You are on page 1of 5

Channel Assignment with Power Allocation for Sum

Rate Maximization in NOMA Cellular Networks


*

1st Priya Gupta 2nd Debashis Ghosh


Electronics and Communication Engineering Electronics and Communication Engineering
Indian Institute of Technology, Roorkee Indian Institute of Technology, Roorkee
Roorkee, India Roorkee, India
pgupta@ec.iitr.ac.in debashis.ghosh@ece.iitr.ac.in

Abstract—The power domain non-orthogonal multiple access very low data rate while another user is demanding a high data
(PD-NOMA) method is seen as one of the advantageous, desirable rate. In case of OMA, the user with IOT user is served with
and enticing strategies for radio access that promises improved more bandwidth than it requires even though the other device
spectral efficiency in the next decade. This paper focuses on
resource utilisation problems for the NOMA downlink network needs more bandwidth. Hence, OMA is unable to fulfill the
in which the Base Station (BS) is placed at the center of requirement of broadband user. However, PD-NOMA supports
the cellular cell. Consequently, uniformly distributed cellular spectrum sharing which allows the broadband user to use the
users communicate with BS using NOMA protocol. Maximizing bandwidth that the IoT user has already used. In addition,
system throughput in such cellular networks under various superposition coding is performed at the transmitter in down-
practical constraints such as power budget and quality of service
may be constructed as a mixed-integer non-linear programming link NOMA, while SIC is performed at the receiver to service
(MINLP) optimization problem. Hence, in this research paper, multiple CU for the overloaded transmission offloading on
we have gone through a channel assignment strategy for cellular the same resource block. Thus, it enhances spectral efficiency
users and subsequently suggest a scheme for optimal power performance of the system [2].
allocation for a given assigned channel. The findings of the Mon-
tecarlo simulation demonstrate the precision and feasibility of the Related work
proposed scheme on the subject of NOMA system throughput.
Index Terms—NOMA cellular system, Channel assignment, Due to exponentially growing demand of higher data rate
Power allocation coefficient, SIC, QoS. applications, sum-rate maximization is a crucial research issue
in the domain of wireless communication. Requirement of
I. I NTRODUCTION massive connectivity and the data quench issues are solved
by resource allocation problem that deals with channel as-
N OMA is most promising and significant technology for
fifth generation (5G) wireless heterogeneous network
which needs low latency, high reliability, massive connectivity
signment and power allocation. It is not possible for OMA to
serve the requested users if the channels are already occupied.
of users with improved fairness in terms of resource distribu- On the other hand, at this emergency, NOMA may quickly
tion [1]. Orthogonal Multiple Access (OMA) techniques are help the requested users by sharing the occupied channels.
already in place but are unable to meet all the requirements Hence, OMA schemes are unable to support large number of
demanded by 5G networks. Since resources are distributed users within a fixed resource budget. A significant amount
orthogonally among users and water-filling algorithm is gen- of research work has already been devoted in this direction
erally involved in allocating the total available power to the while considering various aspects such as channel assignment
users, user fairness is not guaranteed. Water-filling algorithm and power allocation.
does not follow user fairness. Which means that a user with As seen in the literature, total throughput maximization
good channel condition can get more transmiited power in the along with efficiently solving the problem of resource alloca-
water-filling algorithm, while it is preferred in user fairness tion, such as channel and power allocation, is one essential and
that less power is assigned to a cellular user (CU) with good important research objective. Minimizing total transmit power
channel condition. In contrast to OMA, NOMA maintains user while guaranteeing the quality of service (QoS) helps real-time
fairness along with providing best solution to the spectrum services like voice and video streaming, whereas maximization
scarcity problem. The advantage of NOMA over OMA may of the aggregate users’ capacity is required in file transfer
be better understood with one example as follows. In an applications like email and web browsing services. In view of
OFDMA system, let us suppose that one IoT user requires this, maximization of system throughput and weighted sum-
rate while considering the minimum QoS of users is proposed
978-1-7281-9180-5/20/$31.002020IEEE.
c in [3]. Starting with equal distribution of initial power to all

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on December 01,2023 at 20:04:02 UTC from IEEE Xplore. Restrictions apply.
users, channel assignment and power allocation schemes are
investigated herein .

Power
Ziad et al. [4] proposed a power allocation optimization
scheme with Karush-Kuhn-Tucker (KKT) solution to maxi- Frequency
CU1’s signal
decoding
mize the throughput for NOMA system. Constraints imposed CU2’s signal decoding
SIC of
are total transmission power budget and QoS of CU. But, CU2’s signal
by considering CU1’s
signal as interference
ࢎ૚ǡ࡯ࢁ૚
the proposed method is limited to only a two-user scenario. ࢎ࢑ǡ࡯ࢁ૚
CU1
A generalised MIMO working layout for downlink NOMA CU1
and uplink NOMA is developed in [5, 6]. In recent times, ࢎ૚ǡ࡯ࢁ૛ ࢎ࢑ǡ࡯ࢁ૛
NOMA technology has also been applied in millimeter-wave CU2

systems [7]. 1
2
CU2

Channel condition of cellular users for power allocation 3 Downlink communication link
has been a matter of discussion in [8, 9]. It is well known
K
that the channel condition of a user plays an important Channel
role in wireless system. If two users are present in SISO-
NOMA system then sum-rate of NOMA can be written as Figure 1: Resource sharing strategy and interference scenario
2 2
σ 2 +p|h |2
log2 (1 + pa1σ|h
2
1|
) + log2 (1 + pa2σ|h
2
2|
) = log2 ( N σ2 2 ) (downlink sharing).
N N N
where a1 and a2 are the power assignment coefficient and p
is a total transmitted power. If we have assumed the same
channel condition, i.e., h1 = h2 , then above mathematical where transmitted symbol is shown by sn by BS. The received
expression is same as OMA [10]. In case of MIMO-NOMA signal at CUk,n can be shown as
system, different path losses are assumed for different users.
Therefore, similar channel condition can not provide a feasible |N |
region for NOMA user’s rate [11]. In [12, 13], cellular users √  √
ak,n pk hk,n sn + ak,n pk hk,n si + zk,n (2)
are arranged with respect to minimum rate requirement (QoS) i=1,i=n
rather than their channel gain.
Since the need of optimal resource allocation is increas- where hk,n = gk,n dn −α denotes the coefficient of channel
ing due to scarcity of resources in wireless communication from the BS to CU, gk,n is a random variable and follows
system. In this paper, we aim at developing a scheme for Rayleigh distribution, distance measurement between the BS
channel assignment with optimum power allocation for system and CUk,n is taken as dn , whereas the path loss exponent
throughput maximization in NOMA cellular networks with is denoted as α. The additive white Gaussian noise (AWGN)
varying channel gains for different users. The novelty of this is denoted as zk,n ∼ CN (0, σk2 ). W.l.o.g, we consider that
research work lies in optimizing the effectiveness of power users are multiplexed over the assigned channel and follow
allocation coefficient with fair distribution in NOMA system. |hk,1 |2 > |hk,2 |2 > |hk,3 |2 > . . . , > |hk,n |2 .
The remaining paper is structured as shown: Section II Considering Figure 1, we define the SINR of the nth CU
describes downlink NOMA system model description and sharing the k th channel as
suggested a channel and power assignment scheme. Section
III presents the simulation results and discussion. ak,n pk |hk,n |2
γk,n =
2 + |h 2
|N | (3)
II. S YSTEM MODEL σN k,n | pk i=1,i=n ak,i xk,i
We consider multiple CUs in our work within a single
2
cellular structure of circular and SISO system. It is assumed where σN represents noise power and assuming perfect error-
that our model is a single antenna system relying on one base free decoding which helps in reduction of error propagation
station (BS). Each CU occupies at most one channel; let the in SIC, the achievable rate can be defined as
nth CU occupies the k th channel, {n = 1, 2, . . . , |N |}, and
{k = 1, 2, . . . , |K|}. We assume that |K| < |N |, which shows Rk,n (ak,n ) = log2 (1 + γk,n ) (4)
that the cellular cell is fully loaded with active users and due
to scarcity of spectrum resources, free channel is not available. where pk is BS transmitted power on the k th channel. A joint
We also assumed that BS is completely awared about channel problem of simultaneous channel selection along with power
state information (CSI). allocation is described as
Exploiting the NOMA protocol, the transmitted superposed
signal from the BS to the nth user on k th channel can be ⎧ ⎫
⎨|K| |N |
 ⎬
obtained as
max xk,n Rk,n (ak,n ) (5)
ak,n ,xk,n ⎩ ⎭
|N | k=1 n=1
√
ak,n pk sn (1)
n=1
subject to

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on December 01,2023 at 20:04:02 UTC from IEEE Xplore. Restrictions apply.
for channel selection. Maximizing channel-wise SINR brings
|N | about less intereference among the CUs which leads to maxi-

xk,n ak,n = 1 ∀k ∈ K (5a) mization of the data rate of the assigned CUs and subsequent
n=1 total throughput maximization of a system.
ak,n ≥ 0 ∀k ∈ K, ∀n ∈ N (5b) Apart from this, we also solved the channel assignment
using the new version of cvx 2.0 solver mosek tool. This solver
log2 (1 + γk,n ) ≥ min
Rk,n ∀k ∈ K (5c)
solves this type of linear binary problem efficiently. A binary
|K|
 matrix is obtained as an output which provides the information
xk,n = 1 ∀n ∈ N (5d) about the selection of users on the channel.
k=1
xk,n ∈ {0, 1} ∀k ∈ K, ∀n ∈ N (5e) Power allocation on channel assigned CUs
Here, we have channel allocation list Cn for CU on the basis
where xk,n is an binary variable for channel selection, defined
of maximum channel-wise SINR. Thus, we are left with the
as
optimization variable i.e. ak,n for known xk,n . Consequently,
xk,n = 1, if k th channel is shared by the nth CU the problem defined in (5) can be derived as
xk,n = 0, otherwise ⎛ ⎞
Our goal is to optimise overall throughput of our downlink ⎜ ⎟
|N | ⎜ ⎟
NOMA network. Constraint (5a) gives the transmission power   ⎜ a p |h | 2 ⎟
f = max ⎜
log2 ⎜1 +
k,n k k,n ⎟ (6)
budget of the BS on each downlink channel. Eqn. (5b) indi- |N
 | ⎟
ak,n ⎜ 2 + p |h 2 ⎟
cates that each power allocation coefficient cannot be negative, k∈Cn n=1
⎝ σ N k k,n | a k,i ⎠
whereas (5c) ensures the QoS for each CU. Constraints (5d) i=1,
i=n
and (5e) show that every CU is strictly allocate at most one
channel. Summarizing, subject to
th
• CUk,n denotes n cellular user on the k th channel.
|N |
• gk,n shows the random variable which is distributed as 
Rayleigh fading coefficient of the nth CU from the BS ak,n = 1 ∀k ∈ Cn (6a)
on the k th channel. n=1

• hk,n described the channel gain coefficient of the n


th
CU ak,n ≥ 0 ∀k ∈ K, ∀n ∈ N (6b)
th
from the BS on the k channel. log2 (1 + γk,n ) ≥ min
Rk,n ∀k ∈ Cn (6c)
th
• γk,n denotes the SINR of the n CU on the k th channel.
With respect to the power allocation coefficient on each paired
• pk represents the transmitted power from the BS on the
, the1st derivative of f is given as
k th channel.
th
• ak,n shows the coefficient of the power allocation n  
|N |
th
CU on the k channel.  (pk |hk,n |2 )(σN
2
+ |hk,n |2 pk i=1,i=n ak,i )
2

• P denotes the total transmitted power from the BS.


f =
2 +
 |N |
(7)
th pk |hk,n |2 ak,n + σN |hk,n |2 pk i=1,i=n ak,i
• Rk,n denotes the data rate of n CU on k th channel.
• xk,n shows binary variable which represents presence of the 2nd derivative is
user on a channel.
 |N | 
The above eqn. (5) is a joint MINLP optimization problem  −(pk |hk,n |2 )(σN
2
+ |hk,n |2 pk i=1,i=n ak,i )
2
in nature and it is NP hard. For solving the problem, we have f =
2 +
 |N |
(8)
two sequential resource allocation approaches for a NOMA (pk |hk,n |2 ak,n + σN |hk,n |2 pk i=1,i=n ak,i )2
cellular system, as described below. 
Since f is negative. Hence, we can say that the objective
Channel assignment to CUs function is concave in nature. Constraint (7c) may also be
written as
According to the NOMA concept, users share channels non-
orthogonally where large number of CUs are accommodated
|N |
on the same channel. Hence, our task is to assign the best min 
2 2
suited channel to that particular CU in such a manner that ak,n pk |hk,n | ≥ (2 Rk,n
−1)(σN +pk |hk,n |2 ak,i ) (9)
i=1,i=n
the system throughput is maximized. W.l.o.g, we have con-
sidered |K| < |N |. Channel assigned to the CUs follow The above constraint is a linear inequality. From (7) and (8),
|hk,1 |2 > |hk,2 |2 > |hk,3 |2 which shows CU-1 has good it is obvious that our objective function is concave in nature.
channel condition than CU-2 and CU-3 on the k th channel. Hence, fmincon solver can be used to solve such type of
The channel allocation problem is just a binary linear problem efficiently. Moreover, power allocation coefficients
programming problem. Since SINR is directly proportional to are obtained by using interior point method (IPM) while sat-
data rate of CU, a performance parameter γ is considered isfying all the constraints. IPM is considered as one standard

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on December 01,2023 at 20:04:02 UTC from IEEE Xplore. Restrictions apply.
optimization technique which provides an effective method for 84
solving such a non-linear problem. This method is detailed IPM+MOSEK
82
in [14, 15, 16, 17]. It converts the power allocation problem Proposed scheme
80 OMA scheme
into a sequence of problem and then solves it iteratively to

CU Throughput (Bits/sec/Hz)
give a power allocation coefficient vector for each channel. 78

76

74
III. S IMULATION ANALYSIS AND DISCUSSION
72

This section concludes, the performance efficacy of pro- 70

posed resource allocation scheme, its analysis and verification 68

by our experimental results obtained through Monte Carlo 66


2 4 6 8 10 12 14 16 18
simulations. Simulations have been performed in MATLAB Transmit power of BS(dBm)
R2015 for over 1000 iterations. In our simulation experiment,
we have considered that the BS is placed at the center in Figure 2: CU throughput analysis with respect to transmit
a circular region with radius 300 m. The shortest distance power of BS
between the CUs is 30 m. The shortest distance between the
CU and the BS is maintained as 40 m and α is taken as 2.
88
The total power budget varies from 2 dBm to 16 dBm.The
total bandwidth is B = 1 MHz, bandwidth of each channel 86

CU Throughput (Bits/sec/Hz)
B
is |K| and the noise spectral density No = −174 dBm. QoS
min 84
threshold to be Rk,n = 0.5 bps/Hz.
In Figure 2, we plot the throughput of NOMA system with 82
respect to the transmitted power of the BS. It is observed
that the CU’s throughput increases under various schemes. We 80
IPM+MOSEK
can analyse from the figure that mosek channel assignment is
78 Proposed scheme
better than the proposed channel assignment and OMA scheme OMA scheme
but in terms of complexity, mosek solver is more complex as 76
compared to the proposed scheme. Mosek channel assignment 4 6 8 10 12 14 16 18
Number of CUs
is done by cvx tool which is an optimal approach. Througput
(sumrate) of OMA for two users can be expressed as Figure 3: CU throughput analysis with respect to number of
CU

1 Pt |(h1 )|2 1 Pt |(h2 )|2


log2 (1 + 2 ) + log2 (1 + 2 ) (10)
2 σN 2 σN Complexity analysis
The complexity for IPM is defined as γ 0.5 (γ + θ)θ2 [18],
Here this 12 factor arises due to sharing of one resource block where θ shows the number of variables. γ represents the num-
in between two users orthogonally. ber of inequality constraints. Let us assume K is total number
In Figure 3, we see how the number of CUs affect the of channel and N is the total number of users assigned on
system thoughput. As the number of CU increases, throughput channel then, the complexity of our problem can be extracted
also increases for different schemes. Throughput increases due 0.5 N 2
from above expression as O(( K N
+ 2) ( K N
+ 1)( K − 1) ).
to the CU rates getting added with respect to the number of
CUs. However, we have limited the number of CU due to the
IV. C ONCLUSION
high complexity of the cvx application, while our proposed
novel scheme has less complexity as compared to cvx solver. This research paper discussed the resource allocation prob-
From Figure 4, we see that the system throughput is affected lem for CUs in NOMA downlink system while considering
with different number of CUs at different QoS. We maintain that all CUs are in active mode. Our MINLP joint resource
constant QoS, fixed at 0.5 bps/Hz and 0.9 bps/Hz. Here one allocation problem is broken into two disjoint problems –
channel is shared by multi-user, but due to SIC, interference channel assignment followed by power allocation. The channel
among the users are less as compared to the orthogonal assignment problem is solved by selecting channel-wise SINR
schemes. Performance at low QoS is better which shows at equal power whereas power allocation problem is solved
that increasing the number of celular users also satisfying by employing fmincon solver. Downlink power budget of
the QoS requirement. The performance of mosek solver for the BS and QoS of each CU is protected while solving
channel assignment and fmincon solver for corresponding the problem of power allocation. The findings of simulation
power allocation is better for both of QoS but more complex results indicate the effectiveness and significance of system
because cvx solver is used for channel assignment. performance regarding system throughput.

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on December 01,2023 at 20:04:02 UTC from IEEE Xplore. Restrictions apply.
[14] Boyd, S., Vandenberghe, L.: ‘Convex optimization’ (Cambridge Uni-
88
versity Press, Cambridge, UK, 1st edn., 2004), pp. 243–248
87
[15] Andersen, E.D., Andersen, K.D.: ‘The MOSEK interior point optimizer
for linear programming: an implementation of the homogeneous algo-
CU Throughput (Bits/sec/Hz)
86 rithm’, in Frenk, H., Roos, K., Terlaky, T., Zhang, S. (Eds.): ‘High
performance optimization. applied optimization’, vol. 33 (Springer,
85 Boston, MA, 2000), pp. 197–232
[16] R. Gour and A. Tyagi, ”Channel selection and power allocation for
84
device-to-device enabled cellular networks,” in IET Communications,
83
vol. 12, no. 15, pp. 1854-1863, 18 9 2018.
IPM +MOSEK at R min=0.5bps/Hz [17] R. Gour and A. Tyagi, ”Cluster oriented resource allocation and
82 IPM +MOSEK at Rmin=0.9 bps/Hz power optimisation for D2D network in cellular communications,” in
Proposed scheme at Rmin=0.9bps/Hz IET Networks, vol. 9, no. 4, pp. 170-179, 7 2020, doi: 10.1049/iet-
81 net.2019.0091.
[18] D. Ni, L. Hao, Q. T. Tran and X. Qian, ”Transmit Power Minimization
80
4 6 8 10 12 14 16 18 for Downlink MultiCell Multi-Carrier NOMA Networks,” in IEEE
Number of CU Communications Letters, vol. 22, no. 12, pp.2459-2462, Dec. 2018.

Figure 4: CU throughput analysis with respect to number of


CU at different QoS

R EFERENCES

[1] J. G. Andrews et al., “What will 5G be?” IEEE J. Sel. Areas Commun.,
vol. 32, no. 6, pp. 1065–1082, Jun. 2014.
[2] Z. Wei, J. Yuan,D.W.K.Ng, M.Elkashlan, and Z.Ding.(2016). “A survey
of downlink non-orthogonal multiple access for 5G wireless communi-
cation networks.” [Online]. Available: https://arxiv.org/abs/1609.01856
[3] J.Zhu, J. Wang, Y.Huang, S.He, X. You and L.Yang, ”On Optimal
Power Allocation for Downlink Non-Orthogonal Multiple Access Sys-
tems,” in IEEE Journal on Selected Areas in Communications, vol. 35,
no. 12, pp. 2744-2757, Dec. 2017.
[4] Z. Q. Al-Abbasi and D. K. C. So, ”Power allocation for sum rate
maximization in nonorthogonal multiple access system,” 2015 IEEE
26th Annual International Symposium on Personal, Indoor, and Mobile
Radio Communications (PIMRC), Hong Kong, 2015, pp.1649-1653.
[5] Z. Ding, R. Schober, and H. V. Poor, “A general MIMO framework for
NOMA downlink and uplink transmission based on signal alignment,”
IEEE Trans. Wireless Commun., vol. 15, no. 6, pp. 4438–4454, Jun.
2016.
[6] Z. Ding, F. Adachi, and H. V. Poor, “The application of MIMO to
nonorthogonal multiple access,” IEEE Trans. Wireless Commun., vol.
15, no. 1, pp. 537–552, Jan. 2016
[7] Z. Ding, P. Fan, and H. V. Poor, “Random beamforming in millime-
terwave NOMA networks,” IEEE Access, vol. 5, pp. 7667–7681, Feb.
2017
[8] Y. Saito, A. Benjebbour, Y. Kishiyama and T. Nakamura, ”System-
level performance evaluation of downlink non-orthogonal multiple
access (NOMA),” 2013 IEEE 24th Annual International Symposium
on Personal, Indoor, and Mobile Radio Communications (PIMRC),
London, 2013, pp. 611-615.
[9] Z. Ding, Z. Yang, P. Fan and H. V. Poor, ”On the Performance of Non-
Orthogonal Multiple Access in 5G Systems with Randomly Deployed
Users,” in IEEE Signal Processing Letters, vol. 21, no. 12, pp. 1501-
1505, Dec. 2014.
[10] Z. Ding, X. Lei, G. K. Karagiannidis, R. Schober, J. Yuan and V.
K. Bhargava, ”A Survey on Non-Orthogonal Multiple Access for 5G
Networks: Research Challenges and Future Trends,” in IEEE Journal
on Selected Areas in Communications, vol. 35, no. 10, pp. 2181-2195,
Oct. 2017.
[11] M. F. Hanif, Z. Ding, T. Ratnarajah, and G. K. Karagiannidis, “A
minorization-maximization method for optimizing sum rate in the
downlink of non-orthogonal multiple access systems,” IEEE Trans.
Signal Process., vol. 64, no. 1, pp. 76–88, Jan. 2016
[12] Z. Ding, L. Dai, and H. V. Poor, “MIMO-NOMA design for small
packet transmission in the Internet of Things,” IEEE Access, vol. 4,
pp. 1393–1405, Aug. 2016
[13] Z. Ding, Z. Zhao, M. Peng, and H. V. Poor, “On the spectral effi-
ciency and security enhancements of NOMA assisted multicastunicast
streaming,” IEEE Trans. Commun., vol. 65, no. 7, pp. 3151– 3163, Jul.
2017.

Authorized licensed use limited to: MULTIMEDIA UNIVERSITY. Downloaded on December 01,2023 at 20:04:02 UTC from IEEE Xplore. Restrictions apply.

You might also like