PIC-based Receiver Structure For 5G Downlink

NOMA

Asim Anwar, Boon-Chong Seet, Xue Jun Li

Department of Electrical and Electronic Engineering
Auckland University of Technology
Auckland 1010, New Zealand
Email:{aanwar, bseet, xuli}@aut.ac.nz

Abstract—Non-orthogonal multiple access (NOMA) is considered as access (OFDMA) and single carrier frequency division
a promising multiple access scheme for 5G downlink multiple access (SC-FDMA) in downlink and uplink,
transmission. In this paper, we first review the existing downlink respectively. Despite of the fact that LTE presents 2-3 times
NOMA with successive interference cancellation (SIC) and then capacity enhancements over current 3G systems, this gain is
highlight some of the critical performance limiting factors related
inadequate to meet the capacity demands of future 5G mobile
to SIC which can result in the performance degradation of
NOMA. In order to alleviate the problems posed by SIC, we communication systems [7].
propose an alternate receiver structure for downlink NOMA Non-orthogonal multiple access (NOMA) with advanced
based on parallel interference cancellation (PIC), along with some transmission and reception schemes such as dirty paper coding
design consideration factors. The simulation results show that the (DPC) and successive interference cancellation (SIC) receivers is
proposed receiver outperforms that with SIC and hence is a considered as a promising multiple access scheme to enhance the
promising receiver structure for future 5G downlink NOMA. system performance in both uplink and downlink [1]. In
NOMA, the transmitted signals of multiple users are
Keywords-5G communications, non-orthogonal multiple access, multiplexed in the power domain at the transmitter side, and
parallel interference cancellation.
demultiplexed using SIC at the receiver side [1, 3, 5, 7]. The
I. INTRODUCTION non-orthogonality in this context means that multiple data
streams for different users are superimposed on the same
In recent years, the rapid growth in mobile data traffic
frequency band for transmission, but each user’s signal is
coupled with the massive expansion of mobile devices such as
allocated with a different power level by the base station as
smartphones and tablets has compel wireless communications
shown in Figure 1.
industry and academia researchers to define new structures and
The focus of this work is on downlink NOMA transmissions.
models to support the requirements of future 5th generation
There are two objectives of the paper. The first is to review the
(5G) mobile communications systems [1]. These include a
existing downlink NOMA with SIC at user equipment (UE)
1000x increase in cell throughput capacities over current 4th
from which we highlight some critical factors that immensely
generation (4G) mobile communication systems, wide expansion in
degrade the performance of SIC and hence adversely affect the
traffic and number of simultaneous connections, personalized user
NOMA downlink performance. The second objective is to
experience, service access from anywhere, anytime and any
propose an alternate receiver structure at UE side for downlink
device, exceptionally low end-to-end latency, and a 10x
NOMA based on parallel interference cancellation (PIC)
increase in battery life for battery-operated devices [2,3,4]. In
technique that will eradicate the problems posed by the SIC
order to transform a concept of 5G into reality, many enabling
scheme. To the best of our knowledge, PIC-based receiver has
technologies and potential solutions are proposed, among
neither been proposed nor evaluated in the context of 5G
which ultra-densification, massive multiple-input multiple-
downlink NOMA transmissions.
output (MIMO) and millimeter wave have captured the
The rest of the paper is organized as follows: Section II
attention of both industry and academia researchers [4].
reviews the related work on downlink NOMA with SIC. The
Nevertheless, appropriate architecture of multiple access
background on NOMA and the factors limiting the performance of
technology in wireless systems remains a critical factor to
SIC are discussed in Section III. The proposed PIC-based
enhance the system capacity in a cost effective manner [2, 5].
receiver structure is presented in Section IV, which is followed by
The 3rd generation (3G) mobile communication systems
results and discussion in Section V. Finally, Section VI
adopt direct sequence code division multiple access (DS-
concludes the paper.
CDMA) as the radio access technology, with single user
detection at the receiver side [6]. On the other hand, long term
evolution (LTE) established by 3rd generation partnership
project (3GPP) adopts orthogonal frequency division multiple

978-1-4673-7218-3/15/$31.00 ©2015 IEEE ICICS 2015

 II. RELATED WORK
In current literature, the downlink NOMA is investigated
with SIC receiver at the UE side. In [5, 8], the performance of
NOMA technique was evaluated with SIC receiver by
performing system level simulations. The authors reported a
performance gain of 35%-40% over OFDMA in terms of cell
throughput and user rate. However, at the receiver of some
UEs, they assumed a perfect decoding of all the higher order
users and hence neglected the effect of decoding failure and error
propagation in their analysis. Further, they assumed significant
difference in the channel gains of the NOMA users, which
facilitates differential power allocation among multiple users in
order for SIC to function correctly.
In [7], the authors considered downlink NOMA and
investigated the impact of closed loop single-user (SU) MIMO on
its performance. They proposed two error propagation (EP)
models for SIC considering the case of two users. They
presented their results in terms of attaining gains in average
cell and cell edge throughputs as compared to OFDMA, the
degradation in throughputs under the proposed EP models and
the performance gain in throughput relative to UE velocity.
However, the proposed EP models imply that by increasing the
number of users, the error will increase and propagate to the
lower order decoding users, and hence increasing their
probability of unsuccessful decoding.
The performance of downlink NOMA with randomly
deployed users was evaluated in [9] and the fairness of NOMA
scheme in 5G is evaluated in [4]. The authors in [9] showed
that NOMA is able to achieve better outage performance than
OMA, while the work of [4] reported higher achievable
maximum fairness rate in NOMA as compared to OMA for
different values of total power and number of users. The works of
both [4, 9] assumed appropriate rates and power allocation to
the users. This strict condition can limit the performance of
NOMA because SIC at the UE receiver is sensitive to power
and the decoding failure will occur if the power allocation
coefficients for the users are not chosen properly, or if there is
not a significant difference in allocated power levels between
two or more NOMA users.
III. NOMA BACKGROUND
The non-orthogonal DS-CDMA scheme used in 3G is
different from NOMA in that it uses single user detection (e.g.
Rake receiver) at the receiver side, while NOMA uses multi-
user detection based on SIC. The basic signal waveform in
NOMA is assumed to be based on orthogonal frequency
division multiplexing (OFDM), which offers strong resistance
against multipath channels, maintains good compatibility with
MIMO and preserves backward compatibility [6]. Hence,
multiple users are superimposed onto the same frequency band
but each allocated with a different power level, therefore
exploiting the power domain for multiplexing.
A. NOMA System Model with SIC
Let us consider a scenario of downlink cellular transmissions in
which a base station (BS) is simultaneously transmitting signal to
all the UEs in a cell, as shown in Figure 1.
Figure 1. NOMA system model
Assume that there are a maximum of ܰ UEs in a cell. Let
ߙ௡ be the complex channel gain between UE ݊ and the BS.
Further assume that the total transmission power at the BS is
constrained to ܲ , so σே ௡ୀଵ ߨ௡ ܲ ൌ ܲ , where ߨ௡ is the power
allocation coefficient for UE n. Suppose ȁߙଵ ȁଶ ൒ ȁߙଶ ȁଶ ൒ ‫ ڮ‬൒
ȁߙே ȁଶ and hence the power allocation coefficients can be sorted as
ߨଵ ൑ ߨଶ ൑ ‫ ڮ‬൑ ߨே [1, 5, 9, 10]. The superimposed signal of all
the UEs transmitted by the BS is then given by:
ே
‫ ݏ‬ൌ ෍ ඥߨ௜ ܲ ‫ݏ‬௜ (1)
௜ୀଵ
where ‫ݏ‬௜ is the message signal for UE ݅. The received signal at
the ݊th UE can be represented as [10]:
ே
‫ݎ‬௡ ൌ ߙ௡ ‫ ݏ‬൅ ߥ௡ ൌ ߙ௡ ෍ ඥߨ௜ ܲ ‫ݏ‬௜ ൅ ߥ௡ (2)
௜ୀଵ
where ߥ௡ is the additive Gaussian noise.
The optimal decoding order for the users is in the order of
increasing channel gains. Hence, in this case of channel gains
order, the UE ܰ having the least channel gain will decode its
signal in a straightforward manner, as it will treat all the
signals from other UEs as noise. At the receiver of UE
݊ǡ ͳ ൑ ݊ ൑ ܰ , all the message signals for UEs ݊ ൅ ͳǡ ݊ ൅
ʹǡ ǥ ǡ ܰ will be considered as interference and they will be
removed in a successive fashion using SIC. After the removal
of the interference from higher order UEs, it will decode its
own message signal by treating messages for lower order UEs
ͳǡ ʹǡ ǥ ǡ ݊ െ ͳ as noise [1, 10].
There are several potential issues with such SIC receivers
which can adversely affect the performance of NOMA. The
first drawback is that it is power sensitive. This means that its
performance will degrade dramatically if two or more UEs have
similar or same power [11]. For example, if two or more UEs
have comparable channel gains, then NOMA scheme will
assign same power allocation coefficients to them, and thus the
performance of SIC will degrade.
The second shortcoming of the SIC is that its performance is
highly dependent on the correct decoding of the first UE. In
case the first UE is not decoded correctly, this error in
decoding will propagate successively to lower order UEs and
hence deteriorate the system performance [5, 11]. More
specifically, at any stage of decoding, if UE ݇ is not decoded
correctly, the performance of all the UEs having decoding
order less than ݇ will be affected badly.
The third weakness of the SIC receiver is that if we have a
large group of UEs in NOMA, then due to successive decoding
nature of SIC, the decoding delay for lower order UEs will be
increased. In order to reduce decoding delays, NOMA has to
group small number of UEs having significant difference in
channel gains, thus limiting the system capacity [5].
Finally, the users that are served by NOMA transmission
scheme achieve unequal rates because the NOMA protocol is
based on the order of SIC. This issue is extremely important in
situations having strict fairness constraints [4].
The aforementioned critical performance limiting factors
for SIC motivate us to propose an alternate PIC-based receiver
design for NOMA downlink. To the best of our knowledge, it is
the first time that PIC-based receiver structure is studied for
downlink cellular NOMA.
IV. PROPOSED PIC-BASED RECEIVER STRUCTURE
This section describes the structure and operation of the
proposed PIC based receiver at UE for 5G downlink NOMA
transmissions. The block diagram of the proposed PIC-based
receiver for UE ݊ is shown in Figure 2.
The received composite signal at the ݊th UE is given by (2).
The proposed PIC-based receiver will perform decoding in
two steps. In the first step, the proposed receiver will remove
the multiuser interference caused by the transmission from users
ͳǡ ʹǡ Ǥ Ǥ ǡ ݊ െ ͳǡ ݊ ൅ ͳǡ ǥ ǡ ܰǤ In contrast to SIC, the proposed
receiver will mitigate the multi-user interference in a parallel
fashion. After removing the interference from other users, at
the second step, the receiver will decode the message of UE ݊.
In order to illustrate the multi-user interference cancellation
procedure, let us consider an example of UE ݊ with received
signal ‫ݎ‬௡ . Let ‫ݏ‬Ƹ௜ be the estimate of message signal of UE ݅ after
decoding, then, if we sum the estimates of all the ݅ UEs and
subtract it from the original received signal‫ݎ‬௡ , we will obtain an
estimate of the message signal for UE ݊. Hence, the signal at the
input of the decoder for UE ݊ can be written as:
ே
‫ݕ‬௡ ൌ ‫ݎ‬௡ െ ෍ ‫ݏ‬Ƹ௜ ǡ (3)
௜ୀଵ
௜ஷ௡
݅ ൌ ͳǡ ǥ ǡ ݊ െ ͳǡ ݊ ൅ ͳǡ ǥ ǡ ܰ
The decoder of UE ݊ will perform decoding on ‫ݕ‬௡ to get
the estimate of the message signal for UE ݊.
A. Filter Bank Design for Decoder of ݅ Users
Filter bank design is one of the possible approaches for
implementing a block for decoder of ݅ UEs to cancel the multi-
user interference in a parallel manner. The composite signal ‫ݎ‬௡
is fed into the matched filter input of each UE ݅ , where
݅ ൌ ͳǡ ʹǡ ǥ ǡ ݊ െ ͳǡ ݊ ൅ ͳǡ ǥ ǡ ܰ. The output from each matched
filter is then used to get the estimate of the transmitted
message signal for that UE. The resulting probability of bit
error at the receiver of UE ݊ (Ȧ௡ ) in an AWGN channel is
given as [12]:
Figure 2. Proposed receiver structure for UE ݊
Ȧ݊
଴ ͳ െ ሺܰ െ ͳΤ‫ ܭ‬ሻௌାଵ
ൌ ܳ ൭ቈ ቆ ቇ
ʹ‫ܧ‬௕ ͳ െ ሺܰ െ ͳΤ‫ ܭ‬ሻ
(4)
ͳ ሺܰ െ ͳሻௌାଵ െ ሺെͳሻௌାଵ σ௄
௝ୀଵ ܲ௝
൅ ௌାଵ ቆ
‫ܭ‬ ܰ ܲ௡
ିଵȀଶ
൅ ሺെͳሻ ௌାଵ ቇ቉ ቇ
where ܲ௝ is the received power for UE ݆, ܰ is the total number
of UEs, ‫ ܭ‬is a processing gain, ܵ are the number of PIC stages,
଴ is the one sided noise power spectral density of the additive
white Gaussian noise, ‫ܧ‬௕ is the energy per bit and ܳሺǤ ሻ is a
standard ܳ function.
The proposed PIC-based receiver solves the problems of
decoding failure in SIC due to equal or comparable power
allocation for two or more NOMA downlink UEs, successive
error propagation, decoding delay and dependency on the correct
decoding of the first UE. All of these characteristics of the
proposed receiver make it a stronger choice to adopt for the
receiver side of UE in 5G downlink NOMA transmissions.
B. Design Considerations
ƒ Upper limit on the number of UEs in a decoder block
The receiver of a UE has a limited processing capability and
power, which hinders one from processing and removing the
interference from all the (ܰ െ ͳሻ UEs simultaneously, though it is
theoretically feasible to do so. Furthermore, if we increase the
number of UEs, the corresponding complexity of the UE receiver
will also increase. Therefore, there exists a maximum value for
the number of UEs to be processed at a receiver in order to attain
a good trade-off between receiver complexity and system
capacity. We are interested in investigating this upper bound of
system capacity for downlink NOMA transmissions.
ƒ Estimate of user ࢏ message signal at UE ࢔ receiver
The estimate ‫ݏ‬Ƹ௜ of the message signal for user ݅ depends on its
matched filter output at the receiver of UE ݊ . We assumed
unbiased or nearly fair estimate of the UE ݅ message signal,
i.e. ‫ݏ‬Ƹ௜ ൎ ‫ݏ‬௜ . Unfortunately, in reality this estimate is biased
after cancellation and it will increase with system loading [11].
Hence, for practical deployment of this receiver for NOMA
downlink, this issue must be considered and requires further
study.
ƒ Power Control
Power control is a vital component of a cellular network
which is used to manage interference, energy and connectivity by
maintaining the received signal to noise ratio (SNR) in both
uplink and downlink. Therefore, it is of interest in understanding
how NOMA transmission protocol will be integrated with the with PIC receiver at UE performs better than NOMA with SIC in
existing power control strategies (both in uplink and downlink). In all SNR regions.
particular, having implemented uplink power control, the problem The final set of comparison between SIC and PIC techniques is
is how the NOMA transmission scheme will allocate different presented in terms of computational complexity and is shown in
power to the downlink UEs in order to exploit power domain for Figure 5. The computational complexity can be defined as the
multiplexing. One possible solution to solve this problem is number of FLoating point OPerations (FLOP) required to
requiring the UE to send to the BS its local noise information, decode one bit. A FLOP can be an operation implementing a
based on which the NOMA protocol will allocate different power multiplication or addition, while more complex operations can be
levels to the users. Further, it is of interest to investigate how the regarded as multiple operations. For analysis purpose, we use the
power control will help in reducing the multiuser interference at computational complexity per bit of decision derived for SIC
the proposed PIC based receiver. Hence, the investigation of and PIC receivers in [12], given by:
integrating power control with NOMA and its impact on the
proposed receiver structure can be an interesting research topic ‫ܥ‬௉ூ஼ ሺܾሻ ൌ ܰ‫ܮ‬ሾܵሺ͸‫ܰܭ‬௦ ൅ ͹ሻ െ Ͷ‫ܰܭ‬௦ െ ͳሿ (5)
in NOMA.
ƒ Channel Estimation ‫ܥ‬ௌூ஼ ሺܾሻ ൌ ‫ܮ‬ሾܰሺͺ‫ܰܭ‬௦ ൅ ͳʹሻ െ ͸‫ܰܭ‬௦ െ ͹ሿ (6)
ܰ
In all real time scenarios, the assumption of perfect channel ൅ ሾʹܰ௕ ൅ Ž‘‰ ଶ ሺܰሻ ൅ ͳሿ
information will not be valid and hence the channel is ܰ௕
estimated before performing any further processing. Due to the
random nature of wireless channels, the estimate can never be where ‫ܥ‬௉ூ஼ ሺܾሻ and ‫ܥ‬ௌூ஼ ሺܾሻ are computational complexities in
accurate. Therefore, to make the receiver design more practical, it terms of number of FLOPs required for decoding one bit using
is necessary to analyze the performance of the PIC-based PIC, and SIC receiver, respectively, ‫ ܮ‬is the maximum number of
receiver under imperfect channel estimation in the context of 5G multipath signals processed by correlators at the receiver, ‫ ܭ‬is a
NOMA downlink. processing gain, ܰ is the total number of UEs, ܰௌ is the
V. RESULTS AND DISCUSSION number of samples per bit and ܰ௕ is the frame length in bits.
In this section, the analytical performance comparisons
between PIC and SIC are presented. Apart from the fact that
PIC solves the problems posed by SIC, the analysis also serves
to promote the proposed PIC-based receiver structure as an
alternate of SIC for 5G downlink NOMA based on its better
performance.
The first comparison is presented between probability of bit
error and number of UEs, as shown in Figure 3. The results are
obtained with 30 UEs for both SIC and PIC. Each set of results
is collected by keeping constant SNR of 10, 15 and 20 dB. The
results show that by increasing the number of UEs beyond 5, Figure 3. Probability of bit error and number of UEs
there is negligible impact of increasing SNR for SIC receiver,
while for PIC, the performance gain is appreciable up to 15 users
for all three considered values of SNR.
The second set of results is presented in terms of probability
of bit error and SNR, as shown in Figure 4. This figure shows the
analytical bit error rate (BER) performance of SIC and PIC along
with simulation results for NOMA with SIC and PIC receivers for
the case of three users. The power allocation coefficients ߨ௜ǡ ݅ ൌ
ͳǡʹǡ͵ for three NOMA UEs are considered to be random
variables (due to the random nature of the channel), uniformly
distributed between (0,1). In this way, we considered and Figure 4. Downlink NOMA simulation for 3 UEs
4
x 10
incorporated the case of equal or comparable power allocation 5

for two or more NOMA users in our simulations, as there exists 4

PIC
a finite probability that two or all three NOMA users can be
(FLOP/Bit Decision)

SIC
allocated with same power coefficient, which implies that they
Complexity

have same or comparable channel quality. This choice for 2

considering the power allocation coefficient as random
variable makes it more realistic and general. The analytical 1

results for SIC and PIC receivers provide a performance bound 0

0 5 10 15 20 25 30
for NOMA simulation. The simulation results show that NOMA Number of UEs

Figure 5. Computational complexity and number of UEs

 TABLE I. SIMULATION PARAMETERS REFERENCES
Parameter Value [1] A. Benjebbour, Y. Saito, Y. Kishiyama, et al., "Concept and practical
Number of bits, ܰ௕ 100 considerations of non-orthogonal multiple access (NOMA) for future radio
Number of samples per bit, ܰௌ 10 access," in Proc. of International Symposium on Intelligent Signal Processing
Number of multipath signals, ‫ܮ‬ 10 and Communications Systems (ISPACS), pp. 770-774, 12-15 Nov. 2013.
Processing gain, ‫ܭ‬ 1 [2] D. Xiaoming, C. Shanzhi, S. Shaohui, et al., "Successive interference
Number of UEs, ܰ 30 cancelation amenable multiple access (SAMA) for future wireless
Number of PIC stages, ܵ 2 communications," in Proc. of IEEE International Conference on
Communication Systems (ICCS), pp. 222-226, 19-21 Nov. 2014.
TABLE II. COMPUTATIONAL TIME OF PIC FOR EACH MOBILE SOC [3] M. A. Zheng, Z. Z. Quan, D. Z. Guo, et al., "Key techniques for 5G wireless
communications: network architecture, physical layer, and MAC layer
SoC MFLOP/sec ࢀ ሺࣆ࢙ሻ perspectives," Science China Information Society, Apr. 2015.
Samsung Exynos 3110 17.7 2800 [4] S. Timotheou and I. Krikidis, “Fairness for non-orthogonal multiple
Nvidia Tegra 2 54.4 900 access in 5G systems,” IEEE Signal Processing Letters, vol. 22, no. 10,
Texas Instruments OMAP 4460 75 653 pp. 1647-1651, October 2015.
Qualcomm Snapdragon 600 540 90.7 [5] Y. Saito, A. Benjebbour, Y. Kishiyama, and T. Nakamura, "System-level
Samsung Exynos 5 Octa 5410 627 78 performance evaluation of downlink non-orthogonal multiple access (NOMA),"
Qualcomm Snapdragon 800 1034 47.3 in Proc. of IEEE 24th International Symposium on Personal Indoor and Mobile
Radio Communications (PIMRC), pp. 611-615, 8-11 Sept. 2013.
The computational complexity is computed for the parameters [6] N. Nonaka, Y. Kishiyama, and K. Higuchi, "Non-Orthogonal Multiple Access
stated in Table I. The increased computational complexity for Using Intra-Beam Superposition Coding and SIC in Base Station Cooperative
MIMO Cellular Downlink," in Proc. of IEEE 80th Vehicular Technology
PIC, as depicted in Figure 5 is due to the parallel structure of this Conference (VTC Fall), pp.1-5, 14-17 Sept. 2014.
receiver. The performance gains in Figures 3 and 4 were achieved at [7] L. Yang, A. Benjebboiu, C. Xiaohang, et al., "Considerations on
the cost of higher computational complexity. downlink non-orthogonal multiple access (NOMA) combined with
In order to evaluate whether the processing power of closed-loop SU-MIMO," in Proc. of 8th International Conference on
current smartphone (UE) has a capability to implement the PIC Signal Processing and Communication Systems (ICSPCS), pp.1-5, 15-17
Dec. 2014.
receiver with this expected complexity, we consider a certain
[8] A. Benjebbour, Li. Anxin Y. Saito, et al., "System-level performance of
system-on-chip (SoC) used in some smartphones with a downlink NOMA for future LTE enhancements," in Proc. of IEEE Globecom
capability to process ܺ FLOP/sec. The processing power in Workshops (GC Wkshps), pp. 66-70, 9-13 Dec. 2013.
MFLOP/sec (ͳͲ଺ FLOP/sec) of different popular mobile SoCs is [9] Z. Ding, Z. Yang, P. Fan, and H. V. Poor, "On the Performance of Non-
listed in Table II [13-15]. Orthogonal Multiple Access in 5G Systems with Randomly Deployed Users,"
IEEE Signal Processing Letters, vol. 21, no. 12, pp. 1501-1505, Dec. 2014.
The equivalent computational time ܶ (in ߤ‫ ) ݏ‬per bit of
[10] Z. Ding, F. Adachi, and H. V. Poor, "The Application of MIMO to Non-
decision for PIC is calculated for each mobile SoC. The maximum Orthogonal Multiple Access," IEEE Transactions on Wireless
considered computational complexity for PIC with ܰ ൌ ͵Ͳ is Communications, in-press, 2015. doi:10.1109/TWC.2015.2475746
approximately ͶǤͻšͳͲସ  FLOP, as shown in Figure 5. The [11] R. M. Buehrer, Code division multiple access (CDMA), Morgan and
results of ܶ show that the current and upcoming generations of Claypool, 2006.
smartphones should have the capability to implement PIC based [12] R. M. Buehrer, N. S. Correal-Mendoza and B. D. Woerner, "A Simulation
Comparison of Multiuser Receivers for Cellular CDMA", IEEE Transactions
receiver for 5G downlink NOMA. on Vehicular Technology, pp. 1065-1085, Jul 2000.
VI. CONCLUSION [13] F. Busching, S. Schildt, S. and L. Wolf, "DroidCluster: Towards Smartphone
Cluster Computing -- The Streets are Paved with Potential Computer
In this paper, we reviewed the existing downlink NOMA Clusters", in Proc. 32nd International Conference on Distributed Computing
with SIC and highlighted some limitations related to SIC Systems Workshops (ICDCSW), pp.114-117, 18-21 Jun 2012.
which can consequently degrade the performance of NOMA [14] Qualcomm, "Qualcomm Snapdragon Benchmark Report", Sep 30, 2013,
Available: https://www.qualcomm.com/documents/qualcomm-snapdragon-
scheme in downlink. In order to alleviate the problems posed by benchmark-report
SIC, we proposed a PIC based receiver structure for downlink [15] Samsung Galaxy S4 Exynos 5 Octa Benchmarks, Apr 27, 2013, Available:
NOMA. The numerical results prove the superiority of proposed http://www.fonearena.com/blog/67821/samsung-galaxy-s4-exynos-5-octa-
PIC based receiver over SIC and hence is a promising receiver benchmarks.html
choice for UE in 5G downlink NOMA. However, this
performance gain of PIC is achieved at the cost of higher
computational complexity as compared to SIC.
In order to make the proposed PIC based receiver a more
suitable and promising choice for 5G downlink NOMA, one
possible area of future work can be focused on reducing its
computational complexity by attempting to achieve a good
trade-off between receiver performance and complexity. Other
possible directions for future research may include investigating
MIMO downlink NOMA with PIC based UE receiver,
integrating power control in downlink NOMA system with PIC
receiver, and proposing a novel design of decoder block at UE
receiver to reduce the multiuser interference.