate the user’s quality of experience (QoE).It divides the subjective perception of QoE into five levels.And according to Weber-Fechner’s law, we know that therelationship between the degree of physical stimuli and itsperceived intensity presents a logarithmic characteristic inmany scenarios. So that we can use this property to studythe evaluation of QoE [18]. Referring to [19], the expressionof MOS can be expressed in the following form:
MOS ( θ ) = , θ ≤ θ a log θb , θ < θ < θ , θ ≥ θ , (32) where a = 3 . / log( θ /θ ) , b = θ ( θ /θ ) / . . Thus theaverage service quality can be expressed as: MOS ave = MOS PL ( P PL ( α P , T PL ) − P PSL ( α P , T PL , T SL ))+ MOS
PSL P PSL ( α P , T PL , T SL ) . (33) And
M OS
P L = M OS ( T P L ) , M OS P SL = M OS ( T P L + T SL ) . IV. R ESULTS AND D ISCUSSIONS
In this section, the coverage probability of the primarylayer, the coverage probability of both layers, the averageuser rate, and the average QoE under different power allo-cation radios are analyzed. In this simulation, some defaultparameters are configured with reference to [20], [21] :MBS power P tm = 36 dBm, SBS power P ts = 26 dBm,bias factor b = 15, reference path loss C M,N = 10 − . , C M,L = 10 − . , C S,N = 10 − . , C S,L = 10 − . ; pathloss exponents: α M,N = 4 . , α M,L = 2 . , α S,N = 3 . , α S,L = 2 . ; β M = 0 . , β S = 0 . ; BS density λ M = 10 − BSs/m , λ S = 10 − BSs/m ; small-scaleattenuation parameter N ML = 3 , N MN = 2 , N SL = 3 , N SN = 2 ; noise power σ S = − dBm.The effect of the coverage probability with different ratethreshold is evaluated in Fig. 2. The results present that ouranalytical values are basically consistent with the simulation C o v e r age P r oba li t y α p = 0.5, Analytical probality α p = 0.5, Simulation probality α p = 0.7, Analytical probality α p = 0.7, Simulation probality α p = 0.9, Analytical probality α p = 0.9, Simulation probality Fig. 2. The analytical value and simulation value of the coverage probabilityof the primary layer. C o v e r age P r oba li t y α p = 0.5, Analytical probality α p = 0.5, Simulation probality α p = 0.7, Analytical probality α p = 0.7, Simulation probality α p = 0.9, Analytical probality α p = 0.9, Simulation probality Fig. 3. The analytical value and simulation value of the coverage probabilityof both layers when R pl = 0 . . results. As the figure shows, the coverage probability ofprimary layer decreases gradually with the increase of R pl .Furthermore, the more power is allocated to the primary layer,the higher coverage probability can be obtained. However, inFig. 3, when the primary rate R pl = 0.1b/s/Hz, the secondlayer coverage probability decreases with the increase ofpower allocation ratio α p . That is because more power willbe assigned to second Layer if α p decreases so that thesecond layer signal becomes easier to be decoded, whichmakes it easy to get all information of both layers. Therelationship between the coverage probability and the powerallocation ratio in Fig. 2 and Fig. 3 is contrary. Therefore,we should balance the coverage probability of both layers,that is, balance the good channel conditions MUs’ QoE andbad channel condition MUs’ QoE, which is based on the factthat good channel conditions MUs can obtain both layersinformation while poor channel conditions MUs can onlyobtain the PL information.Fig. 4 compares the MU’s average rate in NOMA andOMA considering different power allocation ratios. In OMA,the signal is transmitted as a whole and is not divided intoseveral parts in the power domain. The results show thatNOMA scheme can improve the MU’s average rate, becauseMUs can decode signals as much as they can under differentchannel conditions. Strong MUs can decode all information,while weak MUs can only get basic information. As Fig. 4shows, the average rate decreases with the increase of second .1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.9511.051.11.151.21.25 Second Layer Rate A v e r age R a t e α p = 0.5, NOMA rate α p = 0.7, NOMA rate α p = 0.9, NOMA rateOMA rate Fig. 4. Average rate considering different power allocation ratios when R pl = 0 . . A v e r age qua li t y o f e x pe r i en c e α p = 0.5, NOMA mos α p = 0.7, NOMA mos α p = 0.9, NOMA mosOMA mos Fig. 5. Average service of experience considering different power allocationratios when R pl = 0 . . layer rate, because the second layer signal is harder to bedecoded successfully when the rate threshold is higher. It isalso found that the curve of power allocation ratio α p = 0.5is highest when the rate of SL is less than 0.3 b/s/Hz whereasthe curve of power allocation ratio α p = 0.9 is highest whenthe rate of SL exceeds 0.4 b/s/Hz. That is because when therate of the SL and the power allocation ratio is small, theSL can be allocated more power so the users have a greaterprobability to decode SL signal while the coverage probabilityof the PL may not reduce much. However, when the SL rateis relatively large and the power allocation ratio is small, thecoverage probability of the PL is reduced but the increase ofboth tiers coverage probability is not obvious.Fig. 5 depicts QoE considering different power allocationratios which proves that NOMA can improve the user’sQoE. The green curve is always above other curves whichis different from that in Fig. 4. According to the previousdefinition in Eq. 32, we describe the QoE with a logarithmicrelationship which results in that the effect of rate on the QoEgradually decreases. Besides, the results shows that the morepower is allocated to the primary layer, the QoE will be betterowing to PL, which becomes easier to be decoded and basicservice can be guaranteed. The increase of SL threshold ratemakes it difficult to decode the SL signal while the PL signalcan be easier decoded in NOMA, so the performance slightlydecreased. But in OMA, it’s hard to get all the informationwhich causes bad performance for multicast MUs uses’ QoE. In generally, NOMA scheme will improve multicast MUsaverage QoE and rate because it meets the demand of usersunder different channel conditions.V. C ONCLUSIONS
Considering the broadcast/multicast communications, atwo-tier heterogeneous network with NOMA scheme is pro-posed in this paper. Moreover, the transmission signal poweris divided into the primary layer and second layer by NOMAscheme. Furthermore, the coverage probability, average rateand average QoE are derived for a two-tier heterogeneousnetwork. Simulation results show that proposed method canincrease the quality of experience and the average rate of usersfor two-tier heterogeneous network with NOMA scheme. Ina future work, it would be interesting to explore successiveinterference cancellation technology by applying for NOMAscheme in different wireless networks.ACKNOWLEDGMENTThe authors would like to acknowledge the support fromNational Key R & D Program of China (2016YFE0133000):EU-China study on IoT and 5G (EXICITING-723227)R
EFERENCES[1] X. Ge, S. Tu, G. Mao, et al., “5G Ultra-Dense Cellular Networks,”
IEEEWireless Communications , Vol. 23, No. 1, pp.72–79, Feb. 2016.[2] X. Ge, H. Cheng, et al., “5G Wireless Backhaul Networks: Challengesand Research Advances,”
IEEE Network , Vol. 28, No. 6, pp. 6–11, Nov.2014.[3] X. Ge, B. Yang, et al., “Spatial Spectrum and Energy Efficiency ofRandom Cellular Networks,”
IEEE Transactions on Communications ,vol. 63, no. 3, pp. 1019-1030, March 2015.[4] I. Humar, X. Ge, L. Xiang, M. Jo, M. Chen and J. Zhang, “Rethinkingenergy efficiency models of cellular networks with embodied energy,”
IEEE Network , vol. 25, no. 2, pp. 40-49, March-April 2011.[5] M. Gruber and D. Zeller., “Multimedia broadcast multicast service: newtransmission schemes and related challenges,”
IEEE Commun. Mag. , vol.49, no. 12, pp. 176–181, Dec. 2011.[6] Y. Chi, L. Liu, G. Song, et al., “Practical MIMO-NOMA: Low Complex-ity and Capacity-Approaching Solution,”
IEEE Transactions on WirelessCommunications , 17(9), 6251-6264, Sept. 2018.[7] L. Liu, C. Yuen, Y. L. Guan, and Y. Li, ”Capacity-Achieving IterativeLMMSE Detection for MIMO-NOMA Systems,”
IEEE ICC , KualaLumpur, Malaysia, May 2016.[8] Z. Ding, Z. Yang, P. Fan, et al., “On the performance of nonorthogonalmultiple access in 5G systems with randomly deployed users,”
IEEESignal Process. Lett. , vol. 21, no. 12, pp. 1501–1505, Dec. 2014.[9] L. Xiang, X. Ge, C. Wang, F. Y. Li and F. Reichert, “Energy EfficiencyEvaluation of Cellular Networks Based on Spatial Distributions ofTraffic Load and Power Consumption,”
IEEE Transactions on WirelessCommunications , vol. 12, no. 3, pp. 961-973, March 2013.[10] X. Ge, S. Tu, T. Han, Q. Li and G. Mao, “Energy efficiency of smallcell backhaul networks based on GaussCMarkov mobile models,”
IETNetworks , vol. 4, no. 2, pp. 158-167, 3 2015.[11] L. Liu, Y. Chau, Y. L. Guan, Y. Li and C. Huang, “Gaussian MessagePassing Iterative Detection for MIMO-NOMA Systems with MassiveAccess,”
IEEE Globecom , Washington, DC, USA, Dec 2016.[12] S. R. Mirghaderi, A. Bayesteh, A. K. Khandani, “On the multicastcapacity of the wireless broadcast channel,”
IEEE Trans. Inf. Theory ,vol. 58, no. 5, pp. 2766–2780, May 2012.[13] A. M. C. Correia, J. C. M. Silva, N. M. B. Souto, et al., “Multi-Resolution Broadcast/Multicast Systems for MBMS,”
IEEE Transactionson Broadcasting , vol. 53, no. 1, pp. 224–234, March 2007.[14] L. Dai, B. Wang, Y. Yuan, et al., “Non-orthogonal multiple accessfor 5G: Solutions, challenges, opportunities, and future research trends,”
IEEE Commun. Mag. , vol. 53, no. 9, pp. 74–81, Sep. 2015.15] X. Zhang and M. Haenggi, “Successive interference cancellation indownlink heterogeneous cellular networks,” , Atlanta, GA, 2013, pp. 730–735.[16] J. G. Andrews, F. Baccelli and R. K. Ganti, “A Tractable Approachto Coverage and Rate in Cellular Networks,”
IEEE Transactions onCommunications , vol. 59, no. 11, pp. 3122–3134, November 2011.[17] T. Bai and R. W. Heath, “Coverage and Rate Analysis for Millimeter-Wave Cellular Networks,”
IEEE Transactions on Wireless Communica-tions , vol. 14, no. 2, pp. 1100–1114, Feb. 2015.[18] Lin C, Hu J, Kong X Z. “Survey on Models and Evaluation of Qualityof Experience[J],”
Chinese Journal of Computers , 2012, 35(1):1–15.[19] H. Shao, H. Zhao, Y. Sun, et al., “QoE-Aware Downlink User-CellAssociation in Small Cell Networks: A Transfer-matching Game Theo-retic Solution With Peer Effects,”
IEEE Access , vol. 4, pp. 10029–10041,2016.[20] X. Ge, K. Huang, C. Wang, X. Hong and X. Yang, “Capacity Analysisof a Multi-Cell Multi-Antenna Cooperative Cellular Network with Co-Channel Interference,”
IEEE Transactions on Wireless Communications ,vol. 10, no. 10, pp. 3298-3309, October 2011.[21] X. Ge et al., ”Energy-Efficiency Optimization for MIMO-OFDM Mo-bile Multimedia Communication Systems With QoS Constraints,”