2020 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE)

Resource Allocation in CAT-M and LTE-A

Coexistence: A Joint Contention Bandwidth
Optimization Scheme
Radwa Sultan ∗ , Ahmed Refaey ∗† , and Walaa Hamouda ‡
∗ Manhattan College, Riverdale, New York, USA.
† Western University, London, Ontario, Canada.
‡ Concordia University, Montreal, QC, Canada.

Abstract—There are high expectations for IoT devices by 2022, wireless IoT devices will account for 79% of
and networks concerning reliability, performance, quality, internet traffic in comparison to the 65% share as of
and long-term availability. Indeed, wireless connectivity is 2017 [1]. When coupled with the statistic that global IP
the most critical success factor for the IoT era. Recently, the traffic will triple from 2017 to 2022 [1], the need for
cellular technologies focused on introducing new releases,
infrastructure innovation becomes apparent. This growth
like LTE Cat-M1, to provide global coverage and mobility
has put a high strain on critical resources such as data
for the IoT applications. However, the cellular spectrum
is already congested, and adding new services will defiant bandwidth requirements, spectrum usage, and cluttered
the existing ones. Herein, the network key performance wireless environments [2] [3].
indicator (KPI) should be considered to enhance the re- Low-power wide-area networks (LP-WAN) options
source management for LTE and LTE CAT M1 users.
have been widely considered a promising means to an
Tackling the coexistence between the aforementioned in the
end for spectrally dense cluttered wireless networks [4]–
1.4 Mhz band, three coexistence optimization problems are
formulated. The first and the second coexistence optimiza- [7]. Practically speaking, the traditional cellular options,
tion problems are formulated assuming higher IoT-traffic like 4G and LTE networks, can not be considered as
priority, and higher LTE-traffic priority, respectively. On a fit solution to the IoT devices due to the high-power
the other hand, the third problem is formulated assuming consumption. Furthermore, the design of these cellular
that both the IoT-traffic and the LTE-traffic have the same options is not suitable for application where only a small
priority. Afterward, a scheduling optimization solution algo- amount of data is sporadically transmitted. As a result, the
rithm is proposed using the interior point method. Finally, 3GPP group started to develop new cellular IoT standards
the performance of the proposed scheduling algorithm is
that introduce suitable connectivity options fitting with the
evaluated via numerical analysis.
low-power and long-range applications requirements.
Index Terms—CAT-M 1, Cellular IoT, Coexistence, Re-
source Allocation.
Cat- 0 and 1 were the early successful steps towards
connecting IoT devices using existing LTE networks [8],
[9]. The efforts for Cat 0 and 1 were followed by releasing
I. I NTRODUCTION Rel-13 for LTE Cat-M1 which is considered as the second
Modern wireless communications have been marked generation of LTE chips built for IoT applications [10].
by improvements such as capacity, data rates, and quality In fact, Cat-M1 was capable to reduce the cost and power
of service (QoS). Over the past decade, the diversity of consumption for which Cat-0 and 1 originally set the
practical wireless systems have seen an immense amount precedent. This was the result of capping the maximum
of growth in the industrial sector. Cisco predicts that, system bandwidth at 1.4 MHz rather than 20 MHz in

978-1-7281-5442-8/20/$31.00
Authorized ©2020
licensed use limited to: University IEEE
of Gothenburg. Downloaded on December 21,2020 at 08:19:21 UTC from IEEE Xplore. Restrictions apply.
 2020 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE)

The remainder of this paper is organized as follows.

The system model is presented in Section II. The schedul-
ing optimization problem is formulated in Section III. In
Section (IV), the proposed scheduling optimization al-
gorithm is presented. Numerical results and performance
evaluation are presented in Section V. Finally, the paper
is concluded in Section VI.

II. S YSTEM M ODEL

Assume a wireless network with an LTE system that
is composed of a single LTE base-station (BS), a set of
S service providers (SPs) each serving U users. Addi-
tionally, the network has M IoT devices each is equipped
with a CAT-M modem. The proposed model is shown in
Figure 1.
For the coexistence between the LTE and the CAT-M
IoT to be achieved, it is assumed that there is a minimal
Fig. 1: System Model interference. Accordingly, when assigning the rth resource
block (RB) to the sth service provider’s LTE-user u, the
downlink (DL) throughput is given by
Cat-0 [11]. Consequently, it became a major catalyst for  
s Gs,r
Pr,u BS,u
the next wave of devices and platform architectures by T (u, r, s) = B × log2 1 + , (1)
σ2
allowing a rapid IoT connectivity with countless scale
and fit with a small amount of data transfer applications. where B is the RB bandwidth in Hz, Pr,u s is the DL

In general, the real advantage of Cat-M over other options transmission power between the LTE-BS and the sth
is compatibility with the existing LTE network. service provider’s LTE-user u, Gs,r BS,u is the channel gain
Encouraged by these promising features, the perfor- th
between the LTE-BS, and s service provider’s LTE-user
mance of the CAT-M1 has been recently studied in litera- u on the rth RB, and σ 2 is the noise variance. Accordingly,
ture [12] [3]. In [13], the authors experimentally study the the total LTE throughput is given by
coverage enhancement of LTE Cat-M1. It was observed
S U R
that, vi repetition schemes, a significant coverage gain Tlte = ∑ ∑ ∑ a(s, u, r) × T (u, r, s), (2)
can be achieved, at the expense of a lower throughput. s=1 u=1 r=1
Accordingly, adaptive repetition algorithms are proposed where the binary coefficient a(s, u, r) = 1 when the rth
to achieve higher throughput while maintaining the same resource block is assigned to the sth service provider’s
coverage [14] [15]. Additionally, in [16], an overview of LTE-user u and a(u, r, s) = 0 otherwise. Similarly, the
the LTE-CAT-M1 and its evolution are presented, and it throughput of the IoT device m when connected to the
was shown that LTE-M can also meet 5G requirements. rth RB is given by
In [17], an in depth performance evaluation of LTE CAT-  
M is presented. It has been shown that, it offers additional Pr,m Gr,m
T (r, m) = B × log2 1 + , (3)
advantages to meet machine-to-machine communication σ2
needs in terms of wider coverage, lower throughput, and where Pr,m is the DL transmission power between the
a larger number of machines connected through LTE LTE-BS and the mth IoT device, and Gr,m is the channel
network. gain between the LTE-BS, and the IoT device m on the

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 08:19:21 UTC from IEEE Xplore. Restrictions apply.
 2020 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE)

RB r. Accordingly, the total IoT CAT-M throughput is where ALT E , and AIoT are vectors that include all the
given by LTE scheduling coefficients a(s, u, r), and the IoT devices
scheduling coefficients a(m, r), respectively. The first
M R
TIoT = ∑ ∑ a(r, m) × T (r, m), (4) constraint is to guarantee that a given RB can be only
m=1 r=1 allocated to a single LTE user or IoT user. The second
where the binary coefficient a(r, m) = 1 when the IoT constraint is to guarantee that the each IoT device will
device m is connected to the rth RB, and a(r, m) = 0 oth- be assigned one RB. Finally, the third constraint ensures
erwise. Finally, to calculate the total network throughput, that each LTE user will be assigned a maximum of one
the LTE throughput in (4) is added to the IoT throughput RB. In other words, for large number of IoT devices, no
in (3). Accordingly, the total network throughput is given RBs might be available for the LTE traffic.
by,
B. Scenario 2: Higher LTE-Traffic Priority
T = Tlte + TIoT . (5)
This scenario is the opposite of the coexistence scenario
III. P ROBLEM F ORMULATION presented in Section III-A, in which the LTE-traffic is
given a higher priority than the IoT-traffic. In other words,
As mentioned before, for the coexistence between the
as more LTE devices are introduced to the network, more
LTE-users and the IoT devices, minimal to no interfer-
resource blocks will be assigned to the LTE traffic at the
ence is assumed. Additionally, it is assumed that all the
expense of the IoT traffic. Accordingly, the coexistence
network’s LTE-users and IoT devices always have data to
scheduling optimization problem is modified to
send. The main objective of the coexistence scheduler is
to find the best RBs assignment or scheduling algorithm
that maximizes the total network throughput in (5). For max T
ALT E ,AIoT
that purpose, three different coexistence scenarios, which S U M
depends on different LTE-traffic and IoT-traffic priorities, subject to ∑ ∑ a(s, u, r) + ∑ a(m, r) = 1 ∀r ∈ R
s=1 u=1 m=1
are proposed. R
∑ a(m, r)  1 ∀m ∈ M
r=1
A. Scenario 1: Higher IoT-Traffic Priority
S R
In this scenario, it is assumed that IoT-traffic has a ∑ ∑ a(s, u, r) = 1 ∀u ∈ U,
s=1 r=1
higher priority that the LTE-traffic. In other words, as (P2)
more IoT devices are introduced to the network, more where, the second constraint is to guarantee that the each
resource blocks will be assigned to the IoT traffic at the IoT device will be assigned a maximum of one RB.
expense of the LTE traffic. Accordingly, the coexistence The third constraint ensures that each LTE user will be
scheduling optimization problem is given by assigned one RB. In other words, for large number of
LTE users, no RBs might be available for the IoT traffic.
max T
ALT E ,AIoT
S U M C. Scenario 3: Both IoT-Traffic and LTE-Traffic Have the
subject to ∑ ∑ a(s, u, r) + ∑ a(m, r) = 1 ∀r ∈ R Same Priority
s=1 u=1 m=1
R
This scenario is a generalized case, in which the
∑ a(m, r) = 1 ∀m ∈ M RB scheduling aims at maximizing the total network
r=1
S R throughput while having the same priority for the LTE-
∑ ∑ a(s, u, r)  1 ∀u ∈ U, traffic and the IoT-traffic. Accordingly, the optimization
s=1 r=1
(P1) problem is modified to

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 08:19:21 UTC from IEEE Xplore. Restrictions apply.
 2020 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE)

Algorithm 1: LTE and CAT-M Resource Blocks an exponential complexity in the number of users, number
Scheduling Optimization of service providers, and the number of the resource
s
Data: PBS,u ∀ u ∈ {1 · · ·U}, ∀ s ∈ {1 · · · S}, PBS,m , blocks 1 . Alternatively, the optimization variables are re-
∀ m ∈ {1 · · · M} ,σ 2 , rmin1 , rmin2 , step size laxed to the continuous domain [0, 1]. After relaxation, the
(μ), tolerance (ε), All CSI information, proposed optimization problems given in (P1),(P2), and
s
A0LT E , A0IoT ,ρmin ∀ s ∈ {1 · · · S}
(P3) are linear optimization problems, and hence, can be
Result: A = [ALT E ; A∗IoT ] maximizing T 5
∗ ∗
solved efficiently using the interior point algorithm [18].
1. Define Ac ∈ ∈ [0, 1] ∀ u ∈ {1 · · ·U}, ∀ m ∈ After solving the relaxed problems, the LTE scheduling
{1 · · · M}, ∀ r ∈ {1 · · · R} , ∀ s ∈ {1 · · · S}; coefficients a(s, u, r), and the IoT devices scheduling
2. Determine A∗ (t) minimizing (−tT + Φ(A)) coefficients a(m, r) are re-mapped to the binary domain
3. Update A = A∗ (t) {0, 1}. The solution algorithm is shown in Algorithm 1.
4. if nt = ε then
Increase t = μt In the beginning, a tolerance value ε, a step size μ >
 
Return to step 2. 0, an initial point in the feasible set (A0LT E , A0IoT , the
end log barrier function are determined [18]. The log barrier
else function for n constraints is given by
A∗ ≈ A k
end n
end Φ(A) = − ∑ log(− fi (A)), (6)
i=1

where fi (A) are the constraints functions of the opti-

mization problem. The next step is to apply the interior
max T point algorithm to the relaxed optimization problem by
ALT E ,AIoT
finding A∗ (t) minimizing (−tT + Φ(A)), where t is the
S U M
subject to ∑ ∑ a(s, u, r) + ∑ a(m, r) = 1 ∀r ∈ R approximation parameter, which can be controlled in each
s=1 u=1 m=1 iteration of interior point method. Finally, after satisfying
R
the stopping criteria defined in Step. 4, the variables
∑ a(m, r)  1 ∀m ∈ M are re-mapped to the binary domain. We use standard
r=1
S R MATLAB function to implement the above algorithm.
∑ ∑ a(s, u, r)  1 ∀u ∈ U.
s=1 r=1
(P3) V. N UMERICAL A NALYSIS
Similar to the modification done in (P2), the second
and the third constraints guarantee that each IoT device, TABLE I: Simulation Parameters and Values
and LTE users will be assigned a maximum of one RB,
B 180 KHz SP 2
respectively. It is important to note that the priority of the
IoT traffic or the LTE traffic is determined in terms of the σ2 10−13 R 6
number of RBs assigned to each IoT-device or LTE-user, U 5 M 5
respectively. s
PBS,u 20 W PBS,m 20 W

IV. R ESOURCE B LOCKS S CHEDULING O PTIMIZATION

In this section, the performance of the proposed co-
A LGORITHM
existence scenarios is evaluated in the 1.4 MHz band.
Since the scheduling optimization variables are binary, The proposed system is simulated using Monte Carlo
the proposed scheduling optimization problems in (P1),
1 Theoptimal exhaustive solution is to enumerate all the possible
(P2), and (P3) are of a combinatorial nature. Accordingly, combinations of the RBs scheduling and choose the combination with
obtaining an exact solution using an exhaustive search has the maximum total throughput.

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 08:19:21 UTC from IEEE Xplore. Restrictions apply.
 2020 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE)

107 107
6 6 LTE Throughput
LTE Throughput
IoT Throughput IoT Throughput
Total Throughput
5 5 Total Throughput

Throughput (bits/sec)
4
Throughput (bit/sec)

3 3

2 2

1 1

0 0
1 2 3 4 5 1 2 3 4 5
Number of IoT Devices (M)
Number of LTE Users (U)
Fig. 2: Scenario 1: LTE, CAT-M and Total Throughput Fig. 3: Scenario 2: LTE, CAT-M and Total Throughput
for Different Number of IoT Devices for Different Number of LTE Users

simulations on MATLAB. Unless otherwise mentioned, with keeping M = 5 and increasing the number of the
the simulation parameters are summarized in Table I. LTE users, when considering the last coexistence scenario
Figure 2, shows the variation of the LTE throughput, in (P3). For small numbers of LTE-users, it can be
the CAT-M throughput, and the total network throughput seen that most of the resources is assigned for the IoT
with increasing the number of the IoT devices, when devices. However, as the number of LTE-users increases,
considering the first coexistence scenario in (P1). As at M = 5 with U = 5 and U = 6, the resources are
expected, as the number of IoT devices increases, more shared almost equally between the LTE-users and the
resource blocks are assigned to the IoT traffic at the IoT-devices. Hence, it can be verified from the results,
expense of the LTE-traffic. As a result, the IoT throughput in Figure 4, that both the LTE-traffic and the IoT traffic
increases with the decrease of the LTE traffic. Hence, it have the same priority.
can be verified from the results, in Figure 2, that the IoT From the results shown in Figures 3 and 4, it can
traffic has a higher priority than the LTE traffic. be seen that, when considering the second coexistence
Figure 3, shows the variation of the LTE throughput, scenario, introducing new LTE-users will greatly decrease
the CAT-M throughput, and the total network throughput the IoT throughput. However, when equal priority is
with increasing the number of the LTE users, when considered, the resource sharing is done fairly, and, hence
considering the second coexistence scenario in (P2). As the degradation in the IoT throughput will decrease.
expected, as the number of the LTE users increases, more Finally, it can be concluded that, for large numbers of
resource blocks are assigned to the LTE-traffic at the the LTE-users and the IoT-devices, the scenario proposed
expense of the IoT-traffic. As a result, the LTE throughput in Section III-C will offer the best coexistence between
increases with the decrease of the IoT traffic. Hence, it the LTE-users and the IoT-devices.
can be verified from the results, in Figure 3, that the LTE-
traffic has a higher priority than the IoT traffic. VI. C ONCLUSION
Figure 4, shows the variation of the LTE throughput, In this paper, the coexistence between the LTE and
the CAT-M throughput, and the total network throughput CAT-M IoT, with minimum to no-interference, is con-

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 08:19:21 UTC from IEEE Xplore. Restrictions apply.
 2020 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE)

[5] R. Eletreby, D. Zhang, S. Kumar, and O. Yağan, “Empowering

107
6 low-power wide area networks in urban settings,” in Proceedings
LTE Throughput
IoT Throughput of the Conference of the ACM Special Interest Group on Data
Total Throughput
Communication. Los Angeles, CA: ACM, Aug. 2017, pp. 309–
5
321.
[6] A. Ikpehai, B. Adebisi, K. M. Rabie, K. Anoh, R. E. Ande,
Throughput (bits/sec)

4 M. Hammoudeh, H. Gacanin, and U. M. Mbanaso, “Low-

power wide area network technologies for internet-of-things:
A comparative review,” IEEE Internet of Things Journal, Nov.
3
2018.
[7] U. Raza, P. Kulkarni, and M. Sooriyabandara, “Low power wide
2 area networks: An overview,” IEEE Communications Surveys &
Tutorials, vol. 19, no. 2, pp. 855–873, Jan. 2017.
[8] 3GPP, “Standards for the IoT,” Dec. 2011. [Online]. Available:
1
http://www.3gpp.org/news-events/3gpp-news/1805-iotr14
[9] P. Reininger, “3GPP standards for the internet-of-things,”
0 Huawei, Shenzhen, China, Tech. Rep., report no. 3GPP RAN
1 2 3 4 5 6
WG 3, 2016.
Number of LTE Users (U)
[10] M. Elsaadany and W. Hamouda, “The new enhancements in LTE-
A rel-13 for reliable machine type communications,” in IEEE
Fig. 4: Scenario 3: LTE, CAT-M and Total Throughput
28th Annual International Symposium on Personal, Indoor, and
for Different Number of LTE Users
Mobile Radio Communications (PIMRC), Montreal, QC, Oct.
2017.
[11] R. Ratasuk, N. Mangalvedhe, Y. Zhang, M. Robert, and J.-P.
Koskinen, “Overview of narrowband iot in lte rel-13,” in IEEE
sidered. Three coexistence optimization problems are conference on standards for communications and networking
formulated based on the LTE and CAT-M traffic priority. (CSCN), Berlin, Germany, Nov. 2016.
[12] S. W. et al., “White paper: Coverage analy-
A scheduling optimization solution algorithm is proposed sis of lte-m category m1,” Jan. 2017. [Online].
and evaluated via numerical analysis. It has been shown Available: https://altair-semi.com/wp-content/uploads/2017/02/
that, for large numbers of the LTE-users and the IoT- Coverage-Analysis-of-LTE-CAT-M1-White-Paper.pdf
[13] P.-C. Hsieh, Y. Jia, D. Parra, and P. Aithal, “An experimental
devices, considering equal traffic priority will offer the
study on coverage enhancement of LTE cat-m1 for machine-
best coexistence between the LTE-users and the IoT- type communication,” in IEEE International Conference on
devices. Communications (ICC), Kansas City, MO, May. 2018.
[14] A. Refaey, S. Roy, and P. Fortier, “A new approach for fec
decoding based on the bp algorithm in lte and wimax systems,”
R EFERENCES in 2011 12th Canadian Workshop on Information Theory, May
[1] “White paper: Cisco visual networking index - 2011, pp. 9–14.
global mobile data traffic forecast update, 2017- [15] A. Refaey, W. Hou, and K. Loukhaoukha, “Multilayer authen-
2022,” Tech. Rep. C11-738429-00, Nov. 2017. tication for communication systems based on physical-layer
[Online]. Available: https://www.cisco.com/c/en/us/solutions/ attributes,” Journal of Computer and Communications, vol. 2,
collateral/service-provider/visual-networking-index-vni/ pp. 64–75.
white-paper-c11-741490.html#_Toc529314175. [16] R. Ratasuk, N. Mangalvedhe, D. Bhatoolaul, and A. Ghosh, “Lte-
[2] B. Li and A. P. Petropulu, “Joint transmit designs for coexistence m evolution towards 5g massive mtc,” in 2017 IEEE Globecom
of MIMO wireless communications and sparse sensing radars Workshops (GC Wkshps), Singapore, Dec. 2017.
in clutter,” IEEE Transactions on Aerospace and Electronic [17] S. Dawaliby, A. Bradai, and Y. Pousset, “In depth performance
Systems, vol. 53, no. 6, pp. 2846–2864, Dec. 2017. evaluation of lte-m for M2M communications,” in IEEE 12th
[3] S. Zimmo, A. Moubayed, A. Refaey, and A. Shami, “Coexistence International Conference on Wireless and Mobile Computing,
of wifi and lte in the unlicensed band using time-domain vir- Networking and Communications (WiMob), New York, NY, Oct.
tualization,” in 2018 IEEE Global Communications Conference 2016.
(GLOBECOM), Dec 2018, pp. 1–6. [18] S. Boyd and L. Vandenberghe, Convex optimization. Cambridge
[4] K. Mekki, E. Bajic, F. Chaxel, and F. Meyer, “A comparative university press, Mar. 2004.
study of LPWAN technologies for large-scale iot deployment.”
ICT Express, vol. 5, no. 1, pp. 1–7, Mar. 2019.

Authorized licensed use limited to: University of Gothenburg. Downloaded on December 21,2020 at 08:19:21 UTC from IEEE Xplore. Restrictions apply.