Received March 25, 2021, accepted April 4, 2021, date of publication April 12, 2021, date of current version

April 20, 2021.

Digital Object Identifier 10.1109/ACCESS.2021.3072375

Priority-Aware Fast MAC Protocol for

UAV-Assisted Industrial IoT Systems
SHREYA KHISA AND SANGMAN MOH , (Member, IEEE)
Department of Computer Engineering, Chosun University, Gwangju 61452, Republic of Korea
Corresponding author: Sangman Moh ([email protected])
This work was supported in part by the National Research Foundation of Korea (NRF) Grant by the Korean Government through Ministry
of Science and ICT (MSIT) under Grant 2019R1F1A1060501.

ABSTRACT Many hazardous industrial incidents can occur due to the inadequate and inefficient monitoring
of the offshore plants. Manual inspections of the offshore plants on a regular basis is not only time consuming
but also dangerous regarding to human safety. For considering the safety measurement and alleviating the
burden of the manual inspection, unmanned aerial vehicles (UAVs) can be effectively utilized to collect
data from the remote industrial environment. In an industrial scenario, less delay is required for emergency
packets and high throughput is needed for monitoring packets. This paper proposes a priority-aware fast
MAC (PF-MAC) protocol for UAV-assisted industrial Internet-of-things (IIoT) systems, ensuring fast and
robust data delivery. At first, the IoT devices under the UAV communication range transmit a reservation
frame to the UAV to catch transmission opportunities using CSMA/CA. The devices utilize static traffic
priority and a novel adaptive backoff mechanism during CSMA/CA. After receiving the reservation frames
from the IoT devices, the UAV calculates the dynamic device priority based on their static traffic priority,
communication duration, sampling frequency, and remaining energy. Then, time slot is assigned by the UAV
to each device for data transmission. To ensure fairness, if a device fails in contention during the CSMA/CA
period, the static traffic priority is raised in the next retransmission. There is no prior work in the literature
that considers both the traffic priority and the device priority to ensure Quality of Service in IIoT and related
systems. According to our performance study, the proposed PF-MAC outperforms the conventional protocols
in terms of delay and throughput.

INDEX TERMS Internet of Things, unmanned aerial vehicle, medium access control, traffic priority, device
priority, delay, quality of service.

I. INTRODUCTION Due to the lack of mobility, however, they cannot be deployed

Unmanned aerial vehicles (UAVs) have gained enormous in remote scenarios. In addition, their implementation cost
attention from researcher communities and commercial is very high, which makes them impractical for remote and
industries [1]. Recently, UAVs are deployed for a numerous urgent situations. This situation motivated the implementa-
applications for example surveillance [2], search and recov- tion of non-traditional communication networks such as Loon
ery [3], fire and radiation monitoring [4], [5], sports and project [7] and the drone project of Facebook [8]. Due to their
entertainment [6], and so on. On the other hand, developments capacity to navigate, easy deployment, ability of hovering,
in wireless and mobile networking technology have affected and practical cost, small-size UAVs have gained more atten-
every aspect of our everyday lives. The demand for large tion. UAVs have recently been considered to act as wireless
bandwidth as well as the capability to connect always and relays to cellular network coverage [9] and satellite communi-
everywhere is increasing rapidly. cations [10]. In particular, due to Line-of-Sight (LoS) access,
Conventional networking systems have been expanded to less signal blockage, and less shadowing effects, UAVs can
every corner, relying on infrastructure-dependent networks. facilitate greater communication links between air and land
station.
The associate editor coordinating the review of this manuscript and Modern network technologies such as space–air–ground
approving it for publication was Byung-Seo Kim . integrated network (SAGIN) have recently created appeal

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
VOLUME 9, 2021 57089
 S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

to both academia and industry. Many organizations such as Therefore, limited communication time is one of the chal-
Global Information Grid [11], Oneweb [12], and SpaceX [13] lenges while developing a UAV–IoT communication system.
began their SAGIN ventures in recent years. SAGIN can be At most times, UAVs are equipped with directional anten-
used in many functional areas such as earth observation and nas and the formation of coverage in ground is circular. The
mapping, intelligent transport [14], and disaster rescue [15] ground devices can access UAVs at different times. There-
due to the inherent benefits of broad coverage, high through- fore, fair access to the channel by all ground devices is neces-
put, and good resilience. Satellites, on the other hand, are able sarily important. In addition, a priority-based channel access
to provide consistent coverage to ocean, rural, and moun- mechanism may be needed for Quality of Service (QoS) in
tain areas. UAV-based networks can expand bandwidth for IoT networks. Furthermore, energy efficiency has become
large areas with high service requirements, while Internet a major challenge for remotely located IoT systems. Such
of Things (IoT) devices located on ground can provide the battery-constrained devices can need energy-efficient mecha-
connectivity with high data rate. In the next few years, SAGIN nisms for channel access and transmission control. Therefore,
will carry many facilities and resources from space to the selecting and developing a medium access control (MAC)
earth. protocol for data transmissions to handle the challenges is
Currently, we are experiencing a steady and continuous extremely demanding.
penetration of IoT concepts into the industrial domain, called Contention-based MAC protocols are popular owing to
industrial IoT (IIoT) or Industry 4.0. Every industry is trying their comprehensibility, adaptability, and less overhead char-
to enjoy the benefits of the industrial revolution by adopt- acteristics. Devices have the ability, without undue overhead,
ing IIoT features. Oil and gas industries are also not an to dynamically enter or quit the network. However, collisions
exception in this aspect. However, the processing of off- increase when the number of devices is high. However, a
shore oil and gas (O&G) is highly dynamic and precari- time-division multiple-access (TDMA)-based channel access
ous. O&G companies find it difficult to obtain a timely and mechanism can solve this problem. The TDMA system is
accurate image of their ongoing output due to the remote- divided into time slots, and each device can only trans-
ness and isolation of offshore rigs. Insufficient monitoring mit within its own allocated slots. [19]. The key primary
capacity may lead to catastrophic explosions that take a disadvantage of TDMA is that if there are a few num-
heavy toll on the environment, the lives of employees, and bers of IoT devices, the transmission slot can be wasted.
the reputation of companies. Tragic incidents such as the Therefore, to construct a scalable and versatile communica-
Exxon Valdez [16] and Deepwater Horizon oil spills [17] tion system for UAV-IoT communication network, only the
are examples. contention-based or contention-free mechanism cannot be
O&G manufacturing requires day-and-night observance of suitable.
various equipment (pipes, valves, wellheads, and tanks) and This paper proposes a priority-aware fast MAC (PF-MAC)
parameters (temperature, vibration, friction, flow rates, cor- protocol for UAV-assisted IIoT systems, ensuring fast and
rosion, and gas leaks) to maximize efficiency and protection. robust data delivery. The hybrid PF-MAC protocol integrates
However, conventional communication methods for linking a the benefits of both contention-based and contention-free
large number of different assets on offshore drilling platforms protocols for a remote IoT scenario.
are limited, costly, or cumbersome. Wired technology like This paper’s main contributions are as follows:
supervisory control and data acquisition (SCADA), which is 1) A priority-aware fast MAC (PF-MAC) protocol is pro-
suitable for real-time control management activities, is not posed for UAV-assisted remotely located IIoT systems
equipped for data acquisition from remote locations. In the with QoS requirements, which is a hybrid MAC pro-
center of the ocean, cellular connectivity is most possibly tocol incorporated with the CSMA/CA and TDMA
absent. In addition, setting up mesh networks is also a dif- mechanisms. The UAV acts to gather data from IoT
ficult effort due to the enormous scale, complexity, and dense devices as a wireless relay.
structure of oilrigs. The emerging use of UAVs in remote 2) We introduce an incremental contention priority (ICP)
monitoring applications makes it a suitable candidate for scheme wherein, if an IoT device fails in con-
O&G industries offshore. tention, its static priority is increased by one in the
Several UAV-based air–ground IoT systems have already next retransmission. It ensures the access fairness of
been proposed in the literature, wherein UAV is placed as a devices and prevents the low-priority devices from
mobile base station (BS) [18]. In this manner, UAV-based starvation.
networks eliminate complex routing schemes and greatly 3) We adopt the ABO mechanism wherein the backoff
improve data collection capabilities. Analyzing and devel- period is calculated according to the collision rate with
oping a UAV-based data acquisition system is highly difficult respect to the maximum retransmission of the channel.
due to the mobility and complex nature of the UAV. Devices It helps to reduce the delay significantly.
can reach UAV when the UAV is near to them because of 4) We design the dynamic priority of the devices based
the mobility of UAVs, and the devices lose their wireless on the static traffic priority, communication duration,
link connection if the UAV flies outside of their coverage. sampling rate, and remaining energy. This helps UAV

57090 VOLUME 9, 2021

S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

to allot a timeslot for the devices depending on the optimizing the hovering altitude of UAV and controlling the
significance and emergence of the data frame. power of ground users in SAGIN environment is discussed.
5) Our performance evaluation shows that, the PF-MAC Recently, machine learning-based techniques are becom-
beats the conventional protocols in terms of throughput, ing famous for its adaptability with the dynamic environ-
and average delay. ment. A Q-learning-based resource allocation algorithm is
presented in [30] by handling the channel collision problem in
The paper is organized in the following way. In Section II, dense wireless local area networks. In order to reduce energy
the related works are reviewed. In Section III, the system consumption, a Q-learning-based MAC protocol is proposed
model of our study is introduced. Subsequently, the pro- in [31] which is called greenMAC. Energy consumption
posed PF-MAC protocol is presented in detail in Section IV. decreases by managing the channel collisions properly, which
In Section V, the proposed PF-MAC is analyzed in terms of also enhances system reliability.
major performance metrics. In Section VI, the performance of Some studies have recently focused on the MAC protocol
PF-MAC evaluated via extensive simulation and compared to regarding the communication process between the UAV and
that of conventional schemes. In Section VII, the conclusion ground IoT devices. A comprehensive survey of MAC proto-
of the paper is drawn. cols for UAV-based IoT is presented in [32]. A UAV-based
IoT data collection system for aggressive and inaccessible
areas was proposed by Lin et al. [33]. The key aim of their
II. RELATED WORKS analysis was to improve the entire system’s energy efficiency.
This section presents the related works on UAV-based IoT For data collection, slotted ALOHA-based approach is fol-
communication network and the problems of existing studies, lowed. However, considerable energy is wasted due to col-
which form the motivation of our work. lided slots, empty slots, and overhearing. CSMA/CA-based
Several studies have considered UAV-based IoT networks. MAC protocol for a UAV-based IoT network was explored by
Different types of gateway selection algorithms and cloud- the authors of [34]. For each device in each cluster, the size
based stability-control mechanisms for flying ad-hoc net- of the contention window is modified according to the com-
works (FANETs) is presented in [20], where FANET and its munication period between the IoT and the UAVs. The CW
protocol architectures are also discussed. To provide seamless size must be periodically measured and modified, adding to
network coverage in dense urban areas, an energy-efficient the system’s consumption of energy and time.
UAV deployment strategy is presented in [21], which opti- In [35], a TDMA-based workflow model is introduced
mizes both UAV deployment and UAV recharging strategy where UAVs work both as a data collector and wireless
using particle swarm optimization (PSO). The efficient path power transferor to the ground IoT devices. However, due to
planning of UAVs for data gathering from IoT devices is a the usage of the TDMA for modeling the multi-workflow,
widely popular research area. A previous study [22] investi- high synchronization overhead is predicted. Our PF-MAC
gated the timely delivery of information using UAVs as relays can handle these problems by utilizing both contention-based
by optimizing UAV flight trajectory. Using a bio-inspired and contention-free mechanisms. Only successful devices in
algorithm, UAV path planning is developed in [23]. A joint contention use the timeslots for data transmission during a
optimization technique for UAV trajectory and resource allo- contention-free process. Hence, no energy is wasted due to
cation is considered in [24]. UAV is an essential element for empty slots. We used a predefined CW size for each IoT
creating a future smart city. In [25], a smart city architecture device according to its static priority. Therefore, there is no
is proposed where UAVs form a 5G hierarchical IoT net- need of frequent calculation of the CW size. Moreover, the
work in the sky, linking to a number of BSs on the ground. integration of contention-based and contention-free mecha-
An architecture is presented in [26] considering UAV-aided nisms helps to reduce the large synchronization overhead,
IoT for air quality sensing in smart cities. These studies which is a major problem in TDMA mechanism if we use it
mainly focused on network architecture and UAV trajectory for the entire process. Our proposed study will be presented
optimization. Only few of them have worked on the UAV–IoT exhaustively in Section IV.
data communication protocol. However, it is very important
for efficient data gathering from IoT devices. III. SYSTEM MODEL
The integration of UAVs with terrestrial and satellite net- We consider an industrial offshore environment such as
works has shown a new research path to the industries and oil/gas rig monitoring, situated far away from the main indus-
researchers. In [27], a comprehensive survey on space–air– try that is difficult and dangerous for human access on a daily
ground-integrated (SAGIN) network is presented that covers basis. It is quite impossible to obtain cellular connectivity in
design of network, resource allocation, and optimization. the middle of the ocean. The remote location and isolation of
In [28], a SAGIN-based scheduling approach for task offload- offshore rigs make it difficult to obtain a proper estimation
ing is presented where IoT devices can offload their tasks of the ongoing output for O&G companies. O&G businesses
to closer UAVs. Based on the task’s importance and weight, rely on manual data reading and visual inspection to track
the UAV decides whether to transfer the task to a nearby BS large parts of their processes, equipment and facilities owing
or satellite. In [29], a methodology for resource allocation by to the lack of a cost-effective and scalable communication

VOLUME 9, 2021 57091

 S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

the transmission. Before the transmission of data, if a new

packet arrives, the new packet will replace the former packet.
It ensures that there is always only one packet in the buffer of
each device.
We have assumed an offshore industrial plant situated in
the middle of the ocean. There is no significant obstacles
in the area of interest except the industrial plant itself. Sig-
nal propagation for air-to-ground over the sea area is the
same as the open space with different surface reflectivity and
roughness. Propagation over sea can also be hampered by the
height of waves, which causes anomalous index of refraction
variation with heights and results in propagation loss less
than that of free space [36]. By considering these facts, we
have measured the propagation loss PL with two-ray path
loss model as in [37]. It can be calculated with the following
equation in Decibels:
FIGURE 1. An application scenario of data gathering in a UAV-assisted (  )
λ 2 2πhT hR 2
 
industrial IoT system.
PLtwo−ray = −10 log10 ( ) 2 sin ,
4πd λd
solution. This is highly inefficient, error-prone, and danger- (2)
ous to field workers. where λ is the wavelength in meters, d represents the prop-
In this paper, a SAGIN-based scenario is considered agation distance in meters, and hT and hR are the height of
wherein a UAV manages data gathering from IoT devices transmitter and receiver, respectively. However, the path loss
and relay data to the distant ground station with the help totally depends on the area of interest. Our framework can
of satellites. However, UAVs are not considered to manage be adjusted into a different path loss model under different
the internal process of the industry; UAVs are only used environmental conditions as well.
to inspect, monitor, and gather data from numerous pipes, TABLE 1 shows the notations used in the paper.
valves, wellheads, and tanks, which are dangerous and haz-
ardous for human access. A. UAV, IoT DEVICES, AND ANTENNA TYPE
Fig. 1 shows an application scenario of data gathering
Owing to its flexibility and easy mobility, rotary-wing UAVs
in a UAV-assisted industrial IoT system. A low-earth-orbit
are well suited for inspection applications. For an extended
(LEO) satellite, a multi-rotor UAV, and N number of IoT
duration, rotary-wing UAVs may conduct precision maneu-
devices are considered. Let ϕ = {m1 , m2 , . . . ..mN } denote
vering and hold a visual on a single target. The greatest value
the set of the IoT devices. The limited-capability IoT devices
of rotary-wing UAVs is their ability to vertically take off and
are spread over the given area and the monitored data are
land. In our scenario, both UAV and IoT devices are equipped
collected continuously. To support as many IoT devices as
with directional antenna (phased array) [38], [39], [40],
possible, the UAV is dispatched to the specified area on a
a software-defined radio, and a global positioning system
regular basis. All IoT devices are fixed in their position,
(GPS). Software-defined radio allows the IoT devices and
and we consider no mobility of IoT devices. The flight
the UAV to work on different channels. By concentrating
path and time of a UAV are preplanned before starting its
on not only transmitting energy in one direction but also on
mission by using a central controller system. It is assumed
decreasing interference and fading, directional antennas pro-
that UAV will fly at a fixed height H (H > 0). The
mote communication efficiency. The use of directional anten-
UAV’s location ul at given time t is xuav,t , yuav,t , zuav,t .


nas has other benefits such as lower latency, higher spatial
The IoT devices location mn is given by (xi , yi , zi ). The dis-
reuse, and high quality of links that provide higher through-
tance d(ul , mn ) between UAV and IoT devices is computed
put. We have assumed that UAV is capable of full-duplex
as follows:
q communication. Therefore, it can collect the data from IoT
d(ul , mn ) = (xi − xuav,t )2 + (yi − yuav,t )2 + (zi − zuav,t )2 . devices in downlink and transfer the data to the control
station via satellite in uplink simultaneously. The UAV and
i = 1, 2, . . . .N , t = 1, 2, . . . N (1)
IoT devices exchange their location information via control
We assume that, for every IoT device, data packet arrival packets. Hence, the UAV and the IoT devices can direct
process is considered as a Poisson arrival process and packet their beams towards each other during the communication
arrival rate is defined as λ. As we consider a controlled process, which can reduce the direction alignment problem
scenario, we assume that packet arrival rate for all IoT devices significantly. The phased array based directional antennas can
are the same. In each device, a packet is buffered as long electronically steered to point in different directions without
as it gets the opportunity for transmission and completes moving the antennas physically.

57092 VOLUME 9, 2021

S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

TABLE 1. Notations used in this paper.

FIGURE 2. Channel switching.

ba . The beacon ba carries the data channel information.

After receiving beacon ba , the IoT devices switch to the
data channel Cd immediately and wait for their designated
transmission slot, tslot . After transferring the data frame Ld
to the UAV, the IoT devices switch back to control channel
Cc and wait for the next beacon period. The CC2420 and
the more sophisticated CC2500, with a channel switching
time of only 300 µs and 90 µs, respectively, are typical
transceivers for short-range wireless communication [41].
However, the channel-switching mechanism is not focused in
this paper.

C. PRIORITY OF IoT DEVICES

In this subsection, we describe the priority outline of IoT
devices based on static traffic priority and dynamic device
priority in detail.

1) STATIC TRAFFIC PRIORITY Ps

IoT devices are heterogeneous in nature. The nodes’ static
priority depends on the criticality of the traffic generated by
the node. It is assumed that every IoT device can produce
two types of traffic: emergency traffic te and monitoring
traffic tm . Based on its static traffic priority Ps , the traffics
are prioritized. We can denote the static traffic priority factor
with σ .

a: EMERGENCY TRAFFIC te
This refers to critical and urgent traffic. These packets need
to be sent to the UAV as soon as possible. However, it is
B. MULTICHANNEL STRUCTURE assumed that the generation of emergency traffic te does not
As shown in Fig. 2, the use of two channels is conducted: occur frequently and is only generated when the system does
a control channel Cc and a data channel Cd . The control not perform in a normal manner and something goes wrong
channel Cc is dedicated for exchanging control information internally. This type of situation is life threatening and haz-
such as broadcasting beacons, exchange of control packets ardous to humankind. A few circumstances when emergency
and acknowledgment between the UAV and the IoT devices. traffic can be generated are fire alarm, oil/gas leakage, and
On the contrary, remaining communication is performed high air-pollution level. For emergency traffic te the value of
using data channel Cd . The UAV flies over the target area σ is 1.
and starts to transmit beacon, bn via control channel Cc . The
IoT devices receive the beacon in control channel Cc and b: MONITORING TRAFFIC tm
then send control packets (reservation frame, Lr ) to the UAV. Monitoring traffics tm are generated regularly for mon-
Upon reception of the control packets, the UAV calculates itoring purposes and do not have any deadline bound.
the dynamic device priority and transmits another beacon ba Considering the characteristics of the monitoring traffic,
to the IoT devices, including the data transmission schedul- the static traffic priority factor for monitoring traffic tm is
ing and synchronization information. The UAV switches to σ = 0; this is because it does not have any urgency of
the data channel Cd immediately after sending the beacon transmission.

VOLUME 9, 2021 57093

 S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

2) DYNAMIC DEVICE PRIORITY Pd c: REMAINING ENERGY FACTOR RmE i

Only static traffic priority, Ps of the device cannot fully and We assume that each IoT device is battery-powered and con-
accurately determine the priority of the IoT devices where tains a certain amount of energy depleted over time. We also
QoS requirements are much necessary. Due to UAV mobility assume that after reaching a certain threshold energy level,
and remoteness of the system environment, other few criteria the device is considered unable to communicate with the
play a great role in deciding the total priority Pt of an IoT UAV. The device will therefore not complete the transmission
device. Therefore, we design the dynamic device priority Pd of data. The remaining energy can be calculated as follows:
based on static traffic priority Ps and some other factors such
RmEi = Epresent,i − Etransmission,i , (5)
as communication duration, sampling rate, and remaining
energy. The details discussion of dynamic device priority Pd where Epresent,i and Etransmission,i the present level of energy
is given below. and energy consumed during data transmission by device i
respectively, are represented.
a: COMMUNICATION DURATION PRIORITY FACTOR cp
It is very important for UAV to collect all the data from 3) TOTAL DEVICE PRIORITY Pt
IoT devices during its single flight period. When the UAV In this section, the calculation of total device priority based
is within the contact range of IoT devices, the IoT devices get on the above-discussed criteria will be discussed. The static
the opportunity to connect with the UAV. Therefore, each IoT traffic priority factor σ and sampling rate factor sp are two
device has limited communication time due to the mobility relatively simple assessment criteria of the priority correlated
of UAV. If the UAV is within the communication range of with the device itself. These two factors are related to the
IoT devices, the IoT devices are also within the scope of IoT device itself rather than the communication process. The
communications of the UAV. If the communication range of static traffic priority factor, σ , communication duration fac-
UAV is Ru , the velocity of the UAV is uv . The communication tor, cp and sampling rate factor, sp and their total aggregated
duration, Ti between the UAV and the IoT devices can be amount are utilized as follows to measure the basic level of
measured [34] as: priority of the IoT devices:

2Ru cos θ Pbase,i = floor(cp,i + σi + sp,i ), i = 1, 2, . . . .N , (6)

Ti = , i = 1, 2, . . . .N , (3)
uv Remaining energy factor, RmE i , is the most important fac-
tor in the entire priority system. It is because, if the IoT device
where θi ∈ (0, π 2) denotes the device positions based

does not have a minimum remaining energy, it cannot execute
on created angle between  UAV and IoT device and can be other tasks. Moreover, if RmE i of the IoT device is less than
computed as θi = arcsin(yi Ru ). the threshold level, the communication between UAV and the
The higher the communication duration is, the lower its pri- IoT device will not take place. Therefore, we consider that if
ority is, and vice versa. However, to make the calculation eas- RmE i is less than the threshold level, the total device priority
ier, we assign communication priority value, vc = 1, 2, . . . K becomes zero. By considering RmE i , we can obtain the total
to the values of Ti in descending order. If the communication dynamic priority of device, Pt,i , as follows:
time is the lowest, it will get the highest communication (
priority value. On the contrary, if the communication time is Pbase,i if (RmEi > RmEth,i )
Pt,i = i = 1, 2, . . . .N . (7)
the highest, then it will get the lowest communication priority 0 otherwise,
value.
IV. PRIORITY-AWARE FAST MAC
b: SAMPLING FACTOR sp This section shows the frame structure of PF-MAC and the
The weight of the traffic is measured by the sampling factor, detailed communication process between the IoT devices and
sp , of the IoT devices. The level of the sampling frequency the UAV are discussed. In our scenario, there exists three
variates from the regular sampling frequency indicates the communication processes: communication between the IoT
importance of the IoT device-generated data frames. If the devices and the UAV, communication between the UAV and
value of sp is high, the priority of the device is also high. For the satellite, and communication between the satellite and the
example, the temperature variates from regular values extend ground terminal. We herein focus only on communication
when the internal process is not in the normal condition. The between the IoT devices and the UAV. The entire process of
sampling factor, sp , can be defined as communication is divided into four parts: notification period
(NP), reservation period (RP), announcement period (AP)
(su − s)2 − (s − sl )2 and data collection period (DCP).
sp = , (4)
(|su | + |sl |)2
A. FRAME STRUCTURE OF PF-MAC
where s, su , and sl denote the sampling frequency of the Fig. 3 shows the frame structure of the proposed PF-MAC.
device, standard sampling frequency upper, and lower bounds The notification beacon message bn includes the packet type,
respectively. UAV ID, GPS position of the UAV at a particular time, and

57094 VOLUME 9, 2021

S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

ABO (described in Subsection IV-C.3) mechanism, reserva-

tion frame Lr is retransmitted to the UAV for a limited number
of times. If the transmission is successful, the IoT device will
receive an acknowledgment ACK message from UAV and
hence stops retransmitting Lr and waits for the AP duration.
If the device does not receive any ACK from the UAV and
retransmission time is exceeded, then the frame is dropped.
Algorithm 1 shows the reservation period communication
from IoT side.
Algorithm 1 Reservation Period – IoT Side
FIGURE 3. Frame structure of PF-MAC. Input: Notification beacon message, bn , reservation frame,
Lr
UAV speed. The reservation frame Lr comprises of the type Output: Successful transmission of reservation frame, Lr
of the packet, IoT device ID, GPS location of the IoT device, 1: for each IoT device i ∈ N
static traffic priority indicator, sampling rate of the generated 2: if bn is received in Cc
data, and remaining energy of the device. The announcement 3: try to transmit Lr in m-th transmission
beacon message ba , generated by the UAV after receiving the where m ≤ 7
reservation frame, includes the packet type, device ID, data 4: check the static traffic priority factor, σ
channel information, and scheduling information. Finally, 5: if (σ == 0)
upon receiving the announcement beacon message, the IoT 6: utilize large CW range
devices transmit the data frame Ld that includes information 7: else
on the packet type, IoT device ID, and the message itself. 8: utilize small CW range
9: end if
B. NOTIFICATION PERIOD (NP) 10: if (Lr fails in contention in m-th transmission)
After arriving to the designated location, the UAV broadcasts where m ≤ 7
a notification beacon message bn to notify its presence to all K 11: calculate BO using (8) and perform
number of devices in the field of UAV coverage. After receiv- BO 12: in (m + 1)-th transmis-
ing notification beacon bn message, the devices that have data sion, increment σ
to send will wake up. To conserve energy, the devices with no by 1
data will go to sleep mode. The notification beacon message 13: if (successful)
bn includes UAV speed uv and location ul . 14: σ returns to original value
15: else
C. RESERVATION PERIOD (RP) 16: go to step 11
This period is contention-based period and follows 17: end if
CSMA/CA mechanism. 18: else
19: receive ACK
1) IOT DEVICES IN RP 20: end if
After receiving notification beacon message bn , the active 21: end if
devices will contend with each other for reservation oppor- 22: end for
tunity using the basic access mechanism of CSMA/CA. Two
contention windows CW are selected based on the static
traffic priority Ps of the IoT devices. The IoT devices whose 2) UAV IN RP
σ = 1 will use small CW size, allowing them to access the After receiving the reservation frame from the IoT devices,
channel within a shorter time. Therefore, they can transfer UAV performs a prioritization process, discussed in Subsec-
their reservation frame, Lr with lower delay. The devices tion III-C.2. Algorithm 2 demonstrates the reservation period
with σ = 0 are not delay-sensitive. Hence, they utilize mechanism from UAV side.
larger CW so that they obtain channel access after the higher-
priority devices. The devices within the UAV coverage area, 3) ADAPTIVE BACKOFF (ABO) MECHANISM
which has data to send, will send reservation frame Lr to In the conventional IEEE standard 802.11 CSMA/CA mech-
the UAV. The devices with no data to send will switch to anism, if a packet transmission fails, to determine the BO
sleep mode. The contention becomes successful when only duration, the CW size gets doubled and a random backoff
one device sends the data at a time. A collision happens if value is chosen. It only considers packet transmission failure.
more than one device sends reservation frame, Lr within the Other parameters such as current channel status and collision
same time interval. After collision, if the UAV is still under rate are ignored. Our ABO considers the collision rate with
the communication range of the IoT devices, following an respect to the maximum number of retransmission of the

VOLUME 9, 2021 57095

 S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

FIGURE 4. Flowchart of the ABO mechanism.

Algorithm 2 Reservation Period – UAV Side packet in the channel. In comparison to the traditional backoff
Input: Reservation frame Lr mechanism where the CW size always gets increased in the
Output: TDMA scheduling information same way without considering the current condition of the
1: Dlist ← {} channel, our ABO adjusts the BO length according to the
2: Clist ← {} medium collision rate.
3: K = number of IoT devices, vc = K + 1 Fig. 4 shows the flowchart of the ABO mechanism. The
4: for each IoT device i ∈ K primary backoff period BO0 is set to CW min . The next BO
5: calculate cp using (3) is determined with the following equation after a packet is
6: insert (Clist [i]) dropped due to collisions in the channel:
7: end for BOj = j × (2CWmax − 1) × α, (8)
8: sort_ascending (Clist )
9: for each IoT device i ∈ K where j presents the collision rate in the channel in the j-th
10. Clist [i] = vc − 1 attempt. CW max is the maximum value of CW , α is a random
11: calculate sp using (4) number, the value of which lies in [0, 1].
12: calculate RmE using (5) The collision rate j can be calculated with the number of
13: if (RmE i > RmE th,i ) transmission failures and the maximum retransmission limit
14: calculate Pt using (6) after each transmission of reservation frame. Hence, collision
15: insert (Dlist [i]) rate j in respect to the maximum retransmissions can be
16: else calculated using the following equation:
17: Pt = 0 =max − ψj
18: end if j = , (9)
=max
19: end for
20: sort_descending (Dlist ) where ψj presents the number of collisions in the channel in
21: assign tslot to Dlist the j-th attempt and =max represents the maximum number of
retransmissions (=max = 7).

57096 VOLUME 9, 2021

S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

FIGURE 5. ABO mechanism and ICP model.

The parameter α in (8) is effectively used to avoid colli- devices can fail during the RP in contention. When a device
sions between the same static priority devices with similar σ fails frequently during contention, the transmission efficiency
value. For example: if device A and device B have the same of the device would degrade dramatically. Moreover, due to
static priority and both have encountered the similar number the static traffic priority Ps of the devices, the low-priority
of collisions, then they will have similar backoff value and devices will suffer from starvation. Moreover, when two or
will collide again. To resolve this problem, we consider a more high priority devices try to transmit at the same time,
random value α and multiply it by the collision rate. Hence, they will face collision. Our ICP model helps to ensure
the calculated backoff value will never be the same with each fairness among the same priority devices and protect the
other. Each IoT device locally calculates the number of col- low-priority devices from starvation. In ICP, if reservation
lisions after transmission failure. The next backoff stage can frame of a specific device fails during transmission, the prior-
adjust its length according to the collision rate. In case of high ity of that frame is increased by one to get channel access in
collision rate, the backoff window size can be minimized in the next transmission. After increasing the priority, when the
order to transmit the packet as soon as possible. On the other device transmits the data successfully, the phase of increasing
hand, if the collision rate is low, then the backoff length will priority will be halted and the device priority will return to the
be larger and the packet will get time to get transmitted in the preliminary level. Fig. 5 shows that when a collision occurs
next transmission. between B and C, following ICP, each device increments its
For example, as shown in Fig. 5, four IoT devices A, B, C, σ value by 1 during retransmission. This mechanism helps
and D compete to transmit reservation frames. Devices A, B, fair access among the IoT devices with different priorities.
and C have emergency traffic te and hence have a similar σ However, due to the use of different CW sizes for the different
value of 1. However, device D has monitoring traffic tm and priority of traffic, most of the time, only emergency traffic te
the value of σ is 0. Devices A, B, and C utilize a small CW will compete and no normal monitoring traffic will compete
range and hence have the opportunity to transmit reservation with them. It also guarantees that to access the channel,
frame Lr rather than device D. Fig. 5 shows that after the competition between different types of traffic will not occur.
collision occurred between devices B and C, BO does not get Therefore, emergency traffic te will always be transmitted
doubled. Instead, ABO calculates BO efficiently and reduces before monitoring traffic tm .
extra delay in channel access.

D. ANNOUNCEMENT PERIOD (AP)

4) INCREASING PRIORITY OF IOT DEVICES USING ICP In this period, UAV broadcasts the announcement beacon
MODEL TO MAINTAIN FAIRNESS message to all devices under the communication range of
The PF-MAC’s key function is to ensure QoS for different UAV. The announcement beacon message includes the trans-
traffic types and transmission of emergency traffic te over nor- mission scheduling information of the data packets. Synchro-
mal monitoring traffic tm with minimum delay ∅. However, nization information for TDMA is also included in the beacon
the protocol should also be fair enough for data gathering. The message. Upon receiving announcement beacon, the IoT

VOLUME 9, 2021 57097

 S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

devices that succeeded in RP switch to data channel and control channel. As shown in Fig. 6, after receiving notifi-
prepare to send data packets according to their designated cation beacon, the active devices A, B, and C contend with
time slot. each other for reservation opportunity using the basic access
mechanism of CSMA/CA. In the figure, we assume that IoT
E. DATA COLLECTION PERIOD (DCP) devices A and B have high static priority with σ = 1 and IoT
Algorithm 3 displays the data packet transmission during device C has low static priority with σ = 0. Two contention
data collection period. In this period, the devices that became windows are selected based on the static traffic priority Ps of
successful in the RP start sequentially transmitting their data the IoT devices.
using the TDMA mechanism. The timeslots of TDMA are The IoT devices A and B use small CW size, allowing
divided into M number of equal timeslots, which is indexed them to access the channel within a shorter time. Therefore,
by n = 1, . . . .., M with each of length ∂t . In general, the dura- they can transfer their reservation frame with lower delay. IoT
tion of ∂t tends to be small. Therefore, we can assume that the device C is not delay-sensitive. Hence, it utilizes larger CW
position change of the UAV in ∂t is insignificant. Timeslots so that it can obtain channel access after the high-priority
for each IoT device are selected by the UAV according to the devices. As shown in Fig. 7, the IoT devices A and B have
dynamic device priority so that the device with the highest the same value of σ . Both IoT devices A and B use the
dynamic device priority gets the timeslot allocated faster small CW window and try to transmit the reservation frame.
than the other devices. UAV works as a mobile sink and it However, unfortunately, they face a collision due to the trans-
synchronizes with the IoT devices using the announcement mission at the same time. Therefore, following ICP model (in
beacon message. Section IV.C-4), the σ value of both devices is increased by
1 and becomes 2. Both of the devices calculate the backoff
period using the ABO method (in Section IV.C-3) and then
Algorithm 3 Data Collection Period
again try to retransmit the reservation frame. This time IoT
Input: TDMA scheduling information via announcement device A gets the chance to transmit earlier than IoT device
beacon message, ba B and sends the reservation frame successfully.
Output:Successful data transmission On the other hand, after DIFS time, device B can sense
1: for each IoT device i ∈ K that the channel is busy and hence it waits for some time.
2: if ba is received After waiting for some time, when the channel becomes free,
3: switch to Cd IoT device B transmits its reservation frame successfully.
4: synchronizes with the UAV Then, the priority value σ returns to the initial value. In the
5: IoT devices wait for their time slot meantime, IoT device C uses large CW value and after DIFS
6: IoT devices transmit data in their designated tslot and backoff time, it sends the reservation frame successfully.
7: else The devices with no data to send will switch to sleep mode.
8: go to sleep mode The UAV replies with an ACK packet to the every IoT device,
9: end if which has successful transmission. As we consider a con-
10: end for trolled scenario and UAV is going to collect data on a regular
basis from the IoT devices, the UAV will aware of all the IoT
According to the application scenario, the IoT devices devices after the first round of data collection.
should periodically collect and transmit data. It is very much We do not consider any mobility of the IoT devices and
important to synchronize the IoT devices clocks with the hence their position will not change. The IoT devices will
UAV’s clock. It is because the clocks can shift due to the send their location information only in the first round of the
drift in crystal oscillators and data transmission delay. After data collection. If a new device joins the network, it will
receiving the announcement beacon message, all the IoT send the location information to UAV only during its first
devices are synchronized by taking the UAV’s clock as a data transmission. Therefore, UAV is well aware of all IoT
global time. Therefore, all the IoT devices will have the devices and ends the RP period after collecting all reserva-
same clock as the UAV’s clock and thus all IoT devices are tion frame. After getting all the reservation frame, the UAV
synchronized. As we do not use any other control frames for extracts all the information from the reservation frame such
synchronization, it reduces control overhead in comparison to as location information, residual energy, sampling rate, and
the conventional TDMA mechanism. After proper synchro- static priority value. Then, UAV calculates the dynamic
nization, the devices transmit their data packets according to device priority of the devices, which is mentioned in detail in
the scheduled time slot using the data channel. Section III.C-2. Subsequently, UAV assigns TDMA time slot
to the IoT devices based on the calculated dynamic priority.
F. WHOLE COMMUNICATION SCENARIO The UAV notifies the IoT devices about the timeslots with
Fig. 6 illustrates the whole communication scenario of our the announcement beacon message in AP duration. Upon
proposed PF-MAC protocol. For simplicity, we have consid- receiving the announcement beacon message, the IoT devices
ered three IoT devices and a UAV. When UAV reaches to the immediately switch to the data channel and start to transmit
monitoring area, it starts broadcasting notification beacon via data in the designated timeslot.

57098 VOLUME 9, 2021

S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

FIGURE 6. Communication process of PF-MAC.

G. COMPUTATIONAL COMPLEXITY static priority increases after each transmission failure or

The computational complexity of the proposed PF-MAC is collision. We also assume that the packet generation follows
based on the presented three algorithms. The complexity of a Poisson arrival rate λ for each device. In the buffer of each
the algorithm for reservation period at the IoT side mainly device, there always remains only one packet. If a new packet
depends on the number of transmissions and the number of arrives before transmission, the new one replaces the previous
IoT devices. If the number transmissions is m and the number packet. Furthermore, we assume that if the reservation frame
of IoT devices is n, then the complexity of Algorithm 1 is collides, then the packet is dropped and device retransmits the
O (mn). The most expensive process of Algorithm 2 is the reservation frame with increased priority by 1.
sorting mechanism. By implementing, merge sort, even in the Let us assume that γ represents the duration of the unit
worst-case scenario, the time complexity of Algorithm 2 can backoff period and N (γ ) is the number of packets that arrive
be reduced to O(nlogn). Similar to Algorithm 1, the com- during the γ time interval. Let P0 is the probability that at
plexity of Algorithm 3 can be calculated in O (mn) time. least one new packet is produced during the γ interval. Then,
Finally, the overall computational complexity of the pro- we can calculate the P0 with the following equation:
posed PF-MAC can be calculated as: O (mn) + O (nlogn) +
O (mn) = O(mn) because m > log n. P0 = P(N (γ ) ≥ 1). (10)

During RP, if at least one IoT device with static traffic

V. ANALYSIS OF PF-MAC
priority class q transmits the reservation frame, the channel
A. PROBABILITY OF SUCCESSFUL TRANSMISSION
will be busy. Then, we can measure busy channel Pb as
follows:
We assume that there are k ∈ N number of IoT devices
under the coverage area of UAV and each device belongs Pb = 1 − (1 − τq )k , (11)
to
PQone of Q + 1 static priority classes. More clearly, =
q=0 kq ∈ N , where kq represents the number of IoT devices where τq represents the probability that an IoT device in a
in a q static priority class. The priority of each device class static priority q class transmits during a unit backoff period.
does not remain constant during the whole RP period. The The collision occurs if at least one of remaining k − 1 IoT

VOLUME 9, 2021 57099

 S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

device transmits a packet at the same time. The collision where δgen , is the time to generate Lr , δBO is total BO duration,
probability Pc can be expressed as δDIFS represents the DIFS time, δr is Lr transmission time,
Pc = 1 − (1 − τq )k−1 . (12) δcollision is the time spent due to the collision, δACK is the
time spent for receiving ACK , and δSIFS is the SIFS time. The
On the other hand, Ps represents the successful transmission reservation frame transmission time δr can be calculated by
probability that the reservation packet is transmitted success- Lr
fully, which will only be take place if no IoT device transmits δr = , (19)
}
in the same time. Therefore, successful transmission proba-
bility Ps can be expressed as follows: where Lr is the length of reservation frame and } is the data
transmission rate.
kτq (1 − τq )k−1
Ps,RP = . (13)
Pb 2) DELAY IN DCP
Let the total number of generated reservation packets be M . DCP follows the TDMA mechanism. Therefore, there is no
Then, the total number of successfully received reservation delay occurring due to collisions. δDCP represents the delay
packets can be calculated as follows: in DCP.
ξRP = M × Ps,RP . (14) δDCP = δswitch + δsense + δwait + δmedium + δprop , (20)
During DCP, we consider that all IoT devices are synchro- where δswitch is the channel-switching time, δsense is the data
nized with UAV and no synchronization error occurs dur- sensing time for IoT devices, δwait is the waiting time in queue
ing TDMA. There is no collision during TDMA period and before it is transmitted, δmedium is the time placing a packet
IoT devices transmit data during their designated time slot. into medium, and δprop represents the propagation time. Here,
Transmission failure of packet loss can only occur due to the δmedium and δprop can be represented as
transmission delay of the packets. UAV allocates timeslot to
all of the IoT devices based on their dynamic device priority. Ld
δmedium = (21)
If the total number of data packet is G and the total number td
of timeslots are T , then the number of successfully received and
data packets can be given as χ
δprop = , (22)
ξDCP = G × Ps,DCP , (15) st
where Ps,DCP represents the probability of successful trans- respectively, where Ld represents the length of the data
mission during DCP. Ps,DCP can be calculated as following packet, td is the data transmission time, χ is the distance
equation: between IoT device and UAV, and st is the propagation period.
  The waiting time in queue twait can be calculated as
k
Ps,DCP = pi (1 − pi )k−1 , (16) PL
1 X
twait = Si + w, (23)
where pi represents successful data packets transmission in
i=1
time slot ti .
Therefore, combining (13) and (16), we can calculate the where Si is the service period of each IoT device, w is the
total successful transmission probability during RP and DCP waiting time until it is scheduled, and PL denotes the num-
as follows: ber of priority levels for different IoT devices based on the
dynamic device priority.
Ps = Ps,RP + Ps,DCP Combining (18) and (20), we can get the total delay as
kτq (1 − τq )k−1
 
k follows:
= + pi (1 − pi )k−1 . (17)
Pb 1 N
X N
X
δTotal = δRP + δDCP (24)
B. DELAY i=1 i=1
The delay observed by each IoT devices can be computed by
dividing the process into two steps: reservation period (RP) C. THROUGHPUT
and data collection period (DCP). Let T be the system total throughput. T represents the data
transmitted over a transmission time. The throughput for RP
1) DELAY IN RP and DCP duration can be separately calculated as
RP is contention-based and follows the CSMA/CA mecha- Ps,RP × Lr × 8 × }
nism. Hence, extra delay can be observed due to collisions. TRP = (25)
δRP
So, delay in the RP phase can be calculated as follows:
and
δRP = δgen +δBO +δDIFS + δr + δcollision + δACK + δSIFS , Ps,DCP × Ld × 8 × }
TDCP = , (26)
(18) δDCP
57100 VOLUME 9, 2021
S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

respectively. Combining (25) and (26), we can calculate the UAV at its designated ith time slot, ti .
total throughput of the system: m
X
T = TRP + TDCP Etrans = ETx,i × ti , (32)
Ps,RP × Lr × 8 × } Ps,DCP × Ld × 8 × } i=1
= + . (27)
δRP δDCP Failed devices in contention go to idle mode during DCP.
Thus, it consumes the following energy:
D. NORMALIZED CONTROL OVERHEAD
Normalized control overhead is the ratio of control packet Ein = (k − m) × ti , (33)
transmission for packets being delivered from the source node
Therefore, during DCP, the overall energy intake is
to the destination node. If the number of transmitted control
packets is Nc and the number of successfully transmitted data EDCP = Etrans + Ein , (34)
packets is Nd , then the normalized control overhead (NCO)
can be calculated with the following equation: Therefore, during the whole process, the overall energy con-
P sumption of all devices is
Nc
NCO = P . (28)
Nd Etotal = ENP + ERP + EAP + EDCP , (35)

E. ENERGY CONSUMPTION The energy of transmitter, ETx , can be calculated by two

In this section, the explanation is given for energy- different equations according to communication distance.
consumption of PF-MAC. We consider that UAV power is After the signal is produced by the transmitter, the amplifier
rechargeable, obtains power from the control center, and har- will empower it using different power according to the trans-
vests energy from the sun during the daytime. The IoT devices mission distance. If the distance between IoT device and UAV
are battery-powered and non-replaceable. Thus, we mainly is less than the threshold value d0 , it uses the free space model;
concentrate on the energy consumption of IoT devices. otherwise, multi-path fading model is adopted to calculate the
We calculate the energy consumption of each IoT device energy consumption. If the distance is d between IoT device
based on all phases of communication. and UAV, the transmission energy for l-bit data is calculated
In NP period, all IoT devices inside the UAV coverage as:
area obtain the notification beacon message from the UAV.
(
l.Eelec + l.εfs .d 2 , if d < d0
Then, we can calculate the total energy consumption during ETx (l, d) = (36)
NP period for k number of IoT devices with the following l.Eelec + l.εmp .d 4 , if d ≥ d0 ,
equation: where Eelec denotes the power the transmitter use, εfs
ENP = k × ERx , (29) denotes the amplifier power for free-space model, and εmp
denotes the amplifier power for multi-path fading model.
where ERX represents the receiving energy consumption for We can calculate the threshold value d0 using the following
IoT devices. formula:
During the RP period, m contending devices sends a reser- s
vation frame to the UAV. The total energy consumption can εfs
d0 = . (37)
be calculated by εmp
m
X Then, the energy receiver consumes to receive l-bit data can
ERP = (Eidle,i + Ecollission,i + ETx,i × Lr ), (30)
be measured by
i=1
where Eidle is the energy consumed during the idle time ERx (l) = l.Eelec . (38)
preceding the channel’s busy period (collision or success),
Ecollission is the energy consumed during the collision, ETX is VI. PERFORMANCE EVALUATION
the energy consumed for successful transmission of a packet In this section, the performance of the proposed PF-MAC is
and Lr is the length of reservation packet. evaluated through computer simulation and compared with
During the AP period, the UAV broadcasts announce- the modified CSMA/CA [34] and the conventional TDMA
ment beacon message to all successful devices to provide with UAV mechanism. The modified CSMA/CA [34] is the
the scheduling information for the DCP. Therefore, EAP is most recent MAC protocol proposed for UAV-based IoT
the total energy used for the reception of the announcement systems. On the contrary, conventional TDMA with UAV
beacon message from the UAV during AP is calculated by has been used in most data transfer processes of UAV-based
IoT systems where throughput maximization is the main
EAP = m × ERx . (31)
concern. The five performance metrics of average transmis-
During the DCP period, if the transmission scheduling sion delay, network throughput, normalized control over-
information is received from UAV with announcement bea- head, average energy consumption and network lifetime
con message, the device will transmit its data packet to the are evaluated.

VOLUME 9, 2021 57101

 S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

TABLE 2. Simulation parameters.

FIGURE 7. Simulation area of 1000m × 1000m.

MODIFIED CSMA/CA [34]: In this protocol, the IoT

devices are divided into different clusters. In each cluster,
CW size is dynamically adjusted for each device accord-
ing to the communication duration with the UAV. The
devices with low communication duration get the channel
access earlier than the devices with long communication
duration. If the collision occurs, binary exponential back-
off mechanism is adopted to calculate the backoff value. monitoring and emergency traffic, 150 bytes of payload size
The fixed wing UAV is used which follows a straight is selected.
trajectory. Each simulation is run until the energy level of all
CONVENTIONAL TDMA WITH UAV: We use the con- IoT devices decreases below the threshold level. All other
ventional decentralized TDMA mechanism with UAV. The related parameters regarding to UAV flight, data com-
IoT devices broadcast short beacon packets periodically munication, and simulation conditions are summarized in
after a specified time interval to be synchronized with each TABLE 2. The IoT devices are fixed in their position and
other, which reduces collisions during data transmission to the UAV is moving. We assume that both IoT devices
UAV. The UAV follows S-path mobility model for data and UAV are equipped with directional antennas (phased
collection. array). By exchanging necessary location information with
each other, the beams of both transmitter and receiver
A. SIMULATION ENVIRONMENT can point to each other during communication. These type
Fig. 7 demonstrates the simulation area of 1000 m × 1000 m of antennas have been used in some UAV applications
where IoT devices are randomly distributed. The simulation such as [38].
is performed varying the number of IoT devices with repeated
number of rounds. The UAV altitude is approximately 100 m B. SIMULATION RESULT AND DISCUSSION
and flies with a speed of 20 m/s. We have assumed that 1) TRANSMISSION DELAY
UAV follows a predefined trajectory. UAV adopts an S-shape Fig. 8 presents the transmission delay of the emergency data,
mobility model, which makes the UAV to be in the commu- which is averaged for repeated runs. It is apparent from the
nication area of IoT devices for sufficient amount of time. figure that the average transmission delay for emergency data
We assume that each IoT device can generate two types of is proportionally high for conventional TDMA and modified
traffic: emergency traffic and monitoring traffic. Of the total CSMA/CA in comparison to our proposed MAC. As the mod-
IoT devices, 20% generate emergency traffic; 80% of the total ified CSMA/CA and TDMA with UAV mechanisms do not
traffic cover the normal monitoring traffic. We assume the maintain the QoS requirements during data transmission, the
real-time data collection scenario from IoT devices. Hence, prioritization of emergency data is not performed in these two
no data aggregation is performed in any device. During the protocols. Furthermore, the channel access delay is the key
RP, the CSMA/CA mechanism is followed. Based on the contributor to the transmission delay. If the channel becomes
traffic types, the CW size is dynamically selected. The emer- excessively busy, to complete the channel access, the devices
gency traffic utilizes the small CW size to obtain access with have to back off for more times, creating longer channel
lower delay; the minimum value is 15 and the maximum access delays.
value is 30. As the normal monitoring data are delay-tolerant, In PF-MAC, to reduce channel access delay, different CW
a large CW size is required, ranging from 31 to 1023. For both sizes are considered based on the IoT-device static traffic

57102 VOLUME 9, 2021

S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

FIGURE 8. Transmission delay for emergency traffic. FIGURE 10. Average throughput.

to the modified CSMA/CA and conventional TDMA mech-

anisms. It is because during the CSMA/CA period, it only
exchanges very small reservation packets, thus reducing
the number of collisions. Moreover, our proposed PF-MAC
protocol uses a multichannel directional antenna, reducing
interference and fading by directing the signal in only one
direction. The ICP model helps in delivery of packets after
experiencing collisions, which results in better throughput.
The communication time between IoT devices and the UAV
is hampered by UAV’s mobility.
The proper prioritization process helps the IoT devices
to transfer the data to the UAV during the TDMA period
FIGURE 9. Average transmission delay.
within a short communication time. On the contrary, modified
CSMA/CA uses only CSMA/CA mechanism, so collisions
priority. The devices that have high static traffic priority increase due to the transmission of large data packets. The
utilize lower CW size to obtain channel access earlier than TDMA with UAV suffers from low throughput owing to
the devices, which has lower static traffic priority. Moreover, the absence of the prioritization process. The TDMA with
ABO helps select proper backoff time based on the colli- UAV mechanism does not allocate a timeslot to the IoT
sion probability of the channel, thereby reducing the delay devices according to the traffic urgency and need. There-
in transmission of emergency data to a considerable extent. fore, most of the time, packet loss occurs due to the link
Moreover, the proper prioritization process by UAV helps to disconnection with the UAV. Therefore, PF-MAC gives bet-
transmit the emergency data transmit faster by getting the ter performance than the other two protocols in terms of
earlier timeslot. throughput and achieves an overall 32% increase in average
In Fig. 9, the average transmission delay of the PF-MAC throughput.
is presented. In the proposed PF-MAC, a fixed committed
channel is allotted for the control packets and a dedicated 3) NORMALIZED CONTROL OVERHEAD
channel for data packets. Control channel experiences less As depicted in Fig. 11, when the number of IoT devices
interference and interruptions due to the static use of a sin- is not more than 20, the normalized control overhead of
gle channel. Moreover, the utilization of ABO based on the our proposed PF-MAC is higher than TDMA and similar
collision rate of the channel and priority-based channel access to modified CSMA/CA. However, when the number of IoT
mechanism helps it to achieve less delay during transmission. devices increases, PF-MAC outperforms the TDMA but still
As the number of IoT devices rises, the average transmission has higher control overhead than modified CSMA/CA. Our
delay also gets high. This is because of the fact that as proposed PF-MAC protocol utilizes a uses a reservation
the number of IoT devices increases, more collisions occur frame, ACK and beacons as control packet. On the con-
during RP, causing more delay. However, PF-MAC provides trary, CSMA/CA uses the RTS/CTS/ACK mechanism for
better performance than the other two protocols in terms of establishing a connection between the UAV and the IoT
delay. devices. In the case of the TDMA with UAV mechanism,
it suffers from high synchronization overhead resulting in
2) THROUGHPUT higher control overheads. It is because, the IoT devices need
Fig. 10 displays the network throughput of the proposed pro- to communicate with each other with synchronization packets
tocol. The PF-MAC provides better performance compared in order to avoid collisions in the same time slot. However,

VOLUME 9, 2021 57103

 S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

FIGURE 13. Network lifetime.

FIGURE 11. Normalized control overhead.
5) NETWORK LIFETIME
The network lifetime of the proposed PF-MAC is calculated
in terms of the number of dead nodes after running the
simulation for 120 rounds. Fig. 13 demonstrates the network
lifetime of the PF-MAC. It is clearly observed from the
figure that, after 70 rounds, the IoT devices start to die.
However, until round 90, only less than 10% IoT devices run
out of energy, which is quite low. On the other hand, both the
modified CSMA/CA and the conventional TDMA with UAV
have better network lifetime than PF-MAC. It is because our
proposed PF-MAC emphasizes on more on guaranteed data
delivery to increase throughput rather than the energy aspect.
FIGURE 12. Average energy consumption per IoT device. However, the number of dead nodes until round 90 is almost
similar in all of the three protocols. Also, it should be noticed
that, even though the energy consumption of the IoT devices
though the PF-MAC has higher control overhead than mod- is higher than that of the other two protocols, the network
ified CSMA/CA, it can ensure higher throughput with less lifetime is not much unsatisfactory compared to other two
packet loss and delay. existing protocols.

4) ENERGY CONSUMPTION VII. CONCLUSION

Fig. 12 shows energy consumption comparison among the In this paper, we have proposed a hybrid MAC protocol
PF-MAC, modified CSMA/CA, and conventional TDMA named PF-MAC for UAV-based IIoT networks to achieve
mechanisms. The comparison shows that our proposed the QoS requirements of the target system. In our proto-
PF-MAC has higher energy consumption than modified col, the operation of the whole communication process is
CSMA/CA and TDMA. It is because our proposed PF-MAC divided into four parts: NP, RP, AP, and DCP. Heteroge-
more focused on guaranteed data delivery to achieve high neous devices with two types of static traffic priority con-
throughput rather than energy efficiency. The transmission tend the channel during the RP. During the DCP, time slots
of a reservation frame before transmitting the data frame for data transmission will only be allocated to the success-
ensures the guaranteed packet delivery. However, our pro- ful devices in contention. To maintain fairness among the
posed PF-MAC only permits the devices to transmit a small devices, the static traffic priority of the device failing in
size reservation frame during the RP and utilizes ABO tech- contention at the former transmission will be increased by
niques. Thus, the energy consumption due to collisions is 1 at the next retransmission. In the RP, the ABO mech-
reduced significantly. Moreover, the IoT device that fails anism is implemented based on the collision rate of the
in the reservation period goes into the sleep mode to pre- channel. Moreover, during the DCP, a transmission oppor-
serve energy. On the contrary, modified CSMA/CA and tunity is provided based on the dynamic device priority.
TDMA with UAV requires only one transmission for trans- We evaluated the average transmission delay, throughput,
mitting data. Moreover, TDMA has the lowest energy con- normalized control overhead, energy consumption and net-
sumption among the three protocols because no energy work lifetime to show the performance of our proposed
is wasted due to collisions. However, the proposed pro- PF-MAC protocol in comparison to the existing proto-
tocol’s modest energy consumption contributes to higher cols. The performance study makes it apparent that the IoT
throughput. devices can transfer emergency traffic to the UAV with less

57104 VOLUME 9, 2021

S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

delay and the transmission of normal monitoring traffics [17] M. G. Barron, ‘‘Ecological impacts of the deepwater horizon oil spill:
achieves higher throughput. Implications for immunotoxicity,’’ Toxicologic Pathol., vol. 40, no. 2,
pp. 315–320, Feb. 2012.
In our future study, we are going to exploit an artificial- [18] Q. Wu, Y. Zeng, and R. Zhang, ‘‘Joint trajectory and communication
intelligence-enabled MAC protocol with an optimized UAV design for multi-UAV enabled wireless networks,’’ IEEE Trans. Wireless
trajectory, which can help to reduce the energy consumption Commun., vol. 17, no. 3, pp. 2109–2121, Mar. 2018.
[19] T.-H. Hsu and P.-Y. Yen, ‘‘Adaptive time division multiple access-based
of the whole system. We also plan to incorporate multi-UAV medium access control protocol for energy conserving and data trans-
scenarios and emphasize on the increasing the lifetime of mission in wireless sensor networks,’’ IET Commun., vol. 5, no. 18,
UAV-assisted IoT systems. pp. 2662–2672, Dec. 2011.
[20] J. Wang, C. Jiang, Z. Han, Y. Ren, R. G. Maunder, and L. Hanzo, ‘‘Tak-
ing drones to the next level: Cooperative distributed unmanned-aerial-
ACKNOWLEDGMENT vehicular networks for small and mini drones,’’ IEEE Veh. Technol. Mag.,
The authors thank the editor and the anonymous referees for vol. 12, no. 3, pp. 73–82, Sep. 2017.
[21] X. Li, H. Yao, J. Wang, X. Xu, C. Jiang, and L. Hanzo, ‘‘A near-optimal
their comments that have helped improve the quality of this UAV-aided radio coverage strategy for dense urban areas,’’ IEEE Trans.
paper. Veh. Technol., vol. 68, no. 9, pp. 9098–9109, Sep. 2019.
[22] M. A. Abd-Elmagid and H. S. Dhillon, ‘‘Average peak age-of-information
minimization in UAV-assisted IoT networks,’’ IEEE Trans. Veh. Technol.,
REFERENCES vol. 68, no. 2, pp. 2003–2008, Feb. 2019.
[1] G. Cai, J. Dias, and L. Seneviratne, ‘‘A survey of small-scale unmanned [23] Q. Yang and S.-J. Yoo, ‘‘Optimal UAV path planning: Sensing data
aerial vehicles: Recent advances and future development trends,’’ acquisition over IoT sensor networks using multi-objective bio-inspired
Unmanned Syst., vol. 2, no. 2, pp. 175–199, Apr. 2014. algorithms,’’ IEEE Access, vol. 6, pp. 13671–13684, 2018.
[2] S. Yahyanejad, D. Wischounig-Strucl, M. Quaritsch, and B. Rinner, [24] M. Samir, S. Sharafeddine, C. M. Assi, T. M. Nguyen, and A. Ghrayeb,
‘‘Incremental mosaicking of images from autonomous, small-scale ‘‘UAV trajectory planning for data collection from time-constrained IoT
UAVs,’’ in Proc. 7th IEEE Int. Conf. Adv. Video Signal Based Surveill., devices,’’ IEEE Trans. Wireless Commun., vol. 19, no. 1, pp. 34–46,
Boston, MA, USA, Aug. 2010, pp. 329–336. Jan. 2020.
[3] S. Waharte and N. Trigoni, ‘‘Supporting search and rescue operations [25] F. Qi, X. Zhu, G. Mang, M. Kadoch, and W. Li, ‘‘UAV network and IoT
with UAVs,’’ in Proc. Int. Conf. Emerg. Secur. Technol., Canterbury, U.K., in the sky for future smart cities,’’ IEEE Netw., vol. 33, no. 2, pp. 96–101,
Sep. 2010, pp. 142–147. Mar. 2019.
[4] L. Merino, F. Caballero, J. R. Martínez-de-Dios, I. Maza, and A. Ollero, [26] Z. Hu, Z. Bai, Y. Yang, Z. Zheng, K. Bian, and L. Song, ‘‘UAV aided
‘‘An unmanned aircraft system for automatic forest fire monitoring and aerial-ground IoT for air quality sensing in smart city: Architecture, tech-
measurement,’’ J. Intell. Robotic Syst., vol. 65, nos. 1–4, pp. 533–548, nologies, and implementation,’’ IEEE Netw., vol. 33, no. 2, pp. 14–22,
Jan. 2012. Mar. 2019.
[5] P. G. Martin, S. Kwong, N. T. Smith, Y. Yamashiki, O. D. Payton, [27] J. Liu, Y. Shi, Z. M. Fadlullah, and N. Kato, ‘‘Space-air-ground inte-
F. S. Russell-Pavier, J. S. Fardoulis, D. A. Richards, and T. B. Scott, grated network: A survey,’’ IEEE Commun. Surveys Tuts., vol. 20, no. 4,
‘‘3D unmanned aerial vehicle radiation mapping for assessing contaminant pp. 2714–2741, May 2018.
distribution and mobility,’’ Int. J. Appl. Earth Observ. Geoinf., vol. 52, [28] C. Zhou, W. Wu, H. He, P. Yang, F. Lyu, N. Cheng, and X. Shen,
pp. 12–19, Oct. 2016. ‘‘Delay-aware IoT task scheduling in space-air-ground integrated net-
work,’’ in Proc. IEEE Global Commun. Conf. (GLOBECOM), Dec. 2019,
[6] E. Natalizio, R. Surace, V. Loscri, F. Guerriero, and T. Melodia, ‘‘Filming
pp. 1–6.
sport events with mobile camera drones: Mathematical modeling and
algorithms,’’ HAL-INRIA, Lyon, France, Res. Rep. hal-00801126, 2012. [29] J. Wang, C. Jiang, Z. Wei, C. Pan, H. Zhang, and Y. Ren, ‘‘Joint UAV
hovering altitude and power control for space-air-ground IoT networks,’’
[7] S. Katikala, ‘‘Google project loon,’’ InSight, Rivier Academic J., vol. 10,
IEEE Internet Things J., vol. 6, no. 2, pp. 1741–1753, Apr. 2019.
no. 2, pp. 1–6, 2014.
[30] R. Ali, N. Shahin, Y. B. Zikria, B.-S. Kim, and S. W. Kim, ‘‘Deep
[8] M. L. Gomez and A. Cox, ‘‘Flying Aquila: Early lessons from the first full-
reinforcement learning paradigm for performance optimization of channel
scale test flight and the path ahead,’’ Facebook Code, Tech. Rep., Jul. 2016.
observation–based MAC protocols in dense WLANs,’’ IEEE Access, vol. 7,
[9] W. Guo, C. Devine, and S. Wang, ‘‘Performance analysis of micro pp. 3500–3511, 2019.
unmanned airborne communication relays for cellular networks,’’ in Proc. [31] R. Ali, M. Sohail, A. O. Almagrabi, A. Musaddiq, and B.-S. Kim, ‘‘Green-
9th Int. Symp. Commun. Syst., Netw. Digit. Sign (CSNDSP), Manchester, MAC protocol: A Q-learning-based mechanism to enhance channel reli-
U.K., Jul. 2014, pp. 658–663. ability for WLAN energy savings,’’ Electronics, vol. 9, no. 10, p. 1720,
[10] T. Akiyoshi, E. Okamoto, H. Tsuji, and A. Miura, ‘‘Performance improve- Oct. 2020.
ment of satellite/terrestrial integrated mobile communication system using [32] S. Khisa and S. Moh, ‘‘Medium access control protocols for the Internet
unmanned aerial vehicle cooperative communications,’’ in Proc. Int. Conf. of Things based on unmanned aerial vehicles: A comparative survey,’’
Inf. Netw. (ICOIN), Da Nang, Vietnam, Jan. 2017, pp. 417–422. Sensors, vol. 20, no. 19, p. 5586, Sep. 2020.
[11] V. P. Hubenko, R. A. Raines, R. F. Mills, R. O. Baldwin, B. E. Mullins, [33] X. Lin, G. Su, B. Chen, H. Wang, and M. Dai, ‘‘Striking a balance between
and M. R. Grimaila, ‘‘Improving the global information grid’s performance system throughput and energy efficiency for UAV-IoT systems,’’ IEEE
through satellite communications layer enhancements,’’ IEEE Commun. Internet Things J., vol. 6, no. 6, pp. 10519–10533, Dec. 2019.
Mag., vol. 44, no. 11, pp. 66–72, Nov. 2006. [34] B. Li, X. Guo, R. Zhang, X. Du, and M. Guizani, ‘‘Performance analysis
[12] J. Radtke, C. Kebschull, and E. Stoll, ‘‘Interactions of the space debris and optimization for the MAC protocol in UAV-based IoT network,’’ IEEE
environment with mega constellations—Using the example of the OneWeb Trans. Veh. Technol., vol. 69, no. 8, pp. 8925–8937, Aug. 2020.
constellation,’’ Acta Astronautica, vol. 131, pp. 55–68, Feb. 2017. [35] Y. Du, K. Yang, K. Wang, G. Zhang, Y. Zhao, and D. Chen, ‘‘Joint
[13] T. Pultarova, ‘‘Telecommunications-space tycoons go head to head resources and workflow scheduling in UAV-enabled wirelessly-powered
over mega satellite network,’’ Eng. Technol., vol. 10, no. 2, p. 20, MEC for IoT systems,’’ IEEE Trans. Veh. Technol., vol. 68, no. 10,
Mar. 2015. pp. 10187–10200, Oct. 2019.
[14] G. Xiong, F. Zhu, X. Dong, H. Fan, B. Hu, Q. Kong, W. Kang, and T. Teng, [36] W. Khawaja, I. Guvenc, D. W. Matolak, U.-C. Fiebig, and
‘‘A kind of novel ITS based on space-air-ground big-data,’’ IEEE Intell. N. Schneckenburger, ‘‘A survey of air-to-ground propagation channel
Transp. Syst. Mag., vol. 8, no. 1, pp. 10–22, Jan. 2016. modeling for unmanned aerial vehicles,’’ IEEE Commun. Surveys Tuts.,
[15] M. Casoni, C. A. Grazia, M. Klapez, N. Patriciello, A. Amditis, and vol. 21, no. 3, pp. 2361–2391, 3rd Quart., 2019.
E. Sdongos, ‘‘Integration of satellite and LTE for disaster recovery,’’ IEEE [37] Y. S. Meng and Y. H. Lee, ‘‘Measurements and characterizations of
Commun. Mag., vol. 53, no. 3, pp. 47–53, Mar. 2015. air-to-ground channel over sea surface at C-band with low airborne
[16] P. Daley and D. O’Neill, ‘‘Sad is too mild a word,’ press coverage altitudes,’’ IEEE Trans. Veh. Technol., vol. 60, no. 4, pp. 1943–1948,
Exxon Valdez oil spill,’’ J. Commun., vol. 41, no. 4, pp. 42–57, 1991. May 2011.

VOLUME 9, 2021 57105

 S. Khisa, S. Moh: PF-MAC Protocol for UAV-Assisted IIoT Systems

[38] V. E. Hovstein, A. Saegrov, and T. A. Johansen, ‘‘Experiences with coastal SANGMAN MOH (Member, IEEE) received the
and maritime UAS BLOS operation with phased-array antenna digital M.S. degree in computer science from Yonsei Uni-
payload data link,’’ in Proc. Int. Conf. Unmanned Aircr. Syst. (ICUAS), versity, South Korea, in 1991, and the Ph.D. degree
Orlando, FL, USA, May 2014, pp. 261–266. in computer engineering from the Korea Advanced
[39] G. Wu, C. Dong, A. Li, L. Zhang, and Q. Wu, ‘‘FM-MAC: A multi-channel Institute of Science and Technology (KAIST),
MAC protocol for FANETs with directional antenna,’’ in Proc. IEEE South Korea, in 2002. Since 2002, he has been a
Global Commun. Conf. (GLOBECOM), Abu Dhabi, UAE, Dec. 2018, Professor with the Department of Computer Engi-
pp. 1–7.
neering, Chosun University, South Korea. From
[40] S. Temel and I. Bekmezci, ‘‘LODMAC: Location oriented directional
2006 to 2007, he was on leave from Cleveland
MAC protocol for FANETs,’’ Comput. Netw., vol. 83, pp. 76–84,
Jun. 2015. State University, USA. Until 2002, he was with
[41] H.-S.-W. So, G. Nguyen, and J. Walrand, ‘‘Practical synchronization the Electronics and Telecommunications Research Institute (ETRI), South
techniques for multi-channel MAC,’’ in Proc. 12th Annu. Int. Conf. Korea, where he served as a Project Leader. His research interests include
Mobile Comput. Netw. (MobiCom), Los Angeles, CA, USA, Sep. 2006, mobile computing and networking, ad hoc and sensor networks, cognitive
pp. 134–145. radio networks, and parallel and distributed computing systems. He is a
member of the ACM, the IEICE, the KIISE, the IEIE, the KIPS, the KICS,
the KMMS, the IEMEK, the KISM, and the KPEA.

SHREYA KHISA received the B.S. degree in com-

puter science and engineering from the University
of Chittagong, Bangladesh, in 2017. She is cur-
rently pursuing the M.S. degree with the Mobile
Computing Laboratory, Chosun University, South
Korea. Her current research interests include wire-
less sensor networks, the Internet of Things, and
unmanned aerial vehicle networks with a focus on
network architectures and protocols.

57106 VOLUME 9, 2021