See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/330547058

Sum-Rate Maximization of NOMA Systems under Imperfect Successive

Interference Cancellation

Article  in  IEEE Communications Letters · January 2019

DOI: 10.1109/LCOMM.2019.2893195

CITATIONS READS
0 25

4 authors, including:

Islam Abu Mahady Salama Ikki

Lakehead University Thunder Bay Campus Newcastle University
3 PUBLICATIONS   11 CITATIONS    84 PUBLICATIONS   2,250 CITATIONS   

SEE PROFILE SEE PROFILE

Halim Yanikomeroglu
Carleton University
411 PUBLICATIONS   8,747 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Use of aerial-BSs for increasing the QoS of 5G Network View project

Cognitive radio systems View project

All content following this page was uploaded by Islam Abu Mahady on 26 January 2019.

The user has requested enhancement of the downloaded file.

 IEEE COMMUNICATIONS LETTERS 1

Sum-Rate Maximization of NOMA Systems Under

Imperfect Successive Interference Cancellation
Islam Abu Mahady , Ebrahim Bedeer , Salama Ikki , and Halim Yanikomeroglu

1 Abstract— This letter addresses the sum-rate maximization for the PGS scheme which assumes independent real and imag- 40
2 a downlink non-orthogonal multiple access (NOMA) system in the inary signal components with equal power, the IGS scheme 41
3 presence of imperfect successive interference cancellation (SIC). loosens these constraints and introduce a circularity coefficient 42
4 We assume that the NOMA users adopt improper Gaussian
5 signaling (IGS), and hence derive new expressions of their rates that enables a more general Gaussian signaling scheme [7]. 43

6 under residual interference from imperfect SIC. We optimize To the best of the authors’ knowledge, there is no existing 44

7 the circularity coefficient of the IGS-based NOMA system to work in the literature that exploits IGS in an effort to maximize 45

of
8 maximize its sum-rate subject to quality-of-service requirements. the overall sum-rate of the NOMA system under the practical 46
9 Compared to the NOMA with proper Gaussian signaling, simula- assumption of imperfect SIC, which motivates us to develop 47
10 tion results show that the IGS-based NOMA system demonstrates
11 considerable sum-rate performance gain under imperfect SIC. this work. In particular, new closed-form expressions for 48

the users’ rate are derived for a downlink two-user NOMA 49

12 Index Terms— Improper Gaussian signaling, non-linear opti- system in the presence of imperfect SIC. Using the derived 50
13 mization, NOMA, sum-rate, imperfect successive interference
14 cancellation. expressions, an optimization problem is formulated to optimize 51

the circularity coefficient to maximize the overall NOMA 52

15

16

17

18

19

20

21
N
I. I NTRODUCTION
ON-ORTHOGONAL multiple access (NOMA) proposes

streams from multiple users together and allow them to

transmit simultaneously using the same frequency/time/code
ro
the adoption of power/code domain to multiplex signal

resources [1]. One of the advantages of NOMA systems is that

when the available resource blocks are assigned to weak chan-
sum-rate subject to minimum rate requirements constraints.
Simulation results show a considerable sum-rate performance
gain when using IGS-based NOMA systems compared with
PGS-based NOMA systems.

II. P RELIMINARY: I MPROPER R ANDOM V ECTORS

In this section, preliminary of IGS definitions are presented
to ease the understanding of the derivation of information rates
53

54

55

56

57

58
EP
59
22 nel users, they can still be accessed by other strong channel of NOMA users. 60
23 users, which qualifies NOMA techniques to achieve a higher A complex random variable (RV) is called proper if 61
24 overall spectral efficiency (SE) [2], [3]. However, NOMA its pseudo-variance is equal to zero, otherwise it is called 62
25 techniques achieve this potential higher SE considering perfect improper [8]. For a complex RV xi , we use Cxi and Ĉxi 63
26 successive interference cancellation (SIC) (see, e.g., [1]–[4], to denote the covariance and pseudo-covariance, respectively. 64
27 and the references therein). In real scenarios, the assumption Then for the zero-mean input Gaussian signal xi , ∀i, we have 65
28 of perfect SIC at the receiver might not be practical, since there Cxi = E[xi x∗i ], Ĉxi = E[xi xi ], and the impropriety degree 66
29 still remain several serious implementation problems by using of xi is given as 67
30 SIC, e.g., error propagation and complexity scaling [1]. In [5],
31 a unified framework is presented assuming imperfect SIC, κxi = |Ĉxi |/Cxi , ∀i, (1) 68
IEE

32 which shows that the performance converges to an error floor where 0 ≤ κxi ≤ 1. If κxi = 0, we say that xi is proper, and if 69

33 at the high signal-to-noise ratio (SNR) region and obtain a zero κxi = 1, we have maximally improper signal. Note that Cxi 70

34 diversity order. Hence it is of great interest to compensate the is nonnegative real number equal to the power value of the 71

35 impact of imperfect SIC for the NOMA systems. transmitted signal, while Ĉxi is complex number in general. 72

36 Recent research works have shown that improper Gaussian

37 signaling (IGS) has the potential over proper (conventional) III. S YSTEM M ODEL AND R ATE A NALYSIS 73

38 Gaussian signaling (PGS) to enhance the overall achievable We consider a downlink NOMA system with two users 74

39 rate of systems that suffer from interference [6]. Compared to (strong channel user and weak channel user) and a base 75

station. The channel coefficient between the base station and 76

Manuscript received December 26, 2018; accepted January 11, 2019. The user i is denoted by hi , ∀i = 1, 2, that is modelled as a 77
associate editor coordinating the review of this paper and approving it for
publication was F. Wang. (Corresponding author: Islam Abu Mahady.) complex Gaussian RV with zero-mean and variance σh2 i . The 78

I. Abu Mahady and S. Ikki are with the Electrical Engineering Department, noise at the receivers ends are modelled as zero-mean additive 79
Lakehead University, Thunder Bay, ON P7B 5E1, Canada (e-mail: white Gaussian random variable with variances σn2 . Different 80
iabumah@lakeheadu.ca; sikki@lakeheadu.ca).
E. Bedeer is with the School of Engineering, Ulster University, from the conventional setup where PGS is assumed, in this 81

Newtownabbey BT37 0QB, U.K. (e-mail: e.bedeer.mohamed@ulster.ac.uk). work, user’s 1 signal x1 and user’s 2 signal x2 are zero-mean 82
H. Yanikomeroglu is with the Department of Systems and Computer complex Gaussian RVs which can be improper. Without loss 83
Engineering, Carleton University, Ottawa, ON K1S 5B6, Canada (e-mail:
halim@sce.carleton.ca). of generality, it is assumed that |h1 |2 > |h2 |2 , i.e. user 1 is 84

Digital Object Identifier 10.1109/LCOMM.2019.2893195 with strong channel gain and user 2 is with weak channel gain. 85

1558-2558 © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 2 IEEE COMMUNICATIONS LETTERS

86 According to the NOMA principle, the transmit power of the case of both users adopt IGS, reduces to 130

87 weak user’s signal must be greater than that of the strong user,
R1 (κs1 , κs2 )
i.e., P2 > P1 . Hence, user 2 decodes directly its signal because
131
88  
the interference inflicted by the user 1 is small and can thus be P1 |h1 |2
89
= log2 1 + 2 132
treated as noise. In contrast, user 1 can decode its own signal β P2 |h1 |2 + σ 2
90
 

91 after cancelling the weak user’s decoded signal through a SIC proper
detector [1]. We assume that SIC process at user’s 1 receiver  
92
1 |h21 P1 κs1 |2 + |β 2 P22 |h1 |2 κs2 |2
is imperfect and the residual interference component due to + log2 1 − 133
2 (P1 |h1 |2 + β 2 P2 |h1 |2 + σ 2 )2
93

94 this imperfection is quantified by a factor β (0 ≤ β ≤ 1),  

95 where β = 0 referes to perfect SIC and β = 1 refers to the 

Improper

fully imperfect SIC. 1 |β 2 P 2 |h1 |2 κs2 |2
− log2 1 − 2 2 2
96
. (12) 134
97 The input-output relationship for the two-user SISO system 2 (β P2 |h1 | + σ 2 )2
98 can be expressed as  

Improper

of
  Similarly, by substituting (8), (9), and (10) into (11), the
99 y1 = P1 h1 s1 + β P2 h1 s2 + n1 , (2) 135

  achievable rate expression for the weak user 2 of a NOMA 136

100 y2 = P2 h2 s2 + P1 h2 s1 + n2 , (3) system, in the case of both users use IGS, reduces to 137
 
where si is ith signal and ni is AWGN at the corresponding P2 |h2 |2
101
R2 (κs1 , κs2 ) = log2 1 + 138
receivers. P1 |h2 |2 + σ 2
102
 

103 In the following, we derive the rate expressions for the gen- Proper
104

105

106

107

108

109
√ IGS for for both users, i.e., x1 and x2 are improper.
eral case of
Let xi = Pi si , ∀i = 1, 2 are the independent signals for user
1 and 2, respectively, and denote the covariance and pseudo-
covariance of the transmit signal by

Cxi = Pi Csi ,
Ĉxi = Pi Ĉsi , ∀i = 1, 2,
ro (4)
(5)
1
2


1
2


+ log2 1 −


− log2 1 −
|P2 h22 κs2 |2 + |P1 h22 κs1 |2
(P2 |h2 |2 + P1 |h2 |2 + σ 2 )2


|P1 h22 κs1 |2

(P1 |h2 |2 + σ 2 )2



Improper
Improper

.


(13)
139

140
EP
It is worth noting that each Ri , i = 1, 2, in (12), (13) includes 141
110 where Csi = E[si s∗i ] and Ĉsi = E[si si ]. We assume that two parts; proper and improper. Substituting κsi = 0, ∀i = 1, 2 142
111 Csi = E[si s∗i ] = 1, ∀i = 1, 2, i.e., transmit a symbol with a and β = 0 reduces to the rates of PGS case in perfect SIC, 143
112 unit power. Next, we derive the rate expressions in terms of which proves the correctness of the derived expressions. 144
113 circularity coefficient. The covariance and pseudo-covariance
114 of yi , i = 1, 2, can be obtained from (2) and (3) as IV. O PTIMIZATION P ROBLEM 145

In this section, an optimization problem is formulated to

Cy1 = P1 |h1 |2 + β 2 P2 |h1 |2 + σ12 ,
146
115 (6)
optimize the IGS circularity coefficient in order to maximize 147
116 Ĉy1 = P1 κs1 h21 + β 2 P2 κs2 h21 , (7) the sum-rate of a two-user SISO NOMA system subject 148

Cy2 = P2 |h2 |2 + P1 |h2 |2 + σ22 , (8) to minimum rate requirements of each user. Due to space
IEE

117 149

118 Ĉy2 = P2 κs2 h22 + P1 κs1 h22 . (9) limitations, we focus on the case where we use IGS for 150

strong user (i.e., x1 is improper and κs1 = 0) and PGS 151

119 Define the noise and the interference-plus-noise terms in (2) for weak user (i.e., x2 is proper and κs2 = 0). Other cases 152

120 and (3), as zi , i =√1, 2, at each receiver, respectively, where will be investigated in future work. Also, we assume the 153

121 z1 = n1 and z2 = P1 h2 s1 + n2 , we get powers P1 and P2 are already allocated to user 1 and 2, 154

respectively (i.e., they are not optimization variables). That 155

said, the optimization problem for maximizing the sum-rate

Cz1 = σ12 , Ĉz1 = 0, Cz2 = P1 |h2 | 2
+σ22 , Ĉz2 = P1 κs1 h22 .
156
122 and
under QoS constraints can be formulated as 157
123 (10)
maximize R1 (κs1 ) + R2 (κs1 ) 158
κs 1
124 Following [8], the achievable rate expression for a two-user subject to C1 : R1 (κs1 ) ≥ Rmin1 , 159
125 SISO system is given as [6]
C2 : R2 (κs1 ) ≥ Rmin2 , 160
 
1 Cy2i − |Ĉyi | 2 C3 : 0 ≤ κs1 ≤ 1, (14) 161

126 Ri = log2 . (11)

2 Cz2i − |Ĉzi |2 where R1 (κs1 ) and R2 (κs1 ) are obtained from (12) and (13), 162

respectively, at κs2 = 0. Rmin1 and Rmin2 are the minimum 163

127 By substituting (6), (7), and (10) into (11), and assuming rate requirements of the strong user and the weak user, 164

128 without loss of generality σ12 = σ22 = σ 2 , the achievable rate respectively. The constraint C3 reflects that the circulatory 165

129 expression for the strong user 1 of a NOMA system, in the coefficient is between 0 and 1 as shown in Definition 2. 166
 ABU MAHADY et al.: SUM-RATE MAXIMIZATION OF NOMA SYSTEMS UNDER IMPERFECT SIC 3

167 The optimization problem in (14) can be solved by applying

168 the Karush-Kuhn-Tucker (KKT) conditions; however, it is
169 worthy to mention that the obtained circularity coefficient κ∗s1
170 will be sub-optimal as the problem in (14) is non-convex. The
171 Lagrangian function can be expressed as
172 L(κs1 , λ1 , λ2 )
173 = − (R1 (κs1 ) + R2 (κs1 )) + λ1 (Rmin1 − R1 (κs1 ))
174 + λ2 (Rmin2 − R2 (κs1 )), (15)
175 where λ1 and λ2 are the non-negative Lagrange multipliers
176 associated with the QoS constraints of user 1 and 2, respec-
177 tively. The circularity coefficient constraint not considered
in the Lagrangian function will be satisfied later. That said,

of
178
Fig. 1. Sum-rate vs SNR for IGS-based and PGS-based NOMA systems for
179 the KKT conditions can be written as follows [9] different β.
∂L(κ∗s1 , λ1 , λ2 ) The proposed algorithm to solve the problem in (14) can
180 = 0, (16) 210

∂κs1 be formally summarized as follows. 211

181 λ1 (Rmin1 − R1 (κ∗s1 )) = 0, (17) 1) Input: Rmin1 , Rmin2 , P1 , P2 , h1 , h2 , σ 2 , and β. 212

182 λ2 (Rmin2 − R2 (κ∗s1 )) = 0, (18) 2) Set λ1 = λ2 = 0. Calculate κs1 from (22). Calculate R1 213

Rmin1 − R1 (κ∗s1 ) ≤ 0, (19) and R2 from (12) and (13), respectively. 214

ro
183
3) if R1 ≥ Rmin1 and R2 ≥ Rmin2 , then, the sub-optimal 215
184 Rmin2 − R2 (κs∗1 ) ≤ 0, (20) solution κ∗s1 is reached. 216

185 λ1 , λ2 ≥ 0. (21) 4) else if R1 < Rmin1 and R2 ≥ Rmin1 , then, find non- 217

negative λ1 from (23) such that R1 (κs1 ) = Rmin1 and

κ∗s1
218
186 From (16), we can obtain the circularity coefficient
re-calculate κs∗1 from (22). Repeat until convergence. 219
187 as in (22), shown 2 at the bottom of this page, where
σ2 P2 σ2
2 5) else if R1 ≥ Rmin1 and R2 < Rmin1 , then, find non- 220
188 Φ = 1 + P1 |h2 |2 , Ψ = 1 + P1 + P1 |h2 |2 , and Ω = negative λ2 from (23) such that R2 (κs1 ) = Rmin2 and
 2
221

re-calculate κs∗1 from (22). Repeat until convergence.

EP
σ2
189 1 + β2 P 2
P1 + P1 |h1 | 2 . To consider C3, we need to guarantee 222

6) else R1 < Rmin1 and R2 < Rmin1 , then, find 223

190 that the term under the square root in (22) is positive and also
non-negative λ1 and λ2 from (23) if exists such that 224
191 the first term of (22) is greater than the second term of it.
R1 (κs1 ) = Rmin1 and R2 (κs1 ) = Rmin2 and re-
The values of λ1 and λ2 in (22) can be computed using the
225
192
calculate κs∗1 from (22). Repeat until convergence. 226
193 subgradient method [9] as follows.
7) Output: κs∗1 .
+
227

194 λl+1
i = λli − αli (Ri − Rmini ) , ∀i = 1, 2, (23)
V. S IMULATION R ESULTS 228

195 where [.]+ is defined as max(., 0) √

and αi is a sufficiently small In this section, we simulated a downlink two-user served by 229

196 step size chosen to equals 0.1/ l where l is the iteration a base station in a NOMA system employing IGS and compare 230

197 number [9]. However, one can notice from (17) that either its achieved sum rate (R1 + R2 ) to its counterpart of PGS- 231
IEE

198 λ1 = 0 or R1 (κ∗s1 ) = Rmin1 . Similarly, (18) implies that based NOMA systems. Unless otherwise mentioned, Rmin1 = 232

199 either λ2 = 0 or R2 (κ∗s1 ) = Rmin2 . That said, four possible Rmin2 =1.2 bits/sec/Hz, P1 = 0.3 PT , and P2 = 0.7PT . 233

200 cases exist, as follows. In Fig. 1, the sum-rate is simulated versus SNR = PσT2 , 234
n
201 – Case 1: λ1 = 0 and λ2 = 0 means that both QoS where PT is the total transmit power, at different values of β. 235

202 constraints of user 1 and user 2 are inactive. As can be seen, the IGS-based NOMA system outperforms 236

203 – Case 2: λ1 = 0 and λ2 = 0 implies that the sub-optimal PGS-based NOMA for all levels of imperfect SIC. In particu- 237

204 circularity coefficient exists when R2 (κ∗s1 ) = Rmin2 . lar, as the SIC becomes worse, i.e., β = 0.4, the sum-rate gain 238

205 – Case 3: λ1 = 0 and λ2 = 0 implies that the sub-optimal of using IGS increases over PGS NOMA. IGS also offers a 239

206 circularity coefficient exists when R1 (κ∗s1 ) = Rmin1 . good gain in the low SNR region as the effect of the imperfect 240

207 – Case 4: λ1 = 0 and λ2 = 0 implies that if the problem SIC is significant on the users’ rate. At high SNR, the 241

208 is feasible, the sub-optimal circularity coefficient exists when PGS-based NOMA system approaches the sum-rate perfor- 242

209 both R1 (κ∗s1 ) = Rmin1 and R2 (κ∗s1 ) = Rmin2 . mance of the IGS-based NOMA system. In addition, we use 243

  
1 + λ2
κ2s1 = 0.5 (Φ + Ψ) + (Φ − Ψ)
1 + λ1
   2     12
1 + λ2 1 + λ2
− 0.5 (Φ − Ψ)2 1+ + (Φ − Ψ) (2(Φ + Ψ) − 4Ω) . (22)
1 + λ1 1 + λ1
 4 IEEE COMMUNICATIONS LETTERS

of
Fig. 2. Sum-rate vs SNR for IGS-based and PGS-based NOMA systems for Fig. 4. Sum-rate vs number of iterations for algorithm convergence with
different P1 , P2 values, with β = 0.3. β = 0.2, 0.3, 0.4.

needs small number of iterations to converge and the number 269

increases as β increases. 270

VI. C ONCLUSION

ro
271

In this work, we optimized the sum-rate of a two-user 272

NOMA system subject to minimum QoS requirements under 273

imperfect SIC. An iterative algorithm was developed to find 274

the sub-optimal IGS circularity coefficient that maximizes the 275

sum-rate. Simulation results showed that IGS improves the 276

sum-rate at low-to-medium SNR region. In addition, it was 277

observed that the gain from IGS increases when imperfect

EP
278

SIC gets higher. Results also revealed that the power allocation 279

2 does not affect the gain of IGS-based NOMA systems over its 280
Fig. 3. Sum-rate vs SNR for IGS-based NOMA system for different σh 1
to
2 ratios, with β = 0.3, P = 0.1P , and P = 0.9P .
σh
PGS-based counterpart. Moreover, sum-rate increases when 281
2
1 T 2 T
channel gain ratio between users increases, but this improve- 282

ment saturates at high SNR values. 283

244 exhaustive search method to find the optimal solution and
245 compare it with the proposed KKT sub-optimal solution. The R EFERENCES 284

246 results show that there is a small performance gap between the [1] M. Zeng, A. Yadav, O. A. Dobre, G. I. Tsiropoulos, and H. V. Poor, 285
247 optimal solution and proposed sup-optimal solution in terms “Capacity comparison between MIMO-NOMA and MIMO-OMA with 286

248 of sum-rate at low SNR values and the gap tends to zero at multiple users in a cluster,” IEEE J. Sel. Areas Commun., vol. 35, no. 10, 287
pp. 2413–2424, Oct. 2017. 288
IEE

249 high SNR values. It is worthy note that the proposed solution [2] Y. Liu, Z. Qin, M. Elkashlan, A. Nallanathan, and J. A. McCann, “Non- 289
250 is far less complex than the optimal solution of the exhaustive orthogonal multiple access in large-scale heterogeneous networks,” IEEE 290

251 search. The figure also shows that in case of perfect SIC, J. Sel. Areas Commun., vol. 35, no. 12, pp. 2667–2680, Dec. 2017. 291
[3] L. P. Qian, Y. Wu, H. Zhou, and X. Shen, “Joint uplink base sta- 292
252 i.e., β = 0, both schemes perform similarly. tion association and power control for small-cell networks with non- 293
253 In Fig. 2, the sum-rate vs SNR for different values of P1 , P2 orthogonal multiple access,” IEEE Trans. Wireless Commun., vol. 16, 294

254 at β = 0.3 is simulated for both IGS and PGS NOMA system. no. 9, pp. 5567–5582, Sep. 2017. 295
[4] Y. Sun, D. W. K. Ng, Z. Ding, and R. Schober, “Optimal joint power and 296
255 As the strong user gains more power, i.e., P1 becomes larger, subcarrier allocation for full-duplex multicarrier non-orthogonal multiple 297
256 the sum-rate curves shift up and show higher sum-rate. One access systems,” IEEE Trans. Commun., vol. 65, no. 3, pp. 1077–1091, 298

257 can notice from Fig. 2 that different power allocation ratios do Mar. 2017. 299
[5] X. Yue, Z. Qin, Y. Liu, S. Kang, and Y. Chen, “A unified framework for 300
258 not affect the gain of IGS over PGS based NOMA systems. non-orthogonal multiple access,” IEEE Trans. Commun., vol. 66, no. 11, 301
259 The effect of users’ channel strength on the sum-rate for pp. 5346–5359, Nov. 2018. 302

260 IGS-based NOMA performance is shown in Fig. 3. The sum- [6] Y. Zeng, C. M. Yetis, E. Gunawan, Y. L. Guan, and R. Zhang, “Transmit 303
optimization with improper Gaussian signaling for interference chan-
rate is simulated for the case of σh2 1 = {1, 2, 3, 5, 7, 9}σh2 2 at
304
261
nels,” IEEE Trans. Signal Process., vol. 61, no. 11, pp. 2899–2913, 305
262 P1 = 0.1 PT , P2 = 0.9 PT , and β = 0.3. It is clear that as σh2 1 Jun. 2013. 306

263 increases, the sum-rate enhances, i.e., as the channel of the first [7] O. Amin, W. Abediseid, and M.-S. Alouini, “Overlay spectrum sharing 307
using improper Gaussian signaling,” IEEE J. Sel. Areas Commun., 308
264 user becomes stronger, its rate becomes higher. Meanwhile, vol. 35, no. 1, pp. 50–62, Jan. 2017. 309
265 the rate of the user with weak channel is maximized by the [8] F. D. Neeser and J. L. Massey, “Proper complex random processes with 310

266 proposed approach through the IGS-NOMA concept. applications to information theory,” IEEE Trans. Inf. Theory, vol. 39, 311
no. 4, pp. 1293–1302, Jul. 1993. 312
267 In Fig. 4, we show the convergence of the proposed algo- [9] S. Boyd, L. Xiao, and A. Mutapcic, “Subgradient methods,” Dept. Elect. 313
268 rithm at different values of β. On average, the algorithm Eng., Stanford Univ., Stanford, CA, USA, Lecture Notes EE392o, 2003. 314

View publication stats