Journal of Network and Computer Applications 225 (2024) 103856

Contents lists available at ScienceDirect

Journal of Network and Computer Applications

journal homepage: www.elsevier.com/locate/jnca

Throughput and delay analysis of cognitive M2M communications

Soumen Mondal a , Luca Davoli b ,∗, Sanjay Dhar Roy a , Sumit Kundu a , Gianluigi Ferrari b ,
Riccardo Raheli b
a Department of Electronics and Communication Engineering, National Institute of Technology, Durgapur, India
b Department of Engineering and Architecture, University of Parma, Parma, 43124, Italy

ARTICLE INFO ABSTRACT

Keywords: In this paper, we analyze throughput and delay performance of clustered Machine Type Communication (MTC)
Cognitive radio (CR) devices which access an eNodeB utilizing a primary spectrum in underlay mode. We assume that the MTC
Machine type communication (MTC) devices form two clusters and there is an optimal preamble allocation between the two clusters to maximize the
Machine-to-machine (M2M)
throughput. We further investigate the impact of the tolerable interference threshold on throughput, successful
eNodeB
preamble decoding probability, and delay. Then, the impact of the preamble partition factor and the access
barring factor on throughput and delay is analyzed. Finally, we evaluate the impact of the number of devices,
retransmission requests, and preamble partitions on the delay.

1. Introduction (Di et al., 2019). A dynamic, delay-tolerant, and delay-sensitive scheme

for the allocation of Random Access CHannel (RACH) resources be-
Machine-to-Machine (M2M) communication is evolving as a promis- tween two clusters of devices is proposed in Li et al. (2015). The
ing paradigm in emerging 5G networks, which will support Internet of proposed scheme is shown to improve access probability and reduce
Things (IoT). IoT has gained phenomenal momentum due to significant access delay. While Machine Learning (ML)-based RA schemes have
commercial and research interest. Furthermore, Narrow-Band IoT (NB- gathered considerable attention in recent years (Zhang et al., 2023,
IoT) has been standardized under the Third Generation Partnership 2022), the optimization of access delay in these schemes continues to be
Project (3GPP) to provide reliable connections among large numbers a challenge. In Jiang et al. (2018), repeated preamble retransmissions
of inexpensive low power IoT devices over large areas (Ha et al., in the presence of collisions are considered in the RACH procedure and
2018). M2M communications represent an enabling technology for the RACH success probability is evaluated, in the end demonstrating
IoT networking, as they differ from Human-to-Human (H2H) commu- how the utilization of a repetition scheme has been shown to notably
nications in several distinct features, such as: small amount of data enhance the RACH success probability. The authors in Guo et al.
transmission; sporadic data transmission; delay; tolerance; and group- (2022) propose a collision-aware ACB RA scheme that incorporates
based operations (3GPP, 2011). In this scenario, a substantial number dynamic adjustments to both the ACB factor and preamble resources.
of IoT devices will face a shortage of available spectrum. Cognitive In Chowdhury and De (2022), the authors have presented an approach
Radio (CR) has emerged as a potential approach to solve the conflicting
that involves dynamically assigning higher priority in ACB to Machine
problems of shortage and under-utilization of the available spectrum.
Types Devices (MTDs) whose data queue sizes approach their buffer
It allows unlicensed secondary users to access the spectrum allocated
limits. To mitigate contention during preamble access, in Swain and
to licensed users or Primary Users (PUs) either opportunistically (in
Subudhi (2023) Deep Learning (DL)-based models have been adopted to
the absence of PUs) or without creating unacceptable interference to
design a RACH procedure. This approach aims at predicting incoming
existing PUs (if the presence of PUs) (Mitola and Maguire, 1999).
connection requests in advance, and allowing for proactive allocation
CR-enabled IoT devices will be able to optimize spectrum utiliza-
of uplink resources to UEs.
tion through spectrum sensing and dynamic spectrum access capabili-
A transmission scheme for low volume data is proposed in Oh and
ties (Aijaz and Aghvami, 2015). The throughput performance of M2M
Shin (2017), in the context of 3GPP NB-IoT networks avoiding a Radio
communications is evaluated in Lee et al. (2011), where two different
approaches for dividing the available Random Access (RA) pream- Resource Control (RRC) set up (Debbabi et al., 2022): the access success
bles are considered using an Access Class Barring (ACB) mechanism probability and uplink utilization are shown to improve significantly.

∗ Corresponding author.
E-mail addresses: [email protected] (S. Mondal), [email protected] (L. Davoli), [email protected] (S. Dhar Roy),
[email protected] (S. Kundu), [email protected] (G. Ferrari), [email protected] (R. Raheli).

https://doi.org/10.1016/j.jnca.2024.103856
Received 1 May 2023; Received in revised form 20 December 2023; Accepted 28 February 2024
Available online 2 March 2024
1084-8045/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

However, no cognitive scenario is considered in Oh and Shin (2017). A

cognitive cellular system architecture for M2M communications, along
with possible transmission scenarios, is presented in Ejaz and Ibnkahla
(2015), where energy-efficient resource allocation is also considered.
An advanced power allocation policy in a CR-based M2M network,
where multiple unlicensed M2M devices share the same licensed spec-
trum of a PU, is investigated in Yao et al. (2014). Underlay CR enabling
Machine Type Communication (MTC) devices sharing the spectrum of
PUs, such as cellular UE, is studied in Alhussien and Gulliver (2022),
where power allocation problems are addressed for improving the
energy efficiency of MTC devices by minimizing their power consump-
tion or maximizing their energy efficiency. A cognitive clustered M2M
communication with a joint selection of cellular equipment and M2M
devices is analyzed in Abdullah et al. (2019) to reduce outage of M2M.
A hybrid duplex Base Station (BS) switching between half-duplex and
full-duplex modes to attain the best performance is considered.
In Li et al. (2018), the throughput of a RA narrowband CR IoT
network is maximized through selection of optimal sensing parameter Fig. 1. System model composed by two clusters of MTC devices co-existing with a
via collaborative sensing. In detail, the trade-off between RA NB-CR- transmitter–receiver pair of PUs.
IoT network throughput and sensing accuracy is investigated, and a
set of optimal (throughput maximizing) sensing parameter is derived.
Throughput and delay performance of massive multi-group RA M2M • We evaluate the throughput and delay performance of two clus-
communication in industrial IoT (IIoT) is studied in Zhang et al. (2019).
ters of MTC devices.
MTDs are divided into multiple groups on the basis of their delay
• We develop a novel analytical framework to evaluate the success-
requirement, and a double queue model is used to characterize the
ful decoding probability of MTC devices of two clusters (as shown
access behavior of each MTD device. To this end, throughput of delay
in Fig. 1) at their associated eNBs, with all the devices sharing the
tolerant MTDs is maximized under delay constraints of delay-sensitive
frequency band of a transmitter–receiver pair of PUs in underlay
MTDs by an appropriately tuning back-off parameters of delay-sensitive
mode.
MTDs.
• We derive a novel analytical expression for optimal RA preamble
Given the scarcity in spectrum due to the exponential growth of
partition (Yang et al., 2023) between the two clusters, in order to
the number of MTC devices, the licensed spectrum of a PU needs to
be utilized in cognitive mode. In the process of accessing the primary maximize the throughput.
spectrum in underlay mode, the analysis of throughput and delay of • We derive the impact of the tolerable interference threshold at
MTC devices, as a performance metrics, is important while satisfying the PU receiver and the portion of tolerable interference allowed
the Quality of Service (QoS) of the PU. MTC devices may occur in by each cluster on the decoding probability and throughput.
cluster or groups and may be controlled by secondary BSs or eNBs of Furthermore, we evaluate the impacts of the primary outage
that group. However, while the above-discussed literature on cognitive constraint, the ACB mechanism, and the number of preambles on
M2M communications considers several aspects of M2M communica- the throughput and delay.
tions in cognitive mode, to the best of our knowledge, a comprehensive • We develop a MATLAB-based simulation framework in order to
analysis of throughput and delay of a cluster of MTC devices, with an validate the results predicted by our novel analytical framework.
appropriate division of RA preambles while satisfying QoS of licensed
The remainder of this paper is organized as follows. The proposed
PU in terms of an outage constraint in underlay cognitive mode, is miss-
communication system model is introduced in Section 2. The perfor-
ing. Furthermore, the impact of tolerable interference in an underlay
mance evaluation is carried out in Section 3. In Section 4, simulation
mode on the performance of cluster-based MTC devices and appropriate
division of RA preambles need to be investigated. results are presented and discussed. Finally, in Section 5 we draw our
Given the ‘‘gaps’’ highlighted in the above literature works and conclusions.
compared, for the sake of completeness, in Table 1, in this paper the
performance of cellular-based IoT devices coexisting with a primary 2. System model
network in an underlay mode is examined. In particular, two clusters of
IoT devices, which share the frequency band of a transmitter–receiver We consider two eNBs with two clusters of IoT devices that co-exist
pair of PUs in an underlay mode while communicating to their cor- with a transmitter–receiver pair of PUs. In cluster #1, MTC devices
responding eNB, are considered. We evaluate the successful decoding access eNodeB1 whereas in cluster #2 MTC devices access eNodeB2,
probability of a MTC device belonging to a cluster and satisfying a as depicted in Fig. 1. In the proposed system model, we explore a
cognitive constraint (i.e., the interference at the PU receiver is kept CR scenario in which MTC devices can access the spectrum of the
below a tolerable threshold). Partitioning of RACH preambles between PU as long as the interference generated for the PU, resulting from
two clusters is considered in order to maximize the throughput. An an- the MTC device cluster, remains below the PU’s tolerable interference
alytical expression for the optimal number of preambles to be allocated limit. In cluster #1, MTC devices generate a fraction 𝜂 ∈ (0, 1) of the
to each cluster to maximize the throughput is derived. To the best of our tolerable interference level, while a fraction (1 − 𝜂) of the tolerable
knowledge, the investigation of a transmission scheme of two clusters interference is generated by the MTC devices in cluster #2. Therefore,
of IoT devices sharing a common spectrum of a transmitter–receiver
the transmission power of the MTC devices is constrained so that it does
pair of PUs in an underlay mode and the analysis of throughput and
not exceed the interference threshold at the PU (Zhang et al., 2013).
delay performance for such networks are novel.
MTC devices attempt to access the network following a contention-
Overall, the major contributions of our paper can be highlighted as
based RA procedure between the MTC devices and the eNB, which
follows.
consists (as shown in Fig. 2) of the following four steps: (i) preamble
• We propose a novel network model with two clusters of MTC transmission; (ii) RA response; (iii) connection request; (iv) connection
devices accessing the licensed spectrum of a PU. completion.

2
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

Table 1
Comparison between the proposed work and literature references.
Ref. Device Underlay Overlay Metric Clustering Optimization Preamble partition
Li et al. (2018) NB-IoT ✗ ✓ Throughput ✗ Throughput ✗
Zhang et al. (2019) MTC ✗ ✗ Throughput, delay ✓ Throughput via choice ✗
of back-off parameter Partition of back-off
parameter
Abdullah et al. MTC ✗ ✓ Outage ✓ Outage probability ✗
(2019) Sharing cellular BS
in TDMA mode
Our work MTC ✓ ✗ Throughput, delay ✓ Throughput via optimal ✓
Two clusters preamble partition

is the tolerable interference threshold power (dimension: [mW]). The

probability of outage is bounded as follows:

pu−out1 ≤ 𝛿 . (2)
(pu)
Taking into account the exponential distribution of 𝑖,1 , it follows that:

⎧ ⎫
⎪ (pu) 𝜂𝐼th ⎪
1 − pu−out1 = Pr ⎨𝑖,1 ≤
⎪ 𝑃TX(pu) ⎬
⎪
⎩ 𝑖,1 ⎭

−𝜂𝐼th ∕𝜇1 𝑃 (pu)

TX
=1−𝑒 𝑖,1 . (3)
Fig. 2. Contention-based RA procedure of MTC devices accessing the eNB.
Using (2) in (3), one obtains
−𝜂𝐼th ∕𝜇1 𝑃 (pu)
𝑒
TX
𝑖,1 ≤𝛿 (4)
Our goal is to investigate the probability of successful access request
to an eNB. Whenever an MTC device attempts to access an eNB, it and, then,
first transmits an access request. In particular, an MTC device selects 𝜂𝐼th
≥ − log (𝛿) ,
a preamble of the RACH to transmit an access request according to a 𝜇1 𝑃TX(pu)
𝑖,1
slotted Aloha protocol. In contention-based RA, a slotted Aloha protocol ( )
𝜂𝐼th 1
can be used with the number of available preambles equal to the ≥ log ,
𝜇1 𝑃TX(pu) 𝛿
number of slots (Li et al., 2015). If two or more devices select the same 𝑖,1
preamble during the same slot, thus making the eNB unable to decode 𝜂𝐼th
𝑃TX(pu) ≤ ( ),
any preamble, a collision occurs. After the Random Access Response 𝑖,1
𝜇1 log 𝛿1
(RAR) for each successfully decoded preamble, the eNB generates an
identifier and, then, transmits a RAR to the MTC devices. Next, the MTC −𝜂𝐼th
𝑃TX(pu) ≤ . (5)
device transmits a connection request message with a User Equipment 𝑖,1 𝜇1 log(𝛿)
(UE) identifier to the eNB. Eventually, the eNB transmits a connection Similarly, the outage probability in cluster #2 can be expressed as
completion message to the MTC device. If an MTC device does not { }
(pu)
receive a connection completion message from the eNB, a failure in pu−out2 = Pr 𝑃TX(pu) 𝑖,2 > (1 − 𝜂)𝐼th (6)
𝑖,2
the contention completion occurs and the MTC device attempts a new
(pu)
access. where 𝑖,2 is the channel gain (exponentially distributed with mean
equal to 𝜇2 ) from the 𝑖th device in cluster #2 to the PU. Imposing that
pu−out2 ≤ 𝛿 and carrying out the same analytical steps above, it follows
3. Performance analysis
that:
(𝜂 − 1)𝐼th
In the following, the probability of successful access of an MTC 𝑃TX(pu) ≤ . (7)
𝑖,2 𝜇2 log(𝛿)
device to the eNB, as a function of its transmit power, is derived. We
consider that the allowed PU outage probability at PU is denoted as Therefore, the maximum transmit powers of any MTC devices in the
𝛿 ∈ (0, 1). We assume that all the channels are Rayleigh-faded so that two clusters are the ones specified at the right-hand sides of (5) and (7).
the channel power gains are exponentially distributed. The probability The decoding probability of the 𝑖th MTC device (𝑖 ∈ {1, 2, … , 𝑁1 })
of an outage can be expressed as follows: in cluster #1 at eNodeB1 can be expressed as follows (Oh and Shin,
{ } 2017):
(pu)
pu−out1 = Pr 𝑃TX(pu) 𝑖,1 > 𝜂𝐼th ⎧ ⎫
𝑃TX(e1) (e1)
𝑖,1
⎪ 𝑖,1
𝑖,1 ⎪
⎧ ⎫ dec(e1) = Pr ⎨ > 𝛾1 ⎬ (8)
(e1) (e1)
⎪ (pu) 𝜂𝐼th ⎪ 𝑖,1 ⎪ 𝑃TX(pu) (pu) + 𝑃TX(e1) 𝑖,2 + 𝑛0 ⎪
= 1 − Pr ⎨𝑖,1 ≤ (1) ⎩ ⎭
𝑃TX(pu) ⎬
𝑖,2

⎪ ⎪
⎩ 𝑖,1 ⎭
where: (e1)
𝑖,1
is the channel gain (adimensional and exponentially dis-
where: 𝑃TX(pu) is the transmit power (dimension: [mW]) of the 𝑖th MTC tributed with mean of 1/𝜆𝑣 ) from the 𝑖th MTC device of cluster #1
to eNodeB1; 𝑃TX(pu) is the PU transmit power (dimension: [mW]); (e1)
𝑖,1
(pu)
device from cluster #1 to PU; 𝑖,1
is the channel gain (adimensional), 𝑖,2

exponentially distributed with mean 𝜇1 , from the 𝑖th device of clus- is the channel gain (adimensional and exponentially distributed with
ter #1 to PU; 𝜂 is the reserved fraction (adimensional) of 𝐼th , where 𝐼th mean of 1∕𝜆𝑥 ) from the 𝑖th device of cluster #2 to eNodeB1; (e1)
(pu)
is the

3
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

channel gain (adimensional and exponentially distributed with mean of before letting the MTC devices access the network. In a certain RA slot,
1∕𝜆𝑦 ) from PU to eNodeB1; 𝑃TX(e1) is the transmit power (dimension: we denote: 𝑁1 as the number of MTC devices in cluster #1; 𝑁2 as the
𝑖,1 number of MTC devices in cluster #2; 𝑁tot as the total number of the
[mW]) from the 𝑖th device in cluster #1 to eNodeB1; 𝑃TX(e1) is the MTC devices, i.e., 𝑁tot = 𝑁1 + 𝑁2 . We define the preamble partition 𝛽
𝑖,2
as follows (Li et al., 2015):
transmit power (dimension: [mW]) from the 𝑖th device in cluster #2
to eNodeB1; 𝛾1 is the threshold (adimensional) above which an MTC 𝑀1
𝛽≜ (14)
device in cluster #1 is decodable at eNodeB1; and 𝑛0 is the thermal 𝑀2
noise power (dimension: [mW]), which can be neglected in the scenario
of interest. 𝑀1 + 𝑀2 = 𝑀tot (15)
Let 𝑉 = 𝑎𝑋 + 𝑏𝑌 , where 𝑎 ≜ 𝑃TX(e1) , 𝑏 ≜ 𝑃TX(pu) , 𝑋 ≜ (e1)
𝑖,2
, 𝑌 ≜ (e1)
(pu)
, where: 𝑀1 represents the number of preambles allocated for cluster #1;
𝑖,2
𝑈 ≜ (e1) . From (8), one can write: 𝑀2 represents the number of preambles allocated for cluster #2; and
𝑖,1
𝑀tot is the total number of preambles (Li et al., 2015). From (14) and
⎧ ⎫ (15), it follows that:
⎪𝑈 𝛾1 ⎪
dec(e1) = Pr ⎨ >
𝑃TX(pu) ⎬
𝑀tot 𝛽
𝑖,1 ⎪𝑉 ⎪ 𝑀1 =
1+𝛽
(16)
⎩ 𝑖,1 ⎭
{ } 𝑀tot
= Pr 𝑊 > 𝛼1 𝑀2 = . (17)
1+𝛽
= 1 − 𝐹𝑊 (𝛼1 ) (9) Since every MTC device selects the preamble randomly from the
where 𝑊 = 𝑈 ∕𝑉 and 𝛼1 = 𝛾1 ∕𝑃TX(pu) . available pool, collisions may occur if more than one MTC device
𝑖,1
In order to evaluate dec(e1) , the Probability Density Function (PDF) selects the same preamble (Laya et al., 2013). Moreover, according
𝑖,1 to the contention-based RA procedure, slotted Aloha is the adopted
of 𝑉 = 𝑎𝑋 + 𝑏𝑌 can be expressed as multiple access protocol, with the number of available preambles equal
𝜆𝑥 𝜆𝑦 ( −𝜆 𝑣 𝑣) to the number of slots (Li et al., 2015). In this way, the access success
𝑓𝑉 (𝑣) = 𝑒 𝑥 𝑎 − 𝑒−𝜆𝑦 𝑏 . (10)
𝜆𝑦 𝑎 − 𝜆𝑥 𝑏 probability can be expressed as (Li et al., 2015)
𝑁
The corresponding Cumulative Distribution Function (CDF) of 𝑃𝑠 = 𝑒− 𝑀 (18)
𝐹𝑊 (𝛼1 ) in (9) can then be written as
{ } where 𝑁 is the number of devices and 𝑀 is the number of preambles
𝑈
𝐹𝑊 (𝛼1 ) = Pr < 𝛼1 available within a RA slot. The average number of devices successfully
𝑉
∞ ( ) completing the access attempts from the two clusters can be expressed
= 𝐹𝑈 𝛼1 𝑣 𝑓𝑉 (𝑣)𝑑𝑣 as follows (Li et al., 2015):
∫0
𝜆𝑥 𝜆𝑦 𝑁1 𝑓1
− 𝑀
=1− ( )(𝜆 ). (11) 𝑁ssa,1 = 𝑁1 𝑓1 𝑒 1 (19)
𝜆𝑥 𝑦
𝑎𝑏 𝑎 + 𝛼1 𝜆𝑣 𝑏
+ 𝛼1 𝜆𝑣 𝑁2 𝑓2
− 𝑀
𝑁ssa,2 = 𝑁2 𝑓2 𝑒 2 (20)
Finally, one obtains (see Appendix for details)
where: 𝑓1 is the ACB factor of cluster #1; and 𝑓2 is the ACB factor of
dec(e1) = 1 − 𝐹𝑊 (𝛼1 )
𝑖,1 cluster #2.
𝜆𝑥 𝜆𝑦 (12) At this point, we can express the throughput of a device from
= ( )(𝜆 ). cluster #1, the throughput of a device from cluster #2, and the total
𝜆𝑥 𝑦
𝑎𝑏 + 𝛼1 𝜆𝑣 + 𝛼1 𝜆𝑣
𝑎 𝑏 throughput as follows, respectively (Lee et al., 2011; Li et al., 2015):
Similarly, the decoding probability of the 𝑖th MTC device
1 = 𝑁ssa,1 dec(e1) (21)
(𝑖 ∈ {1, 2, … , 𝑁2 }) in cluster #2 at eNodeB2 can be expressed as: 𝑖,1

⎧ 𝑃TX(e2) (e2) ⎫ 2 = 𝑁ssa,2 dec(e2) (22)

⎪ 𝑖,2
𝑖,2 ⎪ 𝑖,2
dec(e2) = Pr ⎨ > 𝛾2 ⎬
𝑖,2 𝑃 (e2)
⎪ TX(pu) (pu)
 + 𝑃 (e2)  (e2)
+ 𝑛 ⎪ tot = 1 + 2 . (23)
0
⎩ TX𝑖,1 𝑖,1 ⎭
′
𝜆𝑥 𝜆𝑦′ Finally using (12), (13), (19)–(23), we can write:
{ }
= ( ′ ) ( 𝜆′ ) (13) −𝑁1 𝑓1 (1 + 𝛽)
𝜆 𝑦 tot = dec(e1) 𝑁1 𝑓1 exp
𝑎1 𝑏1 𝑎𝑥 + 𝛼2 𝜆′𝑣 + 𝛼 2 𝜆 ′
𝑣 𝑖,1 𝑀 𝛽
1 𝑏 1 { tot }
−𝑁2 𝑓2 (1 + 𝛽)
where: 𝛼2 ≜ 𝛾2 ∕𝑃TX(e2) ; 𝑎1 ≜ 𝑃TX(e2) ; 𝑏1 ≜ 𝑃TX(pu) ; (e2) is the chan- + dec(e2) 𝑁2 𝑓2 exp . (24)
𝑖,2
𝑖,2
𝑖,1
𝑖,2 𝑀tot
nel gain from the 𝑖th MTC device in cluster #2 to eNodeB2, which At this point, our goal is to find the optimal value of 𝛽 which
is exponentially distributed with mean of 1∕𝜆′𝑣 ; (e2) is the channel maximizes tot . Imposing 𝜕tot ∕𝜕𝛽 = 0, where
𝑖,1 { }
gain from the 𝑖th MTC device of cluster #1 to eNodeB2, which is 𝜕tot −𝑁1 𝑓1
= dec(e1) 𝑁1 𝑓1 exp
exponentially distributed with mean 1∕𝜆′𝑥 ; (e2) is channel gain from 𝜕𝛽 𝑖,1 𝑀tot
(pu) { }( )
−𝑁1 𝑓1 𝑁1 𝑓1
PU to eNodeB2, which is exponentially distributed with mean 1∕𝜆′𝑦 ; exp + dec(e2) 𝑁2 𝑓2
𝑃TX(e2) is the transmit power from the 𝑖th MTC device in cluster #1 to 𝑀tot 𝛽 𝑀tot 𝛽 2 𝑖,2
{ } { }( )
𝑖,1 −𝑁2 𝑓2 −𝑁2 𝑓2 𝛽 −𝑁2 𝑓2
eNodeB2; 𝛾2 is the threshold above which an MTC device in cluster #2 exp exp (25)
𝑀tot 𝑀tot 𝑀tot
is decodable at eNodeB2.
it follows:
{ }( )
3.1. Dynamic adjustment of preamble partition −𝑁1 𝑓1 1
dec(e1) 𝑁12 𝑓12 exp (1 + 𝛽)
𝑖,1 𝑀tot 𝛽 𝛽2
{ }
After clustering the attempting devices into two groups, the eNB −𝑁2 𝑓2
dynamically determines the RA preamble partition between the clusters = dec(e2) 𝑁22 𝑓22 exp (1 + 𝛽) . (26)
𝑖,2 𝑀tot

4
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

Considering the logarithm of both sides of (26), one obtains: Indicating as 𝑇𝑠 the duration of a transmission act, the expected
( ) { } ( ) delay E(𝐷1 ) can be written as follows:
𝑁1 𝑓1 1
log dec(e1) + 2 log(𝑁1 ) + 2 log(𝑓1 ) − (1 + 𝛽) + log ( )
𝑖,1 𝑀tot 𝛽 𝛽2 ∑
𝑊 −1
(1 − 𝑓 )𝑗 𝑓
( ) { } E(𝐷1 ) = 𝑇𝑠 𝑗∑ 1 − dec(e1) . (32)
𝑁2 𝑓2 𝑊 −1 𝑘
= log dec(e2) + 2 log(𝑁2 ) + 2 log(𝑓2 ) − (1 + 𝛽) (27) 𝑘=0 (1 − 𝑓 ) 𝑓
𝑖,1
𝑗=0
𝑖,2 𝑀tot
In order to account for the delay in the absence of collisions,
and, finally,
according to 3GPP (2011) we add 𝑇𝑠 ∕2, which is the expected value
⎛ dec(e1) ⎞ ( ) ( ) for a uniform delay in [0, 𝑇𝑠 ] (Tyagi et al., 2012). Thus, the expected
log ⎜
𝑖,1
⎟ + 2 log 𝑁1 + 2 log 𝑓1 delay in (32) can be expressed as follows:
⎜ dec(e2) ⎟ 𝑁2 𝑓2
⎝ 𝑖,2 ⎠ ( )
{ } { } ( ) 𝑇
𝑊∑ −1
(1 − 𝑓 )𝑗 𝑓
=
𝑁1 𝑓1
(1 + 𝛽) −
𝑁2 𝑓2
(1 + 𝛽) − log
1
. (28) E(𝐷1 ) = 𝑠 + 𝑇𝑠 𝑗∑ 1 − dec(e1) (33)
2 𝑊 −1 𝑘
𝑘=0 (1 − 𝑓 ) 𝑓
𝑖,1
𝑀tot 𝛽 𝑀tot 𝛽2 𝑗=0
( )
The optimal value of 𝛽 which maximizes the throughput is obtained 𝑇 𝑇𝑠 ∑
𝑊 −1

by solving (28). Since a closed-form expression of the optimal value of = 𝑠 +∑ 𝑗(1 − 𝑓 )𝑗 1 − dec(e1) . (34)
2 𝑊 −1 𝑘
𝑘=0 (1 − 𝑓 ) 𝑗=0
𝑖,1
𝛽 is not available, we obtain such a value by solving the transcendental
equation (28) numerically. The highest throughput1 is obtained by Assuming 𝓁 < 1, it holds that
inserting the obtained value of 𝛽 in (24). ∑
𝑊 −1
1 − 𝓁𝑊
𝓁𝑘 = (35)
3.2. Delay analysis of M2M communications 𝑘=0
1−𝓁
and
We now investigate the system performance in terms of delay. MTC
∑
𝑊 −1
1 + (𝑊 − 1)𝓁 𝑊 − 𝑊 𝓁 𝑊 −1
devices transmit RA requests to the eNB with a period of duration 𝑇𝑠 . 𝑘𝓁 𝑘 = 𝓁 . (36)
If more than one MTC device sends requests to the eNB using the same 𝑘=0 (1 − 𝓁)2
preamble at the same time, then RA requests collide and the eNB cannot
Using (34) and denoting 𝛾 ≜ 1 − 𝑓 , the average delay for an MTC
decode any RA attempt. Whenever a collision occurs, contention is not
device in cluster #1 can be expressed as follows:
considered to be resolved. MTC devices can identify the contention ( )
resolution results at the last step of the RA. If contention is resolved, 𝑇 𝑇 𝛾(1 + (𝑊 − 1)𝛾 𝑊 − 𝑊 𝛾 𝑊 −1 )
E(𝐷1 ) = 𝑠 + 𝑠 1 −  (e1)
dec𝑖,1
. (37)
then MTC devices will get into RRC connected mode to transmit the 2 (1 − 𝛾)(1 − 𝛾 𝑊 )
data to the eNB. MTC devices repeat the process of transmitting the Similarly, the average delay for an MTC device in cluster #2 be-
preambles when contention is not resolved. In this case, devices wait for comes
a specific period of time, denoted as back-off interval (ranging from 0 ms ( )
𝑇 𝑇 𝛾(1 + (𝑊 − 1)𝛾 𝑊 − 𝑊 𝛾 𝑊 −1 )
to 960 ms Tyagi et al., 2012; Althumali et al., 2020), before retransmit- E(𝐷2 ) = 𝑠 + 𝑠 1 −  (e2)
dec𝑖,2
. (38)
ting to the eNB. For simplicity, we consider a back-off interval equal to 2 (1 − 𝛾)(1 − 𝛾 𝑊 )
0 ms, i.e., no back-off (Tyagi et al., 2012). If more than one MTC device Using expressions (12) and (13) for dec(e1) and dec(e2) in (37) and
𝑖,1 𝑖,2
selects the same preamble at the same time, then they re-transmit in (38), respectively, one obtains the following final expressions of the
the next time. In our system model, we assume that (i) MTC devices average delays:
re-transmitting the preambles randomly choose a preamble from the
available 𝑀tot preambles and (ii) choosing the same preamble again is 𝑇𝑠 𝑇𝑠 𝛾(1 + (𝑊 − 1)𝛾 𝑊 − 𝑊 𝛾 𝑊 −1 )
E(𝐷1 ) = +
allowed. Denoting the preamble transmission success probability as 𝑓 , 2 (1 − 𝛾)(1 − 𝛾 𝑊 )
the probability of 𝑐 collisions before a success can be written as (Tyagi ⎛ ⎞
et al., 2012) ⎜ 𝜆𝑥 𝜆𝑦 ⎟
⋅ ⎜1 − ( )(𝜆 )⎟ (39)
( ) ⎜ 𝜆
1 𝑣 ⎟
𝑦
𝑎𝑏 𝑎𝑥 + 𝛼1 𝜆𝑣 + 𝛼 𝜆
Pr {𝑐 consecutive collisions} = (1 − 𝑓 )𝑐 𝑓 1 − dec(e1) (29) ⎝ 𝑏 ⎠
𝑖,1
𝑇𝑠 𝑇𝑠 𝛾(1 + (𝑊 − 1)𝛾 𝑊 − 𝑊 𝛾 𝑊 −1 )
under the assumption of 𝑐 independent unsuccessful consecutive trans- E(𝐷2 ) = +
2 (1 − 𝛾)(1 − 𝛾 𝑊 )
missions, each with probability (1 − 𝑓 ), followed by a successful trans-
⎛ ⎞
mission, with probability 𝑓 , at the (𝑐 + 1)th attempt. ⎜ 𝜆′𝑥 𝜆′𝑦 ⎟
⎜ )⎟ .
The probability of an MTC device to experience 𝑐 ∈ {0, 1, … , 𝑊 − 1} ⋅ 1−
⎜ ( ′ ) ( 𝜆′ ⎟
(40)
collisions can be expressed as follows: 𝜆 𝑦
⎜ 𝑎1 𝑏1 𝑎𝑥 + 𝛼2 𝜆′𝑣 + 𝛼2 𝜆′𝑣 ⎟
⎝ 1 𝑏1 ⎠
Pr {MTC device to experience 𝑐 consecutive collisions} =
Pr {𝑐 consecutive collisions} 4. Results
= ∑
𝑊 −1
𝑘=0 Pr {𝑘 consecutive collisions}
( ) For the sake of completeness, Table 2 summarizes the notation
(1 − 𝑓 )𝑐 𝑓 adopted in the analytical formulation at a glance, as well as the
= ∑ 1 − dec(e1) (30)
𝑊 −1 𝑘
𝑘=0 (1 − 𝑓 ) 𝑓 specific values chosen for the involved parameters in the experimental
𝑖,1

where 𝑊 is the maximum number of allowed retransmissions. The evaluation. In Fig. 3, the throughput is shown as a function of the
expected number of collisions E(𝐶) is (Tyagi et al., 2012) interference threshold 𝐼th . The number of preambles allocated for clus-
ter #1 (namely, 𝑀1 ) is larger than the number of preambles allocated
∑
𝑊 −1
(1 − 𝑓 )𝑗 𝑓 ( )
for cluster #2 (namely, 𝑀2 ). The throughput performance at eNodeB1
E(𝐶) = 𝑗∑ 1 − dec(e1) . (31)
𝑊 −1 (denoted as 1 ) is better than that at eNodeB2 (denoted as 2 ). As the
𝑗=0 𝑘=0 (1 − 𝑓 )𝑘 𝑓 𝑖,1
interference threshold increases, the throughput of both eNBs increases.
A higher value of the interference threshold leads to an increase of the
1
Considering the throughput expression in (24) it can be verified that tot transmit power of the devices, which, in turn, increases the decoding
is monotonically increasing for 𝛽 between 0 and the value which solves (28), probability and the throughput.
whereas it is monotonically decreasing for larger values. Hence, the value of 𝛽 In Fig. 4, the decoding probability is shown as a function of the
which solves (28) is indeed the maximizer of the throughput. interference threshold 𝐼th . As the value of 𝜂 is higher in cluster #1 than

5
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

Table 2
Notation adopted in the analytical formulation at a glance, and consequent values chosen for those parameters for the experimental
evaluation.
Parameters Value Description
(pu) (pu)
𝜇1 , 𝜇2 , 𝜆𝑣 , 𝜆𝑥 , 𝜆𝑦 , 𝜆′𝑣 , 𝜆′𝑦 , 𝜆′𝑥 1 Mean of 𝑖,1 , 𝑖,2 , (e2)
𝑖,2
, (e1)
𝑖,2
, (e1)
(pu)
, (e2)
𝑖,2
, (e2)
(pu)
, (e2)
𝑖,1
, respectively (Zhang et al., 2013)
𝑃TX(pu) 5 W PU’s transmit power
𝑁1 60 Number of devices in cluster #1
𝑁2 50 Number of devices in cluster #2
𝑀1 30 Number of preambles allocated for cluster #1 (Vardakas et al., 2015)
𝑀2 24 Number of preambles allocated for cluster #2 (Vardakas et al., 2015)
𝑓1 0.7 ACB factor of cluster #1 (Tello-Oquendo et al., 2017)
𝑓2 0.5 ACB factor of cluster #2 (Tello-Oquendo et al., 2017)
𝛾1 1 dB SINR threshold for cluster #1
𝛾2 2 dB SINR threshold for cluster #2
𝜂 0.6 Fraction of 𝐼th
𝛿 0.01 PU outage constraint

Fig. 3. Throughput as a function of the interference threshold 𝐼th .

Fig. 4. Decoding probability dec as a function of the interference threshold 𝐼th .

in cluster #2, the decoding probability of an MTC device in cluster #1 the same preamble during the same slot so that the eNB is unable to
is higher than that of an MTC device in cluster #2. If we increases the decode any preamble: therefore, a larger number of collisions occurs
tolerable interference threshold level at PU, the decoding probability and the throughput at the eNB reduces.
increases at the eNB (in both cluster #1 and cluster #2).
In Fig. 6, the throughput is shown as a function of the interference
In Fig. 5, the throughput at the eNB (for both cluster #1 and
threshold, for various values of the outage constraint. The throughput
cluster #2) is shown as a function of the interference threshold for
various values of the ACB factor. It can be observed that the throughput is higher for higher outage constraint, as higher values of the outage
is higher at both eNBs when the ACB factor is 0.5. For higher values of constraint allow the transmit power of the devices to increase. The
the ACB factor, the throughput reduces. As the ACB factor increases, the results in Table 3 highlight that the throughput is maximized in cor-
traffic intensity increases because of a larger number of access attempts. respondence to the optimal value of the preamble partition 𝛽 (which
Because of this, there is a higher chance that two or more devices select solves (28)) that depends on the values of 𝑁1 and 𝑁2 .

6
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

Fig. 5. Throughput as a function of the interference threshold 𝐼th for various values of the ACB factor 𝑓 .

Fig. 6. Throughput as a function of the interference threshold 𝐼th for various values of the outage constraint 𝛿.

with higher interference threshold leads to an increase of the transmit

Table 3
Optimal value of the preamble partition 𝛽 for various values of 𝑁1 , 𝑁2 , 𝑀1 , and 𝑀2 . power of the devices, which increases the decoding probability. There-
𝑁1 𝑁2 𝑀1 𝑀2 𝛽 fore, increasing the interference threshold reduces the average delay.
10 50 18 36 0.48
20 40 30 24 1.21 In Fig. 10, the average delay is shown as a function of the number of
30 60 32 22 1.45 devices for various values of maximum number of times (𝑊 ) a request
50 50 40 14 2.85 is made to the eNB. The average delay is higher for the cluster with
higher value of 𝑊 . When the number of request attempts increases,
then the collision rate increases as well. If the RA mechanism fails after
a collision, then the device must wait for some time before starting
In Fig. 7, the throughput is shown as a function of the preamble a new RA: this introduces a latency in accessing the channel, thus
partition 𝛽. As the number of preambles allocated for cluster #1 in- increasing the average delay.
creases, the throughput at eNodeB1 increases. Since the total number In Fig. 11, the optimal value of the preamble partition 𝛽 is shown
of preambles is fixed, the number of preambles allocated for cluster #2 as a function of the tolerable interference threshold (𝐼th ). The decoding
reduces, so that the throughput at eNodeB2 decreases. Even in this probability is an increasing function of 𝐼th , as the transmit powers of
case, it can be observed that the total throughput is maximized in MTC devices increase (as shown in Fig. 4). The optimal value of 𝛽,
correspondence to a specific value of 𝛽. which maximizes the throughput, reduces for increasing value of 𝐼th .
In Fig. 8, the average delay is shown as a function of the number Due to the choice of a higher value of 𝜂 in cluster #1, the decoding
of devices, for various values of preambles, while keeping the interfer- probability is higher in cluster #1, which allows more preambles to be
ence threshold identical for both clusters of devices. Therefore, as the assigned to cluster #1.
number of devices increases, the average delay increases because the
collision rate increases. However, if we compare the average delay of 5. Conclusions
clusters with different preambles, then the cluster which has the larger
number of preambles has a lower average delay, if compared to the In this paper, we have investigated the design of CR MTC networks.
other cluster which has a smaller number of preambles. Intuitively, The throughput and delay performance of two clusters of MTC devices
increasing the number of preambles reduces the probability of collision has been investigated in an underlay CR network. A higher value of
and, then, the average delay. tolerable interference threshold of a primary network increases the
In Fig. 9, the average delay is shown as a function of the number preambles’ decoding probability and the throughput, while it reduces
of devices for various values of the interference threshold. The cluster the delay. An optimal value of the preamble partition 𝛽 allows to

7
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

Fig. 7. Throughput as a function of the preamble partition 𝛽.

Fig. 8. Average delay E(𝐷) as a function of the number of MTC devices 𝑁tot , for various values of the number of preambles 𝑀tot .

Fig. 9. Average delay E(𝐷) as a function of the number of MTC devices 𝑁tot with different interference thresholds 𝐼th .

8
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

Fig. 10. Average delay E(𝐷) as a function of the number of MTC devices 𝑁tot with different number of requests 𝑊 made to an eNB.

Fig. 11. Optimal value of the preamble partition 𝛽 as a function of the interference threshold 𝐼th .

maximize the throughput, which also depends on the ACB factor, the Declaration of competing interest
number of MTC devices in each cluster, as well as the interference
threshold via decoding probability. A higher value of the ACB factor The authors declare that they have no known competing finan-
reduces the throughput. The average delay is an increasing function of cial interests or personal relationships that could have appeared to
the number of preambles and re-transmission attempts. influence the work reported in this paper.

Data availability
CRediT authorship contribution statement
The authors are unable or have chosen not to specify which data
has been used.
Soumen Mondal: Conceptualization, Data curation, Formal anal-
ysis, Investigation, Methodology, Resources, Software, Validation, Vi- Appendix. Derivation of (11)
sualization, Writing – original draft, Writing – review & editing. Luca
Davoli: Investigation, Methodology, Visualization, Writing – original The CDF 𝐹𝑊 (𝛼1 ) can be expressed as follows:
draft, Writing – review & editing. Sanjay Dhar Roy: Conceptualiza- { }
𝑈
𝐹𝑊 (𝛼1 ) = Pr < 𝛼1
tion, Data curation, Formal analysis, Funding acquisition, Investigation, 𝑉
{ }
Methodology, Project administration, Resources, Supervision, Valida- = Pr 𝑈 < 𝛼1 𝑉
tion, Writing – original draft, Writing – review & editing. Sumit Kundu: ∞ { }
Conceptualization, Data curation, Formal analysis, Funding acquisition, = Pr 𝑈 < 𝛼1 𝑣 ∣ 𝑉 = 𝑣 𝑓𝑉 (𝑣) 𝑑𝑣
∫0
Investigation, Methodology, Project administration, Resources, Super- ∞ ( )
vision, Validation, Visualization, Writing – original draft, Writing – = 𝐹𝑈 𝛼1 𝑣 𝑓𝑉 (𝑣) 𝑑𝑣
∫0
review & editing. Gianluigi Ferrari: Conceptualization, Data cura- ∞( ) 𝜆𝑥 𝜆𝑥 ( −𝜆 𝑣 𝑣)
= 1 − 𝑒−𝜆𝑣 𝛼1 𝑣 𝑒 𝑥 𝑎 − 𝑒−𝜆𝑥 𝑏 𝑑𝑣
tion, Formal analysis, Funding acquisition, Investigation, Methodology, ∫0 𝜆𝑥 𝑎 − 𝜆𝑥 𝑏
Project administration, Resources, Supervision, Validation, Visualiza- ∞ 𝜆𝑥 𝜆𝑥 ( −𝜆𝑥 𝑣 𝑣)

tion, Writing – original draft, Writing – review & editing. Riccardo =1− 𝑒 𝑎 − 𝑒−𝜆𝑥 𝑏 𝑒−𝜆𝑣 𝛼1 𝑣 𝑑𝑣
∫0 𝜆𝑥 𝑎 − 𝜆𝑥 𝑏
( )
Raheli: Conceptualization, Data curation, Formal analysis, Funding 𝜆𝑥 𝜆𝑥 ∞ 𝜆
− 𝑥 +𝛼 𝜆 𝑣
acquisition, Investigation, Methodology, Project administration, Re- =1− 𝑒 𝑎 1 𝑣 𝑑𝑣
𝜆𝑥 𝑎 − 𝜆𝑥 𝑏 ∫ 0
( )
sources, Supervision, Validation, Visualization, Writing – original draft, 𝜆𝑥 𝜆𝑥 ∞ 𝜆
− 𝑥 +𝛼 𝜆 𝑣
Writing – review & editing. + 𝑒 𝑏 1 𝑣 𝑑𝑣
𝜆𝑥 𝑎 − 𝜆𝑥 𝑏 ∫0

9
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

𝜆𝑥 𝜆𝑥 1 𝜆𝑥 𝜆𝑥 1
=1− + Tello-Oquendo, L., Leyva-Mayorga, I., Pla, V., Martinez-Bauset, J., Vidal, J.-R., Casares-
𝜆𝑥 𝑎 − 𝜆𝑥 𝑏 𝜆𝑥 𝜆𝑥 𝑎 − 𝜆𝑥 𝑏 𝜆𝑥
𝑎
+ 𝛼1 𝜆𝑣 𝑏
+ 𝛼1 𝜆𝑣 Giner, V., Guijarro, L., 2017. Performance analysis and optimal access class barring
parameter configuration in LTE-A networks with massive M2M traffic. IEEE Trans.
𝜆𝑥 𝜆𝑥 ⎡⎢ 1 1
⎤
⎥
Veh. Technol. 67 (4), 3505–3520. http://dx.doi.org/10.1109/TVT.2017.2776868.
=1− − Tyagi, R.R., Lee, K.-D., Aurzada, F., Kim, S., Reisslein, M., 2012. Efficient delivery
𝜆𝑥 𝑎 − 𝜆𝑥 𝑏 ⎢ 𝜆𝑥
+ 𝛼1 𝜆𝑣
𝜆𝑥
+ 𝛼1 𝜆𝑣 ⎥⎦
⎣ 𝑎 𝑏 of frequent small data for U-healthcare applications over LTE-advanced networks.
𝜆𝑥 𝜆𝑥 In: Proceedings of the 2nd ACM International Workshop on Pervasive Wireless
=1− ( )( ). Healthcare. MobileHealth ’12, Hilton Head, South Carolina, USA, pp. 27—-32.
𝜆𝑥 𝜆𝑥
𝑎𝑏 𝑎
+ 𝛼1 𝜆𝑣 𝑏
+ 𝛼1 𝜆𝑣 http://dx.doi.org/10.1145/2248341.2248354.
Vardakas, J.S., Zorba, N., Skianis, C., Verikoukis, C.V., 2015. Performance analysis of
M2M communication networks for QoS-differentiated smart grid applications. In:
2015 IEEE Globecom Workshops. GC Wkshps, IEEE, pp. 1–6. http://dx.doi.org/10.
References 1109/GLOCOMW.2015.7414116.
Yang, B., Wei, F., She, X., Jiang, Z., Zhu, J., Chen, P., Wang, J., 2023. Intelligent ran-
3GPP, 2011. RAN Improvements for Machine-type Communications. Technical dom access for massive-machine type communications in sliced mobile networks.
Specification (TS) 37.868, 3rd Generation Partnership Project (3GPP), URL Electronics 12 (2), 329. http://dx.doi.org/10.3390/electronics12020329.
https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx? Yao, H., Huang, T., Zhao, C., Kang, X., Liu, Z., 2014. Optimal power allocation in
specificationId=2630, Version 1.0.0. cognitive radio based machine-to-machine network. EURASIP J. Wireless Commun.
Abdullah, M.A., Abdullah, Z., Chen, G., Tang, J., Chambers, J., 2019. Performance Networking 2014, 1–9. http://dx.doi.org/10.1186/1687-1499-2014-82.
analysis of cognitive clustered M2M random networks with joint user and machine Zhang, C., Sun, X., Xia, W., Huang, R., Zhou, M., Zhu, H., 2023. Access delay
device selection. IEEE Access 7, 83515–83525. http://dx.doi.org/10.1109/ACCESS. optimization of double-contention random access scheme in machine-to-machine
2019.2924522. communications. IEEE Commun. Lett. http://dx.doi.org/10.1109/LCOMM.2023.
Aijaz, A., Aghvami, A.H., 2015. Cognitive machine-to-machine communications for 3280920.
internet-of-things: A protocol stack perspective. IEEE Internet Things J. 2 (2), Zhang, C., Sun, X., Xia, W., Zhang, J., Zhu, H., Wang, X., 2022. Deep learning
103–112. http://dx.doi.org/10.1109/JIOT.2015.2390775. based double-contention random access for massive machine-type communication.
Alhussien, N., Gulliver, T.A., 2022. Energy-efficient interference-aware cognitive IEEE Trans. Wireless Commun. 22 (3), 1794–1807. http://dx.doi.org/10.1109/
machine-to-machine communications underlaying cellular networks. IEEE Access TWC.2022.3206769.
10, 33932–33942. http://dx.doi.org/10.1109/ACCESS.2022.3162252. Zhang, C., Sun, X., Zhang, J., Wang, X., Jin, S., Zhu, H., 2019. Throughput optimization
Althumali, H.D., Othman, M., Noordin, N.K., Hanapi, Z.M., 2020. Dynamic backoff with delay guarantee for massive random access of M2M communications in
collision resolution for massive M2M random access in cellular IoT networks. IEEE industrial IoT. IEEE Internet Things J. 6 (6), 10077–10092. http://dx.doi.org/10.
Access 8, 201345–201359. http://dx.doi.org/10.1109/ACCESS.2020.3036398. 1109/JIOT.2019.2935548.
Chowdhury, M.R., De, S., 2022. Queue-aware access prioritization for massive machine- Zhang, X., Xing, J., Yan, Z., Gao, Y., Wang, W., 2013. Outage performance study of
type communication. IEEE Internet Things J. 9 (17), 15858–15873. http://dx.doi. cognitive relay networks with imperfect channel knowledge. IEEE Commun. Lett.
org/10.1109/JIOT.2022.3151408. 17 (1), 27–30. http://dx.doi.org/10.1109/LCOMM.2012.112812.121371.
Debbabi, F., Jmal, R., Fourati, L.C., Aguiar, R.L., 2022. An overview of interslice
and intraslice resource allocation in B5G telecommunication networks. IEEE Trans.
Netw. Serv. Manag. 19 (4), 5120–5132. http://dx.doi.org/10.1109/TNSM.2022.
Soumen Mondal received his B.Tech degree in Electron-
3189925.
ics and Communication Engineering in 2008 from Haldia
Di, C., Zhang, B., Liang, Q., Li, S., Guo, Y., 2019. Learning automata-based access class
Institute of Technology, Haldia, India, and M.Tech. de-
barring scheme for massive random access in machine-to-machine communications.
gree in Telecommunication Engineering in 2010 from NIT,
IEEE Internet Things J. 6 (4), 6007–6017. http://dx.doi.org/10.1109/JIOT.2018.
Durgapur. After that he joined Central Mechanical Engi-
2867937.
neering Research Institute, Durgapur as a Junior Research
Ejaz, W., Ibnkahla, M., 2015. Machine-to-machine communications in cognitive cellular
Fellow in 2010. After that he joined Bengal College of
systems. In: 2015 IEEE International Conference on Ubiquitous Wireless Broadband.
Engineering and Technology, Durgapur as an Assistant Pro-
ICUWB, Montreal, QC, Canada, pp. 1–5. http://dx.doi.org/10.1109/ICUWB.2015.
fessor in Electronics and Communication Dept on 2012. He
7324472.
completed his Ph.D. under Vivesvaraya Ph.D. scheme in
Guo, Z., Zhu, X., Wei, Z., Jiang, Y., Wang, Y., 2022. Collision-aware random access
department of Electronics and Communication Engineering,
control with preamble reuse for industrial IoT. In: 2022 IEEE 95th Vehicular
National Institute of Technology Durgapur in 2021. His
Technology Conference:(VTC2022-Spring). IEEE, pp. 1–6. http://dx.doi.org/10.
research interests include Cognitive Radio Networks, Energy
1109/VTC2022-Spring54318.2022.9860484.
Harvesting, NOMA, MIMO, and FSO. He is serving as
Ha, S., Seo, H., Moon, Y., Lee, D., Jeong, J., 2018. A novel solution for NB-IoT cell
reviewer of IEEE Wireless Communication Letter and IEEE
coverage expansion. In: 2018 Global Internet of Things Summit. GIoTS, Bilbao,
Communication Letter.
Spain, pp. 1–5. http://dx.doi.org/10.1109/GIOTS.2018.8534519.
Jiang, N., Deng, Y., Condoluci, M., Guo, W., Nallanathan, A., Dohler, M., 2018. RACH
preamble repetition in NB-IoT network. IEEE Commun. Lett. 22 (6), 1244–1247. Luca Davoli is a non-tenured Assistant Professor at the
http://dx.doi.org/10.1109/LCOMM.2018.2793274. Internet of Things (IoT) Laboratory, Department of Engi-
Laya, A., Alonso, L., Alonso-Zarate, J., 2013. Is the random access channel of LTE and neering and Architecture, University of Parma, Italy. He
LTE-A suitable for M2M communications? A survey of alternatives. IEEE Commun. obtained his Dr. Ing. degree in Computer Engineering and
Surv. Tutor. 16 (1), 4–16. http://dx.doi.org/10.1109/SURV.2013.111313.00244. his Ph.D. in Information Technologies at the Department
Lee, K.-D., Kim, S., Yi, B., 2011. Throughput comparison of random access methods for of Information Engineering of the same university, in 2013
M2M service over LTE networks. In: GLOBECOM Workshops (GC Wkshps), 2011 and 2017, respectively. He is an IEEE member. His research
IEEE. Houston, TX, USA, pp. 373–377. http://dx.doi.org/10.1109/GLOCOMW.2011. interests focus on Internet of Things (IoT), Pervasive Com-
6162474. puting, Software-Defined Networking, and smart systems.
Li, W., Du, Q., Liu, L., Ren, P., Wang, Y., Sun, L., 2015. Dynamic allocation of RACH He has served as Editorial Board Member and Technical
resource for clustered M2M communications in LTE networks. In: Identification, Program Committee Member of international journals and
Information, and Knowledge in the Internet of Things (IIKI), 2015 International conferences.
Conference on. Beijing, China, pp. 140–145. http://dx.doi.org/10.1109/IIKI.2015.
38.
Li, T., Yuan, J., Torlak, M., 2018. Network throughput optimization for random access Sanjay Dhar Roy received his B. E. (Hons.) degree in
narrowband cognitive radio Internet of Things (NB-CR-IoT). IEEE Internet Things Electronics & Telecommunication Engineering in 1997 from
J. 5 (3), 1436–1448. http://dx.doi.org/10.1109/JIOT.2017.2789217. Jadavpur University, Kolkata, India, and M. Tech. in
Mitola, J., Maguire, G.Q., 1999. Cognitive radio: Making software radios more personal. Telecommunication Engineering in 2008 from NIT Durga-
IEEE Pers. Commun. 6 (4), 13–18. http://dx.doi.org/10.1109/98.788210. pur, and Ph.D. in Wireless Communication in 2011 from NIT
Oh, S.-M., Shin, J., 2017. An efficient small data transmission scheme in the 3GPP Durgapur, respectively. He worked for UshaFone, a mobile
NB-IoT system. IEEE Commun. Lett. 21 (3), 660–663. http://dx.doi.org/10.1109/ communication company, from 1997 to 2000. He joined the
LCOMM.2016.2632128. Department of Electronics & Communication Engineering,
Swain, S.N., Subudhi, A., 2023. Recurrent neural network based RACH scheme NIT Durgapur as a lecturer in 2000 and is currently an
minimizing collisions in 5G and beyond networks. In: IEEE INFOCOM 2023-IEEE Associate Professor there. He is an IEEE senior member.
Conference on Computer Communications Workshops. INFOCOM WKSHPS, IEEE, His research interests include Radio Resource Management,
pp. 1–7. http://dx.doi.org/10.1109/INFOCOMWKSHPS57453.2023.10226096. Handoff, Device to Device Communication, Cognitive Radio

10
S. Mondal et al. Journal of Network and Computer Applications 225 (2024) 103856

Networks, CR Femto Cell Networks, Physical Layer Secu- senior member. He is reviewer of several papers for IEEE,
rity and Energy Harvesting in Cognitive Radio Networks, Elsevier, Wiley journals, etc., and many important national
Cooperative Cognitive Relay Networks, Cellular IoT and 5G and international IEEE conferences.
networks. As of today, he has published more than two hun-
dred (200) research papers in various reputed journals and
conferences. He has reviewed for various IEEE journals and Gianluigi Ferrari received the Laurea (summa cum laude)
different journal papers and books from Elsevier, Springer, and Ph.D. degrees in Electrical Engineering from the Univer-
and Wiley, etc. He has reviewed for IEEE Globecom, IEEE sity of Parma, Parma, Italy, in 1998 and 2002, respectively.
WCNC, IEEE VTC, IEEE PIMRC, IEEE WIMob and NCC on Since 2002, he has been with the University of Parma,
several occasions. where he is currently a Full Professor of Telecommunica-
tions and also the coordinator of the Internet of Things (IoT)
Laboratory, Department of Engineering and Architecture.
Sumit Kundu received Bachelor of Engineering (Hons) in He is co-founder and President of things2i ltd., a spin-off
Electronics & Communication Engineering from National of the University of Parma dedicated to IoT and smart
Institute of Technology, Durgapur (Erstwhile Regional Engg systems. He is an IEEE senior member. His current research
College, Durgapur, University of Burdwan) with University interests include signal processing, advanced communication
Gold medal in 1991, Master of Technology (M.Tech) in and networking, IoT and smart systems.
Telecommunication Systems Engineering from IIT Kharag-
pur in 1993–94 and Ph.D (Wireless Communication) from
IIT Kharagpur in 2004. He has been a faculty in the Riccardo Raheli is Professor of Communication Engineering
Department of ECE, NIT Durgapur since 1995 where he at the University of Parma, Italy, which he joined in 1991.
is currently a Full Time Professor (Higher Administrative From 1988 to 1991, he was with the Sant’Anna School of
Grade). He also served as an Assistant Professor in the Advanced Studies, Pisa, Italy. From 1986 to 1988 he was
G.S.Sanyal School of Telecommunications, IIT Kharagpur with Siemens Telecommunications, Milan, Italy. In 1990 and
for a year in 2007. He has supervised 11 PhDs so far 1993, he spent leaves as a Visiting Assistant Professor at
in the domain of Wireless Communication and Networking the University of Southern California, Los Angeles, USA. His
and continuing supervision for several more students. He scientific interests are in the general area of systems for
has published more than 200 research papers in interna- communication, processing and storage of information, in
tional journal, national and international conferences, and which he has published diversely and extensively. He is an
contributed to several book chapters. His current areas of IEEE life member. He has served as Editorial Board Member
research includes Cooperative Communication, NOMA for and Technical Program Committee Co-Chair of prestigious
5G, Intelligent Reflecting Surfaces (IRS), Cognitive Radio international journals and conferences.
Network, Energy Harvesting in Wireless Networks, Physical
Layer Security, Wireless Sensor Networks. He is an IEEE

11