IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO.

2, APRIL 2019 2273

An Energy Efficient Hybrid MAC Protocol With

Dynamic Sleep-Based Scheduling for High
Density IoT Networks
Thair A. Al-Janabi , Member, IEEE, and Hamed S. Al-Raweshidy , Senior Member, IEEE

Abstract—Advances in the development of low powered sensors jobs by enabling them to communicate to share information
has meant they can now provide solutions to Internet of Things and coordinate decisions [2]. Things, in the IoT, refer to
(IoT) networks that suffer from restricted power supply and a a multitude of heterogeneous devices with attached sensors
lack of resource facilities. In this paper, a hybrid time division
multiple access (TDMA)-carrier sense multiple access with a col- working across diverse applications. Moreover, wireless sensor
lision avoidance mechanism (CSMA/CA) medium access control networks (WSNs) are considered as an essential component
(MAC) protocol has been proposed that efficiently utilizes the of these networks [3]. Power consumption, however, is the
energy of the nodes and dynamically adapts the sleep/wake-up primary restriction regarding IoT devices. Consequently, the
periods according to the variance in the network loads. This IEEE 802.15.4 standard [4] for low power medium access con-
hybrid protocol first schedules the TDMA time slots (TDMAslots ),
and then allocates each slot to a group of devices that compete trol (MAC) and physical network specifications have been used
for the medium using the CSMA/CA. This case is different from with them to address insufficiency in the power supply [5].
the traditional CSMA/CA-TDMA hybrid protocol, in which all In the IEEE 802.15.4 standard [4], a central controller
the devices compete to access the channel, following which, each known as the personal area network (PAN) coordinator over-
successful device is allocated an individual time slot. At the com- sees the construction, upgrading, and monitoring across the
mencement of each superframe, the base station broadcasts a
scheduler table, which includes network grouping information entire network. Furthermore, the standard assigns two mech-
that is used by the IoT devices to categorize themselves into anisms: 1) beacon-enabled and 2) beaconless, to access the
wake-up and sleep groups. To reduce the number of collisions or channel. The beacon-enabled mechanism uses a slotted carrier
channel access failures, this information permits only one group sense multiple access with a collision avoidance mechanism
to compete for each TDMAslot . A 3-D Markov model is used to (CSMA/CA) to share the medium. In this mechanism, the slot
develop a per user stochastic behavior for the proposed hybrid
MAC protocol-based adaptable sleep mode. The simulation margins of each device are aligned with those margins of the
results demonstrate the effectiveness of the proposed protocol, PAN coordinator using beacons and conversely, in the beacon-
which improves the network throughput and enhances energy less mechanism, a simpler unslotted CSMA/CA is applied. The
conservation by 40%–60% more than the IEEE 802.15.4-based beacons contain information used to describe the construction
MAC protocol. of superframes and synchronize the devices in the transmis-
Index Terms—Hybrid medium access control (MAC) protocol, sion range to recognize the PAN coordinator. The superframe
IEEE 802.15.4, Internet of Things (IoT), personal area network comprises active and inactive portions, with former consisting
(PAN) coordinator, scheduler table (ST), sleep mode. of the contention access period (CP) and the contention free
period (CFP). Throughout the inactive portion, the device does
I. I NTRODUCTION not interact with its coordinator and may move to a low power
mode to conserve energy [4].
HE INCREASED globalization of information and com-
T munication technologies has transformed the world into
the Internet of Things (IoT). The IoT refers to the network
In general, the energy consumption of the network is highly
influenced by the selection of the MAC protocol [6], which is
divided into three main categories: 1) contention free proto-
connectivity of physical or nonhuman things to the Internet col; 2) contention-based protocol; and 3) hybrid protocol [7].
which, along with a level of artificial intelligence (AI), Regarding the contention free protocol, in IoT networks it
enables these things to collect and exchange data [1]. The usually uses the time division multiple access (TDMA) mech-
IoT enables physical objects to see, hear, think, and perform anism to control the channel [8]. TDMA is a channel access
technique that authorizes a set of devices to access a certain
Manuscript received June 5, 2018; revised November 14, 2018; accepted
March 12, 2019. Date of publication March 19, 2019; date of current version carrier frequency without intervention. Furthermore, it is use-
May 8, 2019. (Corresponding author: Thair A. Al-Janabi.) ful for extending the battery life of the devices and reducing
T. A. Al-Janabi is with the Department of Information and Communication the number of collisions. However, TDMA has some draw-
Technology, Middle Technical University, Baghdad, Iraq (e-mail:
[email protected]). backs, for example, it is difficult to adapt to variant network
H. S. Al-Raweshidy is with the Wireless Networks and Communications loads, i.e., low or high, leading to either inefficient channel
Centre, Department of Electronic and Computer Engineering, College of utilization or limited network scalability. Another drawback is
Engineering, Design, and Physical Sciences, Brunel University London,
London UB8 3PH, U.K. (e-mail: [email protected]). that the transmission is delayed when the beacon interval is
Digital Object Identifier 10.1109/JIOT.2019.2905952 increased due to the fixed size of the beacon intervals [9].
2327-4662 c 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 2274 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

CSMA/CA is a contention-based mechanism used to share 4) It evaluates and verifies the performance of the proposed
the medium randomly between the IoT devices [10]. This model for small and high-density networks by compar-
random technique is applied to low power devices, such as ing it with the traditional IEEE 802.15.4 standard.
those associated with IEEE 802.11 [11] or IEEE 802.15.4 [12]. The remainder of this paper is structured as follows.
However, contention-based protocols suffer from high conges- Section II briefly discusses previous research related to the
tion associated with either a high density of IoT devices or a IoT networks, whilst Section III describes the formulation of
hidden node problem [9]. This high congestion leads to an the problem behind the development of the proposed HSW-
increase in the number of retransmissions, which affects the 802.15.4. The frame structure and the methodology for the
overall network performance in terms of channel utilization proposed protocol are provided in Section IV. Section V briefly
and power consumption. A hybrid CSMA/CA-TDMA protocol discusses the proposed IEEE 802.15.4-based dynamic sleep
is, therefore, useful for avoiding the weaknesses of the above mode and presents an analytical model for a 3-D Markov
protocols and combining their strengths to enhance channel chain compatible with the proposed model for calculation
utilization, whilst also reducing power consumption in wire- of the network metrics. Section VI validates the proposed
less networks [13]. However, the performance of the hybrid protocol by comparing the results with the existing IEEE
protocol still seems to suffers from enormous energy depletion 802.15.4 standard. The numerical results are presented and dis-
in high-density networks. Thus, adaptable scheduling, involv- cussed in Section VII, whilst final conclusions are presented
ing dynamic switching between sleep and wake-up periods, in Section VIII.
is required to accomplish superior channel access in such
networks.
In this paper, an on-demand sleep/wakeup extension is II. R ELATED R ESEARCH
proposed to the IEEE 802.15.4 standard to deal with the This literature divides the MAC studies into two categories:
above-mentioned IoT issues. The sleeping periods (LSs) are 1) the analytical-based studies and 2) the simulation-based
dynamically adapted according to the network density, col- ones. Furthermore, it splits the former based on a Markov
lision status and channel failure, thus leading to efficient mode into two classes: those that mimic and evaluate the per-
throughput and power savings as well as high channel uti- formances and properties of the IEEE 802.15.4 standard, and
lization. The base station (BS) arranges and schedules the those that attempt to improve the performance of this standard.
TDMAslots and the scheduling table (ST), whereas each indi-
vidual device, to save energy, has to calculate its LSi and
switch to a sleep/wakeup period. The information in the ST A. Analytical-Based MAC Protocols
message enables a certain group of devices to compete for Many studies have been aimed at assessing and under-
entering the transmission period. In contrast, other groups are standing the analytical MAC model of the IEEE 802.15.4
forced to enter the sleep mode each with a different calcu- standard. These studies have attempted to mimic aspects and
lated LSi . However, when the network load is low or there mechanisms widely used by the IoT and WSN devices to
is no collision, then all the devices can continue accessing investigate and identify the properties and performances of
the channel at any TDMAslot after a competition to access those networks. Most of the analytical models for the IEEE
the channel using CSMA/CA. Moreover, the devices move 802.15.4 standard are built according to the Markov model,
into sleep mode only when the medium experiences a high presented by Bianchi [11] for the IEEE 802.11 standard. That
collision rate or channel failure. This has the advantage of is, Bianchi designed a Markov chain model that mimics the
avoiding a costly channel access pattern by reducing the num- functionalities of the IEEE 802.11 standard for ideal channel
ber of collisions occurring due to a high density of devices. conditions and saturated networks. In this section, the ana-
Furthermore, the proposed hybrid protocol-based adaptable lytical studies based on a Markov mode is split into two
sleep model can be theoretically displayed using a 3-D Markov categories: those that mimic and evaluate the performances
chain and then compared to the original IEEE 802.15.4. and properties of the IEEE 802.15.4 standard, and those that
Hence, this paper makes the following contributions to current attempt to improve the performances of the IEEE 802.15.4
knowledge. standard.
1) It proposes a dynamic hybrid MAC (HyMAC) protocol 1) Performance Evaluation Studies: From a performance
for high density IoT communications with a scheduler evaluation perspective, a comprehensive Markov chain model
sleep/wakeup mode called HSW-802.15.4. This mode and analysis for the IEEE 802.15.4 [4] MAC and PHY layers
has been designed to be activated only after the device has been conducted by many authors [9], [12], [14]–[20]. For
experiences a high collision rate and/or channel failure example, Pollin et al. [12] proposed a 2-D Markov model,
situation. with an extension of saturated and unsaturated mechanisms
2) It develops a frame structure scheme for the for the slotted CSMA/CA mechanism that mimics the MAC
proposed protocol, which offers enhanced exploitation behavior of the IEEE 802.15.4. The performance of the IEEE
of TDMAslot and compatibility with the IEEE 802.15.4 802.15.4 standard in terms of power depletion and throughput
standard. was also analyzed. In contrast, Sahoo and Sheu [15] proposed
3) It provides a comprehensive analytical evaluation using a modified model including retransmission with determinate
a 3-D Markov model that predicts the performance of the retry limits for the IEEE 802.15.4. However, it appears that
proposed hybrid protocol-based dynamic sleep mode. there was an inexact match between the analytical and the

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 AL-JANABI AND AL-RAWESHIDY: ENERGY EFFICIENT HyMAC PROTOCOL WITH DYNAMIC SLEEP-BASED SCHEDULING 2275

IEEE standard results for both energy depletion and through- the collisions caused by the hidden node problem in
put. Park et al. [16] proposed a generalized analytical model of IEEE 802.15.4 networks using the GTS slots. Recently,
IEEE 802.15.4 with retry limits and acknowledgments under Al-Janabi and Al-Raweshidy [30] introduced a solution that
unsaturated star networks. This was an accurate Markov model expands network scalability and prolongs the lifetime of the
that could evaluate the performance metrics in terms of delay, IoT network protocol by developing a scheduling sleep/wake-
reliability and energy depletion. up model based on AI. In addition, Ayadi et al. [31] provided
On the other hand, Zayani et al. [17] developed an IEEE two methods that aim to better manage the energy exhausted
802.15.4 model that takes into consideration the functionalities by the WSNs. The first method concerned efficiently iden-
of this standard on both the PHY and MAC layers. On the tifying the energy depleted by the device in its various
PHY layer, the model is similar to the mathematical framework transmission stages, whilst the other method provided an
used by Zuniga and Krishnamachari [18] for calculating link approach that based on the residual energy of the devices to
unreliability. Moreover, the MAC layer is inspired from an identify their most compatible duty cycles. Although, both of
improved Markov model proposed by Park et al. [16]. This the above protocols provided energy efficient solution for the
model considers the effect of underlying packet discards at IoT network, they did not consider the high density of the IoT
either the PHY or MAC layers [17]. Performance estimation network.
in all the above studies has been aimed at providing optimal
Markov models for the CSMA/CA mechanism that mimic the B. Simulation-Based MAC Protocols
behavior of IEEE 802.15.4 standard; however, they do not It is generally agreed that MAC protocols are the sim-
seem to have involved taking the sleep/wakeup mode or any plest techniques in terms of system set-up and implementation,
other improvement into consideration. some of which are described below.
2) Performance Improvement Studies: To improve energy Amongst the most basic MAC protocols are the ran-
preservation and the performance of the IEEE 802.15.4 stan- dom access protocols, such as ALOHA and slotted-ALOHA,
dard, several researchers [21]–[27] have proposed models with stated by [32] and [33], respectively. In these protocols, the
periodic sleep or idle periods, and device contention, as per- devices either send the packet as soon as it is generated or
fect solutions. Mišic and Mišic [21], for example, suggested start the transmission at the commencement of the follow-
an M/G/1/K model of a beacon-enabled IEEE 802.15.4 with a ing slot. The key downside of these random-based protocols
sleep mechanism. With the proposed mechanism, it is assumed is the high rate of collisions, which restricts the value of
that a device turns to sleep for a specific amount of time the throughput and eliminates the utilization of the channel
when there are no packets in its buffer [21]. Alternatively, bandwidth [33]. CSMA protocols are considered an advance-
Chiasserini and Garetto [22] considered the performance of ment toward decreasing the collisions suffered by ALOHA
sleeping devices in WSNs; however, they did not focus on the methods [34]. However, these protocols do not significantly
performance evaluation of the IEEE 802.15.4 MAC layer. diminish the collisions and there can be degradation in the
From the performance evaluation aspect, throughput caused by the hidden and collided node problems.
Ghazvini et al. [23] suggested an energy efficient MAC HyMAC protocols that associate CSMA with TDMA, which
protocol, based on IEEE 802.15.4, for WSNs that reduces are described by Rhee et al. [35] and Zhang et al. [36], act as a
energy depletion and network collisions by permitting devices CSMA with a light contention state and change to the TDMA
to move between sleep and listening modes during the active MAC time sort procedure at a higher contention state. On the
period, when they have no packets to transmit. After that, other hand, Salajegheh et al. [37] proposed an HyMAC that
Xiao et al. [24] also proposed a Markov model that took into combines CSMA with both TDMA and FDMA. In a compa-
consideration the sleep-based mode of the beacon-enabled rable manner, other HyMAC protocols have been proposed
IEEE 802.15.4 standard. Jurdak et al. [25], in contrast, imple- in [38] and [39]. In those protocols, the devices combine
mented the nonbeacon IEEE 802.15.4-based RFID devices. CDMA with TDMA and utilize CSMA to send their request
To conserve energy, a device remains in sleep mode, if it has for a bandwidth message to reserve time slots and codes
no packet to send or the RFID tag has not received any data. for their transmission successfully. Another HyMAC protocol,
Khanafer et al. [26] suggested a method that saves energy by called zebra MAC (ZMAC), is proposed by Rhee et al. [35].
forcing some devices to move into sleep mode after successful The hybrid technique of this protocol works as CSMA in a
communication. Furthermore, Xiao et al. [27] proposed an low network traffic and switches to TDMA when it is high.
energy efficient MAC protocol with a sleep enabling mode However, if the devices require more than one-time slot, they
based on the length of the device queue. However, all these can try to exploit other neighbors’ unutilized time slots. To
studies focused only on the contention-based channel access, exploit other time slots, the device initiates a long enough
and not on network density, the numbers of collisions or even random backoff period and starts its transmission at the end
channel failure. of this period, but the device is not able to start its transmis-
Consequently, the performance of the hybrid IEEE 802.15.4 sion if the primary user needs its own time slot. Moreover,
MAC protocol has also been studied by [28] and [29]. The ZMAC is required to implement TDMA scheduling in a peri-
research conducted in [28] offered a general Markov model odic manner to recover the deviation in the time clock among
that takes into consideration both the slotted CSMA/CA the devices.
and GTS in a heterogeneous network set-up and nonsat- A scalable hybrid CSMAMAC protocol is suggested in [1],
urated conditions. Sheu et al. [29] focused on reducing which works by splitting each frame period into four

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 2276 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

portions: 1) notification period (NP); 2) contention only period to be considered when undertaking protocol design in such
(COP); 3) announcement period (AP); and 4) transmission networks.
only period (TOP). Throughout the COP, the devices with data Regarding the protocol design problem, the MAC proto-
to send employ the p-persistent CSMA mechanism to transmit col responsibility should be in handling the channel access
the request to send (RTS) packet to the BS and those that suc- mechanisms, failure control, and validation of the superframes.
ceed in doing so, are allocated time slots in the TOP to transmit However, MAC protocols, such as CSMA/CA and TDMA usu-
their data. This HyMAC protocol is improved by Liu et al. [40] ally focus on some network parameters and ignore others.
by the implementation of the fairness and quality of service For example, TDMA takes into consideration energy sav-
(QoS) concepts. The improved protocol categorizes the devices ing and neglects throughput, delay and channel utilization.
into different classes and permits the devices to pick up the Furthermore, with TDMA, it is difficult to adapt the channel
contention probability rendering to their observed through- to the variations in the network density, whether this is low or
put and urgency. However, both the above protocols proposed high load, thus leading to either inefficient channel utilization
in [1] and [40] suffer from supplementary energy depletion or limited network scalability. On the other hand, the HyMAC
caused by the requirement for the contention between devices protocol is useful for addressing the weaknesses of the above
during the COP. protocols and also combines their strengths. However, it can-
Despite the above hybrid protocols trying to be more scal- not ensure efficient energy saving in a high-density network.
able, they are still not adequate for deployment in most IoT Hence, a dynamic MAC protocol that minimizes the energy
applications. Also, they still seem to suffer from a higher consumed by IoT devices during their connection process to
number of collisions and channel failure in saturated and high- the telecom is essential and needs to be designed. Furthermore,
density networks due to their failure to include a sleep/wake-up extended periods of sleeping devices and the use of energy
mechanism in their operation. Consequently, congestion in the efficient techniques can all help to extend the battery life of
IoT network increases as the network density rises. Moreover, these devices.
the hybrid protocols-based on FDMA and CDMA have the Accordingly, in this paper, an effective power saving struc-
issues of high hardware expenses and high complexity, respec- ture using a dynamic scheduling MAC technique is proposed.
tively. Hence, the HyMAC protocols based on both TDMA and This technique dictates the task to be achieved at a particu-
CSMA are the most encouraging from the perspective of IoT lar time and obliges groups of devices to sleep. Furthermore,
applications. Accordingly, the proposed HSW-802.15.4 pro- the scheduling sleeping technique is adapted according to
tocol provides an amenable hybrid TDMA-CSMA/CA MAC conditions, such as high collision or channel failure, to deter-
mechanism with a dynamic sleep/wake-up extension that effi- mine which group of devices will be active in a certain
ciently accesses the channel with both low and high-density TDMAslot . The proposed protocol will accomplish superior
networks. The proposed protocol tunes up the sleep/wake-up channel access enhancement in a high density network and
periods by forcing groups of devices to sleep for various peri- extend the lifespan of IoT devices.
ods once they experience a collision or channel failure. As
a result, the proposed model works by reducing the hidden
node problem, and thus reduces the probability of collisions, IV. P ROPOSED S CHEDULER H YBRID MAC P ROTOCOL
thereby leading to an improvement in the performance of the The novelty of this paper is represented by suggesting a
network in terms of power conservation, channel utilization, new sleep/wake-up extension to the mechanism of the IEEE
and throughput. 802.15.4 standard that forces groups of devices, when the
network experiences high collisions or channel failure, to sleep
mode with different sleep periods. Therefore, this extension
III. P ROBLEM F ORMULATION makes the proposed HSW-802.15.4 MAC work efficiently with
IoT devices generally suffer from two types of problems, both high and small density IoT networks. However, to pro-
namely: device resource restriction and protocol design. vide an analytical model for the proposed protocol, a three
Regarding device resource restriction, IoT devices are all dimensions Markov model has been presented that includes
severely constrained in terms of power supply, communica- the new sleep/wake-up mode in its operation. Further details
tion capability and memory units. In fact, power consumption about the Markov model are presented by [14] and [17].
is the primary restriction placed upon them. Moreover, the Power consumption is a vital issue for battery power sup-
vast majority are not rechargeable because they are usually ported devices in IoT and WSNs. To address this problem, an
designed in a small-sized system that constrains the capac- effective scheduling message, known as the ST (see Fig. 1),
ity of the battery and memory. Additionally, the periodic has been proposed to force groups of them into sleep mode.
battery replacement for those devices is infeasible due to The proposed scheduling method is extremely efficient in
the device deployment in a vital environment and the large terms of energy conservation, especially for battery-powered
number of IoT devices. Furthermore, the complex compu- devices. It selects the mission to be performed at a given
tations that necessitate unique processing capabilities and moment by determining which group of devices will be active
extended memory units require massive power resources and in each TDMAslot , and by allocating physical layer resources.
thus, dramatically reduce the overall lifespan of the IoT in this section, the structure of the proposed protocol in terms
network. The strict resource-constrained situations in which of network architecture, frame structure, and methodology is
IoT devices operate make energy conservation a primary factor discussed.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 AL-JANABI AND AL-RAWESHIDY: ENERGY EFFICIENT HyMAC PROTOCOL WITH DYNAMIC SLEEP-BASED SCHEDULING 2277

Fig. 1. Structure of the proposed ST.

A. Network Architecture Fig. 2. Proposed superframe structure that shows the CP and TP periods
while TDMAslots = 16.
A star wireless PAN with a single-hop is assumed to be
used in the simulation. It is composed of a fixed number (N)
of devices and a PAN coordinator. Furthermore, it is assumed network collisions and improve performance. It uses a super-
that, as a high-density network, each device always has a frame architecture, where network beacons are periodically
packet available for sending, which means it is working in broadcast by the PAN coordinator to bind each superframe.
a saturated area. The devices also exhibit location awareness The network beacons are used to describe the architecture of
pro-actively during the device installation process without any the superframes, and to synchronize as well as scheduling the
the device cooperation, and then can send their location along attached devices by using the ST message [see Fig. 3(a)].
with their identification address (ID) to the PAN coordinator The superframe of the proposed protocol is, first, divided
to be utilized in avoiding the hidden node problem. into 16 equal sized TDMAslots by the PAN coordinator, in a
In the proposed protocol, the PAN coordinator divides the similar way to the IEEE 802.15.4 standard. Each TDMAslot is
network devices into multiple subgroups (SGi ) according to allocated to a specific SGi . A device that wishes to access the
the ID and location of the device. At the commencement of channel can only compete with other devices in the same SGi
each superframe, the coordinator sends the ST message as a using a CSMA/CA mechanism [see Fig. 3(a)]. However, there
notification during the beacon period. As Fig. 1 shows, the is an exception to this rule, which is that the device can enter
ST message provides information about the number of sub- the CP with other subgroups when the channel has not suffered
groups (NG) and the IDs of the devices belonging to each SGi . from any previous failure or collision. This mechanism has the
Additionally, it contains information about the sleep/wake-up dual advantage of offering efficient utilization of the channel
times for each SG, which is used to conserve energy by speci- in low-density networks, and reducing the number of collisions
fying the maximum number of TDMAslots each device should by following the ST message with high-density networks.
have. These are known as KiTDMAslots . The present IEEE 802.15.4 protocol is based on a sim-
The KiTDMAslots for each SGi are chosen by the coordina- ple time unit called a UnitBackoffslots period (BP) and three
tor according to device priority, so that devices at a SGi with random-based variables: 1) backoff exponent (BE); 2) con-
higher priority are allocated fewer KiTDMAslots than other tention window (CW); and 3) number of backoff stages (NBF).
subgroups. The ST message is thus sent after the beacon (see The proposed protocol introduces other substantial variables,
Fig. 2) and used to schedule devices to enter a longer sleep which are the length of sleep period (LSi ), and the number
mode, rather than a longer CP. The length of the ST message is of SGi of the devices (NG). As shown in Fig. 3(b), after
constant and equivalent to multiple messages according to the receiving the Beacon and the ST message from the PAN coor-
network requirements and node density; however, each device dinator and agreeing with the proposed sleep/wake-up mode
needs to be aware of the size of the ST message. During the (see Section V), the device either immediately starts sending
CP, the devices in the specified SGi are eligible to compete its data or follows the sleep mode, and then starts sending
using the CSMA/CA method, to obtain access to the channel its data. Furthermore, each device has to calculate its sleep
and transmit data during TDMAslots , also known as transmis- period continuously by utilizing the information from the ST
sion periods (TPtransmission ). However, when the network does message.
not experience any channel collisions or failure, all the other
devices can compete and obtain access to the channel as a C. Methodology
normal IEEE 802.15.4, without an adaptable sleep mode.
The proposed protocol is designed according to the spec-
ifications of IoT networks and the transmission process for
B. Frame Structure the proposed HSW-802.15.4 protocol is shown in Fig. 3(c).
In this paper, an HyMAC protocol is proposed in which a In the proposed protocol, the MAC layer first initializes the
specified group of devices are permitted to enter the CP at a following variables: CW, backoff exponent (BE = MinBE),
specific TDMAslot , rather than all the devices. Fig. 2 shows macMaxCSMABackoff (M = 2), number of backoffs counter
the modified superframe structure for the proposed model, (NBF = 0), maximum frame retries (R = 3), retransmission
which is compatible with the traditional IEEE 802.15.4. As counter (r = 0), maximum number of CCAs (CMax = 2),
such, the proposed hybrid protocol can significantly reduce CCAs counter (i = 0), and number of SG = NG.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 2278 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

(a)

(b) (c)

Fig. 3. (a) Network star topology with SGi s, (b) overall communication scenario for a device using HSW-802.15.4, and (c) transmission process for the
proposed HSW-802.15.4.

The device then waits for a random backoff number (BF) in this protocol is to save power and prolong the overall lifes-
the range [0; 2BE − 1] multiplied by the time of the BP. When pan of the IoT network. The energy consumption of the IEEE
the backoff counter reaches zero, the device senses the chan- 802.15.4 standard is primarily influenced by the probability of
nel using clear channel assessment (CCA). If two successive channel access failure (CAF) or collisions, and MAC param-
CCAs are idle, the device initiates the transmission, whilst if eters. Hence, the proposed HSW-802.15.4 protocol employs
is busy, the MAC layer will increase the value of both NBF a strategy to force some groups of nodes to sleep for a cal-
and BE by one up to a maximum value of M and MaxBE, culated LSi period, whenever there are channel collisions or
respectively. If the channel access is successful, the device channel failure.
starts transferring its packet and waits for acknowledgment Using this method, the proposed protocol becomes more
(ACK). If the device receives the ACK, the transmission has adaptable and flexible as it dynamically adapts the sleep nodes,
succeeded and if not, a collision is detected and the devices working with both high and low-density networks. The behav-
involved should increase the variable r up to the value R. If ior of this new adaptable protocol can be easily observed
the channel experiences a channel failure or a collision, the in a high-density network by allowing more nodes to sleep
device calculates the length of the sleep period LSi and either with different periods. The main benefit of this lies in a sub-
continues trying to access the channel if it belongs to a per- stantial reduction in the probability of channel failure and
mitted SGi and LSi = 0 or turns to sleep mode, according to collisions. The following sections explain the sleep period cal-
the ST and the calculated LSi [see (1)]. After the LSi period, culation used by each device, and the 3-D Markov model
the device, depending on its NBF and r, decides either to dis- which is based on a dynamic sleep mode. Additionally, the
card its packet and generate a new one or to retransmit its influence of performance metrics and MAC parameters on
packet by reinitializing the NBF = 0, BE = minBE, r = 0, overall network operations are also analyzed. Table III in
and i = 0. The parameters used for the proposed protocol Appendix A summarizes the notation used in this paper.
are listed in Table I, which are the same as the default IEEE
802.15.4 MAC parameters.
A. Dynamic Sleep/Wake-Up Period Calculation
The grouping method that divides the nodes into SGs ,
V. A NALYTICAL M ODEL AND N OTATIONS whilst dynamically allowing the SGs to calculate and enter
This section describes the proposed HSW-802.15.4 MAC the Sleep/Wake-up periods, is described. In each specified
protocol-based schedule sleep extension. The main aim of TDMAslot , only one SG is permitted to compete for channel

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 AL-JANABI AND AL-RAWESHIDY: ENERGY EFFICIENT HyMAC PROTOCOL WITH DYNAMIC SLEEP-BASED SCHEDULING 2279

TABLE I TABLE II
PARAMETERS U SED BY THE HSW-802.15.4 P ERFORMANCE C OMPARISON B ETWEEN HSW-802.15.4 AND IEEE
802.15.4 IN S MALL AND H IGH D ENSITY N ETWORKS

TDMAslot . Slotseq is the sequence of the current TDMAslot

and lies between 0 and 15. NG represents the number of sub-
groups. Furthermore, the ratio of the active and sleep nodes is
adjusted by adding the calculated number of sleep TDMAslots
to the model based on LSi .
The role of LSi is to decide whether the device that has
experienced a collision or channel failure, turns to sleep, if
LSi > 0 or continues its attempt to access the channel with
the same TDMAslot . In this case, the energy required to access
the channel for the whole network is reduced, because some
of the devices move to sleep mode, each for a different period,
according to its calculated LSi . It is worth noting that the sleep
period identification process is similar to the random process
used to identify the BF, which depends on the previous stage
only and not on any others. In fact, the only difference between
BF and LSi is that the former is selected randomly, whilst the
value of the latter is calculated according to (1).
However, as presented in Table II, the device with the
proposed protocol, especially in a small density network, has
to consume an additional small amount of energy for the pur-
pose of receiving the ST message at the commencement of
each superframe. This message is essential in predicting the
Fig. 4. Subgroups separation among time slots. LSi period, while it is not applicable for the case of the IEEE
802.15.4 standard.

access (see Fig. 4). The aim of this technique is to distribute B. Markov Model
the active nodes evenly among the TDMAslots and reduce the
This section describes the analytical model for the proposed
power consumed by each node as it attempts to access the
adaptive HyMAC protocol-based synchronized sleep mode.
channel. This is because, depending on the node density in
For the purposes of the analysis, a saturated network with bea-
the network, a collision may be more likely to occur at one
con enabled and retry restrictions for each packet transmission
TDMAslot than at others. Hence, when the device in subgroup
is considered. The behavior of the proposed HSW-802.15.4-
(SGi ) experiences a channel failure or collision, it calculates
based schedule sleep is modeled using a Markov model that
the length of its sleep period (LSi ) as follows:
takes into consideration the probabilities of collision and fail-

Yi × NBP , if Yi ≤ Ki ure. Fig. 5 shows the Markov model-based sleep mechanism
LSi = (1)
Ki × NBP , if Yi > Ki of the HSW-802.15.4.
The analysis is split into two distinct steps, the aim of which
where, Ki refers to the number of sleeping TDMAslots allocated is to identify a set of equations that exclusively describes
to the SGi that is initially identified by the PAN coordinator the functionality of the network. During the first step, the
(see Fig. 1), while Yi is calculated as actions of a single device are modeled using a 3-D Markov
   chain to obtain the stationary probability () of the device
Yi = |SGi − Slotseq mod NG + 1 | (2)
attempting to access the channel for the first CCA within a
where, SGi is the subgroup i of devices to which the cur- given time slot. The per user Markov chains are then cou-
rent device belongs. NBP refers to the number of BPs in each pled to derive an extra set of equations to solve the system

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 2280 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

Fig. 5. Markov model for the proposed HSW-802.15.4.

model. A significant assumption made during this coupling is respectively. The transition probabilities for the proposed 3-D
that the probability of sensing the carrier channel is constant Markov model can therefore be derived as follows:
and independent of other devices. Three stochastic processes
P(i, j, r|i, j + 1, r) = 1, for j ≥ 0 (3)
describe the different states of the device during the transmis-
ω + ξ − ωξ
sion of a packet, namely: the backoff stage Gb (t), the state of P(i, j, r|i − 1, 0, r) = , for i ≤ M (4)
the backoff counter Gc (t), and the state of the retransmission CWi
Gr (t). 1 − ω − ξ + ωξ
P(0, j, 0|i, 0, r) = , for i ≤ M, r < R
The integer time (t) matches the beginning of the time slot CW0
and is declared as the delay stages [when Gb (t) ∈ (0, NBF)] (5)
or the transmission stage (Gr (t) = −1). Furthermore, the (1 − ω − ξ + ωξ )PSW
P(0, j, r|i, 0, r − 1) = , for r < R
stages (Gb (t) = −2, −3) introduce the collision stage and CW0
the sleep/wake-up stage, respectively. The Gc (t) stage repre- 
(6)
sents the delay counter of the device for a random duration 1 − ω − ξ + ωξ
P(0, j, 0|M, 0, r) =
between (0 and CWi−1). The values (Gc (t) = −1, −2) reflect CW0
to the first CCA and second CCA, respectively. The parameters (1 − ω − ξ + ωξ )PSW
M and R, in turn, are set to the values macMaxCSMABack- +
CW0
offs and macMaxFrameRetries. All other MAC parameters 
(ω + ξ − ωξ )PSW
are shown in Table I. The probability ω refers to the like- + , for r = R.
CW0
lihood of the channel assessment finding the channel busy
(7)
during CCA1, Whilst the probability ξ pertains to chan-
nel assessment of finding the channel busy during CCA2. Equation (3) represents the delay counter for a random num-
The stages (−1, k, j), (−2, k, j), and (−3, k, j) characterize the ber of slots. Equation (4) provides the probability that the
transmission state, the packet collision state, and the dynamic device picks a state in the next delay stage due to sensing
sleep state, respectively. The parameters Ls , and Lc denote the channel is busy during either the first or the second CCA,
the durations of packet transmission and packet collision, with probabilities ω and ξ , respectively. Equation (5) refers

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 AL-JANABI AND AL-RAWESHIDY: ENERGY EFFICIENT HyMAC PROTOCOL WITH DYNAMIC SLEEP-BASED SCHEDULING 2281

to the probability of successfully sending the packet with- where

out collision after sensing the channel is idle for both CCAs   R+1
1 − PSW 1 − (ω + ξ − ωξ )M+1
and (r < R). Equation (6) denotes the packet being sent for =   (14)
retransmission due to collision. In this case, the device should 1 − PSW 1 − (ω + ξ − ωξ )M+1
calculate the LSi before retransmitting it under the condition and  is equivalent to
of (r < R). Equation (7) defines the probability of beginning 1 − (ω + ξ − ωξ )M+1
a new transmission when the device is in the final delay stage. =× . (15)
1 − ω − ξ + ωξ
This is either due to: sending the packet successfully; send-
ing it with a collision after sensing the channel is idle for For simplicity, we assume in the analysis that each SGi has
both CCAs, whereby the device needs to calculate LSi before a fixed value of sleep period (LS). Then the last expression
starting a new transmission; or CAF during either CCA1 or of (11) is expressed as
CCA2, which means the device needs to calculate the LSi at NG−1 R
a specific TDMAslot . b−3,j,r
The steady state probability for the proposed Markov model j=0 r=0
is denoted by bi,j,r = Pr {(Gb (t), Gc (t), Gr (t)) = (i, j, r)}, LS × NG + LS
R
where i ∈ {−3, M}, and j ∈ {−1, max(CWi − 1, Lt − 1, = b(−3, 0, r)
2
Lc − 1, LSi − 1)}, and r ∈ {0, R} r=0
LS × NG + LS
CWi − j =
bi,j,r = bi,0,r (8)  2
CWi
× [Lc Pc + (1 − Lc Pc )(ω + ξ − ωξ )M+1 ] ×  b0,0,0 .
bi,0,r = (ω + (1 − ω)ξ )i b0,0,r (9)
  r (16)
b0,0,r = PSW 1 − (ω + ξ − ωξ )M+1 b0,0,0 . (10)
PSW signifies the probability that the device is in the sleep
From (6), for i = M and r = R, b0,0,0 is located from the stage and equivalent to
following:
PSW = Lc Pc + Pf . (17)
M CWi −1 R M R
bi,j,r + bi,0,r Pf is calculated in Section VII-A, and Pc which is the prob-
ability of more than one node that sense an idle channel and
i=0 j=0 r=0 i=0 r=0
start their transmission at the same time. The Pc is given by
M R Lt −1 R
+ bi,−1,r + b−1,j,r N(1 − )N−1
Pc = 1 − (18)
i=0 r=0 j=0 r=0 1 − (1 − )N
Lc −1 R NG−1 R in which  is
+ b−2,j,r + b−3,j,r = 1. (11) M−1 R−1
j=0 r=0 j=0 r=0 = bi,0,r (19)
Next each expression in (11) is derived separately accord- i=0 r=0

ing to (8)–(10). Furthermore, the delay line CWi is started and after analyzing (19) the value of  will be represented by
with CW0 = 2MinBE and CWMax = 2MaxBE , where ( × b0,0,0 ).
(MaxBE–MinBE = Dif) ≥ i ≤ M to get
VI. M ODEL VALIDATION
M CWi −1 R
bi,j,r To confirm the accuracy of the proposed analytical
i=0 j=0 r=0
model, Monte-Carlo simulations of the IEEE 802.15.4-based-
contention formulation are compared with the proposed model
b0,0,0 1 − (2(ω + ξ − ωξ ))Dif +1 and included in the results. The simulation model attempts to
=  +  × CW0
2 1 − 2(ω + ξ − ωξ ) imitate, as closely as possible, the actual transmission process
 based on the CSMA/CA of each device (see Appendix-B for
Dif 1 − (ω + ξ − ωξ )
M
+ CW0 2 (12) more details about the simulation algorithm). In general, the
1 − ω − ξ + ωξ CSMA/CA allows all the devices to compete for the channel
and access as a single group rather than as multiple ones. Hence,
only NG = 1 will be considered when validating the accuracy
M R M R Lt −1 R
of the proposed model; however, all other NG values have been
bi,0,r + bi,−1,r + b−1,j,r
provided to show the improvement in the proposed model over
i=0 r=0 i=0 r=0 j=0 r=0
the traditional IEEE 802.15.4-based CSMA/CA protocol. The
Lc −1 R
values of the parameters used to implement both the analytical
+ b−2,j,r model and the simulation runs are listed in Table I, and are
j=0 r=0 identical to the default values of the IEEE 802.15.4 standard.
=  × ((2 − ω) + (1 − ω − ξ + ωξ ) In Fig. 6, the analytical model is compared to the simula-
× (Lt (1 − Pc ) + Pc Lc ))b0,0,0 (13) tion results with one value of NG(NG = 1) to validate the

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 2282 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

Fig. 6. Validation of the proposed Markov model by comparing the numerical

results of ω, ξ , and  with the results obtained from the simulation when Fig. 8. ξ validation for the proposed HSW-802.15.4 with different values of
NG = 1. NG.

the channel sensing process, the HSW-802.15.4 performs the

same, or better, when NG = 4 or 8, and MaxBE = 3, than
when NG = 1 and MaxBE = 5. This means that the value
of NG with lower power consumption due to sleep mode
can compensate the MaxBE when the device consumes more
power, especially in a high-density network. Fig. 8 also shows
the effect of the NG and MaxBE on channel sensing for the
second CCA with the probability ξ .

VII. M ETRICS R ESULTS

In this section, the analytical model is used to describe
numerically the performance of the HSW-802.15.4 model and
the results from this are compared to those of the IEEE 802.15.4
standard in terms of: failure probability (Pf ), throughput (S),
channel utilization (ρ), and power consumption (PW ). The
performance of the proposed protocol has been examined with
different numbers of subgroups (NG = 1; 2; 3; 4; 8 and 16).
It should be noted that the proposed HSW-802.15.4 behaves
Fig. 7. ω validation for the proposed HSW-802.15.4 when MaxBE = 3 and identically to the IEEE 802.15.4 standard when NG = 1.
5 and with different values of NG.

accuracy of the proposed model. This is because the proposed A. Failure Probability
analytical model behaves identically to the traditional IEEE Failure probability (Pf ) refers to the probability of a packet
802.15.4 standard when NG = 1, which means there is no being discarded as the result of either CAF or retransmission
sleep mode and all the devices compete to access the channel limits (RLs). According to the Markov model in Fig. 5, the
as one group. Pf can be mathematically represented as follows:
The figure presents the validation for the derived expres-
sions ξ, ω and  when compared to the simulation results. Pf = PRL + PCAF
 
As shown, the values of the analytical prediction results are R+1
accurate and very close to those of the simulation results. = PSW 1 − (ω + ξ − ωξ )M+1
Additionally, when the number of devices is increased, the 
values of ω and ξ increase, whilst the value of  decreases. +  × (ω + ξ − ωξ ) M+1
b0,0,0 (20)
Thus, there is an increase in the number of collisions and
channel failures when the number of devices is increased. where, the PCAF and PRL are the probabilities of the packet
Fig. 7 displays the ω probability against the number of getting discarded due to CAF and RL, respectively.
devices. The value of ω has been tested when NG = 1, 4 Fig. 9 shows the Pf against the number of devices with
and 8, and MaxBE = 3 and 5. This figure shows that, during different values of NG and it is clear from the figure that

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 AL-JANABI AND AL-RAWESHIDY: ENERGY EFFICIENT HyMAC PROTOCOL WITH DYNAMIC SLEEP-BASED SCHEDULING 2283

Fig. 9. Probability of failure PF with different values of NG and with MAC Fig. 10. Average throughput S with different values of NG.
parameters: MaxBE = 5 and M = 3.
C. Channel Utilization
failure probabilities decrease with higher values. This justi- The channel utilization (ρ) for one superframe is defined as
fies the effectiveness of the proposed sleep/wakeup protocol, the summation of the time during the successful transmission
especially when the network begins to suffer from a higher of packets to the total time during one superframe (SFduration ).
number of collisions or channel failure. When the number of The ρ in the simulation can be mathematically evaluated as
subgroups increase, the number of sleep devices also increases,
(SFduration − TBN − TST ) × ϕ
which leads to a decrease in the number of contending devices ρ= (23)
during the CP and consequently, the number of collisions fall SFduration
and channel failure is more likely to be avoided. and SFduration can be clarified as
Ns
SFduration = TDMAslot i + TBN + TST (24)
B. Throughput
i=0
The throughput [S(bit/s)] is defined as the aggregated where, Ns represents the number of TDMAslots , TBN refers
amount of data that is successfully transmitted during a time to the beacon notification interval. TST represents the time
specified. Successful transmission means that the device needs required to broadcast the ST message and it is comparable to
to sense the channel successfully for both CCAs and then send 2 × TBN . The value of TBN , in the results, is taken equivalent
the packet without channel failure or collision. The S can be to at least one PCKL . PCKL refers to the total length of the
mathematically represented as follows: packet including overhead and payload (see Table I). Then,
TBN is estimated as
S = DR × ϕ (21)
PCKL
TBN =  . (25)
where, ϕ is calculated as follows: DR
Fig. 11 investigates ρ in terms of the number of devices
ϕ = (1 − ω)(1 − ξ )(1 − PSW ) (22) using different values of NG. As shown in the figure, the
proposed HSW-802.15.4 provides a better ρ when NG =
where, DR represents the data rate (see Table I). The through- 2, 3, 4, 8, or 16. Moreover, the value of ρ is high when the
put generated using the proposed model is explored using number of devices is less than 50, for from there it then starts
different values of NG. Fig. 10 shows the throughput against decreasing when the number of devices is increased. Finally,
different numbers of devices when NG = 1, 2, 3, 4, 8, and 16. the figure also shows that the proposed protocol means that
It shows that the throughput is increased when NG is ρ remains high even with a higher number of devices, which
increased. However, the throughput starts decreasing with all indicates the effectiveness of proposed protocol in a higher
values of NG when the number of devices increases, which network density. Furthermore, the value of ρ stays high for a
is because the number of collisions starts to increase when longer time when NG is increased.
there are more devices. To compensate, the proposed protocol
expands to cover a larger number of devices in the network. D. Delay
In the conclusion, there is an inverse relationship between The average delay (Davg ) is defined as the summation of the
the throughput and the number of devices, and vice versa periods that the devices elapse in each transmission stage to
depending on NG. obtain a successful one, i.e., receiving the ACK message. The

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 2284 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

as the network density is grows and illustrates almost the same

performance as the standard with a lower density network.
Finally, the proposed protocol displays that Davg remains lower
even with a higher number of devices, which indicates its effec-
tiveness with higher network density. This efficiency in the Davg
metric is the result of increasing the network performance owing
to device scheduling to share the channel and improvement in
the number of successful transmissions.

E. Power Consumption
In this section, the average energy consumed by a
device attempting to transmit a single packet using the
proposed HSW-802.15.4 protocol is considered. Using the
IEEE 802.15.4 power parameters listed in Table I, the power
consumed in idle, channel sensing, transmission, receiving,
and sleep modes is denoted by PWidle , PWCCA , PWTx , PWRx ,
and PWSL , respectively. Furthermore, taking into consideration
the Markov probabilities of the proposed protocol, the mean
Fig. 11. Channel utilization ρ with different values of NG.
energy ETot can be obtained as follows:
ETot = PWidle Pridle +PWCCA PrCCA
+ PWTx PrTx +PWRx PrRx +PWSL PrSL (27)
where, PrSL is given in (16). Pridle is the probability of the
device being idle either at the backoff, or at the time of waiting
for the ACK (LACK−out ), and can be calculated as follows:
M CWi −1 R LACK−out R
Pridle = bi,j,r + b−2,j,r . (28)
i=0 j=0 r=0 j=0 r=0

PrCCA is the probability of channel sensing for two consec-

utive CCAs, and it is calculated as follows:
PrCCA =  + (1 − ω) (29)
and PrTx is the transmission probability and calculated as
PCKL  
PrTx = × 1 − (1 − )N (1 − ω)(1 − ξ ). (30)
DR
PrRx is the probability of successfully receiving the ACK
Fig. 12. Total average delay (Davg ) with NG values = 1, 4, 8, and 16, and after a transmission and it is given as follows:
with MAC parameters: MaxBE = 5 and M = 3.
LACK R
PrRx = b−1,j,r . (31)
transmission stages are: the backoff counter, channel sensing j=0 r=0
(CCAs), sleep/wake-up and packet transmission (either col- Fig. 13 compares the energy consumed by the HSW-
lision or successful). Hence, the Davg spent by the network 802.15.4 MAC for the following values of NG: 1, 2, 3, 4, 8,
during one entire superframe duration is equivalent to the and 16. This shows that average energy depletion has been
summation of all the periods excluded the time during the improved by 40%–60% with the proposed protocol depend-
successful transmission one. Then, it can be expressed as ing on the number of devices and NG. This is because the
Davg = SFduration − (SFduration − TBN − TST ) × ϕ (26) devices avoid the high-energy depletion caused by attempt-
ing to access the channel and thus, there is no energy lost
where, the values of ϕ, SFduration and TBN are shown in idle stages waiting for the channel state to be detected, or
in (22), (24), and (25), respectively. when receiving unwanted packets due to collisions or channel
Fig. 12 inspects Davg in terms of the number of devices overhead. Instead, the devices in the proposed protocol spend
and values of NG = 1, 4, 8, and 16. The figure shows that the more time in a low energy sleep stage when network density
analytical results when NG = 1 is well estimated in compari- is high and the NG is increased, which results in a decrease
son with the simulation results. Furthermore, it reveals that the in the number of competing devices depleting their energy in
proposed HSW-802.15.4 protocol runs with a lower Davg when successful transmission or reception.
NG = 4, 8, or 16 than the IEEE 802.15.4 standard. Moreover, as Fig. 14 shows the energy consumed by the devices during
expected, the figure presents that Davg exponentially increases the sleep stage. This shows that the device spends more energy

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 AL-JANABI AND AL-RAWESHIDY: ENERGY EFFICIENT HyMAC PROTOCOL WITH DYNAMIC SLEEP-BASED SCHEDULING 2285

This HyMAC protocol enhances the channel access mecha-

nism by merging the strong aspects of the CSMA/CA with
those of the TDMA. The PAN coordinator broadcasts an ST
message that categorizes the network into NG sleep groups
of devices called SGi . With the proposed extension, certain
groups of devices are forced to sleep after each channel failure
or collision. Each device that experiences channel collision or
failure calculates its sleep period according to the scheduling
information in the ST message and the TDMAslot sequence.
However, all the groups of devices are permitted to access the
channel with low-density networks when there are no colli-
sions or channel failures. Moreover, the number of IoT devices
is increasing, and most of these are battery-powered. Hence,
the proposed protocol enables IoT devices to conserve more
power to access the channel, thereby prolonging the lifes-
pan of the network regardless of the number of devices it
contains. Moreover, dividing the devices into multiple sleep
groups reduces the hidden node problem. A frame structure
Fig. 13. Average power consumption with different values of NG and for the proposed protocol compatible with the IEEE 802.15.4
MaxBE = 5 as a function of saturated network.
standard has also been developed. To provide a mathematical
model for the proposed protocol, a theoretical analysis-based
Markov chain has been constructed. The results show that
the proposed HSW-802.15.4 protocol can efficiently utilize the
energy of the battery by 40%–60% depending on the NG, and
this demonstrates very efficient channel utilization with high
throughput when compared to the performance of the IEEE
802.15.4 standard.

B. Future Work
The IEEE 802.15.4 MAC is subject to many other issues
that also need to be investigated and incorporated to improve
the energy conservation of the devices, some of these being
the: back-off methodology, optimal duty cycle, unreliability,
network complexity, packet size limitation and the hidden
node problem. Finding efficient algorithms for these issues
can prevent the high energy consumed by the devices, due to
data transmission, reception, collision as well as the idle and
sleep modes.
Fig. 14. Power consumption during the sleep stage with different values
of NG.
For instance, tuning the values of the maximum and min-
imum number of binary exponential back-off in an effective
on the sleep stage when the NG value is increased. Energy way will have a direct impact on the energy efficiency of any
consumption is, therefore, increased when the network density device. This is considered to be a better solution as will have
(number of devices) is increased. Furthermore, there is no energy a greater impact on the communication reliability, rather than
consumption when NG = 1, because there is no sleep stage, increasing the number of retransmissions, for this will require
while the highest energy consumption occurs when NG = 16. higher energy consumption. Furthermore, depending on the
Thus, this figure illustrates why there is efficient utilization of application constraints and specifications, some devices are
overall energy consumption, as the device consumes its energy allocated a fixed duty cycle (long or short period), while oth-
during the sleep stage or in successful transmission and channel ers require adaptable periods. These two factors can obviously
access, rather than in idle stages or higher energy depletion affect the unreliability issue in IEEE 802.15.4 MAC. This is
stages, such as collision or channel failure. because it employs the CSMA/CA mechanism to access the
channel, which is considered as being an inefficient algorithm
for controlling the number of collisions, especially in high
VIII. C ONCLUSION
density networks.
A. Conclusion Another important solution to conserving the energy con-
This paper proposes a dynamic HyMAC protocol known sumed by IoT devices and reducing the network complexity
as HSW-802.15.4, which adds an adaptable sleep/wake-up is the involvement of SDN and network virtualization tech-
extension to the mechanism of the IEEE 802.15.4 standard. nologies in the MAC process. These can be implemented in

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 2286 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

TABLE III
N OTATION U SED T HROUGHOUT T HIS PAPER
Algorithm 1 Transmission-Based CSMA/CA
1: procedure T RANSMISSION ATTEMPT FOR EACH DEVICE
2: N, MinBE, MaxBE, R
3: r(1 : N) ← 0;
4: NBF(1 : N) ← 0;
5: CCA(1 : N) ← 2;
6: TxSucc (1 : N) ← 0;
7: M ← (MaxBE − MinBE);
8: for each Node i do
9: Initialisation of Backoff Exp.
10: CW(i, 1) ← 2MinBE − 1;
11: if R > M then
12: if M > 0 then
13: CW(i, 2 : M + 1) ← 2(1 : M) ∗ CW(i, 1);
14: CW(i, M + 1 : R + 1) ← CW(i, M + 1);
15: else
16: if M = 0 then
17: CW(i, 2 : R + 1) ← CW(i, 1);
18: end if
19: end if
20: else
21: CW(i, 2 : M + 1) ← 2(1:M) ∗ CW(i, 1);
22: end if
23: while (NBF(i) < M) & (TxSucc (i) = 0) do
24: BE = MinBE + NBF(i);
25: Backoff-Delay(i) ← Random number in [1,..., CW(i, BE)]
26: Wait for the Backoff-Delay(i)
27: Note: Each node has different Backoff-Delay
28: if Backoff-Delay(i) = 0 then
29: Check First CCA(i)
30: if Channel idle then
31: CCA(i) ← CCA(i) -1;
different ways, such as proactively, reactively (on demand), 32: Check the second CCA(i)
33: if Another node is transmitting then
or by mixing between the two. With these approaches, the 34: Collision!
SDN controller has the responsibility of monitoring the num- 35: else
ber of network traffic collisions and deciding upon the optimal 36: Successful Transmission
37: TxSucc (i) ← 1;
sleep/wake-up period for each device in a way that elimi- 38: end if
nates the effect of the hidden node issue. These methods can 39: else
be easily combined with the proposed HSW-802.15.4 MAC 40: Channel Busy
41: NBF(i) ← NBF(i)+1; CCA(i) ← 2;
protocol, because the SDN controller and the virtualization 42: end if
technology over the cloud can categorize the devices into sev- 43: end if
eral groups and identify the TDMA time slots as well as the 44: end while
45: if r(i) < R & TxSucc (i) = 0 then
group associated with each. Furthermore, the involvement of 46: r(i) ← r(i) + 1; CCA(i) ← 2;
such technologies in the MAC protocol will facilitate the pro- 47: BE ← MaxBE;
tocol updating process, according to the network requirements, 48: Repeat the from line 25
49: else
such as in the case of an emergency. In this case, the device 50: if TxSucc (i) = 0 then
will need to be accompanied by the GPS facility in order to 51: Transmission Failure
be able to send an emergency request, when required, to the 52: end if
53: end if
SDN controller over the cloud, which in turn will update the 54: end for
network with an emergent route for the requested device. 55: end procedure

A PPENDIX A
N OTATION R EFERENCES
The notation used throughout this paper are described in [1] Y. Liu, C. Yuen, J. Chen, and X. Cao, “A scalable hybrid MAC protocol
Table III. for massive M2M networks,” in Proc. IEEE Wireless Commun. Netw.
Conf. (WCNC), 2013, pp. 250–255.
[2] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, and
A PPENDIX B M. Ayyash, “Internet of Things: A survey on enabling technologies, pro-
CSMA/CA A LGORITHM tocols, and applications,” IEEE Commun. Surveys Tuts., vol. 17, no. 4,
pp. 2347–2376, 4th Quart., 2015.
Algorithm 1 describes the behavior of the CSMA/CA mech- [3] A. Aijaz and A. H. Aghvami, “Cognitive machine-to-machine commu-
anism, which is similar to the simulator used by [14] with nications for Internet-of-Things: A protocol stack perspective,” IEEE
Internet Things J., vol. 2, no. 2, pp. 103–112, Apr. 2015.
some enhancements. After guaranty the accuracy of the sim- [4] IEEE Standard for Information Technology—Local and Metropolitan
ulator by gotten the same results presented by [14], its results Area Networks—Specific Requirements—Part 15.4: Wireless Medium
have been compared with the proposed numerical model. Access Control (MAC) and Physical Layer (PHY) Specifications for
Low Rate Wireless Personal Area Networks (WPANs), IEEE Standard
Furthermore, the notation used in Algorithm 1 have been 802.15.4-2006, 2006. Accessed: Mar. 1, 2016. [Online]. Available:
described in Tables I and II (in Appendix A). https://standards.ieee.org

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.
 AL-JANABI AND AL-RAWESHIDY: ENERGY EFFICIENT HyMAC PROTOCOL WITH DYNAMIC SLEEP-BASED SCHEDULING 2287

[5] B. Bandyopadhyay, S. J. Ahmed, D. Das, and A. Chatterjee, “AS [27] Z. Xiao, C. He, and L. Jiang, “An analytical model for IEEE 802.15.4
802.15.4: A modified IEEE 802.15.4 standard for more reliable commu- with sleep mode based on time-varying queue,” in Proc. IEEE Int. Conf.
nication and utilization of inactive period using optimized sleep period,” Commun. (ICC), 2011, pp. 1–5.
in Proc. IEEE Annu. India Conf. (INDICON), 2014, pp. 1–6. [28] B. Shrestha, E. Hossain, and S. Camorlinga, “IEEE 802.15.4 MAC
[6] C. G. Peces, J. Eriksson, and N. Tsiftes, “Sleepy devices vs. radio duty with GTS transmission for heterogeneous devices with application to
cycling: The case of lightweight M2M,” IEEE Internet Things J., to be wheelchair body area sensor networks,” IEEE Trans. Inf. Technol.
published. Biomed., vol. 15, no. 5, pp. 767–777, Sep. 2011.
[7] A. Rajandekar and B. Sikdar, “A survey of MAC layer issues and pro- [29] S.-T. Sheu, Y.-Y. Shih, and W.-T. Lee, “CSMA/CF protocol for
tocols for machine-to-machine communications,” IEEE Internet Things IEEE 802.15.4 WPANs,” IEEE Trans. Veh. Technol., vol. 58, no. 3,
J., vol. 2, no. 2, pp. 175–186, Apr. 2015. pp. 1501–1516, Mar. 2009.
[8] O. Tickoo and B. Sikdar, “Modeling queueing and channel access [30] T. A. Al-Janabi and H. S. Al-Raweshidy, “A centralized routing protocol
delay in unsaturated IEEE 802.11 random access MAC based wire- with a scheduled mobile sink-based AI for large scale I-IoT,” IEEE
less networks,” IEEE/ACM Trans. Netw., vol. 16, no. 4, pp. 878–891, Sensors J., vol. 18, no. 24, pp. 10248–10261, Dec. 2018.
Aug. 2008. [31] H. Ayadi, A. Zouinkhi, T. Val, A. van den Bossche, and
[9] B. Shrestha, E. Hossain, and K. W. Choi, “Distributed and centralized M. N. Abdelkrim, “Network lifetime management in wireless sensor
hybrid CSMA/CA-TDMA schemes for single-hop wireless networks,” networks,” IEEE Sensors J., vol. 18, no. 15, pp. 6438–6445, Aug. 2018.
IEEE Trans. Wireless Commun., vol. 13, no. 7, pp. 4050–4065, [32] N. Abramson, “The ALOHA system: Another alternative for computer
Jul. 2014. communications,” in Proc. ACM Fall Joint Comput. Conf., Nov. 1970,
[10] R. Bruno, M. Conti, and E. Gregori, “Optimal capacity of p-persistent pp. 281–285.
CSMA protocols,” IEEE Commun. Lett., vol. 7, no. 3, pp. 139–141, [33] L. G. Roberts, “ALOHA packet system with and without slots and cap-
Mar. 2003. ture,” ACM SIGCOMM Comput. Commun. Rev., vol. 5, no. 2, pp. 28–42,
[11] G. Bianchi, “Performance analysis of the IEEE 802.11 distributed 1975.
coordination function,” IEEE J. Sel. Areas Commun., vol. 18, no. 3, [34] L. Kleinrock and F. Tobagi, “Packet switching in radio channels:
pp. 535–547, Mar. 2000. Part I—Carrier sense multiple-access modes and their throughput-
[12] S. Pollin et al., “Performance analysis of slotted carrier sense IEEE delay characteristics,” IEEE Trans. Commun., vol. COM-23, no. 12,
802.15.4 medium access layer,” IEEE Trans. Wireless Commun., vol. 7, pp. 1400–1416, Dec. 1975.
no. 9, pp. 3359–3371, Sep. 2008. [35] I. Rhee, A. Warrier, M. Aia, J. Min, and M. L. Sichitiu, “Z-MAC: A
[13] V. Nguyen, T. Z. Oo, P. Chuan, and C. S. Hong, “An efficient time hybrid MAC for wireless sensor networks,” IEEE/ACM Trans. Netw.,
slot acquisition on the hybrid TDMA/CSMA multichannel MAC in vol. 16, no. 3, pp. 511–524, Jun. 2008.
VANETs,” IEEE Commun. Lett., vol. 20, no. 5, pp. 970–973, May 2016. [36] R. Zhang, R. Ruby, J. Pan, L. Cai, and X. Shen, “A hybrid
[14] A. Faridi et al., “Comprehensive evaluation of the IEEE 802.15.4 MAC reservation/contention-based MAC for video streaming over wireless
layer performance with retransmissions,” IEEE Trans. Veh. Technol., networks,” IEEE J. Sel. Areas Commun., vol. 28, no. 3, pp. 389–398,
vol. 59, no. 8, pp. 3917–3932, Oct. 2010. Apr. 2010.
[15] P. K. Sahoo and J.-P. Sheu, “Modeling IEEE 802.15.4 based wire- [37] M. Salajegheh, H. Soroush, and A. Kalis, “HYMAC: Hybrid
less sensor network with packet retry limits,” in Proc. 5th ACM Symp. TDMA/FDMA medium access control protocol for wireless sensor
Perform. Eval. Wireless Adhoc Sensor Ubiquitous Netw., 2008, pp. networks,” in Proc. IEEE 18th Int. Symp. Pers. Indoor Mobile Radio
63–70. Commun. (PIMRC), 2007, pp. 1–5.
[16] P. Park, P. Di Marco, P. Soldati, C. Fischione, and K. H. johansson, [38] T. Shu and M. Krunz, “Energy-efficient power/rate control and schedul-
“A generalized Markov chain model for effective analysis of slotted ing in hybrid TDMA/CDMA wireless sensor networks,” Comput. Netw.,
IEEE 802.15.4,” in Proc. IEEE 6th Int. Conf. Mobile Adhoc Sensor vol. 53, no. 9, pp. 1395–1408, 2009.
Syst. (MASS), 2009, pp. 130–139. [39] M. Chatterjee and S. K. Das, “Performance evaluation of a request-
[17] M.-H. Zayani, V. Gauthier, and D. Zeghlache, “A joint model for IEEE TDMA/CDMA protocol for wireless networks,” J. Interconnection
802.15.4 physical and medium access control layers,” in Proc. IEEE Netw., vol. 2, no. 1, pp. 49–67, 2001.
7th Int. Wireless Commun. Mobile Comput. Conf. (IWCMC), 2011, [40] Y. Liu, C. Yuen, X. Cao, N. U. Hassan, and J. Chen, “Design of a
pp. 814–819. scalable hybrid MAC protocol for heterogeneous M2M networks,” IEEE
[18] M. Zuniga and B. Krishnamachari, “Analyzing the transitional region in Internet Things J., vol. 1, no. 1, pp. 99–111, Feb. 2014.
low power wireless links,” in Proc. 1st Annu. IEEE Commun. Soc. Conf.
Sensor Ad Hoc Commun. Netw. (IEEE SECON), 2004, pp. 517–526.
[19] B. Bandyopadhyay, S. J. Ahmed, M. K. Naskar, and A. Mukherjee,
“Improving the performance metrics of beacon enabled IEEE 802.15.4 Thair A. Al-Janabi (GS’16–M’18) received the
using adaptive synchronization algorithm for reliable communication,” B.Sc. and M.Sc. degrees in information and com-
in Proc. IEEE Int. Conf. Adv. Netw. Telecommun. Syst. (ANTS), 2013, munication engineering from Nahrain University,
pp. 1–6. Baghdad, Iraq, in 2008 and 2011, respectively. He
[20] Y. Zhuang and L. Ma, “An energy-efficient and low-collision IEEE is currently pursuing the Ph.D. degree in wireless
802.15.4-based MAC for data gathering in wireless sensor networks,” network engineering at Brunel University London,
in Proc. IEEE Int. Conf. Comput. Sci. Service Syst. (CSSS), 2012, London, U.K.
pp. 1244–1247. His current research interests include IoT, artificial
[21] J. Mišic and V. B. Mišic, “Queuing analysis of sleep management in an intelligence, wireless sensor networks, development
802.15.4 beacon enabled PAN,” in Proc. IEEE/ACM 1st Int. Workshop and optimization of routing and MAC protocols,
Broadband Wireless Services Appl. (BroadWISE), 2004. SDN, NFV, and cloud computing.
[22] C.-F. Chiasserini and M. Garetto, “An analytical model for wireless
sensor networks with sleeping nodes,” IEEE Trans. Mobile Comput.,
vol. 5, no. 12, pp. 1706–1718, Dec. 2006.
[23] M. H. F. Ghazvini, M. Vahabi, M. F. A. Rasid, and R. S. A. R. Abdullah, Hamed S. Al-Raweshidy (M’92–SM’97) received
“Energy efficiency in MAC 802.15.4 for wireless sensor networks,” in the B.Eng. and M.Sc. degrees from the University
Proc. IEEE 6th Nat. Conf. Telecommun. Technol. 2nd Malaysia Conf. of Technology Iraq, Baghdad, Iraq, in 1977 and
Photon. NCTT-MCP, 2008, pp. 289–294. 1980, respectively, the Post Graduate Diploma
[24] Z. Xiao, C. He, and L. Jiang, “Slot-based model for IEEE 802.15.4 degree from Glasgow University, Glasgow, U.K.,
MAC with sleep mechanism,” IEEE Commun. Lett., vol. 14, no. 2, in 1987, and the Ph.D. degree from Strathclyde
pp. 154–156, Feb. 2010. University, Glasgow, in 1991.
[25] R. Jurdak, A. G. Ruzzelli, and G. M. O’Hare, “Radio sleep mode He was with Space and Astronomy Research
optimization in wireless sensor networks,” IEEE Trans. Mobile Comput., Centre, Iraq, PerkinElmer, Waltham, MA, USA,
vol. 9, no. 7, pp. 955–968, Jul. 2010. British Telecom, Edinburgh, U.K., Oxford
[26] M. Khanafer, M. Guennoun, and H. T. Mouftah, “Adaptive sleeping University, Oxford, U.K., Manchester Metropolitan
periods in IEEE 802.15.4 for efficient energy savings: Markov-based University, Manchester, U.K., and Kent University, Canterbury, U.K. He is
theoretical analysis,” in Proc. IEEE Int. Conf. Commun. (ICC), 2011, currently the Director of Wireless Network Communications Centre, Brunel
pp. 1–6. University London, London, U.K.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on August 27,2021 at 08:22:17 UTC from IEEE Xplore. Restrictions apply.