Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Power Allocation Optimization for Uplink

Non-Orthogonal Multiple Access Systems


Huiling Zuo, Xiaofeng Tao
National Engineering Laboratory for Mobile Network Technologies
Beijing University of Posts and Telecommunications, Beijing, 100876, China
Email: zuohuiling@bupt.edu.cn

Abstract—Non-Orthogonal Multiple Access (NOMA) is a was studied in [6]. In [7], the authors have provided theo-
promising multiple access technique for 5G wireless communi- retical insights and algorithmic solutions to jointly optimize
cation networks due to its higher spectral efficiency and user power and channel allocation for downlink NOMA systems.
fairness. Power allocation is the key to improve the perfor-
mance of NOMA systems. In this paper, we focus on a certain As for uplink NOMA systems, subcarrier and power allocation
NOMA group of a hybrid NOMA system. We formulate a algorithm has been developed in [4], [8], and a user-pairing
sum-throughput maximization problem for uplink under the scheme has been proposed in [9], nevertheless, all of them
constraints of the total transmission power within a NOMA are suboptimal due to the non-convexity of the formulated
group and the minimum rate requirements of the users, which optimization problem. The work in [10] investigated a dynam-
is a convex optimization problem. The closed-form solution of
the optimal power allocation within a NOMA group is derived ic power allocation scheme to flexibly meet various quality
using the Karush-Kuhn-Tucker (KKT) conditions. According to of service requirements under the total transmission power
the numerical results, the throughput performance of a certain constraint, while it only considered NOMA scenarios with
NOMA group with our proposed power allocation scheme is two users and did not take the throughput maximization into
much better than its OMA counterpart. account.
Index Terms—Non-Orthogonal Multiple Access, power alloca-
tion, convex optimization, throughput. In the optimal NOMA scheme, there is no control on the
number of users that share each resource block, which makes
the multi-user detection (MUD) at the receiver infeasible.
I. I NTRODUCTION
Hence, considering the implementation complexity of MUD, a
The explosive traffic growth in mobile communications promising alternative is to construct a hybrid NOMA system,
has motivated research activities to design the next genera- in which users in one cell can be divided into small groups,
tion (5G) of mobile communication networks that can offer where NOMA is used within each group and orthogonal
significant improvement in coverage, spectral efficiency and resource blocks, such as subbands or time slots, can be used
user experience [1]. In conventional wireless communication to distinguish NOMA groups. The user grouping scheme for
networks, the multiple access schemes are mainly based on the hybrid NOMA system has been studied in [6], [9], [11],
orthogonal multiple access (OMA), e.g. time-division multiple [12], showing that the performance gain of hybrid NOMA over
access (TDMA) and orthogonal frequency division multiple conventional OMA can be further enlarged by pairing users
access (OFDMA) in current 4G [2]. However, towards 5G, with more distinctive channel conditions into a NOMA group,
non-orthogonal multiple access (NOMA) has attracted a lot i.e., the distinctness of channel gains among different users
of attention in both academia and industry as one of the within a NOMA group is crucial. The total transmission power
key techniques. The key idea of NOMA is to exploit the constraint is an important criterion in hybrid NOMA systems,
channel gain differences among users so that multiple users where users are paired to perform NOMA and inter-pair
could be multiplexed in transmission power domain while interference is eliminated through conventional interference
using the same resource blocks [3]. Successive interference management techniques. The use of the total transmission
cancellation (SIC) is applied at the receiver to separate the power constraint within a NOMA group is therefore useful
superimposed signals of different users. Therefore, NOMA can to control interference between NOMA groups.
offer a significant improvement in both spectrum efficiency In this paper, we mainly consider a certain NOMA group
and user fairness compared to the conventional OMA [4]. of a hybrid NOMA system. We formulate a sum-throughput
Recently, a wide range of research has been carried out maximization problem for uplink under the constraints of
about NOMA, while the majority of the existing works focused the total transmission power within a NOMA group and the
on downlink. In [5], closed-form expressions for ergodic minimum rate requirements of the users, which is a convex
sum-rate and outage probability were derived for a two-user optimization problem. And then, the closed-form solution
downlink NOMA system considering static power allocation. of the optimal power allocation within a NOMA group is
The impact of user pairing on the performance of two down- derived using the KKT conditions. We find that in order to
link NOMA systems, i.e., NOMA with fixed power allocation maximize the sum-throughput of the NOMA group under the
(F-NOMA) and cognitive-radio-inspired NOMA (CR-NOMA) constraints of the total transmission power and the minimum

978-1-5386-2062-5/17/$31.00 ©2017 IEEE


rate requirement of each user, the optimal power allocation B. Problem Formulation
scheme within the NOMA group is to allocate power for all
In this paper, we aim to maximize the sum-throughput
users except the user with the highest channel gain, so that
through power control under the constraints of the total
the rates of those users are just equal to their required data
transmission power within a NOMA group and the minimum
rates. Then, the rest of the maximum total transmission power
rate requirements of the users. Assume that the maximum total
is allocated to the user with the highest channel gain. The
transmission power within the m-user NOMA group for uplink
numerical results show that the throughput performance of
is Pt . The total transmission power constraint is useful to
a certain NOMA group with our proposed power allocation
manage the inter-group interference. In addition, the minimum
scheme is much better than its OMA counterpart under any
data rate requirement of UEi is assumed to be Ri . Then, the
channel conditions and users’ data rate requirements.
power allocation optimization problem can be given as
The rest of this paper is organized as follows: Section
II introduces the system model for uplink NOMA and for- m
 Pi γi
mulates the optimization problem. The optimal solution of max ωB log2 (1 + 
m ), (2)
P
power allocation is derived in section III. Some numerical i=1 Pj γ j + ω
results are shown in section IV to evaluate the performance j=i+1
m

of the proposed power allocation scheme. Finally, section V
concludes the paper. subject to : Pi ≤ Pt , (3)
i=1

II. S YSTEM M ODEL A ND P ROBLEM F ORMULATION Pi γ i


ωBlog2 (1 + 
m ) ≥ Ri , i = 1, 2, ..., m (4)
Consider a single-cell hybrid NOMA system consisting of Pj γ j + ω
j=i+1
one single-antenna base station (BS) and N single-antenna
users. Note that the focus of this paper is to study the optimal III. P OWER A LLOCATION O PTIMIZATION
power allocation scheme within a certain NOMA group in
order to maximize the sum-throughput of the NOMA group, The Hessian matrix of the objective function can be written
so we assume that user grouping has been done and the detail as the product of a negative constant and a positive semi-
of user grouping scheme is out of the scope of this paper. definite matrix, and the standardized constraint functions are
Suppose that there is a NOMA group composed of m users. convex functions of P, so the aforementioned maximization
Without loss of generality, assume that the users are ordered problem is a convex problem [13]. The closed-form solution
as h1 > h2 > · · · > hm , where hi is the channel gain of UEi . of the optimal power allocation within the m-user NOMA
group is given in this section using the KKT conditions.
A. System Model The corresponding Lagrangian function for uplink power
For uplink NOMA, with SIC at the BS side, the user that is allocation optimization can be expressed as
decoded first needs to endure the inter-user interference from
all other users in the same NOMA group, while users with L(P, λ, μ)
later decoding order can benefit by cancelling the interference m m
Pi γi
from the former decoded users. For the purpose of minimizing = −ωB log2 (1 + 
m ) + λ( Pi − Pt )
the negative impact from SIC, the order of decoding should be i=1 Pj γj + ω i=1
j=i+1
in the order of decreasing channel gain of users’ links to the
m

serving BS. Therefore, users with higher channel gain will be Pi γi
+ μi [Ri − ωBlog2 (1 + 
m )]
decoded earlier and it could achieve high enough performance
i=1 Pj γ j + ω
although inter-user interference exists. As a result, the user j=i+1
with the highest channel gain experiences interference from m
 Pi γi
all other users in the same NOMA group, while the user with = −ωB (1 + μi )log2 (1 + 
m )
the lowest channel gain enjoys interference-free transmission. i=1 Pj γ j + ω
Consequently, for uplink, the achievable throughput for UEi j=i+1
m
 m

in an m-user NOMA group can be expressed as
+λ( Pi − P t ) + μ i Ri (5)
Pi γi i=1 i=1
ri = ωBlog2 (1 + 
m ), ∀i = 1, 2, ..., m (1)
Pj γj + ω We can write the KKT conditions as follows
j=i+1

where γi = Nh0iB is the normalized channel gain of UEi , N0 λ∗ ≥ 0, (6)


is the power spectral density of the additive white Gaussian μ∗i ≥ 0, i = 1, ..., m (7)
noise (AWGN), ω is the number of subbands assigned for m

the NOMA group, B is the transmission bandwidth of each λ ( Pi∗ − Pt ) = 0, (8)
subband, Pi is the transmission power for UEi . i=1
Pi∗ γi
μ∗i [Ri − ωBlog2 (1 + 
m )] = 0, ∀i = 1, ..., m Based on the above expression, we can derive that μ∗3 > 0.
Pj∗ γj +ω Following similar steps as above, we can come to a conclusion
j=i+1 as
(9) μ∗i > 0, i = 2, ..., m, (16)
∂L ∗ 1 + μ∗i
= λ − ωBγ i [ 
m So, (12) is derived according to (9) and (16).
∂Pi∗
Pj∗ γj + ω
j=i Theorem 1: The closed-form solution of the optimal power
i−1
 allocation for uplink within the m-user NOMA group can be
Pj∗ γj (1 + μ∗j )
− 
m 
m ] = 0, ∀i = 1, ..., m given as follows
j=1 ( Pk∗ γk + ω)( Pk∗ γk + ω) m

Rj
k=j k=j+1 ω j=i+1 Ri

(10) Pi∗ = 2 ωB (2 ωB − 1), i = 2, ..., m, (17)


γi
m

Lemma 1: The optimal solution satisfies the following P1∗ = Pt − Pi∗ (18)
formulas i=2
m
 ∗
Proof. Based on (12), when i=m, the optimal solution of Pm
Pi∗ = Pt , (11)
can be expressed as
i=1
Pi∗ γi ∗ ω Rm
ωBlog2 (1 +  ) = Ri , i = 2, ..., m, (12) Pm = (2 ωB − 1), (19)
m γm
Pj∗ γj + ω ∗
j=i+1 When i=m-1, the optimal solution of Pm−1 can be expressed
as
Proof. Based on (10), when i=1, we have ∗ ω Rm Rm−1
Pm−1 = 2 ωB (2 ωB − 1), (20)
γm−1
ωBγ1 (1 + μ∗1 )
λ∗ = 
m > 0, (13) Following similar steps as above, (17) and (18) can be derived
Pi∗ γi + ω according to (11).
i=1
According to Lemma 1, we can find that in order to
So, (11) is derived according to (8) and (13).
∂L ∂L ∗ ∗ maximize the sum-throughput of the NOMA group under the
Since ∂P ∗ = ∂P ∗ = 0, the relation between μ1 and μ2 can
2 1 constraints of the total transmission power and the minimum
be expressed as rate requirement of each user, the optimal power allocation
m
 scheme within the NOMA group is to allocate power for all
( Pi∗ γi + ω)γ2 μ∗2 users except the user with the highest channel gain, so that
i=1 the rates of those users are just equal to their required data
m
 rates. Then, the rest of the maximum total transmission power
= [P1∗ γ1 γ2 + γ1 ( Pi∗ γi + ω)]μ∗1 is allocated to the user with the highest channel gain. Thus,
i=2
m it can be seen that the user with the highest channel gain

+(γ1 + γ2 )( Pi∗ γi + ω), (14) contributes most to the sum-throughput of a NOMA group,
i=2 the power of which should be as high as possible.
In fact, the above conclusion not only applies to uplink
Based on the above expression, we can derive that μ∗2 > 0. but also applies to downlink with the same scenario and
∂L ∂L ∗
Similarly, since ∂P ∗ = ∂P ∗ = 0, the relation between μ1 and
3 1 constraints. For downlink case, the same conclusion could be
μ∗3 can be expressed as derived following similar steps as the uplink case, and we do
m
 m
 not describe the downlink case in detail due to limited space.
( Pi∗ γi + ω)( Pi∗ γi + ω)γ3 μ∗3
i=1 i=2
IV. N UMERICAL R ESULTS
m
 m
 In this section, we take a 3-user NOMA group for example
= [P1∗ γ3 ( Pi∗ γi + ω) + P2∗ γ3 ( Pi∗ γi + ω) to evaluate the sum-throughput performance of the proposed
i=3 i=1 power allocation scheme. The 3-user OFDMA-based system
m
 m
 without power control is chosen for performance comparison.
+( Pi∗ γi + ω)( Pi∗ γi + ω)]γ1 μ∗1
In our simulation, the bandwidth of each subband is 180 KHz,
i=2 i=3
m m and there are 3 units of subband assigned for the 3-user NOMA
 
+(γ1 − γ3 )( Pi∗ γi + ω)( Pi∗ γi + ω) group. The maximum total transmission power for uplink is
i=2 i=3 28 dBm.
m
 As described in [6], the channel quality of the user with a
+P2∗ γ3 (γ1 + γ2 )( Pi∗ γi + ω), (15) poor channel condition is critical to the performance of NOMA
i=2 systems, so first, we suppose that the channel condition of
It can be observed from the simulation results that, the
uplink sum-throughput performance of NOMA is always better
than that of OMA under any channel conditions and users’
data rate requirements, and a larger sum-throughput gain over
OMA can be achieved for more distinct channel conditions
of users in the same NOMA group. NOMA could adapt to
scenarios with more dynamic channel conditions and various
quality of service requirements. In addition, the user with the
lowest channel gain does not considerably affect the sum-
throughput performance of NOMA, unless its channel gain
is so low that a huge proportion of total transmission power
is allocated for it to meet its rate requirements, which results
in obvious decay of sum-throughput. Analogously, the channel
variations of the users except the user with the highest channel
gain and the user with the lowest channel gain have little
impact on the sum-throughput performance of NOMA, unless
Fig. 1. Sum-throughput performance of 3-user NOMA group. Normalized their required rates are so high that less power could be
channel gains of UE1 and UE2 are 40 dB and 30 dB, respectively. allocated for the user with the highest channel gain which
contributes most to the sum-throughput of the NOMA group.
Note that the conventional OMA is unable to operate under
such circumstances, especially when the rate requirements of
the user with the lowest channel gain is high.
V. C ONCLUSION
Power allocation is the key issue to improve the perfor-
mance of NOMA systems. In this paper, we focus on a certain
NOMA group of a hybrid NOMA system. We investigate the
sum-throughput maximization problem for uplink under the
constraints of the total transmission power within a NOMA
group and the minimum rate requirements of the users. The re-
sulting problem is convex, and the closed-form solution of the
optimal power allocation within a NOMA group is given. The
numerical results shows that the sum-throughput performance
of a certain NOMA group with our proposed power allocation
scheme is much better than its OMA counterpart under various
channel conditions and users’ data rate requirements.
Fig. 2. Sum-throughput performance of 3-user NOMA group. Normalized ACKNOWLEDGMENT
channel gains of UE1 and UE3 are 40 dB and 20 dB, respectively.
This work is supported by Natural Science Foundation of
China (No.61461136002).
UE3 is variable. Fig.1 shows the sum-throughput of a 3- R EFERENCES
user NOMA group and its corresponding OMA system under
[1] ”5G radio access: Requirements, concepts and technologies,” NTT DO-
different constraints of the minimum rate requirements. The COMO, Inc., Tokyo, Japan, 5G White Paper, Jul. 2014.
normalized channel gain of UE1 and UE2 are 40 dB and [2] E. Dahlman, S. Parkvall, and J. Skold, 4G: LTE/LTE-advanced for mobile
30 dB respectively, when the normalized channel gains of broadband. Academic Press, 2013.
[3] X. Chen, A. Benjebbour, A. Li and A. Harada, ”Multi-User Proportional
UE1 and UE2 are 40 dB and 30 dB respectively, and the Fair Scheduling for Uplink Non-Orthogonal Multiple Access (NOMA),”
normalized channel gain of UE3 is less than 30 dB. When 2014 IEEE 79th Vehicular Technology Conference (VTC Spring), Seoul,
not all the minimum rate requirements of the three users 2014, pp. 1-5.
[4] M. Al-Imari, P. Xiao and M. A. Imran, ”Receiver and resource allo-
could be satisfied, the sum-throughput is set as zero. Then, cation optimization for uplink NOMA in 5G wireless networks,” 2015
we suppose that the channel condition of UE2 is variable. International Symposium on Wireless Communication Systems (ISWCS),
Fig.2 shows the sum-throughput of a 3-user NOMA group and Brussels, 2015, pp. 151-155.
[5] Z. Ding, Z. Yang, P. Fan and H. V. Poor, ”On the Performance of Non-
its corresponding OMA system under different constraints of Orthogonal Multiple Access in 5G Systems with Randomly Deployed
the minimum rate requirements, when the normalized channel Users,” in IEEE Signal Processing Letters, vol. 21, no. 12, pp. 1501-
gains of UE1 and UE3 are 40 dB and 20 dB respectively, and 1505, Dec. 2014.
[6] Z. Ding, P. Fan and H. V. Poor, ”Impact of User Pairing on 5G Nonorthog-
the normalized channel gain of UE2 is less than 40 dB and onal Multiple-Access Downlink Transmissions,” in IEEE Transactions on
greater than 20 dB. Vehicular Technology, vol. 65, no. 8, pp. 6010-6023, Aug. 2016.
[7] L. Lei, D. Yuan, C. K. Ho and S. Sun, ”Power and Channel Allocation
for Non-Orthogonal Multiple Access in 5G Systems: Tractability and
Computation,” in IEEE Transactions on Wireless Communications, vol.
15, no. 12, pp. 8580-8594, Dec. 2016.
[8] M. Al-Imari, P. Xiao, M. A. Imran and R. Tafazolli, ”Uplink non-
orthogonal multiple access for 5G wireless networks,” 2014 11th In-
ternational Symposium on Wireless Communications Systems (ISWCS),
Barcelona, 2014, pp. 781-785.
[9] S. Chen, K. Peng and H. Jin, ”A suboptimal scheme for uplink NOMA
in 5G systems,” 2015 International Wireless Communications and Mobile
Computing Conference (IWCMC), Dubrovnik, 2015, pp. 1429-1434.
[10] Z. Yang, Z. Ding, P. Fan and N. Al-Dhahir, ”A General Power Allocation
Scheme to Guarantee Quality of Service in Downlink and Uplink NOMA
Systems,” in IEEE Transactions on Wireless Communications, vol. 15, no.
11, pp. 7244-7257, Nov. 2016.
[11] M. S. Ali, H. Tabassum and E. Hossain, ”Dynamic User Clustering
and Power Allocation for Uplink and Downlink Non-Orthogonal Multiple
Access (NOMA) Systems,” in IEEE Access, vol. 4, no. , pp. 6325-6343,
2016.
[12] J. Guo, X. Wang, J. Yang, J. Zheng and B. Zhao, ”User Pairing and
Power Allocation for Downlink Non-Orthogonal Multiple Access,” 2016
IEEE Globecom Workshops (GC Wkshps), Washington, DC, 2016, pp.
1-6.
[13] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge, U.K.:
Cambridge Univ. Press, 2004.

You might also like