LoRa-DuCy: Duty Cycling for LoRa-Enabled Internet of Things Devices

Full text

LoRa-DuCy: Duty Cycling for LoRa-Enabled
Internet of Things Devices
1st Thenuka Karunathilake
University of Bremen
Bremen, Germany
thenuka@comnets.uni-bremen.de
2nd Asanga Udugama
University of Bremen
Bremen, Germany
adu@comnets.uni-bremen.de
3rd Anna F¨
orster
University of Bremen
Bremen, Germany
anna.foerster@uni-bremen.de
Abstract—The LoRa (Long Range) communication technology
has gained a lot of interest recently. Typically, Internet of
Things applications would use the complete LoRaWAN stack for
their purposes. However, LoRaWAN supports only three types
of communication, called classes A, B, and C. Each of these
classes addresses a different application scenario and resource
usage goals, but all of them target star topologies, where devices
communicate only to the gateway. On the other hand, wireless
sensor networks (WSNs) have been successfully developing and
using the concept of self-organised networks and duty cycling.
In this paper, we combine these fields and implement a WSN
duty-cycling medium access protocol, known as Contiki-MAC,
on LoRa-enabled LoPy4 devices. We experimentally evaluate our
implementation and we compare its power consumption with that
of LoRaWAN-enabled devices. We show that our implementation
is more power-efficient while achieving higher traffic rates and
offering the freedom of self-organised networking for various
applications.
I. INTRODUCTION
The exponential growth of IoT (Internet of Things) in the
past decade has spread its applications to every field of every-
day life from industrial applications to art and culture. Various
communication technologies can be used to interconnect IoT
devices, but LoRa (Long Range) has experienced a growing
interest. The main advantages of LoRa are the extended
wireless range and low power consumption. Normally, IoT
devices are deployed without a continuous power supply and
most of them are battery-powered or equipped with an energy
harvesting method. In either case, the power consumption
is an important factor to consider because it determines the
node’s lifetime and thus also the costs of the system. In an
IoT device, the radio transceiver is among the highest power-
consuming components, and optimising its power consumption
will significantly improve the lifetime.
Duty cycling has been traditionally used as the main driver
for saving energy in wireless sensor networks (WSN) [16].
The current LoRaWAN protocol stack already employs duty
cycling to save energy by introducing three different classes
(Class A,B, and C). However, the main drawback in Lo-
RaWAN architecture is that all end-nodes have to communi-
cate directly with a LoRa gateway. Therefore, such a network
with multiple gateways is more expensive and furthermore,
generates safety vulnerabilities at the gateway and those as-
pects also have to be considered while deploying [23].
To reduce the number of LoRa gateways, a viable solution
is to apply a multi-hop communication where LoRa nodes
are used to route information in an AdHoc manner. The
importance of such a multi-hop communication is emphasized
by authors in [19]. In such a communication network, to
avoid collisions and to increase channel utilization, a Medium
Access Control (MAC) protocol is a must to have. Further-
more, an efficient MAC protocol can provide much-needed
radio duty cycling to reduce idle listening resulting in better
consumption in battery-powered devices.
Therefore, in this paper, we propose to implement one well-
known WSN MAC protocol, namely Contiki-MAC [5] over
LoRa physical layer. Our contributions are:
•Implementation of Contiki-MAC for LoRa devices
•An experimental performance evaluation
•Energy consumption comparison between the new imple-
mentation and LoRaWAN
The rest of the paper is organised as follows. Section II
discusses the state of the art WSN MAC protocols and LoRa
related works. Section III provides background on Contiki-
MAC. Section IV presents our performance evaluation on
LoPy devices, while Section V focuses on the power con-
sumption comparison against LoRaWAN. Finally, Section VI
concludes the paper.
II. RE LATE D WORK
There exist plenty of MAC protocols for WSNs. In general,
they are all based on duty cycling the radio and can be divided
into synchronous and asynchronous. Synchronous protocols
include S-MAC [22], T-MAC [20] and TSMP [15]. However,
they require time synchronisation, which is not always avail-
able or incurs high costs.
In asynchronous methods, there is no synchronisation be-
tween nodes, which makes them less complex. Further, asyn-
chronous protocols can be divided into sender-initiated and
receiver-initiated protocols [5]. In sender-initiated protocols,
nodes wake-up periodically and sense the channel for ongoing
transmission. Most commonly used examples are B-MAC
[16], X-MAC [4], BoX-MAC [12], WiseMAC [7] and Contiki-
MAC [5] [11]. In receiver-initiated protocols, the responsibility
of initiating the transmission is given to the receiver. Some
examples are Koala [13], RI-MAC [17] and A-MAC [6].

However, sender-initiated protocols are more popular and
have been widely used in the community. Therefore, we
selected Contiki-MAC for our implementation because of
following reasons. Apart from traditional benefits in asyn-
chronous protocol such as less complexity and low overhead,
Contiki-MAC consists of one of the most efficient wake
up mechanisms in the radio duty cycling family. Further-
more, Contiki-MAC in-cooperates most efficient techniques
introduced in earlier protocols (phase lock optimisation from
WiseMAC [7] and replacing wake up strobe with multiple data
copies from BoX-MAC [12]).
During our implementation, we selected LoRa communica-
tion because it has become an ideal candidate for many IoT
applications mainly due to the long-range and low power fea-
tures. One such application is wildlife monitoring and in [21]
authors have compared several low-cost radio transceivers and
found out that LoRa was able to outperform other transceivers.
As mentioned in Section I, the current LoRaWAN stack
is equipped with duty cycling and it is restricted to 1% by
the standard. Therefore, in some applications, the occasional
transmission of a large amount of data is not possible. As
a solution, [3] proposes an activity time-sharing mechanism
to allow a node to use additional activity time of other
nodes. The limitation of scalability in current ALOHA based
LoRaWAN MAC is explored in [10] and a solution for this
drawback is provided in [24] with time-slotted communication.
Additionally, to overcome the collisions in ALOHA based
schemes, several studies have explored the Carrier Sense
Multiple Access (CSMA) technique in LoRa networks [18]
[8]. Similar work is carried out in [2] to avoid a higher
number of collisions in ALOHA based LoRaWAN with a
modified listen-before-talk (LBT) mechanism. The discussed
approaches do not provide a full solution and most of them
are focused on LoRaWAN and thus, we propose to use a WSN
MAC protocol such as Contiki-MAC over LoRa physical layer.
In a MAC protocol, avoiding collisions is among the highest
priorities. To achieve this, nodes must be able to perform
carrier sensing to detect channel activity. LoRa modules can
use two different techniques for carrier sensing [14]. First,
with Receive Signal Strength Indicator (RSSI) the received
RF energy at the antenna can be used to measure the channel
occupancy. The second method is called Channel Activity
Detection (CAD) which is a LoRa specific feature. It is
designed to identify LoRa preamble signals and the main
advantage of this mechanism is being able to work below the
noise floor. A key drawback of this mechanism is not being
able to detect the payload part of the signal [14]. Therefore,
the payload can still cause collisions. However, the most recent
LoRa modules (SX126x) claim to handle the payload detection
with CAD. During our implementation we used SX127x and
therefore, we opted to select the first mechanism as our carrier
sensing method.
III. CONTIKI-MAC
Contiki-MAC is a successor and a combination of various
other WSN MAC protocols, such as B-MAC [16], X-MAC [4],
BoX-MAC [12], and WiseMAC [7], and additionally intro-
ducing some optimisations to further save energy. It is part
of the ContikiOS for embedded devices1and it is one of the
best explored and mostly used MAC protocols for WSNs. An
overview of the Contiki-MAC protocol is depicted in Figure 1,
based on [5]. This Section briefly discuss the Contiki-MAC
mechanism and detailed descriptions are available in [5] [11].
The driving parameter of any duty cycling MAC protocol
is the wake-up interval. At regular intervals, Contiki-MAC
wakes up and senses the channel for activity (clear channel
assessment - CCA). In case the CCA detected an activity,
it starts receiving the packet and if no activity is detected,
Contiki-MAC goes into a transmit mode and differentiates
between unicast and broadcast transmissions.
There are several important optimisations in Contiki-MAC
namely fast sleep, phase lock loop (PLL), and broadcast
optimisation. The latter two optimizations are depicted in
Figure 1 with red colour.
Fast sleep optimisation is used to identify the false positives
from CCA and prevents long awake periods in case of radio
noise. PLL optimisation is used to reduce the unnecessary
transmissions to the same neighbour during unicast transmis-
sion by using previous acknowledgements from that particular
neighbour. Finally, broadcast optimisation suggests switching
off the radio between two repetitions of the packet (Packet in-
terval) to save power because in broadcast, acknowledgements
are not expected.
IV. EXP ERIME NT S AN D PERFORMANCE ANALYSI S
We have implemented Contiki-MAC on pycom LoPy4 de-
vices [1]. These devices are popular among the IoT developing
community as they are equipped with multiple communication
technologies (LoRa, WiFi, Bluetooth, and Sigfox), easy to
program with MicroPython and well documented. We call our
implementation LoRa-DuCy (LoRa on Duty Cycles) and have
made it publicly available under GitHub2.
The used Contiki-MAC parameter values with LoRa are presented in
Table I with original Contiki-MAC work parameters as described in [5].
The LoRa parameters were set to spreading factor 7 and coding rate
4/5. We selected a packet size of 255 bytes because it is the maximum
allowed LoRa payload. Since this 255 bytes packet is much larger than
the 23 bytes used in the original work and additionally, to cater to slow
data rates in LoRa, all other parameters (CCA duration, CCA interval,
and Packet interval) had to be adjusted from the original values. These
timing parameters are selected and set up after initial experiments to
make sure that communication is mostly successful.
The conducted experiments were focused on two main
areas: a performance analysis to show the overall effectiveness
of LoRa-DuCy and a power consumption analysis (Section V)
to compare the power consumption of the LoRa-DuCy against
LoRaWAN.
Delivery ratio and duty cycle values were used as metrics
during the performance analysis. The delivery ratio was calcu-
lated as how many packets were successfully delivered from all
the transmitted packets. The duty cycle value meant how much
1http://www.contiki-os.org/
2https://github.com/ComNets-Bremen/LoRa-DuCy

Sleep Mode
CCA
Receive Mode Transmit Mode
Unicast Packet
Reception
Broadcast
Transmission
Unicast
Transmission
Radio switch off
during packet
interval
Broadcast Packet
Reception
Send
Acknowledgement
Scheduled Wakeup No available packets to transmit
Noise detected and
fast sleep activated
Channel is Busy Channel is Free
Sleep mode until the
neighbour's wake-up
phase
Send Unicast
Packet
Send Unicast
Packet
Is PLL
threshold
reached
Is Unicast
threshold
reached
Remove
neighbour from
neighbourlist
Remove the
known wake up
phase
Send Broadcast
Packet
Is
Wakeup interval
reached
Update wake-up
phase
Update wake-up
phase
Is the
node a intended
receiver
Radio kept alive
during packet
interval
No
No
No No
Yes
Yes Yes Yes
Without Broadcast
Optimization
With Broadcast
Optimization
With PLLWithout PLL
Is ACK
received
Is ACK
received
No
Yes Yes
No
Fig. 1. Contiki-MAC mechanism as described in [5]
TABLE I
CONTIKI-MAC PARA ME TER VAL UE S US ED DU RI NG T HE
IMPLEMENTATION
Parameter Values with LoRa Original work values
as described in [5]
Wakeup interval 5 – 40 sec 30 - 200 ms
Packet size 255 bytes 23 bytes
CCA duration 80 ms 0.192 ms
CCA interval 400 ms 0.5 ms
RSSI threshold 100 dBm N.A
Packet interval 500 ms 0.4 ms
ACK packet size 15 bytes N.A
time the node’s radio was kept in active mode during the full
experiment period, which depends not only on the parameter
values such as wakeup interval but also on the actual traffic
and the transmission success. All results have been presented
as averages from several experiments with confidence intervals
computed at a 95% confidence level.
We conducted experiments with 10 LoPy4 devices. The
nodes contained all possible expansion board options for the
LoPy4 device (Pysense, Pytrack, and Expansion board 3.0) and
during our tabletop experiments all the LoPy4 devices were
connected to a laptop. The experiment duration was set to 1
hour. During this time, each node was configured to generate
packets randomly.
A. Traffic impact
First, we evaluate the impact of different types of traffic
on the performance of LoRa-DuCy. For this, we use three
scenarios and we activated the respective optimisations:
Scenario 1: Nodes do not transmit any data
Scenario 2: Nodes transmit broadcast data
Scenario 3: Nodes transmit unicast data
The obtained duty cycle values for the three scenarios are
depicted in Figure 2(a). As expected the lowest duty cycles
were achieved in scenario 1 because in this configuration
nodes do not transmit any data. The highest duty cycle values
were obtained in scenario 2 because in broadcast mode the
transmission must continue for the full wake-up interval.
In contrast, in unicast mode nodes only transmit until the
acknowledgement. Therefore, scenario 3 shows lower duty
cycle values compared to scenario 2.
In scenario 2 and scenario 3 LoRa-DuCy reacted similarly
to higher traffic amount by increasing the overall duty cycle.
This clearly implies that the node needs to spend more time
in active mode to transmit more data packets. However, the
increment in duty cycle values was significantly different in
unicast mode and broadcast mode because of their separate
active optimisations.
The obtained delivery ratios are illustrated in Figure 2(b) for
the broadcast scenarios and Figure 2(c) for unicast scenarios.
The wake-up intervals above 20 seconds were able to achieve
more than 90% delivery rates. However, smaller wake-up
intervals below 20 seconds depicted relatively lower delivery
rates due to overlapping CCA.
B. Impact of phase lock loop optimisation
Here, we evaluate the impact of the PLL optimisation on
the performance of LoRa-DuCy. For this, we use the unicast
scenario (scenario 3) from above, since PLL only influences
unicast transmissions. For better comparison, we also plot the
no transmissions scenario (scenario 1).
The results are provided in Figure 3 for different traffic
volumes. The dotted lines represent the duty cycle values
without the PLL optimisation and corresponding solid lines
indicate the duty cycle values with the PLL optimisation. The

!" !# $" $# %" %# &"
$"
%"
&"
#"
'"
("
)"
*"
!""
+,-./012%23/.4510
+,-./012)23/.4510
+,-./012!#23/.4510
6/457829,15:;/<2=0>
?5<-;5:@2A/1-B2=C>
!" !# $" $# %" %# &"
"
!
$
%
&
#
'
(
)
*+,-./0123113+01
4.+/56/17,%,8/69:71,;37<,=/53+,>??
4.+/56/17,),8/69:71,;37<,=/53+,>??
@036/17,%,8/69:71,;37<,8AA
@036/17,),8/69:71,;37<,8AA
@036/17,!#,8/69:71,;37<,8AA
;/9:BC,D07:.E/F,G1H
IB7J,KJ6F:,GLH
!" #" $" %"
#"
$"
%"
&"
'"
("
)"
*"
!""
+,-./0.123$34.05621
+,-./0.123)34.05621
7.56893:;26,<.=3>1?
@6=A<6,B3C.2A-3>D?
(a)
(b)
(c)
with broadcast optimisation
with broadcast optimisation
Fig. 2. Obtained duty cycles and delivery rates with different traffic volumes.
PLL was able to perform significantly better in every wake-
up interval compared to non optimised versions. The achieved
benefit is increasing with longer wake-up intervals because
PLL can permit a node to stay longer in the sleep mode. This
behaviour is illustrated in Figure 3 with an increasing gap
between the dotted line and the solid line as the wake-up in-
terval increases. Additionally, with PLL the cost of delivering
a packet is similar in every wake-up interval. Therefore, the
gradient of the solid line was significantly decreased. It can be
summarised that PLL is a very effective measure to minimise
the duty cycles of LoRa-DuCy.
10 15 20 25 30 35 40
0
2
4
6
8
10
12
14
16
No Transmissions
Unicast 3 Packets Without PLL
Unicast 3 Packets With PLL
Unicast 8 Packets Without PLL
Unicast 8 Packets With PLL
Unicast 15 Packets Without PLL
Unicast 15 Packets With PLL
Wakeup Interval (s)
Duty Cycle (%)
Fig. 3. Impact of PLL optimisation with different traffic volumes.
10 20 30 40
0
2
4
6
8
10
12
No Transmissions
Broadcast 3 Packets Without Radio Off
Broadcast 3 Packets With Radio Off
Broadcast 8 Packets Without Radio Off
Broadcast 8 Packets With Radio Off
Wakeup Interval (s)
Duty Cycle (%)
Fig. 4. Impact of broadcast optimisation with different traffic volumes.
C. Impact of broadcast optimisation
Figure 4 shows the impact of the broadcast optimisation and
the dotted lines represents the duty cycle values without the
optimisation and the solid line depicts the performance with
the broadcast optimisation.
In broadcast, with shorter wake-up intervals the time spent
on transmitting is shorter than for longer wake-up intervals.
Hence, the resulting duty cycle values were increased with
longer wake-up intervals. A significant improvement can be
achieved by the broadcast optimisation and the obtained ben-
efit depends on the length of the wake-up interval. Because
in longer wake-up intervals, broadcast optimisation is used
more frequently due to a higher number of transmissions.
Furthermore, similar to unicast mode all the optimised and
non optimised duty cycle values were increased with higher
traffic amounts. Thus, also the broadcast optimisation has a
significant positive effect on the performance of the protocol.
V. POWE R CONSUMPTION ANALY SI S
We conducted power measurements with an oscilloscope
and a power supply. During all the power measuring experi-
ments, a LoPy4 device was used with a pysense expansion

0 5 10 15 20
50
100
150
0 5 10 15 20
20
40
60
80
Sender Receiver
Time (s)
Time (s)
Current (mA)
Current (mA)
Wakeup
Packet Transmission
Wakeup
Packet Reception
Fig. 5. LoRa-DuCy broadcast current consumption.
board. The device was connected in series with a known
resistor and the voltage drop across the resistor was used to
calculate the current in the network. The power consumption
analysis was conducted in three different modes: with LoRa-
DuCy, LoRaWAN, and without any MAC protocol.
A. Power Consumption of LoRa-DuCy
First, the node was configured to operate in broadcast mode
and the wake-up interval was set to 10 seconds, while a
second node was a receiver. The obtained power measurements
are depicted in Figure 5. In the broadcast mode, the sender
continuously transmitted multiple copies during the full wake-
up interval and this behaviour is clearly visible in the figure.
In the transmit mode the device used a current of around
115 mA. As depicted, the receiver wakes up at the scheduled
time, performed the CCA and was able to identify the ongoing
transmission. The receive state consumed a current of around
70 mA and after packet reception, the node returned to sleep
state which consumed around 52 mA
Next, we configured the device to work in the unicast mode
with 10 seconds wake-up interval. As illustrated in Figure 6,
the first transmission was done normally without the PLL
optimisation and the neighbour’s wake-up phase was identified
by the acknowledgement. After receiving the data packet, the
receiver transmitted an acknowledgement packet and this is
depicted in the figure with a short peak after the receive
state. The following transmission to the same receiver was
now optimized by the PLL and the transmission only begins
just before the neighbour’s scheduled wake up. Until that time
sender remained in the sleep state and consumed around 52
mA. Similar to the broadcast mode, the unicast transmit state
and the receive state consumed around 115 mA and 70 mA
respectively.
B. Power Consumption Comparison
Next, we compared the power consumption of LoRa-DuCy
with LoRaWAN mode and a “without MAC” mode. For
LoRaWAN we have configured the nodes to work in class A
or class C configurations. While LoRaWAN always employs
0 5 10 15 20
50
100
0 5 10 15 20
50
100
Sender Receiver
Time (s)
Time (s)
Current (mA)
Current (mA)
Wakeup and PLL activated
Packet Transmission
Wakeup
Acknowledgement
Fig. 6. LoRa-DuCy unicast current consumption.
TABLE II
CUR REN T CO NS UMP TI ON VALU ES F OR D IFFE RE NT S TATES WI TH
LORAWAN , LORA-DUCY AND WITHOUT MAC
Mode Transmit State Receive State Sleep State
LoRa-DuCy 115 mA 70 mA 52mA
Without MAC 120 mA 68 mA -
LoRaWAN Class A 115 mA 68 mA 57mA
LoRaWAN Class C 115 mA 68 mA -
a gateway, class A devices are configured asynchronously and
can go back to sleep at any time. Class C devices are always
on, while class B devices operate with periodic beacons from
the gateway [9]. However, class B configuration is not avail-
able for our Pycom devices. In the mode “without MAC”, the
node is always on and blindly transmits packets to other nodes
without sensing the channel for any ongoing transmissions.
Therefore, collisions are expected and retransmissions are not
possible.
All measured power consumption values are summarised in
Table II. LoRaWAN class C and the mode “without MAC”
consume the most power because they operate without a sleep
state. LoRa-DuCy and the LoRaWAN class A were the most
power-efficient modes because of the sleep state. However,
during the sleep state LoRaWAN class A consumed slightly
higher current than LoRa-DuCy.
Finally, we conducted an experiment with 5 LoPy4 devices.
Each device was equipped with a pysense expansion board
and a 1000 mAH battery. In each configuration, a packet
transmission was scheduled every 120 seconds. The LoRa-
DuCy mode was set to operate with 20 seconds wake-up
interval. The overall aim of the experiment was to find out
how much time can a node survive with the available battery
capacity. Each mode was tested in five experiments and the
obtained results are illustrated in Figure 7 (averages with
confidence intervals at 95% confidence level).
As expected the mode without the MAC protocol and the
LoRaWAN class C configurations depicted the shortest life-

800 900 1000 1100 1200 1300 1400
LoRa-DuCy Unicast
LoRaWAN Class A
LoRa-DuCy Broadcast
LoRaWAN Class C
Without MAC
Time (min)
Mode
Fig. 7. Average number of minutes the nodes lasted with 1000 mAH battery
for LoRaWAN, LoRa-DuCy and without MAC
times because these modes operate without a sleep state. The
LoRa-DuCy broadcast was the next most power-hungry mode
because in broadcasting same packet must be transmitted with
multiple copies. As explained in earlier sections the unicast
was able to out-perform broadcast because PLL optimisation
was able to reduce the required number of transmissions. The
LoRaWAN class A depicted a slightly shorter lifetime than
LoRa-DuCy unicast mode. The reason for this shorter lifetime
in class A was slightly higher power consumption in sleep state
as shown in Table II.
VI. SUMMARY AND CONCLUSIONS
In this paper, we presented our implementation of Contiki-
MAC over the LoRa physical layer on LoPy devices, called
LoRa-DuCy. We have not only shown that the implementation
is feasible, but we also validated it with real-world experiments
and have explored the impact of its various optimisations on
the performance with different traffic volumes.
From our conducted experiments, we showed that the imple-
mented LoRa-DuCy was able to achieve more than 90% deliv-
ery ratio. Additionally, we measured the power consumption of
LoPy4 devices working with LoRa-Ducy in different modes.
Furthermore, we have compared its power consumption to
LoRaWAN and have demonstrated its potential to save energy.
With this implementation, which is also available open-source,
the research community can achieve higher traffic volumes
and enable multi-hop communications. We plan to use this
implementation in our ongoing work in underground sensor
networks [23].
REFERENCES
[1] Pycom - Next Generation Internet of Things Platform.
https://docs.pycom.io/. Library Catalog: pycom.io.
[2] S. Ahsan, S. A. Hassan, A. Adeel, and H. K. Qureshi. Improving
Channel Utilization of LoRaWAN by using Novel Channel Access
Mechanism. In 2019 15th International Wireless Communications
Mobile Computing Conference (IWCMC), pages 1656–1661, June 2019.
ISSN: 2376-6506.
[3] N. Benkahla, H. Tounsi, Y.-Q. Song, and M. Frikha. Enhanced Dynamic
Duty Cycle in LoRaWAN Network. In N. Montavont and G. Z.
Papadopoulos, editors, Ad-hoc, Mobile, and Wireless Networks, Lecture
Notes in Computer Science, pages 147–162, Cham, 2018. Springer
International Publishing.
[4] M. Buettner, G. Yee, E. Anderson, and R. Han. X-MAC: A short
preamble MAC protocol for duty-cycled wireless sensor networks.
volume 4, pages 307–320, Jan. 2006.
[5] A. Dunkels. The ContikiMAC radio duty cycling protocol. May 2012.
[6] P. Dutta, S. Dawson-Haggerty, Y. Chen, C.-J. M. Liang, and A. Terzis.
Design and evaluation of a versatile and efficient receiver-initiated link
layer for low-power wireless. In Proceedings of the 8th ACM Conference
on Embedded Networked Sensor Systems, SenSys ’10, pages 1–14,
Z¨
urich, Switzerland, Nov. 2010. Association for Computing Machinery.
[7] A. El-Hoiydi and J.-D. Decotignie. WiseMAC: An Ultra Low Power
MAC Protocol for Multi-hop Wireless Sensor Networks. In S. E.
Nikoletseas and J. D. P. Rolim, editors, Algorithmic Aspects of Wireless
Sensor Networks, Lecture Notes in Computer Science, pages 18–31,
Berlin, Heidelberg, 2004. Springer.
[8] M. O. Farooq and D. Pesch. A Search into a Suitable Channel
Access Control Protocol for LoRa-Based Networks. In 2018 IEEE 43rd
Conference on Local Computer Networks (LCN), pages 283–286, Oct.
2018. ISSN: 0742-1303.
[9] C. E. Fehri, M. Kassab, S. Abdellatif, P. Berthou, and A. Belghith.
LoRa technology MAC layer operations and Research issues. Procedia
Computer Science, 130:1096–1101, 2018.
[10] A. Lavric and V. Popa. Performance Evaluation of LoRaWAN Com-
munication Scalability in Large-Scale Wireless Sensor Networks, June
2018. ISSN: 1530-8669 Pages: e6730719 Publisher: Hindawi Volume:
2018.
[11] P. J. Marr´
on and K. Whitehouse. Wireless Sensor Networks: 8th
European Conference, EWSN 2011, Bonn, Germany, February 23-25,
2011, Proceedings. Springer Science & Business Media, Feb. 2011.
Google-Books-ID: D3OR7uCztGwC.
[12] D. Moss and P. Levis. BoX-MACs: Exploiting Physical and Link Layer
Boundaries in LowPower Networking. Technical report, 2008.
[13] R. Musaloiu-E., C.-J. M. Liang, and A. Terzis. Koala: Ultra-Low Power
Data Retrieval in Wireless Sensor Networks. In 2008 International
Conference on Information Processing in Sensor Networks (ipsn 2008),
pages 421–432, Apr. 2008.
[14] M. O’Kennedy, T. Niesler, R. Wolhuter, and N. Mitton. Practical
evaluation of carrier sensing for a LoRa wildlife monitoring network.
page 8.
[15] K. Pister and L. Doherty. TSMP: Time synchronized mesh protocol.
IASTED International Symposium on Distributed Sensor Networks, DSN
2008, Jan. 2008.
[16] J. Polastre, J. Hill, and D. Culler. Versatile low power media access
for wireless sensor networks. In Proceedings of the 2nd international
conference on Embedded networked sensor systems - SenSys ’04,
page 95, Baltimore, MD, USA, 2004. ACM Press.
[17] Y. Sun, O. Gurewitz, and D. B. Johnson. RI-MAC: a receiver-initiated
asynchronous duty cycle mac protocol for dynamic traffic loads in
wireless sensor networks. In Proceedings of the 6th ACM conference on
Embedded network sensor systems - SenSys ’08, page 1, Raleigh, NC,
USA, 2008. ACM Press.
[18] T. To and A. Duda. Simulation of LoRa in NS-3: Improving LoRa
Performance with CSMA. In 2018 IEEE International Conference on
Communications (ICC), pages 1–7, May 2018. ISSN: 1938-1883.
[19] V. Toldov, L. Clavier, and N. Mitton. Multi-channel Distributed MAC
protocol for WSN-based wildlife monitoring. In 2018 14th Interna-
tional Conference on Wireless and Mobile Computing, Networking and
Communications (WiMob), pages 1–8, Oct. 2018. ISSN: 2160-4886.
[20] T. van Dam and K. Langendoen. An adaptive energy-efficient MAC
protocol for wireless sensor networks. In Proceedings of the 1st interna-
tional conference on Embedded networked sensor systems, SenSys ’03,
pages 171–180, Los Angeles, California, USA, Nov. 2003. Association
for Computing Machinery.
[21] J. Wotherspoon, R. Wolhuter, and T. Niesler. Choosing an integrated
radio-frequency module for a wildlife monitoring wireless sensor net-
work. In 2017 IEEE AFRICON, pages 314–319, Sept. 2017. ISSN:
2153-0033.
[22] W. Ye, J. Heidemann, and D. Estrin. An energy-efficient MAC protocol
for wireless sensor networks. In Proceedings.Twenty-First Annual
Joint Conference of the IEEE Computer and Communications Societies,
volume 3, pages 1567–1576 vol.3, June 2002. ISSN: 0743-166X.
[23] I. Zaman and A. F ¨
orster. Challenges and opportunities of wireless
underground sensor networks. In Fachgespr¨
ache sensornetze, 2018.
[24] D. Zorbas, K. Abdelfadeel, P. Kotzanikolaou, and D. Pesch. TS-LoRa:
Time-slotted LoRaWAN for the Industrial Internet of Things. Computer
Communications, 153:1–10, Mar. 2020.