Conference PaperPDF Available

Exploiting NOMA for Radio Resource Efficient Traffic Steering Use-case in O-RAN



In this work, we consider the design of a radio resource management (RRM) solution for traffic steering (TS) use-case in the open radio access network (O-RAN). The O-RAN TS deals with the quality-of-service (QoS)-aware steering of the traffic by connectivity management (e.g., device-to-cell association, radio spectrum, and power allocation) for emerging heterogeneous networks (HetNets) in 5G-and-beyond systems. However, TS in HetNets is a complex problem in terms of efficiently assigning/utilizing the radio resources while satisfying the diverse QoS requirements of especially the cell-edge users due to their poor signal-to-interference-plus-noise ratio (SINR). In this respect, we propose an intelligent non-orthogonal multiple access (NOMA)-based RRM technique for a small cell base station (SBS) within a macro gNB. A Q-learning-assisted algorithm is designed to allocate the transmit power and frequency sub-bands at the O-RAN control layer such that interference from macro gNB to SBS devices is minimized while ensuring the QoS of the maximum number of devices. The numerical results show that the proposed method enhances the overall spectral efficiency of the NOMA-based TS use case without adding to the system's complexity or cost compared to traditional HetNet topologies such as co-channel deployments and dedicated channel deployments.
Exploiting NOMA for Radio Resource Efficient
Traffic Steering Use-case in O-RAN
Muhammad Waseem Akhtar1, Aamir Mahmood1, Sarder Fakhrul Abedin1, Syed Ali Hassan2, Mikael Gidlund1
1Department of Information Systems &Technology, Mid Sweden University, 851 70 Sundsvall, Sweden
2School of Electrical Engineering and Computer Science (SEECS)
National University of Sciences and Technology (NUST) H-12, Islamabad, Pakistan
Email: 1{muhammadwaseem.akhtar, aamir.mahmood, sarder.fakhrulabedin, mikael.gidlund},
Abstract—In this work, we consider the design of a radio
resource management (RRM) solution for traffic steering (TS)
use-case in the open radio access network (O-RAN). The O-
RAN TS deals with the quality-of-service (QoS)-aware steering
of the traffic by connectivity management (e.g., device-to-cell
association, radio spectrum, and power allocation) for emerging
heterogeneous networks (HetNets) in 5G-and-beyond systems.
However, TS in HetNets is a complex problem in terms of effi-
ciently assigning/utilizing the radio resources while satisfying the
diverse QoS requirements of especially the cell-edge users due to
their poor signal-to-interference-plus-noise ratio (SINR). In this
respect, we propose an intelligent non-orthogonal multiple access
(NOMA)-based RRM technique for a small cell base station (SBS)
within a macro gNB. A Q-learning-assisted algorithm is designed
to allocate the transmit power and frequency sub-bands at the O-
RAN control layer such that interference from macro gNB to SBS
devices is minimized while ensuring the QoS of the maximum
number of devices. The numerical results show that the proposed
method enhances the overall spectral efficiency of the NOMA-
based TS use case without adding to the system’s complexity or
cost compared to traditional HetNet topologies such as co-channel
deployments and dedicated channel deployments.
Index Terms—Internet-of-things (IoT), HetNets, Open RAN
(O-RAN), non-orthogonal multiple access (NOMA), Q-learning,
spectral efficiency, traffic steering.
Satisfying varying data rate requirements of emerging
Internet-of-thing (IoT)-based use cases are challenging, espe-
cially for cell-edge users, with limited radio resources [1]–[3].
However, considering the accelerated deployment pace of IoT
applications in 5G-and-beyond communication systems [4],
intelligent solutions are required to achieve high area spectral
efficiency, massive connectivity, and high data rates [5]–[8].
Such requirements have invoked the idea of non-orthogonal
multiple access (NOMA) in future wireless networks [9],
allowing multiple devices to utilize the same spectrum si-
multaneously. However, in high connection density scenarios,
the performance of weak devices located far away from the
next generation-NodeB (gNB) or in a low coverage area is
still unsatisfactory. As a result, various techniques, including
cooperative NOMA and HetNets, are under investigation in
the literature to improve the performance of cell-edge devices.
In particular, when compared to non-cooperative NOMA, in-
tegrating NOMA and cooperative communications can enable
higher power- and spectral-efficiency while improving fairness
for IoT networks with massive connectivity.
Cooperative relaying in NOMA systems is studied exten-
sively in the literature [10]–[13]. For example, in [10], the
authors proposed cooperative NOMA in which the strong
devices can serve as relays for the weak devices (i.e., hav-
ing bad channel conditions), increasing their diversity gain.
However, this scheme increased the computational and power
requirements by significantly increasing successive interfer-
ence cancellation (SIC) operations at the device. Users may
experience high energy costs because of many SIC in-network
with limited energy, computation, and time resources. To
reduce the number of SIC operations, a space-time block code
(STBC)-based cooperative mechanism is proposed in [14],
[15], which improves the throughput of the weak devices with
a low number of SIC but at the cost of timing offsets due to
distributed nature of wireless devices.
Meanwhile, the idea of using a small-cell base station (SBS)
in the same spectral band in the low coverage area is known as
heterogeneous network (HetNet) architecture, in which a high
power gNB is used to cover the large area. In contrast, low-
power gNBs are used to improve the coverage at the cell edge,
or hot-spots [16], [17]. However, radio resource management
(RRM) in HetNets is a complex problem in terms of efficiently
assigning/utilizing the radio resources while satisfying the
diverse quality-of-service (QoS) requirements of especially
the cell-edge users for their poor SINR. In this context, the
open RAN (O-RAN) traffic steering (TS) use case envisions
machine learning (ML)-enabled TS–xApp at RAN intelligent
controller (RIC) for QoS-aware steering of the traffic by con-
nectivity management (e.g., device-to-cell association, radio
spectrum, and power allocation) [1]–[3], [18]. xApps can be
deployed by multiple sources (e.g., network operator, factory
management) to include new features as 5G network control
applications (e.g., industrial IoT applications [19]). Yet, the
literature lacks a QoS-aware radio resource-efficient design of
TS–xAPP in O-RAN control for HetNets.
The properties of this cross-tier interference in NOMA-
based HetNets differ from those of conventional orthogonal
multiple access (OMA)-based macro-cellular network interfer-
ence. Because of the limited association, macro IoT devices
(MIDs) may be unable to connect to a SBS, even if it is
the closest serving gNB, resulting in significant interference.
Cross-tier cellular architecture is also studied in [9], [20], [21].
The authors in [9] used the cognitive radio-based NOMA
approach for cross-tier architecture. The author applied the
greedy approach to maintain the QoS of the macro and small
cell IoT devices (SIDs). However, this approach becomes
more complex and limits the SID spectral efficiency when the
number of devices increases. Similarly, classical approaches
such as SIC and parallel interference cancellation (PIC) are not
recommended due to their high computational requirements.
An iterative joint resource allocation and device pairing in
NOMA is proposed in [22], while [23] studied the Dinkelbach
method to optimize the cross-tier interference in a heteroge-
neous NOMA network.
Most previous research has concentrated on cooperation
mechanisms that enhance complexity and power consumption,
especially for higher device densities. Also, the diverse data
rate requirements are not considered, which ultimately does
not guarantee the QoS of the devices with low data rate
requirements. We design an intelligent interference control and
efficient resource allocation for the traffic steering use case in
O-RAN, termed as RRE-NOMA. In this scheme, the near-RT
RIC handles cross-tier interference to SIDs so that the resulting
interference can be efficiently mitigated, which means that the
data rate requirements of the maximum number of devices are
fulfilled by limiting interference to the low data rate devices.
Also, the assignment of low power to the devices at overlapped
spectral portion does not affect their data rate outage due
to their low data rate requirements. Furthermore, MIDs with
high data rate requirements are assigned to the non-overlapped
spectral portion, limiting the interference from SBS. In this
respect, our main contributions can be summarized as follows.
Leveraging the O-RAN architecture and interfaces for
reinforcement learning (RL)-based model training to en-
force the long-term policy-based sub-band and power
allocation to MIDs grouped in two sub-bands based on
their data rate demands while applying NOMA in each
sub-band independently. It leads to QoS-aware cross-tier
and inter-devices (NOMA) interference reduction.
Performing an extensive experimental analysis to demon-
strate the efficacy of the proposed approach. The results
indicate the overall network spectral and power-efficiency
enhanced while achieving the QoS of the maximum
number of devices.
The rest of the paper is organized as follows. In Section II,
we describe the system model and explain the three HetNet
deployment scenarios (i.e., co-channel, cross channel, and
dedicated channel). Section III provides the proposed rein-
forcement learning approach, specially Q-learning, for interfer-
ence management. In Section IV, we present the experimental
results, and finally conclusions are given in Section V.
We consider a single antenna macro gNB and a single-
antenna SBS, with the latter deployed at the cell edge to
improve the coverage, as shown in Fig. 1. Macro gNB and SBS
are assumed to be connected through a dedicated backhaul to
TS–xAPP at the O-RAN controller (i.e., near-real-time—near-
RT RIC) for applying the intelligent resource allocation policy
Macro gNB SBS
Direct NOMA link Interfering NOMA link Devices with diverse
data rate requirements
time RIC)
Fig. 1. A system model for NOMA-based traffic steering (TS) in O-RAN.
guidance received from orchestration and automation O-RAN
layer. Meanwhile, we assume 3GPP RAN functional split 7.2
for the control unit (CU), distributed unit (DU), and radio
unit (RU) [18]. Further, there are MMID devices randomly
distributed in the coverage area of macro gNB and FSID
devices under the coverage area of SBS. Both macro gNB and
SBS employ the NOMA transmission scheme in the downlink.
The MIDs and SIDs are arranged as per descending order of
their channel conditions. Let U {U1, U2, . . . , Um, . . . , UM}
be the set of all MIDs and K {K1, K2, . . . , Kf, . . . , KF}
be the set of all SIDs. Therefore, U1and K1are, respectively,
the strongest macro and SIDs, while UMand KFare the
weakest macro and SIDs. The power assigned by the macro
gNB to the MID mis pm, whereas the power assigned by the
SBS to the SID fis pf. We also consider all the channels
as independent, identically distributed (i.i.d.) Rayleigh model
and with block fading channel conditions. Using this model,
we describe the possible HetNet deployment scenarios in the
following subsections.
A. Deployment Scenarios
There are three possible deployment scenarios for spectrum
allocation for the considered cross-tier O-RAN architecture.
1) Co-channel deployment: In this scenario, complete
bandwidth of BHz is concurrently assigned to both macro
gNB and SBS, and consequently, underlayed SIDs face a
strong interference from the overlayed macro gNB. The SINR
at a SID fis given as
Imacro +Pf1
j=1 |hf|2pj+σ2,(1)
where hfis the channel between f-th SID and the SBS, and
Imacro =PM
i=1 |hf|2piis the interference received from the
macro gNB to the SIDs. Using (1), the capacity of SID is
f=Blog21 + γco
Similarly, sum data rate of small cell (Rco
f) and macro (Rco
devices (Rco
m) can be, respectively, determined as
f=1 Cco
m=1 Blog21 + γco
High data rate
High data rate
devices Low data rate
Low data rate
B -
B -
(a) (b) (c)
Fig. 2. Spectrum management techniques for Macro-SBS RAN architecture using NOMA: (a) co-channel deployment (b) dedicated channel deployment (c)
proposed radio resource management technique.
2) Dedicated channel deployment: In this scenario, orthog-
onal bandwidth is allocated to both macro gNB and SBS, such
that macro gNB is assigned Bθand SBS is assigned θof
the bandwidth, where θB/2, as shown in Fig. 2(b). In the
dedicated channel deployment, SIDs do not face interference
from the macro gNB, however, at the cost of less bandwidth.
For the dedicated channel deployment, we have Imacro = 0 ,
and therefore, the SINR and achievable data rate of the f-th
SID is given as
j=1 |hf|2pj+σ2,(5)
f=θlog21 + γded
3) Proposed spectrum management technique: In this case,
the sub-band θis assigned to the SBS, whereas the whole
bandwidth Bis assigned to macro gNB. The macro gNB
further divides bandwidth Binto two sub-bands( i.e., θand
Bθ), where θis a portion of the spectrum overlapping with
SBS, and Bθis the portion of the spectrum orthogonal to
SBS. We consider that the sub-band θcontains the set of low
data rate MIDs whereas the sub-band Bθcontains the set of
high data rate MIDs. Now let bthe generalized term for each
sub-band, such that b {θ, B θ}, then the set of MIDs in
each sub-band is Ψb,l ={lb,1, lb,2, lb,3, . . . , lb,N }, where Nthe
total number of MIDs in sub-band b, such that P2
b=1 N=M.
The devices in each sub-band follow the NOMA principle.
Our objective is to reduce the interference from macro gNB
to the SIDs, which ensures the QoS demands of a large number
of SIDs, as well as the MIDs, as shown in Fig. 2(c). If the
MIDs in the non-overlapped spectrum is allocated with P
(W), and Pη(W) is the total power allocated to MIDs in
the overlapped spectrum, where η < P , then the interference
received at the SIDs from macro gNB is given as
macro =
where hθ,l is the channel between macro gNB and and the
device lin the sub-band θ,hθ,l is the channel between macro
gNB and device l,pθ,l is the power allocated to the device l
in the sub-band θ. Therefore, the SINR and achievable data
rate of SIDs in this case is
macro +Pf1
j=1 |hf|2pj+σ2,(8)
f=θlog21 + γprop
respectively. The sum desired data rate (Φsum) of SIDs is
calculated as follows
Φsum =
where Φfis the f-th SID’s desired data rate. Similarly, the
sum achievable data rate (Υsum) of SIDs is
Υsum =
The macro gNB in the proposed NOMA system has to allocate
sub-band and power to the devices to maximize the sum
data rate of the overall heterogeneous RAN. However, a
stochastic transition model for the macro gNB between states
as a function of interfering power cannot be defined due to
the intrinsic dynamics of a wireless channel. Therefore, we
propose a radio resource-efficient (RRE) NOMA system using
a reinforcement learning model technique, specifically the Q-
learning method, to reduce interference. The control strategy
for the interference mitigation problem described in the next
section is a self-managing process that belongs to the domain
of self-organizing future networks.
We propose an RL technique, especially Q-learning, for
interference management of the SIDs. We apply RL to this
problem due to its suitability for handling such issues that
involve deciding on multiple options. Macro gNB gets the
information about SIDs from the backhaul that connects the
macro gNB and SBS and then decides about the power
assignment to the SIDs that are operated on the overlapped
spectral portion of the small cell. Hence, with the help of a
reinforcement learning algorithm, macro gNB can enhance the
QoS of the SIDs located at the cell edge.
We apply the Q-learning algorithm due to its proven abil-
ity to perform in a highly dynamic environment. The main
components of Q-learning algorithms are i) Agent with an
environment, ii) state, iii) action, iv) Cost (a reward or penalty),
and v) Q-table that stores the cost-values for all possible
actions in any state. In the following, we provide a detailed
discussion on the Q-learning model that includes the design
of the state and action spaces, and cost/reward function.
State: A state is the current required data rate of macro
IoT device, channel conditions, and sub-band allocated
to that IoT device and is given as
sS{R, CQI , b},(12)
where Ris the device required data rate, CQI is the
channel quality indicator.
Action: Action is the assignment of power levels to the
devices that is given by
where ais the action taken in the current state s,pA
is a set of power levels with Aas the maximum number
of power levels. We use three discrete power levels as
actions depending upon the data rate demands of the
Cost or Reward: In our optimization problem, the goal
of any device is to achieve the minimum required data
rate. If the devices meet the required data rate, they are
rewarded in terms of better spectral efficiency; otherwise,
they receive penalty. The reward function is given as
r(s, a) = Φsum
Υsum Cprop
where Φsum,Υsum , and Cprop
fare given in (10), (11),
and (9), respectively. where Φfis the data rate threshold
for f-th SID. Our goal is to maximize the reward so that
the SID can achieve the minimum data rate requirement
and small cell, on the other hand, tries to reduce interfer-
ence which can ultimately enhance the spectral efficiency
of the overall network.
The proposed Q-table for RRE-NOMA is updated according
to the following rule
Q(s, a)Q(s, a) + α·(r(s, a)
aQ(s, a)Q(s, a)),(15)
where 0α1is the learning rate and ϑ[0,1] is the
discount factor. The terms sand arefer to the current state and
current action, whereas sand adenote the next state and the
next action, respectively. The action-selection policy causes the
system to shift into a new state. This policy is developed during
the exploration phase and implemented during the exploitation
phase. The policy is known as the ε-greedy action selection
policy. It involves observing a reward for all potential actions
for a given state, then updating the state-action pair with the
highest reward in the Q-table until all state-action pairs have
been updated.
Algorithm 1 Q-learning algorithm for O-RAN traffic steering
Input: Information about small cell and MIDs, B(Hz), θ
(Hz), and Pnoma.
Output: Total number of MIDs on the overlapped spectral
portion. Power and bandwidth assignment to the SIDs on
the overlapped spectral portion.
Initialization :
1: Organize devices into two groups based on their data rate
requirements (i.e., low and high data rate devices) w.r.t a
certain data rate requirement threshold.
2: Assign the overlapped spectral part of the bandwidth to
low data rate devices and the non-overlapped spectral part
to high data rate devices and organize the devices in each
sub-band using the NOMA principle.
3: For overlapped spectral portion θ, start Q-value Q(0,0) =
4: for each episode,
select state sSrandomly do
5: for for each step of episode do
6: if (rand(·)< ε)then
7: Observe the current state sSin the overlapped
spectral portion.
Assign fixed power as an action ato the devices.
Take the immediate reward as in (3).
Update the Q-values in the Q-table as per (15).
Move to the new state sS.
Change the action, i.e., assign more or less
power to the devices based on the new state.
8: end if
Repeat the process until the algorithm determines
the best power level for each device.
9: end for
10: end for
The goal of the agent is to find the optimal action selec-
tion/transition policy π(s)for each state [4], which maxi-
mizes the reward with time, which is given by
Wπ(s)=Wπ(s) = max
ar(s, a)+ϑX
where r(s, a)is the reward for performing action ain state s,
aAis the next feasible action that maximizes the reward,
and P s, sis the transition probability from state ssas
We describe the strategy for the interference management
for future NOMA-based heterogeneous RAN as in Algorithm
1. In the next section, we analyze the outage probability of
SID and compare its performance for different deployment
0 50 100 150 200 250 300 350 400
Time Slot
Average Reward
= 0.2
= 0.4
= 0.6
= 0.8
= 0.9
Fig. 3. Impact of learning rate on the convergence of the algorithm, with
B= 5 MHz, θ= 2 MHz, and F= 4.
This section provides and discusses the results for our
proposed system design and compares these with the baseline
schemes. We consider a macrocell with a radius of 1 km, and
Mdevices are uniformly distributed in its area. We further
assume that these devices’ data rates are uniformly distributed
in the range [5, 8000] kbps unless stated otherwise. There is
one SBS located at the edge of the macrocell, with a radius
of 200 m. The transmit power of the macro gNB is 45 dBm
(unless it is decided to change to 35 dBm or 25 dBm using Q-
learning), whereas the transmit power of each SBS is 20 dBm.
Thermal noise density is considered as -173 dBm/Hz, and path
loss exponent as 4.
Fig. 3 shows the learning rate’s influence on the algorithm’s
convergence in the proposed scheme. It can be observed that
the algorithm converges quickly for higher learning rate values
(i.e., as α1) and vice versa. Also, the convergence time
is almost the same for the learning rate values of 0.4.
Therefore, we choose the learning rate of 0.4 in the following
simulation results.
Fig 4 shows the impact on the spectral efficiency of the
weakest SID by using different power levels (i.e., 45 dBm,
35 dBm, and 25 dBm) from the macro gNB in the sub-band θ.
Higher transmit power at the macro gNB will generate higher
interference at the SIDs and vice versa. Spectral efficiency
(SE) of the weakest SID is simulated, corresponding to the
signal-to-noise ratio (SNR) on the dB scale and interference-
to-signal ratio (β) on the linear scale. We can observe that for
the values of β= 1, and SNR = 15 dB, it can achieve a SE
of 5 bps/Hz with the macro gNB transmit power of 25 dBm,
whereas its SE is degraded to 3 bps/Hz, and 1 bps/Hz for the
transmit power of 35 dBm, and 45 dBm, respectively.
Fig. 5 shows the average capacity of the weakest SID for
an increasing number of SIDs in different NOMA deployment
topologies, such as co-channel, dedicated channel, and the
proposed RRE-NOMA. It is observed that the average capacity
of the SID decreases with the increase in the total number of
SIDs. Since the inter-devices, NOMA interference is increased
with the increase in SIDs. For F= 2, the weakest SID
achieves 7.9 bps/Hz, 5.5 bps/Hz, and 3.3 bps/Hz for proposed
Fig. 4. Link-level simulation for the proposed RAN architecture. βis the
interference-to-signal ratio (ISR) on linear scale, and SINR is optimized using
Q-learning to achieve desired QoS requirement of the SID when F= 2, and
the macro gNB transmit power is: (a) 45 dBm, (b) 35 dBm, and (c) 25 dBm.
deployment, dedicated channel deployment, and co-channel
deployment, respectively. However, the values of average
capacity are decreased to 5.8 bps/Hz, 3 bps/Hz, and 0.6 bps/Hz
with F= 12. Also, these results show that the weakest SID
in the proposed RRE-NOMA achieves significantly higher
capacity for any number of SIDs.
Fig. 6 depicts the average number of SIDs in an outage
for an increase in the total number of SIDs with θ= 6 MHz,
and the target data rate requirements range [5, 800] kbps. Our
proposed RRE-NOMA outperforms the conventional NOMA
and orthogonal multiple access (OMA) schemes for all cases
of the total number of SIDs. However, we can also observe
2 4 10 126 8
Total number of SIDs [F]
Av. capacity of the weakest SID (bps/Hz)
RRE-NOMA (proposed)
NOMA with dedicated channel deployment
NOMA with co-channel deployment
Fig. 5. Average capacity of the weakest SID with respect to the total number
of SIDs served.
F = 6 F = 24F = 12 F = 18
Total number of SIDs
Av. number of SIDs in outage
RRE-NOMA (proposed)
Fig. 6. Average number of SIDs in outage with respect to increase in the
total number of SIDs with data rate range [5, 800] kbps for OMA, NOMA,
and the proposed Q-learning-based NOMA with θ= 6 MHz, and 20 dBm of
transmit power at SBS in the proposed deployment.
that the performance gap between our proposed scheme and
that of NOMA and OMA also increases with the increase in
the total number of SIDs. This is because, with the increase
in F, OMA and NOMA schemes can not manage the SIDs’
resources (i.e., power and bandwidth). On the other hand,
the proposed RRE-NOMA scheme manages the resources
intelligently by assigning the power and sub-bands such that
interference is minimized at the SIDs, ultimately maintaining
the QoS requirements of maximum SIDs and enhancing the
overall SE of the network.
This paper proposed a novel Q-learning-based interference
control and radio resource management technique for NOMA-
enabled macro-SBS HetNet architecture. Instead of using the
conventional interference cancellation techniques, we intelli-
gently minimize the interference at the SIDs without adding
extra cost or complexity to the network. The overlapped
spectral portion is allocated to the MIDs with low data rate
requirements, whereas the MIDs with high required data rates
are allocated to the non-overlapped spectral portion. Macro
gNB assigns the power, as per the NOMA principle, to the
SIDs located at the overlapped portion so that their QoS is
not affected. As a result, the interference received at the SIDs
from macro gNB is significantly reduced, which ultimately
enhances the spectral efficiency of the weakest SID, and the
overall QoS and sum data rate of the network is improved.
In the future, we will study the user association, NOMA user
grouping, and resource efficiency as a joint problem in the
O-RAN traffic steering use case.
[1] L. Bonati, S. D’Oro, et al., “Intelligence and learning in O-RAN for
data-driven NextG cellular networks, IEEE Communications Magazine,
vol. 59, no. 10, pp. 21–27, 2021.
[2] I. Chih-Lin, S. Kuklinsk´
ı, and T. Chen, A perspective of O-RAN
integration with MEC, SON, and network slicing in the 5G era, IEEE
Network, vol. 34, no. 6, pp. 3–4, 2020.
[3] S. K. Singh et al., “The evolution of radio access network towards open-
RAN: Challenges and opportunities,” in IEEE WCNC Workshops, 2020,
pp. 1–6.
[4] M. W. Akhtar et al., “Q2A-NOMA: A Q-learning-based QoS-aware
NOMA system design for diverse data rate requirements, IEEE Trans.
Ind. Informat., 2022.
[5] H. Munir et al., “Resource optimization in multi-tier HetNets exploiting
multi-slope path loss model,” IEEE Access, vol. 5, pp. 8714–8726, 2017.
[6] A. Mahmood et al., “Industrial IoT in 5G-and-beyond networks: Vision,
architecture, and design trends,” IEEE Trans. Ind. Informat., vol. 18,
no. 6, pp. 4122–4137, 2022.
[7] J. Kim et al., “A novel resource allocation scheme for NOMA-V2X-
Femtocell with channel aggregation,” in IEEE GLOBECOM, 2020, pp.
[8] Z. Liu et al., “Robust resource allocation in two-tier NOMA heteroge-
neous networks toward 5G, Computer Networks, vol. 176, p. 107299,
[9] I. Budhiraja et al., “Cross layer NOMA interference mitigation for
femtocell users in 5G environment, IEEE Trans. Veh. Technol., vol. 68,
no. 5, pp. 4721–4733, 2019.
[10] Z. Ding et al., “Cooperative non-orthogonal multiple access in 5G
systems,” IEEE Commun. Lett., vol. 19, no. 8, pp. 1462–1465, 2015.
[11] Z. Yang et al., “Novel relay selection strategies for cooperative noma,
IEEE Trans. Veh. Technol., vol. 66, no. 11, pp. 10 114–10 123, 2017.
[12] W. Mei and R. Zhang, “Uplink cooperative noma for cellular-connected
uav, IEEE J. Sel. Topics Signal Process, vol. 13, no. 3, pp. 644–656,
[13] P. Xu et al., “Optimal relay selection schemes for cooperative NOMA,”
IEEE Trans. Veh. Technol., vol. 67, no. 8, pp. 7851–7855, 2018.
[14] M. W. Akhtar et al., “STBC-aided cooperative NOMA with timing
offsets, imperfect successive interference cancellation, and imperfect
channel state information,” IEEE Trans. Veh. Technol., vol. 69, no. 10,
pp. 11 712–11 727, 2020.
[15] ——, “On the symbol error probability of STBC-NOMA with timing
offsets and imperfect successive interference cancellation, Electronics,
vol. 10, no. 12, p. 1386, 2021.
[16] A. J. Muhammed et al., “Energy-efficient resource allocation for NOMA
based small cell networks with wireless backhauls,” IEEE Trans. Com-
mun., vol. 68, no. 6, pp. 3766–3781, 2020.
[17] Q.-T. Trinh et al., “Performance analysis of NOMA for wireless down-
link in multi-tiers heterogeneous network,” in International Conference
on System Science and Engineering (ICSSE), 2019, pp. 329–334.
[18] O-RAN Alliance, “O-RAN use cases and deployment scenarios,” White
Paper, Feb. 2020.
[19] S. F. Abedin et al., “Elastic O-RAN slicing for industrial monitoring and
control: A distributed matching game and deep reinforcement learning
approach,” IEEE Trans. Veh. Technol., pp. 1–15, 2022.
[20] I. Budhiraja et al., “An energy-efficient resource allocation scheme for
SWIPT-NOMA based femtocells users with imperfect CSI, IEEE Trans.
Veh. Technol., vol. 69, no. 7, pp. 7790–7805, 2020.
[21] F. Zhou et al., “Energy-efficient NOMA enabled heterogeneous cloud
radio access networks,” IEEE Network, vol. 32, no. 2, pp. 152–160,
[22] I. Budhiraja et al., “An energy-efficient resource allocation scheme for
swipt-noma based femtocells users with imperfect csi,” IEEE Trans. Veh.
Technol., vol. 69, no. 7, pp. 7790–7805, 2020.
[23] H. F. Saeed et al., “Joint pairing and resource allocation for backhaul of
small cells using NOMA,” Journal of Computational Science, vol. 45,
p. 101197, 2020.
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Wireless use cases in industrial internet-of-thing (IIoT) networks often require guaranteed data rates ranging from a few kilobits per second to a few gigabits per second. Supporting such a requirement in a single radio access technique is difficult, especially when bandwidth is limited. Although non-orthogonal multiple access (NOMA) can improve the system capacity by simultaneously serving multiple devices, its performance suffers from strong device interference. In this paper, we propose a Q- learning-based algorithm for handling many-to-many matching problems such as bandwidth partitioning, device assignment to sub-bands, interference-aware access mode selection (orthogonal multiple access (OMA), or NOMA), and power allocation to each device. The learning technique maximizes system throughput and spectral efficiency (SE) while maintaining quality-of-service (QoS) for a maximum number of devices. The simulation results show that the proposed technique can significantly increase overall system throughput and SE while meeting heterogeneous QoS criteria.
Full-text available
Cellular networks are envisioned to be a cornerstone in future industrial IoT (IIoT) wireless connectivity in terms of fulfilling the industrial-grade coverage, capacity, robustness, and timeliness requirements. This vision has led to the design of verticals-centric service-based architecture of 5G radio access and core networks. The design incorporates the capabilities to include 5G-AI-Edge ecosystem for computing, intelligence, and flexible deployment and integration options (e.g., centralized and distributed, physical and virtual) while eliminating the privacy/security concerns of mission-critical systems. In this paper, driven by the industrial interest in enabling large-scale wireless IIoT deployments for operational agility, flexible, and cost-efficient production, we present the state-of-the-art 5G architecture, transformative technologies, and recent design trends, which we also selectively supplemented with new results. We also identify several research challenges in these promising design trends that beyond-5G systems must overcome to support rapidly unfolding transition in creating value-centric industrial wireless networks.
Full-text available
Small cells is a promising technique to increase the spectral efficiency or quality-of-service (QoS) of a cellular network as it reduces the distance between the user and the base station (MBS). It has been proposed for indoor as well as outdoor users. For the outdoor scenario, the backhaul (the link from a small cell to the core network) of the small cell has always been a challenging issue. Different techniques have been proposed in the literature such as optical fiber, satellite and wirelessly through MBS. Wireless backhaul through MBS has gained a lot of attention in the literature and different resource allocation with orthogonal resources strategies have been proposed for it as well. On the other hand non-orthogonal, multiple access (NOMA) is a multiple access strategy with the key idea of sharing the same bandwidth between different users with different power levels. The user's data is superimposed and transmitted and successive interference cancellation (SIC) is employed on the receiver end to extract the data. It increases the spectral efficiency of the network as compared to orthogonal schemes. This makes it an ideal candidate to be employed on the backhaul of small cells. With the employment of NOMA on the backhaul of small cells, the pairing of small cells and the allocation of resources need to be handled in an efficient way to exploit the complete advantages of NOMA. In this paper, we investigate the problem of pairing of small cells and the allocation of resources. Small cells’ load information plays vital role to solve this problem. An optimization problem with an objective of maximizing the spectral efficiency of the backhaul network is formulated. The joint problem is decomposed and solved using a heuristic scheme for the backhaul of small cells. Furthermore, our performance evaluation shows that the proposed scheme is able to achieve the nearly-same spectral efficiency as that of exhaustive search with much less complexity.
Full-text available
The combination of non-orthogonal multiple access (NOMA) and cooperative communications can be a suitable solution for fifth generation (5G) and beyond 5G (B5G) wireless systems with massive connectivity, because it can provide higher spectral efficiency, lower energy consumption, and improved fairness compared to the non-cooperative NOMA. However, the receiver complexity in the conventional cooperative NOMA increases with increasing number of users owing to successive interference cancellation (SIC) at each user. Space time block code-aided cooperative NOMA (STBC-CNOMA) offers less numbers of SIC as compared to that of conventional cooperative NOMA. In this paper, we evaluate the performance of STBC-CNOMA under practical challenges such as imperfect SIC, imperfect timing synchronization between distributed cooperating users, and imperfect channel state information (CSI). We derive closed-form expressions of the received signals in the presence of such realistic impairments and then use them to evaluate outage probability. Further, we provide intuitive insights into the impact of each impairment on the outage performance through asymptotic analysis at high transmit signal-to-noise ratio. We also compare the complexity of STBC-CNOMA with existing cooperative NOMA protocols for a given number of users. In addition, through analysis and simulation, we observe that the impact of the imperfect SIC on the outage performance of STBC-CNOMA is more significant compared to the other two imperfections. Therefore, considering the smaller number of SIC in STBC-CNOMA compared to the other cooperative NOMA protocols, STBC-CNOMA is an effective solution to achieve high reliability for the same SIC imperfection condition.
In this work, we design an elastic open radio access network (O-RAN) slicing for the Industrial Internet of things (IIoT). Due to the rapid spread of IoT in the industrial usecases such as safety and mobile robot communications, the IIoT landscape has been shifted from static manufacturing processes towards dynamic manufacturing workflows (e.g., Modular Production System). But unlike IoT, IIoT poses additional challenges such as severe communication environment, networkslice resource demand variations, and on-time information update from the IIoT devices during industrial production. First, we formulate the O-RAN slicing problem for on-time industrial monitoring and control where the objective is to minimize the cost of fresh information updates (i.e., age of information (AoI)) from the IIoT devices (i.e., sensors) with the device energy consumption and O-RAN slice isolation constraints. Second, we propose the intelligent O-RAN framework based on game theory and machine learning to mitigate the problem’s complexity. We propose a twosided distributed matching game in the O-RAN control layer that captures the IIoT channel characteristics and the IIoT service priorities to create IIoT device and small cell base station (SBS) preference lists. We then employ an actor-critic model with a deep deterministic policy gradient (DDPG) in the O-RAN service management layer to solve the resource allocation problem for optimizing the network slice configuration policy under timevarying slicing demand. Furthermore, the proposed matching game within the actor-critic model training helps to enforce the long-term policy-based guidance for resource allocation that reflects the trends of all IIoT devices and SBSs satisfactions with the assignment. Finally, the simulation results show that the proposed solution enhances the performance gain for the IIoT services by serving an average of 50% and 43.64% more IIoT devices than the baseline approaches.
Next generation (NextG) cellular networks will be natively cloud-based and built on programmable, virtualized, and disaggregated architectures. The separation of control functions from the hardware fabric and the introduction of standardized control interfaces will enable the definition of custom closed-control loops, which will ultimately enable embedded intelligence and real-time analytics, thus effectively realizing the vision of autonomous and self-optimizing networks. This article explores the disaggregated network architecture proposed by the O-RAN Alliance as a key enabler of NextG networks. Within this architectural context, we discuss the potential, the challenges, and the limitations of data-driven optimization approaches to network control over different timescales. We also present the first large-scale integration of O-RAN-compliant software components with an open source full-stack softwarized cellular network. Experiments conducted on Colosseum, the world's largest wireless network emulator, demonstrate closed-loop integration of real-time analytics and control through deep reinforcement learning agents. We also show the feasibility of radio access network (RAN) control through xApps running on the near-real-time RAN intelligent controller to optimize the scheduling policies of coexisting network slices, leveraging the O-RAN open interfaces to collect data at the edge of the network.
The deployment of 5G mobile networks has begun in earnest. The 5G network is built on a service-based architecture (SBA) which enables programmability of the control plane of 5G Core (5GC) and supports network slicing (NS) in both core and access networks. NS enables the creation of multiple, isolated network slices tailored for specific services with diverse KPI objectives. SBA of 5GC facilitates core network user plane functions (UPF) being deployed near the network edge, which has triggered intense interest in edge activities such as Multi-access Edge Computing (MEC), whereas the traditional Self-Organizing Network (SON) functions are being enhanced continuously in 5G RAN. Meanwhile, the development of open and smart RAN led by an industry alliance, O-RAN, has received great attention. In this short paper we would like to sort out if and how all of the above would fit together in a consistent and efficient manner regarding their architecture aspects and functionalities.
This paper proposes a framework of users association and power allocation in the two-tier heterogeneous networks with nonorthogonal multiple access (NOMA). Taking account of practical wireless communication environment, the channel uncertainties are considered and described in the form of probability constraint. In order to maximize the total energy efficiency of the system and reduce the intra-cell and inter-cell interference, the joint optimization problem is formulated with the uncertain channel gain. The probability constraint is transformed the deterministic one based on the integral transformation. With the aid of relaxation variable and successive convex approximation method, the original integer non-convex optimization problem is divided into two solvable convex subproblems. The user association algorithm and power control algorithm are presented to determine the optimal resource allocations. The simulation results show that the proposed algorithm is effective and robust to the dynamic communication environment.