Escolar Documentos
Profissional Documentos
Cultura Documentos
Peru
November 08, 2017
Motivation and Objective
System Model
We consider a downlink MU-MIMO cellular network that jointly supports NOMA and
D2D underlaid users as shown in Fig. 1.
The BS has N antennas with power PMBS . M cellular users (CUs) are randomly deployed,
each equipped with one antenna. Furthermore, a total number of P underlaid D2D users
(DUs) are also randomly deployed.
Beam N
Beam 1
CU .. . ... ...
CU
N ,1
1,1
CU N Antennas K
1, K ... CU
N,
N
w1
w
h1,1 hp , p
D 2 D Link 1
D 2 D Link P CUs
h2,2 D 2 D TX
D 2 D Link 2 MBS D 2 D RX
System Model
For beam n, NOMA allows a set of n = {u(n, 1), u(n, 2), . . . , u(n, K )} CUs to be
scheduled on the same radio resource simultaneously, K 2. We use u(n, k) to denote
the CU that is served by beam n with NOMA sequence k in that beam. Assume xn is the
transmitted signal in the n-th beam, and according to NOMA, xn is a superimposed
signal of a total K users in beam n,
K q
X
xn = u(n,k) Pn su(n,k) . (1)
k=1
Beam N
Beam 1
CU .. . ... ...
CU
N ,1
1,1
CU N Antennas K
1, K ... CU
N,
N
w1
h1,1 hp , p
D 2 D Link 1
D 2 D Link P CUs
h2,2 D 2 D TX
D 2 D Link 2 MBS D 2 D RX
System Model
In the above equation, E(|su(n,k) |2 ) = 1, E(.) is the expectation function. u(n,k) is the
fraction of the allocated power to user u(n, k), Kk=1 u(n,k) = 1. Pn is the total
P
transmitted power for beam n. The total transmission power of a BS is equally
partitioned among N beams, i.e., Pn = PMBS N
, where PMBS is the total BS transmission
power.
At the MBS, a precoding scheme is applied to support MU-MIMO. We denote the
precoding matrix as W, which consists of N vectors, i.e.,
W = [w1 , w2 , . . . , wN ], (2)
System Model
P p
X N
X
yDUp = PD hp0 ,p sp0 + hp wn xn + np , (4)
p 0 =1 n=1
SIC process
1 Within a NOMA group, CU with a weaker channel is normally allocated a higher
downlink transmission power to ensure QoS.
2 UEs with strong channel gain can always decode the weaker UEs message then
subtract from the composite signal.
3 The decoding process will continue until the UE get its desired signal. This is called
successive interference cancellation (SIC).
where the second term on the right side is the interference from users in the same
NOMA group. The third term represents inter-beam interference.
N
X
U
Iu(n,k) = Pn0 |hu(n,k) wn0 |2 , (8)
n0 =1,n0 6=n
P
X
D
Iu(n,k) = PD |hp,u(n,k) |2 , (9)
p=1
(11)
Prof. Rose Qingyang Hu IEEE Communications Society Distinguished
Key Wireless
Lecturer
Access Technologies in 5G and IoT Systems
A NOMA and MU-MIMO Supported Cellular Network with Underlay D2D
Communications Problem Formation
The design objective is to maximize the total system sum throughput from both CUs and
DUs. To this end, we need to determine 1) the NOMA set of each beam, i.e., n ; 2) the
power allocation factor u(n,k) for each user k in the NOMA set of beam n; and 3) the
precoding vector wn . Therefore, the problem can be formulated as follows.
N X
X K P
X
max f (E{u(n,k) }) + f (E{DUp }) (12)
n ,wn ,u(n,k)
n=1 k=1 p=1
subject to
K
X
u(n,k) = 1, n = 1, 2, . . . , N, (13)
k=1
First ZF Precoding
In this scheme, we first select one user from each beam and then generate the
beamforming matrix based on N selected users. Specifically, users with the largest
channel gain in each beam are selected. The channel gain vector for these N selected
CUs are denote as H = [hu(1,K ) , hu(2,K ) . . . hu(N,K ) ]. The zero-forcing beamforming vector
is calculated based on:
hu(n,K ) wm = 0, if m 6= n. (17)
Thus, wm should lie in the null space of Hn . Here, Hn is defined as
which consists of downlink channel vectors for CUs from all beams except from beam n.
Second ZF Precoding
U
The first ZF based method helps reduce inter-beam interference Iu(n,K ) = 0 in (6). Since
we aim to maximize the total sum rate in the system, the total throughput from DUs
contributes to the total throughput as well. Therefore, thePsecond precoding method
helps reduce the interference between CUs and DUs, i.e., N 2
n=1 Pn |hp wn | = 0 in (10).
Hence we should set hp wn = 0, for all n. Or equivalently,
wn = null(HD ), (19)
User Selection
Criteria
I NOMA would prefer to group users with greater channel differences.
I Precoding matrix W is designed to minimize inter-beam interference or CU to DU
interference.
I NOMA groups users with highly correlated channels so that using the precoding
matrix generated by the representative CU in each beam can achieve a small
inter-beam or CU-DU interference
Therefore, the criteria for NOMA user grouping is to choose CUs with highly correlated
channels but with big channel gain differences in each beam. For simplicity, we set
K = 2, which means each NOMA group consists of 2 users. In each NOMA pair, we
denote the user with a weaker channel gain as the first user while the stronger one as the
second user.
For the first ZF scheme, the bemforming matrix is designed based on the null space of
the second users in all N beams, second users will not receive any inter-beam
interference. Thus the SINR is
u(n,2) Pn |hu(n,2) |2
u(n,2) = D
. (20)
Iu(n,2) + n2
The first users, on the other hand, will receive non-zero inter-beam interference as the
precoded signals from other beams will have components projected into the first user
signal space. The SINR is expressed as
subject to
0 < u(n,2) < 1, (23)
f (E{u(n,1) }) R0 . (24)
The problem defined above is convex with respect to u(n,2) and its KKT conditions are
given as follows.
P2
k=1 f (E{ u(n,k) }) R 0 f (E{ u(n,1) })
=
, (25)
u(n,2) u(n,2)
R0 f (E{u(n,1) })|u(n,2) 0, (26)
0, (27)
R0 f (E{u(n,1) })|u(n,2) = 0. (28)
Equation (25) is the stationarity condition and is KKT multiplier, (26) is the primal
feasibility, (27) is dual feasibility and (28) is the complementary slackness. Solving for
(25), we can get
(ID2 + 1) (ID1 + 1 + )H2 (1 + )H1
u(n,2) = , (29)
H1 H2 ( ID2 )
H1 + ID1 + 1 + ID1 + 1 +
u(n,2) = . (30)
2R0 H1 H1
u(n,1) = 1 u(n,2) . (31)
Simulation Results
We present the performance results from simulation. The coverage area of MBS is a
circle with a radius of 500 m. The number of transmit antennas is N = 3. The total
numbers of CUs and DUs are M = [8, 16, 32, 60, 90] and P = 2 respectively. M varies in
order to study the multi-user diversity effect. The distance with each DU pair is fixed at
30 m. PMBS and PD are set to 30 Watt and 1 Watt, respectively.
For comparison purpose, instead of using NOMA in each beam, we apply a traditional
TDMA scheme here to support these 2 users in each beam. Specifically, we allocate an
equal number of time slots to 2 TDMA users. The scheme is also referred as Naive
TDMA.
1
RTDMA = log(1 + 1 ) + log(1 + 2 ) . (32)
2
Fig. 2 presents the system capacity of two proposed ZF precoding methods as the
number of users grows, the results are scaled over the highest achievable rate. Here we
set R0 = 0.5 b/s/Hz.
100
90
80
System Throughput (%)
70
60
50
40
30 NOMA, ZF1
Naive TDMA, ZF1
20 NOMA, ZF2
Naive TDMA, ZF2
10
15 30 45 60 75 90
Number of CUs
Figure 2: System capacity of two proposed ZF precoding methods vs. Naive TDMA as the
number of user grows (R0 = 0.5 b/s/Hz).
1 NOMA outperforms naive TDMA in both precoding schemes when the number of
CUs is large. However, when the number is small, limited CUs can be chosen to
perform NOMA, thus, the performance gain is not obvious, even worse than TDMA.
2 ZF2 leads to a higher overall system throughput than ZF1. With ZF2, DUs
experience a much lower interference than with ZF1.
3 The system benefits more from NOMA+MU-MIMO due to a higher multiuser
diversity gain.
100
90
80
System Throughput (%)
70
60
50
40
30 NOMA, ZF1
Naive TDMA, ZF1
20 NOMA, ZF2
Naive TDMA, ZF2
10
15 30 45 60 75 90
Number of CUs
In Fig. 3, the throughput of CUs is calculated. ZF1 has a much better performance than
ZF2 since ZF1 precoding eliminates inter-beam interference for CUs while ZF2 aims to
eliminate interference from CUs to DUs. But if we combine results from both Fig. 2 and
Fig. 3, we can see that the overall throughput is higher with ZF2 since DUs are
configured with a very good channel setting so that they contribute to overall throughput
significantly.
100
90
80
70
CUs Throughput (%)
60
50
40
30
20
CUs with NOMA, ZF1
CUs with Naive TDMA, ZF1
10 CUs with NOMA, ZF2
CUs with Naive TDMA, ZF2
0
15 30 45 60 75 90
Number of CUs
Figure 3: CUs capacity of two proposed ZF precoding methods vs. Naive TDMA as the number
of user grows (R0 = 0.5 b/s/Hz).
System Model
In this paper we consider a multi-cell uplink cellular network with underlaid D2D
communications, as shown in figure below. We assume that CUEs are uniformly
distributed in each cell and DUEs follows a PPP distribution.
CUE
D2D pair
Signal
Interference
Power Control
For CUE
Denoted as r the distance from a typical CUE to its associated BS, the uplink transmit
power can be expressed as Pc = r , where > 2 is pathloss exponent
For DUE
Denote the distance from a typical DUE to its associated BS as D. The transmit power
of DUE is Pd = D , where is a control parameter. The value of should be very
small to avoid generating excessive interference to cellular links and other D2D links.
The interference to CUE is composed of interference from CUEs in other cells and all
DUEs. We assume all interfering CUE form a PPP and their transmit power are i.i.d. To
analyze the interference from DUEs, we further partition the interfering DUEs into two
groups, i.e., same cell DUEs and other cell DUEs. The SINR and coverage probability for
CUE can be expressed as
hr Pc
SINRc = . (33)
2 + Ic + Idin + Idout
LIc (S), LI in (S), and LIdout (S) are the Laplace transform of random variable Ic , Idin and Idout
d
evaluated at S, respectively.
After some appropriate approximation, we can get the close form expression for coverage
probability of CUE
2 R2
LIc (S) = exp b (SE[Pc ]) atan( 2 ) . (35)
2 (SE[Pc ])
2R
E[Pc ] = . (36)
2+
SR 2
LI in (S) = exp 2d . (37)
d 2(1 + S)
R 2 2R
LIdout (S) = SE[Pd ] , E[Pd ] = (38)
2 +2
The interference (from other DUEs and all CUEs) to a DUE is dependent on the distance
from it to associated BS. However, the accurate dependency is hard to track. To deal
with this issue, we assume the transmit power of interfering DUEs and CUEs are i.i.d.
Although there may be a notable gap between analytical result and simulation result,
numerical results show that our analytical result still can very well provide a tight upper
bound for the DUE coverage probability. The SINR and coverage probability of DUE can
be expressed as
hPd d
SINRd = (39)
2+ Ic + Id
R
Td 2
Z
P[SINRd > T ] = P[h > ( + Ic + Id )|D]fD (D)dD. (40)
0 D
Td 2 Td Td
P[h >
( + Ic + Id )|D] = LIc ( )LId ( ). (41)
D D D
After some appropriate approximation, we can get the close form expression for coverage
probability of DUE
R 2
1 A A
P[SINRd > T ] = x exp( ) AE i ( ) . (42)
R2 x x 1
Td 2 2 2
A= 2
b E[Pc ] + d E[Pd ] . (43)
sinc( )
2
2 R2 2 R2
E[Pd ] = , E[Pc ] = .
2 2
Numerical Evalution
Under the proposed scheme the CUE coverage in the system with underlay DUEs (blue
dots) is almost the same as the CUE coverage with no underlay DUEs (red dots), which
clearly shows that the existence of DUEs only has a slight impact on CUEs if distance
based power control is applied.
A higher value allows for a higher transmit power from DUEs, leading to a better DUE
coverage.
1 1
sim,uniform sim-=0.001
0.9 ana 0.9 ana-=0.001
sim,no d2d,uniform sim-=0.01
sim,no d2d,ppp
0.8 0.8 ana-=0.01
Coverage Probability
Coverage Probability
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
-10 -5 0 5 10 15 20 -10 -5 0 5 10 15 20 25 30
SINR (dB) SINR (dB)
Numerical Evalution
The distance based power control scheme can effectively protect CUEs from being
interfered by DUEs. With = 0.001, there is barely any impact on CUEs from DUEs with
power control while at the same time about 80 percent DUEs have SINR above 0 dB.
1
PC-=0.001
noPC-=0.001
0.9 PC-=0.005
noPC-=0.005
0.8 PC-=0.01
noPC-=0.01
noD2D
Coverage Probability
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-10 -5 0 5 10 15 20
SINR (dB)
Figure 7: Coverage Probability for CUE. The result for no power control cases and no D2D case
are provided to compare with our proposed scheme
Numerical Evaluation
0.9
0.8
Coverage Probability
0.7
0.6
0.5
0.4
0.3 PC-=0.001
noPC-=0.001
0.2 PC-=0.005
noPC-=0.005
0.1 PC-=0.01
noPC-=0.01
0
-10 -5 0 5 10 15 20 25 30
SINR (dB)
Figure 8: Coverage Probability for DUE. The result for no power control cases are provided to
compare with our proposed scheme
Motivation
I Spectrum below 5 GHz is extremely crowded
I Support skyrocketing traffic growth in 5G wireless cellular networks
I Exploring higher radio spectrum is imperative
Advantages of mmWave
I Spectrum resource is abundant
I Interference from nearby mmWave nodes can be small due to shorter transmission
distance and the directional beamforming technique
I mmWave Communications can be more secure than radio frequency (RF)
communications since the mmWave signals cannot penetrate walls and other
non-transparent objects
Challenges
I Coverage could be limited due to high pathloss
I mmWave signals are very sensitive to blockage effects compare with the low
spectrum RF signals
System Model
System Model
BK
LO
SL
INK
NLOS LINK RN
BS
UE
Blockage model
I Given a link between a UE and a mBS or a RN with distance r , the probability that
the link has LOS is given as
P(LOS|r ) = e r , (44)
where
2i (E [L] + E [W ])
= , i {m, r }. (45)
E [L] and E [W ] are the average length and width of blockages, respectively
I The NLOS probability given a distance is simply P(NLOS|r ) = 1 e r
I The probability of LOS decreases when the length of link increases, while the
probability of NLOS increases when the length of link increases
Pathloss model
Association rule
I UE is associated to the nearest mBS or RN. The pdf of the distance r between a UE
and its nearest mBS or RN can be written as
2
friu (r ) = 2i re i r , i {m, r } (48)
I When a UE experiences NLOS link to its associated BS, it can switch to the nearest
RN to establish a two-hop route to the BS
I A RN is connected the BS that is closest to this RN
Directional beamforming
I All the mBSs have the same transmit power Pm and all the RNs have the same
transmit power Pr
I Rayleigh fading (g ) with zero mean is assumed for channel fading
I General downlink received signal power is P G g PL1
Pi Gi Gu giu PL1
LOS,riu
SNRLOS,riu = , i {m, r }, (50)
N
where riu represents the distance between a BS or a RN and a UE, giu is the
Rayleigh fading, PLLOS,riu is the LOS pathloss as in (46) but in linear scale, and N is
the noise power
I Given a UE served by a NLOS mBS or a NLOS RN, its SNR is similar to (50), but
replacing PLLOS,riu with the NLOS pathloss PLNLOS,riu
I For a UE that is associated to its nearest mBS or RN, the probability that a LOS
link exists can be expressed as
Z Z
2
Piu (LOS) = PLOS|r fr (r )dr = e r 2i re i r dr , i {m, r } (51)
Pi Gi Gu giu PL1
!
LOS,riu
pc,LOS,iu = Er [P(SNRLOS,riu > T )|rLOS ] = Er [P > T |rLOS ]
N
Z TNPLLOS,r
TNPLLOS,riu P G G iu
= Er [P giu > |rLOS ] = e i i u friu ,LOS (r )dr , i {m, r },
Pi Gi Gu r >0
(55)
I The overall SNR distribution of a single tier mmWave network with no relays is
I In an ultra dense mmWave network, interference from other mBSs or RNs cannot be
ignored
I Assume there is no interference between mBS and RN and there is no interference
between different RNs
I For a UE associated to a mBS, only these co-channel mBSs cause interference on
the downlink. For a UE that is connected to a RN, only co-channel RNs in other
cells can cause interference
I RNs in the same cell does not cause interference by properly provisioning orthogonal
radio resources for the RNs in the same cell. The density of the interfering RNs is m
I Given a UE served by the nearest mBS or RN and the link is LOS, its SINR can be
expressed as
giu riu L
SINRLOS,riu = 2 , i {m, r }. (59)
N ( 4 )
Ii + Pi Gi Gu
Ii is the normalized interference either from other mBSs or from other RNs.
!1
X PLrju
Ii = PI gju 2 , i {m, r }. (60)
4
ji
m u
PI = 4 2
, i is the set of all interfering mBSs or all interfering RNs.
I Given a typical UE served by the nearest mBS or RN and the link is LOS, the
coverage probability when considering interference can also be derived as
giu riu L
pc,I ,LOS,iu = Er [P(SINRLOS,riu > T )|rLOS ] = Er [P 2 > T |rLOS ]
N ( 4 )
I + Pi Gi Gu
2
!
N 4
= Er [P giu > T (I + )r L |rLOS ]
Pi Gi Gu iu
2 L
Z
TN ( 4
) iu
r
= e Pi Gi Gu
LI (Triu L )friu ,LOS (r )dr , i {m, r }, (62)
r >0
where LI (Triu L ) is the Laplace transform of random variable I evaluated at Triu L ,
and LI (Tr L ) can be written as
R
(11/(1+Tr L PI v L ))e v vdv
LI (Tr L ) = LILOS (Tr L )LINLOS (Tr L ) = e 2m r
R
(11/(1+Tr L PI v N ))(1e v )vdv
e 2m r (63)
I Given a typical UE served by the nearest mBS or RN and the link is NLOS, the
coverage probability with considering interference can be expressed as
I The overall SINR distribution of a single tier mmWave network with no relays can
be expressed as
where pc,I ,LOS,mu and pc,I ,NLOS,mu are from (62) and (64), respectively. Pmu (LOS)
and Pmu (NLOS) are from (51) and (52), respectively
I The overall coverage of a relay-assisted mmWave network can be expressed as
Notation Value
Pm , Pr 27 dBm
Gm , Gr 30 dBi
Gu 5 dBi
m , r , u 10o
LOS 2.1
NLOS 3.4
LOS 3.6 dB
NLOS 9.7 dB
Thermal noise -174 dB per Hz
Noise figure 7 dB
BW 1 GHz
1
No relay
0.9 With relays
0.8
0.7
0.6
Probability
0.5
0.4
0.3
0.2
0.1
0
20 0 20 40 60 80
SNR Threshold (dB)
Figure 10: SNR performance comparison between the single tier mmWave network with no relays
and the relay-assisted mmWave network.
I m = 1
1002
= 3.1831 105 , b = 100 m , r = 10 m , and u = 200 m
I Blockages are assumed to be squares with E [L] = E [W ] = 2 m
I Pmu (LOS) = 0.52 and Pru (LOS) = 0.8
I The performance of the network with RNs is much better than the performance of
the network without RNs
Prof. Rose Qingyang Hu IEEE Communications Society Distinguished
Key Wireless
Lecturer
Access Technologies in 5G and IoT Systems
Performance Study on Relay-Assisted Multi-hop Millimeter Wave Networks Numerical Results
1
Theoretical result
0.9 Simulation result
0.8
0.7
0.6
Probability
0.5
0.4
0.3
0.2
0.1
0
20 0 20 40 60 80
SNR Threshold (dB)
Figure 11: Theoretical and simulation SNR curves of the relay-assisted network.
0.9
0.8
0.7
0.6
Probability
0.5 No RN, b = 50m
Figure 12: SNR comparison between the relay-assisted network and the single tier network with
different blockage density.
I Without RNs, the SNR performance degrades significantly when the blockage
density increases
I With RNs, when the blockages increase, the number of LOS links decreases in a
much slower pace
I Adding RNs into the mmWave network can significantly improve the overall system
coverage and SNR performance, as well as greatly enhance the system robustness
against blockages
Prof. Rose Qingyang Hu IEEE Communications Society Distinguished
Key Wireless
Lecturer
Access Technologies in 5G and IoT Systems
Performance Study on Relay-Assisted Multi-hop Millimeter Wave Networks Numerical Results
1 No relay
With relays, r = 5m
With relays, r = 10m
0.8
With relays, r = 20m
With relays, r = 100m
0.6
Probability
0.4
0.2
0
20 0 20 40 60 80
SNR Threshold (dB)
Figure 13: SNR comparison among relay-assisted network with different RN density.
1
No relay
0.9 With relays
0.8
0.7
0.6
Probability
0.5
0.4
0.3
0.2
0.1
0
20 0 20 40 60 80
SINR Threshold (dB)
Figure 14: SINR performance comparison between no relay network and relay-assisted network.
I m = 1
202
= 7.9577 104 , b = 50 m , r = 10 m , and u = 200 m
I Blockages are assumed to be squares with E [L] = E [W ] = 1 m
I Pmu (LOS) = 0.4502 and Pru (LOS) = 0.7609
I The performance of the network with RNs is much better than the performance of
the network without RNs
1
Theoretical result
0.9 Simulation result
0.8
0.7
0.6
Probability
0.5
0.4
0.3
0.2
0.1
0
20 0 20 40 60 80
SINR Threshold (dB)
Conclusions
I NOMA and MU-MIMO together can greatly improve spectral efficiency when proper
cooperation (precoding) and NOMA power are selected
I Interference from D2D underlay cellular system may be significantly reduced through
proper power control and beam-forming
I NOMA, D2D, MU-MIMO, mmWave are considered promising key technologies for
future 5G/IoT systems
References
References
[1] B. Xie, Z. Zhang, Rose Qingyang Hu, G. Wu, Joint Spectral Efficiency and Energy
Efficiency in FFR based Wireless Heterogeneous Networks, submitted to IEEE
Transactions on Vehicular Technology, 2016.
[2] F. Zhou, Y. Wu, Rose Qingyang Hu, Y. Wang, K. Wong, Energy-Efficient NOMA
Heterogeneous Cloud Radio Access Networks: Enabling Techniques and Challenges,
submitted to IEEE Wireless Communications Magazine, 2016.
[3] L. Feng, H. Yang, Rose Qingyang Hu, J. Wang, mmWave and VLC based Indoor
Channel Models in 5G Wireless Networks, submitted to IEEE Wireless
Communications Magazine, 2016.
[4] C. Yang, J. Li, Rose Qingyang Hu, J. Xiao, Distributed Optimal Cooperation for
Spectral and Energy Efficiency in Hyper-Dense Small Cell Networks, to appear in
IEEE Wireless Communications Magazine, 2016.
[5] L. Feng, Rose Qiangyang Hu, J. Wang, P. Xu, and Y. Qian, VLC in 5G Wireless
Networks: Architecture and Key Technologies, IEEE Network Magazine, November
2016.
[6] Y. Xu, Rose Qingyang Hu, Y. Qian, T. Znati, Quality-based Spectrum and Energy
Efficient Mobile Association in Wireless Heterogeneous Networks, IEEE
Transactions on Communications, Vol.64, No.2, pp.805-817, 2016.
[7] H. Zhang, C. Jiang, Rose Qingyang Hu, Y. Qian, Self-Organization in Disaster
Resilient Heterogeneous Small Cell Networks, IEEE Network Magazine, Vol.30, No.
2, pp.116-121, 2016.
Prof. Rose Qingyang Hu IEEE Communications Society Distinguished
Key Wireless
Lecturer
Access Technologies in 5G and IoT Systems
References
[15] Rose Qingyang Hu, Y. Qian, Heterogeneous Cellular Networks, John Wiley & Sons,
Ltd., 2013.
[16] Q. Li, Rose Qingyang Hu, Y. Qian, and G. Wu, Cooperative communications for
wireless networks: techniques and applications in LTE-advanced systems, IEEE
Trans. Wireless Commun., vol. 19, No. 2, pp. 21-29, 2012.
[17] Z. Zhang, Rose Qingyang Hu, Uplink Non-Orthogonal Multiple Access with
Fractional Power Control, submitted to IEEE WCNC 2017.
[18] Z. Zhang, H. Sun, Rose Qingyang Hu, Y. Qian, Stochastic Geometry Based
Performance Study on 5G Non-Orthogonal Multiple Access Scheme, in Proceedings
of IEEE Globecom 2016.
[19] Z. Zhang, Rose Qingyang Hu, Y. Qian, D2D Communication Underlay in Uplink
Cellular Networks with Distance Based Power Control, in Proceedings of IEEE ICC
2016.
[20] B. Xie, Z. Zhang, Rose Qingyang Hu, Performance Study on Relay-Assisted
Millimeter Wave Cellular Networks, in Proceedings of IEEE VTC 2016 Spring.
[21] H. Sun, Y. Xu, Rose Qingyang Hu, A NOMA and MU-MIMO Supported Cellular
Network with Underlaid D2D Communications, in Proceedings of IEEE VTC 2016
Spring.
[22] H. Sun, B. Xie, Rose Qingyang Hu, Y. Qian, G. Wu, Non-orthogonal Multiple
Access with SIC Error Propagation in Downlink Wireless MIMO Networks, in
Proceedings of IEEE VTC 2016 Fall.
Prof. Rose Qingyang Hu IEEE Communications Society Distinguished
Key Wireless
Lecturer
Access Technologies in 5G and IoT Systems
References
Thank You !