IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO.

1, FIRST QUARTER 2013 101

The Evolution of MAC Protocols in Wireless

Sensor Networks: A Survey
Pei Huang, Li Xiao, Senior Member, IEEE, Soroor Soltani, Student Member, IEEE,
Matt W. Mutka, and Ning Xi, Fellow, IEEE,

Abstract—Wireless Sensor Networks (WSNs) have become a Therefore, an efficient MAC protocol increases the lifetime
leading solution in many important applications such as intrusion of a sensor network to a great extent. In addition, the MAC
detection, target tracking, industrial automation, smart building layer controls how nodes share the wireless medium. An
and so on. Typically, a WSN consists of a large number of small,
low-cost sensor nodes that are distributed in the target area for efficient MAC protocol can reduce collisions and increase
collecting data of interest. For a WSN to provide high throughput the achievable throughput, providing flexibility for various
in an energy-efficient way, designing an efficient Medium Access applications.
Control (MAC) protocol is of paramount importance because When developing new protocols, it is necessary to inves-
the MAC layer coordinates nodes’ access to the shared wireless
medium. To show the evolution of WSN MAC protocols, this
tigate all the previous studies thoroughly. There has been
article surveys the latest progresses in WSN MAC protocol a tremendous amount of research on the design and imple-
designs over the period 2002-2011. In the early development mentation of MAC protocols in WSNs. Hence, surveys of
stages, designers were mostly concerned with energy efficiency WSN MAC protocols are conducted to summarize the varieties
because sensor nodes are usually limited in power supply. of designs and implementations. The survey in [3] reviews
Recently, new protocols are being developed to provide multi-
task support and efficient delivery of bursty traffic. Therefore,
several early designed MAC protocols based on their medium
research attention has turned back to throughput and delay. access strategies: random access or static access. Later, the
This article details the evolution of WSN MAC protocols in work in [4] analyzes advantages and disadvantages of some
four categories: asynchronous, synchronous, frame-slotted, and recently proposed WSN MAC protocols. A more comprehen-
multichannel. These designs are evaluated in terms of energy sive study in [5] classifies WSN MAC protocols into four
efficiency, data delivery performance, and overhead needed to
maintain a protocol’s mechanisms. With extensive analysis of
categories: random, slots, frames, and hybrid. These categories
the protocols many future directions are stated at the end of are generally based on the similarities and differences in
this survey. The performance of different classes of protocols the medium access methods. A recently published survey [6]
could be substantially improved in future designs by taking into classifies protocols based on various problems they intend to
consideration the recent advances in technologies and application address. In this paper, we do not just classify protocols into
demands.
different categories. We further provide a clue about why a
Index Terms—Wireless sensor networks, MAC protocols, en- protocol is proposed for a problem and what remaining issues
ergy efficiency, time-critical, TDMA, multichannel. lead to another solution, identifying the intrinsic development
flow.
I. I NTRODUCTION In the early stages, efficient data delivery was not the first
priority. Designers traded throughput and delay for energy ef-
A DVANCES in microelectronics led to the development
of low-cost tiny sensor nodes that are equipped with
sensing, processing, and communication units. A broad range
ficiency. However, to support multi-task and efficient delivery
of bursty traffic, new protocols are being developed. In this
of applications such as precision agriculture, environment survey, we provide an overview of WSN MAC protocols from
monitoring, intrusion detection, target tracking, and etc. are perspectives of both energy efficiency and data delivery perfor-
facilitated through networking these sensor nodes [1]. Many mance with more recently proposed work. We cover the most
valuable applications have been demonstrated in the SENSEI recent progression from energy efficiency towards efficient
[2] project, which is developed by the European FP7-ICT. data delivery. Based on research issues, we divide WSN MAC
Since the sensor nodes are expected to operate autonomously protocols into four branches: asynchronous, synchronous,
with small batteries for a number of months or years, energy frame-slotted, and multichannel. Compared with prior clas-
efficiency is a fundamental criterion in the design of WSN sifications, this classification makes the major problems that
protocols. A major power consuming component of a sensor each branch intends to solve clearer because each branch has
node is the radio, which is controlled by the MAC protocol. its own special obstacles to overcome. It is beneficial for re-
searchers to know the major challenges they face in the design
Manuscript received 15 July 2011; revised 26 February 2012. of MAC protocols in each branch. For instance, asynchronous
P. Huang, L. Xiao, S. Soltani, and M. W. Mutka are with the Department and synchronous are related to the mechanism of duty cycling
of Computer Science and Engineering, Michigan State University, 3115
Engineering Building, East Lansing, MI 48824 (e-mail: {huangpe3, lxiao, in WSNs. To save energy, duty cycling is widely adopted
soltani1}@cse.msu.edu). in WSNs. In this technique, each node alternates between
N. Xi is with the Department of Electrical and Computer Engineering, active and sleep states. Two nodes can communicate only
Michigan State University, 2120 Engineering Building, East Lansing, MI
48824 (e-mail: [email protected]). when they are both active. In synchronous MAC protocols,
Digital Object Identifier 10.1109/SURV.2012.040412.00105 neighboring nodes are synchronized to wake up at the same
1553-877X/13/$31.00 
c 2013 IEEE
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 102 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

RICER(2004),
PTIP(2005), RC-MAC(2010)
STEM(2002), CSMA-MPS(2004), RI-MAC(2008)
no Preamble TICER(2004), X-MAC(2006),
synchronization Asynchronous Sampling(2002), DPS-MAC(2007), CMAC(2007)
B-MAC(2004)
WiseMAC(2004), AS-MAC(2008),
A-MAC(2010)
SyncWUF(2007) PW-MAC(2011)

sender-initiated low receiver-initiated low

power listening (LPL) power polling (LPP)
SCP-MAC(2006)

DSMAC(2004)
local
synchronization S-MAC(2002),
RMAC(2007),
Synchronous S-MAC adaptive T-MAC(2003)
DW-MAC(2008)
listening (2004) synchronize to
receiver’s
DMAC(2004), wakeup
Q-MAC(2006)

reduce delay
TDMA- Crankshaft(2007)
ZMAC(2005) TreeMAC(2009)
ASAP(2009) PMAC(2005)

global Frame- TRAMA(2003)

synchronization Slotted based on data
gathering tree
AI-LMAC(2004)
structure

assign time slots for sending assign slots for wake up

Static MMSN(2006) TMCP(2008) GBCA(2010) MC-LMAC(2011)

multi-channel

Dynamic K-way Cut(2008) Y-MAC(2008) MuChMAC(2010)

FDMA hybrid design of FDMA and TDMA

Fig. 1. Taxonomy of WSN MAC protocols.

time. Therefore, the communication is facilitated and the need to maintain multiple schedules if it belongs to multiple
focus is on the delay reduction and throughput improvement. clusters. With slightly more strict global time synchronization,
Asynchronous MAC protocols, on the other hand, focus on frame-slotted schemes are able to address collision and hidden
how to efficiently establish communication between two nodes terminal problem by assigning a node an unique time slot
that have different active/sleep schedules. To provide high within its two-hop communication neighborhood to transmit.
throughput, frame-slotted mechanisms allocate time slots in Recently, the frame-slotted structure is also used for assigning
a way that no two nodes within the two-hop communication time slots to nodes for receiving instead of transmission.
neighborhood are assigned to the same slot. This addresses We discuss benefits and problems introduced by this change.
collision and hidden terminal problem, providing a collision- Finally, as current sensor platforms support multiple channels,
free data transmission environment. However, a major concern multichannel MAC protocols also become a hot topic. The
is that the channel utilization is low when few nodes have idea is to increase parallel transmissions by assigning nodes
data to send because time slots assigned to their neighbors different channels. The future trend is to combine Time
are wasted. Therefore, the focus of this branch is to im- Division Multiple Access (TDMA) with Frequency Division
prove channel utilization under low contention. In the final Multiple Access (FDMA) to address cross-channel commu-
branch, multichannel is employed to further boost network nication deficiency. Fig. 1 links MAC protocols that share
capacity. Distributed channel assignment and efficient cross- similarity in utilizing some underlying structures or exhibit
channel communication are two major challenging issues in similar behaviors.
multichannel MAC protocols. The remainder of this article is structured as follows. We
A taxonomy of WSN MAC protocols is illustrated in Fig. 1. first summarize the studies on asynchronous MAC protocols
The four horizontal branches represent our classifications. in Section II. After that, we focus on delay reduction in
We start by introducing asynchronous MAC protocols. In the synchronous MAC protocols in Section III. In Section IV
beginning, asynchronous MAC protocols use sender-initiated and Section V, we show how TDMA and FDMA can boost
low power listening (LPL) to reduce cost on the receiver side. network capacity. In Section VI, we review how standards
Later, receiver-initiated low power probing (LPP) is introduced incorporate the various techniques. Finally, we provide conclu-
to improve throughput and reduce cost on the sender side. We sions and identify potential research directions in Section VII.
analyze underlying reasons behind this transition. For syn-
chronous MAC protocols, we review the efforts that have been
II. A SYNCHRONOUS MAC PROTOCOLS
made to reduce delay. Synchronous MAC protocols group
nodes into clusters in order to set up a common active/sleep In this section we cover the trend in the design of asyn-
schedule within a cluster. An inconvenience is that a node may chronous MAC protocols, where each node chooses its active
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 HUANG et al.: THE EVOLUTION OF MAC PROTOCOLS IN WIRELESS SENSOR NETWORKS: A SURVEY 103

A
schedule autonomously. Without paying the price for synchro- C
nizing neighbors’ schedules, asynchronous MAC protocols can B

achieve ultra-low duty cycle but have to search efficient ways

to establish communication between two nodes. D
In WSNs, a node spends most of the time in the low-
power sleep mode and wakes up periodically to check whether Data Wakeup
transmission preamble
there are packets for it. To indicate that there is an impending
data transmission, a sender precedes its data with a preamble Fig. 2. Interference problem of preamble transmission.
that is long enough to be detected by all potential receivers.
This design is suitable for low traffic load applications where
occasional data transmissions do not incur too much overhead the channel and compare it to the noise floor. The channel
and channel contention is not severe. Because the preamble is considered clear only if the measured channel energy is
transmission occupies the channel and prevents neighboring below the noise floor. Because noise has significant variance
nodes from transmission, the achievable throughput is limited. in channel energy [10], the thresholding method introduces
To support higher throughput, asynchronous MAC protocols a large number of false positives (i.e., channel is active)
began to adopt receiver-initiated probing to release more room that lower the effective channel bandwidth because nodes
for data transmission. We will review various techniques that cannot send, and reduce energy efficiency as nodes believe
have been proposed to reduce the preamble length and the that there is an impending data transmission. Motivated by
recent transition from sender-initiated transmission to receiver- the observation that channel energy is fairly constant during
initiated transmission. packet reception, B-MAC (Berkeley MAC) [10] introduces an
outlier detection method to improve the quality of CCA.
A. Preamble Sampling Each node first takes signal strength samples at times when
the channel is supposed to be clear (e.g., immediately after
Assuming traffic is light in WSNs, the primary source
transmitting a frame). These samples are entered into a FIFO
of energy waste is idle listening. To reduce idle listening,
queue and the median of the queue is used as the input of
duty cycling is widely adopted in WSNs. In this technique,
an exponentially weighted moving average function for noise
each node alternates between active and sleep states. To let
floor estimate. Then, when a node needs to perform the CCA,
most nodes sleep as long as possible, preamble sampling is
it searches for outliers in the received signal such that the
proposed to minimize the active duration of receivers. The
channel energy is significantly below the noise floor. If an
method is to let a node periodically wake up for a short
outlier is detected during channel sampling, B-MAC declares
duration to sample the channel. If the channel is idle, the
the channel is clear because a valid transmission could never
node goes back to sleep immediately. If channel activities are
have an outlier significantly below the noise floor. If no outlier
detected, the node keeps listening until the subsequent data
is found within five samples, B-MAC declares the channel
frame is received or a timeout occurs. The method requires
is busy. With the more accurate CCA, B-MAC names its
that the first data frame is preceded by a preamble that is at
preamble sampling as Low Power Listening (LPL) because
least as long as the channel sampling interval. Assuming all
fewer false positives lead to lower duty cycle.
nodes have the same channel sampling interval, such a long
preamble ensures that all potential receivers can detect the
preamble and stay awake to receive the subsequent data frame. C. Decouple Data Transmission and Preamble Sampling
The bad consequence is that in unicast transmissions nontarget
In preamble sampling, a node turns its radio on when it has
nodes are unable to find out that the packet is not destined
data to send. It is unaware of neighboring nodes’ activities.
for them until the end of the long preamble transmission.
Due to the hidden terminal problem, even with the CCA
The overhearing problem thus incurs energy waste that is
algorithm, its preamble transmission may collide with ongoing
proportional to the node density. The advantage of preamble
data transmissions of neighboring nodes. An example is shown
sampling is that the channel sampling duration is short and
in Fig. 2, the preamble transmission of node C may collide
thus the channel sampling can be made frequent. This makes
with the ongoing data transmission between A and B. STEM
preamble sampling protocols be able to achieve satisfactory
[11] thus separates data transmission channel from wake-up
delay with ultra-low duty cycle.
channel by using two radios.
The preamble sampling technique has been introduced
STEM provides two preamble variants: STEM-T (STEM
along with the Mica wireless platform [7]. The performance
Tone) and STEM-B (STEM Beacon). STEM-T uses a bit
of Aloha with preamble sampling and CSMA with preamble
stream as the preamble. A node can detect the preamble by
sampling have been analyzed in [8] and [9] respectively.
comparing measured channel energy with the noise floor. Es-
sentially, STEM-T is similar to traditional preamble sampling
B. Low Power Listening (LPL) technique except for moving data transmission to a separate
In CSMA with preamble sampling, a transmitter per- channel. In STEM-B, a series of beacon packets are used
forms Clear Channel Assessment (CCA) before transmitting as the preamble. Each beacon contains the MAC address of
a preamble so as to avoid collisions. A receiver also needs to both the sender and the target receiver. A node thus can
perform the CCA to detect the preamble. The initial proposal determine whether it is the target receiver of an impending data
in the Mica wireless platform [7] is to sample the energy on transmission. Compared with traditional preamble sampling
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 104 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

protocols, nontarget nodes can go back to sleep earlier in only be mimicked by sending a packet repeatedly. A gap exists
STEM-B. In addition, inter-beacon intervals are inserted to let between two transmissions and the outlier detection based
the target receiver be able to reply with an acknowledgment CCA searches for outliers to determine channel state. The gap
(ACK). In this way, the sender can start data transmission between two transmissions may introduce false negatives (i.e.,
earlier, resulting in a reduction in setup latency. It is possible channel is not active) that increase setup latency. Therefore,
that collisions between beacons occur. A node then cannot a node must sample the channel long enough to accurately
determine whether it is the target receiver or not. It turns determine the channel state.
its data radio on without sending an acknowledgment back. To be compatible with packetizing radios, it is justified to
Since the sender cannot get any acknowledgment, it has to uses a series of short packets containing useful information
transmit the beacon stream for a sufficient amount of time as the wake-up preamble and utilize the gap to accommodate
(approximately equal to the channel sampling interval) before an early ACK as shown in Fig. 3. When there is no collision,
it sends the data frame on the data channel. If a node wakes up the design addresses the overhearing problem and cuts back
due to beacon collisions and it does not receive any data after the preamble length. Protocols that share the similar design
some time, it times out and goes back to sleep. Obviously, include CSMA-MPS [18], TICER [19], X-MAC [20], MH-
when there is no beacon collision, STEM-B exhibits its merit MAC [21], DPS-MAC (Dual Preamble Sampling MAC)
of shortening preamble length, leading to lower setup latency [22], and CMAC [23]. DPS-MAC [22] and MH-MAC [21]
and better energy efficiency. However, if data transmission is also include timing information for broadcast messages, allow-
rare, STEM-B is less energy efficient than STEM-T because ing receivers to go back to sleep and wake up at the beginning
the channel sampling period must be longer than the inter- of the data transmission. Strobed preamble sampling, however,
beacon interval in STEM-B while detecting a wake-up tone incurs more energy waste in channel sampling because the
in STEM-T takes much less time. channel sampling duration must be larger than the gap between
two short packets. Some protocols such as SpeckMAC-D
[15] and MX-MAC [24] repeat an actual data packet as the
D. Reduce Preamble Length by Packetization preamble. The channel sampling duration is even longer. As
Apparently, it is beneficial to divide a long preamble into shown in Fig. 3, if a node detects channel activity when it
a series of short packets that take some useful information. wakes up, it must stay awake until it decodes the destination
If the destination address is included, nontarget nodes can address information from the next short preamble packet.
immediately go back to sleep when they receive a short Using data packet as the short preamble packet increases the
preamble packet. This improves energy efficiency. ENBMAC idle listening period.
[12] uses the method to address the aforementioned overhear- In continuous preamble sampling, the channel sampling
ing problem. If the timing information about when the data duration is short and thus the channel sampling can be made
transmission will begin is also included, the target receiver frequent. This reduces per-hop latency while keeping a low
can return to sleep once it decodes the timing information duty cycle. Strobed preamble sampling has the merit of cutting
and wake up later to receive the data as shown in Fig. 3. back preamble length. To combine the two features together,
Neighboring senders that fail on channel contention can go DPS-MAC (Dual Preamble Sampling MAC) [22] and CMAC
back to sleep as they know when the current transmission [23] introduce a dual preamble sampling approach. Nodes
will end. Many protocols such as MFP [13], B-MAC+ [14], periodically check the channel state by polling the channel
SpeckMAC [15], DPS-MAC (Divided Preamble Sampling- with the basic LPL method. Because the polling may fall
MAC) [16], and SyncWUF [17] share the similarity although into the gap between two short packets, a second channel
they utilize different information for different problems. Note polling is initiated if the channel is found free in the first
that although they divide a long preamble into chunks, there polling. To ensure that channel activity can be detected, two
is no gap between chunks. We classify them as Continuous conditions must be met: (1) the interval between two instants
Preamble Sampling protocols. of polling is larger than the gap between two short preamble
For unicast transmissions, it is desirable to terminate the packets; (2) the time duration of a short preamble packet
preamble transmission once the target receiver is awakened. transmission is larger than the interval between two instants
The similar idea of STEM-B has been adopted in single chan- of polling. Although this method reduces idle listening of
nel designs: a gap between two short packets is deliberately strobed preamble sampling, it is sensitive to the switching
inserted to allow the target receiver to reply with an early time of radio. Since the time duration of a short preamble
ACK. We call it Strobed Preamble Sampling. This design packet transmission must be larger than the interval between
is partially pushed by hardware advance toward packetizing two instants of polling, switching time affects the size of a
radios. Instead of transmitting a raw bit stream, a packetizing short preamble packet.
radio takes the input as the payload of a packet and inserts its In addition to dual channel sampling, CMAC [23] also
own preamble, header information, and CRC. When a packet introduces anycast to reduce per-hop latency. Instead of let-
is received, the radio passes the payload of the packet to the ting nontarget nodes go back to sleep to reduce overhead
microprocessor if the packet passes the CRC check. While this of overhearing, CMAC allows them to reply if they can
radio technology reduces the burden on the microprocessor, it make progress (defined by geographical distance) towards the
introduces a challenge to LPL protocols that assume a wake- destination. To alleviate collisions among multiple candidates,
up preamble of a continuous bit stream. The reason is that the a contention-based mechanism is adopted to let the best one
TX FIFO is limited in low cost radios and a long preamble can reply first. This results in a slightly larger gap between two
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 HUANG et al.: THE EVOLUTION OF MAC PROTOCOLS IN WIRELESS SENSOR NETWORKS: A SURVEY 105

Basic LPL Tx Long Preamble Data

t
Continuous Preamble Data
Rx
Preamble
Sampling t

Packetization Tx P P P P P P P P P P Data
t
P Data
Rx
t
Tp Tgap A
Packetization C
Tx P P P K Data
Early ACK
t
Strobed A
Preamble C
Rx P K Data
Sampling
Ts Tgap A t
Packetization
Early ACK C
Rx P K Data
Dual preamble sampling
t
Tp + Tgap

Transmit Receive Active Tp - Ts

Fig. 3. Evolution of preamble sampling.

short packets. As shown in Fig. 3, if the channel is detected as The shortened preamble alleviates the overhearing problem
busy in the second polling, a node can nap for at most Tgap and increases the channel capacity. Some protocols utilize
if it does not want to miss the next short preamble packet a similar idea with some improvements. For example, in
for address decoding. The maximum idle listening period is SyncWUF [17], if the receiver’s schedule is known and the
(Tgap + Tp ) − (Ts + Tgap ) = Tp − Ts where Tgap is the time computed preamble length is short enough, a simple wake-up
period of the gap, Tp is the time period of a short preamble tone is used same as in WiseMAC. However, if the receiver’s
packet, and Ts is the channel sampling period. Since Tp must schedule is unknown or out-of-state or it is a broadcast
be greater than Tgap , the larger gap used in CMAC introduces message, a long preamble is divided into chunks that contain
slightly higher idle listening overhead. In addition, the anycast timing information to reduce unnecessary waiting time. In
incurs higher overhead compared with unicast due to collisions CSMA-MPS [18], strobed preamble sampling is combined
among multiple candidates. Therefore, CMAC switches a node with schedule learning. Because CSMA-MPS allows for early
to unicast if the best next hop is found or no better next hop ACK, a node may not need to finish a preamble of the
can be found after a duty cycle length. worst case length. Although schedule learning squeezes more
room for data transmission, the cost of maintaining up-to-
date schedules of neighbors pushes some studies to shift the
E. Reduce Preamble Length by Schedule Learning responsibility of establishing communication from the sender
Although strobed preamble sampling reduces the preamble side to the receiver side.
length on average, the preamble transmission still occupies
the channel until the target receiver wakes up. The preamble
transmission prevents neighboring nodes from transmission, F. Improve Throughput with Receiver Initiated Transmission
leading to low channel capacity. In addition, transmitters waste Fig. 4 illustrates the disadvantage of preamble transmission.
energy in preamble transmission and nontarget nodes are When there are contending flows (e.g., A sends to B and C
awakened unnecessarily, causing energy deficiency at both the sends to D), the preamble transmission prevents neighboring
sender and the receiver. nodes from delivering their packets. When nodes C and D
For an infrastructure network, WiseMAC [25] lets access wake up, they go back to sleep immediately after they find out
points stay awake to learn sensor nodes’ wake-up schedules. that they are not the target receiver of current transmission.
Access points are not power constrained and each access point This leads to low channel capacity and high delay when there
serves a number of sensor nodes. If the access point does are contending flows.
not know the channel sampling schedule of a sensor node, To avoid occupying the channel before a pair of nodes
it sends a long preamble. When a node receives a packet, are ready to exchange data, a number of protocols muffle
it include its channel sampling schedule in the ACK. With the sender and rely on the target receiver’s beacon message
the up-to-date channel sampling schedules of sensor nodes, to initiate data transmission. As shown in Fig. 5, a node
the access point only needs to start preamble transmission broadcasts a beacon to announce that it is ready for receiving
slightly before the target receiver wakes up. The duration of when it wakes up. Because a node stays awake if it has
the preamble is computed to guarantee that the senor node data to send, it can hear the target receiver’s beacon. Upon
will not miss the preamble at a worst case clock drift offset. receiving the beacon from the target receiver, the sender starts
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 106 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

A B B DATA B
C A t
P P P P P P K DATA
A t
A B DATA B
C B t
P K DATA
B t B: beacon Active

P Transmit Receive
C t

Fig. 5. In low power probing, a sender does not occupy the wireless medium
P
D t until the receiver is ready to receive.

P: short preamble with Active

target address the beacon message, a node is able to quickly determine
whether there is a packet for it.
Transmit Receive

Fig. 4. In strobed preamble sampling, short preambles occupy the wireless

G. Reduce Collision in Receiver Initiated Transmission
medium until the data packet is delivered. Since senders are required to wake up before receivers,
RC-MAC [31] coordinates multiple senders’ transmissions by
piggybacking a scheduling message, which specifies the node
to transmit data packets to the receiver. Multiple nodes may that can transmit upon receiving the message, to an ACK.
have data for the receiver. An efficient collision resolving The method relies on that a data gathering tree exists or is
mechanism is thus needed. If a node does not receive anything formed during route discovery. On the tree structure, a parent
for a certain amount of time after it broadcasts a beacon, it receives packets from multiple children and forwards packets
goes back to sleep. The advantage of the receiver-initiated to its own parent. A parent acknowledges its children with an
Low Power Probing (LPP) design is that the channel is vacant ACK message that indicates which child is the next sender.
for use before the target receiver is ready to receive. The Therefore, the scheduled child can transmit immediately to
idea of shifting communication initiation from the sender side reduce latency and sibling nodes will back off to avoid
to the receiver side is early presented in RICER [19] and collisions. By learning the bandwidth demands of children, a
employed in PTIP [26] for infrastructure networks, Koala parent adjusts the channel access opportunities in accordance
[27], AS-MAC [28], RI-MAC [29], and A-MAC [30] for with their demands so that fairness is ensured. RC-MAC
general sensor networks. When a node broadcasts a beacon, also explicitly interrupts the scheduling of a parent-children
more than one of its neighbors may respond with a data packet. unit so as to ensure that no unit can occupy the channel
RI-MAC resolves collisions by reusing the beacon messages exclusively. The scheduling round of a unit is dynamically
and allows back-to-back data transmission. adjusted according to the parent’s remaining buffer size, which
Considering that the traffic is light most of the time in a balances the channel access opportunities among units.
sensor network, RI-MAC lets a node broadcast a base beacon The local scheduling method reduces collisions and in-
message with no backoff field when it wakes up. As a result, creases throughput in a basic parent-children unit. The con-
neighboring nodes can start data transmission to the node tention among units, however, significantly affects the per-
immediately. This is optimized for light traffic load scenarios formance. RC-MAC relies on explicit back-off to reduce
where there is only one sender most of the time. Under heavy interruption of scheduling caused by contention among units.
traffic load, collisions may occur and the receiver broadcasts Coexistence of units in a dense network, however, still remains
another beacon message that includes a backoff window field. a key issue. Multichannel design is a promising solution for
Neighboring nodes will back off randomly according to the addressing interference among units.
random backoff window size indicated in the new beacon
message. Whenever a collision is detected after a beacon, H. Estimate the Wake-up Time of Receiver
the receiver increases the backoff window size in a Binary Although LPP increases channel utilization, only the re-
Exponential Backoff (BEB) way until the maximum window ceiver benefits from the receiver-initiated design. The sender
size is reached. Because neighboring nodes’ contentions are has to stay awake until the data packets are delivered. To
synchronized by each beacon, collision is still the major reduce energy consumption of senders, PW-MAC [32] intro-
factor that bounds the throughput. In addition, overhearing a duces a method to predict the target receiver’s wake-up time
transmission may incorrectly trigger the recovery mechanism, so that a sender only needs to wake up slightly before the
leading to more redundant beacons and then more collisions. target receiver. The motivation is same as that of WiseMAC
To reduce collisions, RC-MAC [31] utilizes the tree structure [25], but PW-MAC allows nodes to independently generate
formed during data gathering to schedule the transmission pseudo-random schedules, which avoids persistent collisions
of multiple nodes. Another improvement upon RI-MAC is caused by two similar wake-up schedules. The collision can
to devise a prediction mechanism for estimating neighbors’ also be avoided by distributing nodes’ schedules as introduced
wake-up time as introduced in PW-MAC [32]. Recently, A- in AS-MAC [28], but PW-MAC allows to use variable wake-
MAC [30] shows that by enabling auto-acknowledgement to up intervals.
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 HUANG et al.: THE EVOLUTION OF MAC PROTOCOLS IN WIRELESS SENSOR NETWORKS: A SURVEY 107

Listen (Active) Sleep

Since a pseudo-random function generates the same se-
quence of pseudo-random numbers for the same parameters, SYNC DATA SLEEP

PW-MAC utilizes the beacon message used in RI-MAC to A CW RTS CTS DATA ACK

relay these parameters. By learning neighbors’ pseudo-random SIFS CW

B RTS CTS DATA ACK RTS CTS
function parameters, a node can easily calculate the future
wake-up time of any neighboring node. Clock drift, hardware CTS RTS CTS
C
and operating system latency may cause prediction errors. wake up
A sender in PW-MAC thus advances its wake-up by a time D
period that is adjusted according to the clock drift rate. In
addition, if a sender finds that the prediction error (defined Transmit Receive Overhear
as the difference between the estimated wake-up time of the
CW: contention window RTS: Request-to-Send CTS: Clear-to-Send
target receiver and the actual wake-up time of the receiver)
is greater than a threshold, the sender requests an update of Fig. 6. The adaptive listening in S-MAC.
the prediction state. This ensures that the prediction error is
within the sender wake-up advance time so that a sender will
not miss the receiver’s beacon. III. S YNCHRONOUS MAC PROTOCOLS
Currently, the pseudo-random function parameters are fixed In this section we summarize the recent developments in
for each node. If these parameters need to be adjusted accord- synchronous MAC protocols. Synchronizing active time of
ing to some optimization mechanisms, maintaining informa- neighboring nodes is a natural solution to establish commu-
tion consistence among neighbors would be a problem. In ad- nication between two nodes. However, it brings the cost of
dition, constant calculation of neighbors’ schedules introduces additional synchronization overhead.
unnecessary computing overhead, which causes additional In synchronous MAC protocols, a node listens to the
energy consumption and introduces delay for other system channel for a certain amount of time. If it does not hear
operations. A periodic wake-up schedule in AS-MAC [28] is any schedule from other nodes, it determines its next wake-
easier to follow but nodes’ schedules must be distributed to up time and broadcasts its schedule. This makes the node
avoid constant collisions. become a synchronizer. If a node receives a schedule from
a neighbor before choosing its own schedule, it follows the
received schedule, which makes it a follower. Generally, nodes
I. Summary form clusters of several hops. All nodes in a cluster are
In this section, we have surveyed various ways to establish synchronized to a synchronizer that is one or a few hops
communication between two nodes in asynchronous MAC away. If a node receives a different schedule after it sets its
protocols. The research starts with using long preamble to own schedule (i.e., either has broadcasted its own schedule
wake nodes up. The preamble sampling method is first im- or followed a previously received schedule), it adopts both so
proved with a better outlier detection based CCA algorithm that it can be a bridge between two clusters. In other words,
and has been suggested to decouple data transmission from the node wakes up at times of both its neighboring cluster and
preamble sampling. Soon four approaches have been pro- its own cluster.
posed to reduce cost at receivers and senders as shown in As schedule learning is introduced to asynchronous MAC
Fig. 1. The first solution is to include useful information protocols, the boundary between synchronous MAC and asyn-
in preamble. This addresses the overhearing problem but chronous MAC gradually becomes fuzzy. A main differ-
still inherits the drawback of consuming channel bandwidth ence is that only senders wake up at the target receiver’s
for preamble transmission. The other three methods include polling/probing time in prediction integrated asynchronous
schedule learning, synchronized polling, and receiver-initiated MAC protocols while a cluster of nodes wake up at the same
LPP. All of the three methods reduce the amount of time time in synchronous MAC protocols. The common active
a pair of nodes occupy the wireless medium before they period is not reserved for a single receiver. It is shared by all
actually exchange data. In the following section we will nodes in the cluster and they must contend for channel access.
discuss the synchronized polling, which shares similarity with In other words, synchronous MAC protocols require local time
schedule learning in that nodes are synchronized to a common synchronization while a node in asynchronous MAC protocols
schedule. However, contention in synchronized polling is more chooses its schedule independently without awareness of its
severe than that in schedule learning because all nodes that neighbors’ schedules.
are synchronized to a common schedule share the wireless Synchronous MAC protocols do not face the problem of
medium in synchronized polling whereas only several senders establishing communication between nodes as asynchronous
are synchronized to the target receiver’s schedule in schedule MAC protocols. Therefore, most designs focus on improving
learning. A good combination in asynchronous MAC protocols throughput and reducing delay. They are more appropriate for
would be receiver-initiated LPP with auto-ACK to the beacon applications of periodic traffic where the wake-up schedule is
and distributed wake-up schedule, but how to ensure that the easy to determine.
beacon will not be missed while minimizing the idle listening
is platform dependent. Once nodes can reach each other A. Adaptive Listening
efficiently, more efforts can be put on throughput improvement A classical synchronous MAC protocol is S-MAC [33],
and delay reduction. where nodes are organized into clusters by local time synchro-
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 108 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

contend
nization. Each cluster has an independent schedule composed RTS CTS DS DATA ACK
A
of three periods: SYNC, DATA, and SLEEP. All nodes of
the same cluster wake up at the beginning of the SYNC B RTS CTS DATA ACK RTS
period to synchronize clocks with each other. Then, nodes
with packets to send contend for exchange of Request-to- C contend CTS FRTS contend RTS CTS
Send (RTS) and Clear-to-Send (CTS) frames in the DATA
period. Nodes that are not involved in communication, return D active FRTS active RTS CTS
to sleep at the start of the SLEEP period; other nodes return TA
to sleep after they finish transmission of data packets and Collision FRTS: Future-Request-to-Send DS: Data-Send
acknowledgement (ACK) frames. It is noticeable that in each TA: the minimal amount of idle listening before going back to sleep
cycle a packet can only be forwarded by one hop. Although
adaptive listening [34] is later introduced into S-MAC to Fig. 7. The FRTS mechanism in T-MAC.
overcome this deficiency, the improvement is limited. A packet
can only be forwarded by at most 2 hops per cycle.
S-MAC with adaptive listening [34] is illustrated in Fig. 6. C. Shifting Data Transmission to the Sleep Period
Nodes can only hear their immediate neighbors (e.g., node Inspired by the FRTS mechanism, RMAC [36] presents a
B can only hear node A and node C). Because node C can novel approach to reduce latency in multi-hop forwarding.
overhear the CTS sent by node B, it goes back to sleep at the Same as S-MAC, time is divided into repeated cycles and
beginning of the SLEEP period but wakes up at the end of each is further divided into three periods: SYNC, DATA, and
the current transmission. Node B can therefore immediately SLEEP. Instead of exchanging data during the DATA period,
forward the data packet to node C instead of waiting for the a control frame called Pioneer Frame (PION) is forwarded by
next cycle. The cost is that not only the target receiver but multiple hops. The control frame informs nodes on a routing
all nodes that overhear the RTS and the CTS will wake up. path of when to wake up in the SLEEP period. A PION not
The benefit is that S-MAC with adaptive listening can relay only serves as a RTS frame to request communication, but also
a packet by 2 hops per cycle. However, the delivery cannot confirms a request like a CTS frame. RMAC and DW-MAC
go beyond this because the next hop (i.e., node D) will not [37] are similar as shown in Fig. 8. The differences are that
wake up since it cannot overhear the CTS from its two-hop the Scheduling Frame (SCH) in DW-MAC is called PION in
neighbor (i.e., node B) and does not know when the current RMAC, and the time T1s is 0 in RMAC, which implies that
transmission will end. The long end-to-end delay makes S- the first hop starts data transmission when the SLEEP period
MAC inappropriate for applications that require strict timing starts. The drawback of RMAC is that two hidden terminals
constraints. may always cause collisions of data transmission although they
succeed in scheduling through PIONs. For example, node C in
B. Future Request-to-Send
Fig. 8 may initiate a PION later than node A. Although they
Instead of fixing the length of the active period, T-MAC can succeed in PION transmission, they start data transmission
[35] makes a node stay awake until no activation event has at the same time when the SLEEP period starts as they are both
occurred for a certain amount of time. This design aims to the first hop. They will collide with each other and downstream
achieve optimal active periods under various traffic loads. nodes will wake up unnecessarily if the expected data packets
However, overhearing is introduced because a node has to stay cannot arrive due to collisions at previous hops.
awake while it is not involved in data transmission. T-MAC DW-MAC [37] thus introduces a one-to-one mapping func-
justifies overhearing with improvement on the throughput. tion to this design to ensure collision-free data transmission in
Another contribution of T-MAC is the idea of Future the SLEEP period. In the example illustrated in Fig. 8, node
Request-to-Send (FRTS) packet, which allows a node to notify B calculates its wake up time T1S as
its target receiver that it cannot access the medium at the
current time. The mechanism has the potential to extend the T1D T1S
=
data delivery to up to 3 hops per cycle. As shown in Fig. 7, TDAT A TSLEEP
when node C loses contention and overhears a CTS packet, TSLEEP
⇒ T1S = T1D ·
it sends a FRTS packet to its target receiver D. The FRTS TDAT A
packet contains the length of the current data transmission
In addition,
from node A to B, hence the target receiver D can learn its
T3D T3S
wake-up time. However, nodes that are further downstream = (1)
cannot overhear the FRTS and thus will not wake up. The TDAT A TSLEEP
mechanism thus cannot be performed iteratively and a packet Equation 1 implies that if node A can receive a confirmation
can only be forwarded by at most 3 hops per cycle. During SCH from node B, the time length of T3S is assured to be
the time it postpones its data transmission for the FRTS, node collision-free because node B can receive a SCH correctly
A must transmit a Data-Send (DS) packet of the same size during T3D . They are one-to-one scaled based on the ratio be-
of FRTS to prevent any neighboring node from taking the tween TSLEEP and TDAT A . Therefore, successfully reserved
channel. The collision of FRTS and DS at node B is not a data transmission will never collide at a target receiver in
problem because they are useless for node B. The potential the SLEEP period. If the SCH sent to node B is collided,
collision of multiple FRTS packets, however, is not discussed. node B will not respond a confirmation SCH and thus node
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 HUANG et al.: THE EVOLUTION OF MAC PROTOCOLS IN WIRELESS SENSOR NETWORKS: A SURVEY 109

T2D T2S
wake up
T1D T3D T1S T3S

CW SCH SCH data ACK

A
DIFS SIFS SIFS
SCH SCH SCH data ACK data ACK
B

SCH SCH data ACK data

C
SYNC DATA SLEEP
D D D
T1 T T T DATA
S
= 2
S
= 3
S
= Transmit Receive
T1 T 2 T3 T SLEEP
CW: contention window DIFS: DCF Interframe Space SIFS: Short Interframe Space

Fig. 8. Multi-hop forwarding of a unicast packet in DW-MAC.

A and B will not reserve a time for data transmission during another RECV slot 3μ later than the current RECV slot to
T3S . However, if the confirmation SCH is collided at node A, check whether node A has data to send to it. The length of 3μ
node A will not wake up to send during T3S but node B will is selected to ensure that the previous packet has be forwarded
wake up to receive. The difficulty is that the sender cannot 3 hops away. In addition, node D also needs to send a MTS
distinguish the collision of a request SCH and the collision of packet to node E to make node E wake up periodically. This
a confirmation SCH. It can only conservatively regard that the ensures that when node D wins channel access, node E is
reservation failed. Consequently, all downstream nodes will ready to receive. The overhead is increased along with the
wake up unnecessarily to receive the expected data packet traffic load because nodes on routing paths have to wake up
that will not arrive due to the false alarm. repeatedly for contention of sending or receiving.
The cost of waking up unnecessarily may be tolerable, but Similarly, Q-MAC [39] also defines staggered schedule but
any failure of a sequential SCH reservation is a waste of time the active periods are shifted in a way that facilitates downlink
and it significantly limits the number of packets that can be traffic, carrying queries from the sink to sources. The MAC
delivered in a cycle. A sequence of time slots are reserved is designed for query based sensor networks where the sink
for a data packet but no data will be transmitted, whereas initiates data gathering by sending a query to the source. For
many other data packets have no time to reserve time slots for uplink traffic, Q-MAC has two options. The simplest way is to
transmission because the DATA period is limited. In a complex let each node remain active from the instant of query reception
multiple flow scenario, the performance of throughput, delay, to the time the queried data are forwarded. If the sink knows
and power consumption is significantly compromised due to the route length in advance, an uplink staggered schedule can
the incorrect channel reservation. be constructed. Although Q-MAC reduces latency, it incurs
long idle listening if the route length is unknown. Low-power
wireless communication is vulnerable to interference and thus
D. Staggered Schedule
routes in WSNs are usually not fixed. Therefore, the energy
For specific applications, some underlying structures can efficiency of Q-MAC is questionable when the route length is
be utilized. DMAC [38] targets at data gathering applications hard to determine.
where data are delivered from multiple sources to a sink.
Based on the naturally formed data gathering tree, the ac-
tive/sleep schedules of nodes are staggered so that packets E. Adaptive Duty Cycle
can flow continuously toward the sink. The basic preamble sampling method requires a long pream-
As shown in Fig. 9, each node skews its wake-up time dμ ble before data, following the design of synchronous MAC
ahead of the sink’s schedule in accordance with its depth d on protocols, SCP-MAC [40] synchronizes neighbors’ channel
the data gathering tree. The variable μ represents the length polling so that only a short preamble is required to wake
of the time that is needed for one packet transmission and the receiver up. Because a cluster of nodes share the same
reception. The ordered offsets of schedules ensure sequential schedule, contention is more severe compared with schedule
transmission and lead to low delay. However, because nodes learning in asynchronous MAC protocols. However, SCP-
of the same depth have the same offset, they contend for MAC incurs less schedule maintenance overhead as nodes do
sending to their receivers. When one node wins channel not need to maintain multiple schedules.
access, its neighboring nodes of the same level lose their To reduce multi-hop latency, SCP-MAC develops an adap-
chance of transmission. In order to increase the number of tive channel polling mechanism to add additional polling slots
active slots, a data prediction mechanism and a More-to-Send along the path. As shown in Fig. 10, when node B receives a
(MTS) notification are introduced. In Fig. 9, if node B wins packet at the first regular polling slot, it adds n high frequency
channel access opportunity to send to C, node C needs to add polling slots to receive additional data packets from node A. If
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 110 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

3µ Recv sleep Recv

C E
Recv Send sleep Recv Send
D
A
B Recv Send sleep Recv Send

Recv Send sleep Recv Send

Fig. 9. DMAC in a data gathering tree.

any of the additional polling slots was successfully used, node in each cluster. However, if the active periods of two clusters
B extends the adaptive polling with another n slots; otherwise, overlap, the collision-free data transmission is no longer
it returns to the regular polling frequency. In order to quickly guaranteed. Therefore, although local time synchronization is
shift nodes on a routing path to adaptive polling, node A more scalable, TDMA is usually defined with slightly more
intentionally pauses its transmission at the second regular strict global time synchronization. Although TDMA requires
polling slot, allowing node B to send to C, but nodes A and global time synchronization, the adoption of frame-slotted
B have to contend for sending in the following adaptive slots. schemes will not introduce additional overhead in applications
The procedure repeats so that a new node can enter the high that require time synchronization to provide spatial-temporal
frequency polling at each regular polling slot. For adaptive correlation between reports generated by multiple sensors. For
polling slots, n is set to 3 considering that each node contends example, in the SENSEI project [2] developed by European
with its previous and next hop. However, it is possible that FP7-ICT, heterogeneous wireless sensor and actuator networks
a node loses contention for all of the 3 slots and returns have to be integrated into a common framework. In order to
to the regular low frequency polling, which will interrupt obtain information from various sensors located in the physical
the continuous forwarding. In addition, inter-flow contentions world, the spatial and temporal results from the sensors should
may also degrade the adaptive polling performance. DSMAC be synchronized. Frame-slotted MAC protocols would be a
[41] is another WSN MAC protocol that changes duty cycle promising candidate for these kinds of applications.
dynamically. A node doubles its duty cycle when it detects Frame-slotted MAC protocols are also popular in small
the increase of its neighbors’ bandwidth demands and the new scale networks. A special case of WSN is the wireless body
duty cycle is broadcasted along with the SYNC packet. These area network (WBAN). A WBAN consists of a number of sen-
methods reduce delay when the traffic load increases. sors that are either connected with a person’s body or are small
enough to be implanted. Sensors transmit data at a relatively
F. Summary wide range of data rates from 1 Kbps to 1Mbps for body tem-
perature, electrocardiography (ECG, heart), electromyography
In this section, we reviewed the efforts that have been
(EMG, muscular contractions), electroencephalography (EEG,
made to reduce multi-hop forwarding delay in synchronous
brain), movement, and etc. The WBANs have relatively con-
MAC protocols. With the help of data gathering tree structure,
stant network structures and fixed sensor functions. Therefore,
schedules can be staggered to create a continuous flow to
many recently proposed MAC protocols are TDMA-based
the sink. However, nodes wake up unnecessarily if their
[42] [43] [44] [45]. In these networks the synchronization
upstream nodes lose contention in channel access. Note that
procedure can be simplified due to the hierarchical structure.
multiple flows converge at the sink and thus the contention
Master nodes that have more power act as coordinators and
becomes more and more severe near the sink. Without the
this removes the need of idle listening for other nodes [46].
traffic pattern constraint, most protocols try to notify as many
nodes as possible on active routes of incoming data as soon
as possible. One trend is to shift data transmission from the A. Increase Channel Utilization by Slot Stealing
DATA period to the SLEEP period and use the DATA period A major drawback of TDMA is its low channel utilization
for notification and scheduling. While this paradigm mini- when only few nodes have data to send because a node can
mizes delivery latency for individual packets, some packets transmit only in its assigned time slots. A sensor network,
may experience long delay as they cannot be scheduled for however, generates no or few data packets most of the time. To
transmission within limited DATA period. Therefore, more improve channel utilization of TDMA under low contention,
efficient scheduling methods are needed. Without a subtle Z-MAC [47] incorporates Carrier Sense Multiple Access
scheduling design, the delay can be reduced by dynamically (CSMA) into TDMA.
increasing the duty cycles of nodes on the active routing paths. When a WSN is deployed, Z-MAC applies DRAND [48] to
The performance improvement, however, comes with the cost do time slot assignment. DRAND ensures that no two nodes
of increased energy consumption. within the two-hop communication neighborhood are assigned
to the same slot. To further increase the channel utilization,
IV. F RAME - SLOTTED MAC PROTOCOLS a time frame rule is designed to adapt the local frame size
Frame-slotted schemes derive from Time Division Multiple to a node’s local neighborhood size. To combine CSMA with
Access (TDMA). The advantage of TDMA is that the through- TDMA, a node is allowed to contend for sending if a slot is not
put is high with maximized channel utilization under high con- used by the owner. Consequently, Z-MAC performs as CSMA
tention. TDMA can be used in synchronous MAC protocols under low contention and possesses high channel utilization
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 HUANG et al.: THE EVOLUTION OF MAC PROTOCOLS IN WIRELESS SENSOR NETWORKS: A SURVEY 111

A B C D
n adaptive dynamic extension
1st regular 2nd regular
channel
polling

A→B A→B A→B

packet or or
TX A→B A→B A→B B→C B→C B→C

Fig. 10. Adaptive channel polling in SCP-MAC.

slot. Although the time slot assignment incurs no communi-

cation overhead, the spatial reuse of time slots is low because
nodes’ priorities may be sequential. As shown in Fig. 11, node
2 2 is the node with the highest priority among node 0’s two-
1 3
0 132 hop neighborhood for a particular time slot, and node 3 is the
4 51 winner of the slot in node 1’s view. Consequently, only node
70 103
24 2 can occupy the slot, leading to low spatial reuse of slots.
TRAMA also devises an adaptive election algorithm to
reuse slots that are discarded by the owners. When a winner
of a slot does not have data to send, it scans its one-hop
Fig. 11. Low channel utilization problem of TRAMA. The number under
each node is its priority for a particular time slot t.
neighbor set. If a one-hop neighbor has the highest priority
among its own two-hop neighbors, the one-hop neighbor is
as TDMA under high contention. However, the hybrid design added in the Possible Transmitter Set (PTS). Note that the
faces challenges in a dense network. When some nodes have node does not know the entire two-hop neighborhood of the
data to send, they have to contend for slots that are assigned one-hop neighbor and thus collision-free data transmission
to their neighbors who have no traffic and these contentions cannot be guaranteed. Finally, the node of the highest priority
are synchronized in each slot. In every slot, a sender has to in the PTS is authorized to send. The procedure introduces
wait for a certain amount of time to ensure that the slot is additional computing and control overhead, which sacrifices
abandoned by the owner. Each receiver also has to stay awake energy efficiency and introduces system delay. In addition,
to check whether it is the target receiver. As a result, the packets experience long delay as downstream nodes do not
slot stealing method introduces nontrivial additional energy know that they will have data to send and thus will not reserve
consumption. time slots for data transmission.
Another work that studies the slot stealing problem is AI-LMAC [52] adaptively assigns time slots to nodes when
TDMA-ASAP [49]. It utilizes the data gathering tree structure the sink initiates a query for data. AI-LMAC assumes that a
to reduce collisions in slot stealing. The advantage of TDMA- parent-child relationship exists between all the nodes in the
ASAP is that when a node abandons a slot, only nodes that network and the root is the highest parent in the hierarchy.
have data for the same parent can steal the slot and thus other Since the parent knows the proportion of data that will be
nodes do not need to wake up to check whether they have contributed by each of its immediate children, it sends a
data to receive. However, the overhead for detecting whether message to each of its children to advise the ideal number
a slot is abandoned still exists. of slots that they should take up. A node then searches for
free slots within its two-hop communication neighborhood and
takes up any available slot to meet its bandwidth demand.
B. Increase Channel Utilization by Adaptive Assignment However, if the data rates of sources are not fixed, the sink
TRAMA [50] increases channel utilization of TDMA in cannot accurately estimate the number of slots that each of its
another way. It switches between random access period and children should take up. In addition, there is no mechanism to
scheduled access period. In the random access period, nodes ensure that a node can grab enough slots to meet the desired
that have data to send will claim slots for use. The protocol bandwidth demand.
avoids assigning time slots to nodes of no data to send and
thus increases the channel utilization from a new perspective. C. Maximize Throughput at the Sink
To determine the ownership of a slot, it adopts the The time slot assignment of most TDMA protocols is based
neighborhood-aware contention resolution (NCR) algorithm of on graph coloring, which aims at maximizing spatial reuse of
NAMA [51], which defines a node u’s priority at time slot t slots. For data gathering applications, multiple flows converge
as a hash of the concatenation of u and t: at the sink. Fair assignment of slots among nodes cannot result
priority(u, t) = hash(u ⊕ t). in throughput maximization at the sink. Therefore, TreeMAC
[53] devises a time slot assignment algorithm that is well tuned
For any particular time slot, the node with the highest priority for throughput maximization at the sink by utilizing the data
within the two-hop communication neighborhood wins the gathering tree structure.
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 112 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

sink unicast slots. Therefore, the receiver of a slot also contends

with nodes that intend to send data to it. To reduce contention,
0 0 a node that loses contention in current frame retry transmission
with a probability of 70% in the next frame.
The time slot assignment can be more elaborate as in
1 1 1 PMAC [55], which takes traffic load into consideration. In
PMAC, a node is encouraged to increase its sleep time
exponentially until the upper bound is reached if it does not
2 2 have data to send. The sleep/active schedule of a node is
represented by a pattern consists of n-sleep/1-awake slots.
Nodes’ patterns are exchanged in the end of each frame. The
0 period is set long enough such that all nodes can broadcast
their patterns. Based on received neighbors’ patterns, a node
Fig. 12. In TreeMAC, nodes of the same depth on the tree get the same can determine whether the target receiver will wake up at
transmission slot in different frames. The number above each node denotes the a particular time slot or not. The adaptivity to traffic load,
node’s transmission slot, and the shaded squares illustrate the frames assigned
to the node. however, comes with the overhead of exchanging patterns
every time frame. In addition, contention is severe as multiple
The basic idea of TreeMac is to eliminate horizontal two- receivers exist in a time slot. A node in PMAC reverts to a
hop interference by frame assignment and vertical two-hop pattern of repeated ‘1’ once it has data to send. This design
interference by slot assignment. Time is divided into cycles enables a node to quickly respond to a new flow, but it also
and each cycle is divided into N frames of 3 slots. As shown increases energy consumption unnecessarily if only several
in Fig. 12, nodes of the same depth on the tree are assigned packets need to be delivered. The node has to wake up at
different frames for transmission. In a frame, the transmission every slot in the next frame as its pattern is reset to repeated
slot is calculated as (u − 1) mod 3 where u is the depth of ‘1’.
the node. Using three slots, a node can avoid contention with
its previous and next hop. Moreover, the number of frames
E. Summary
assigned to a node is proportional to its traffic load. Therefore,
fairness is ensured in terms of flow instead of individual nodes. In this section, we have introduced two hybrid designs
Since a node may not have data to send, some slots are wasted. of CSMA and TDMA for improving channel utilization:
In addition, a node needs time to join the tree as it adopts slot stealing and adaptive time slot assignment. Slot stealing
CSMA and has a lower channel access priority than those introduces additional overhead for detecting whether a slot
already on the tree. The protocol thus prefers a relatively stable is abandoned. In addition, each node must wake up in every
tree structure. slot to check whether there is a packet for it. This design
sacrifices energy efficiency and deviates away from the goal
of low power operation in WSNs. Adaptive slot assignment
D. Reduce Duty Cycle by Switching Sending Slots to Receiv-
incurs less overhead but also yields lower spatial reuse of
ing Slots slots. A better distributed slot assignment algorithm is highly
TDMA schemes specify time slots that a node can transmit. demanded. The target of time slot assignment can be quite
If a node receives data from multiple nodes, it has to wake diverse as we have seen. Instead of maximizing the channel
up multiple times no matter whether there is a packet for utilization, we can try to maximize the throughput at the sink.
it. Some protocols thus use the frame-slotted structure for In addition, time slots can be assigned for receiving instead of
specifying time slots in which a node should wake up for sending. The benefit is that if a node has no data for the owner
listening. The benefit is that only nodes that have data for of the slot, it does not need to wake up and hence saves energy.
the owner of the slot need to wake up. Since time slots are However, the challenges in this type of MAC protocols are
assigned to receivers, collision-free data transmission is no same as prediction integrated asynchronous MAC protocols,
longer guaranteed. This type of MAC protocols adopt frame- which include collision resolving and deterministic delay
slotted structure of TDMA but are functionally more similar guarantee.
to prediction integrated asynchronous MAC protocols (e.g.,
WiseMAC [25], AS-MAC [28], and PW-MAC [32]).
V. M ULTICHANNEL MAC PROTOCOLS
In Crankshaft [54], time is divided into frames and each
frame is further divided into slots. Each node listens for one Recently parallel data transmission also attracts intense
unicast slot in every frame. A key feature of Crankshaft is attention as many WSN platforms emerged with multichannel
that several broadcast slots are added to the end of unicast support. The radio bandwidth in WSNs is limited and thus it
slots. Nodes can contend for broadcast in the broadcast slots. is desirable to devise multichannel MAC protocols to handle
Crankshaft assigns unicast slots to nodes based on node ID bursty traffic or provide multi-task support. The next genera-
modulo frame size. The simple assignment saves effort for tion of networks highlighted by the SENSEI [2] project focus
learning schedules of neighbors, but two nodes may own on efficient utilization of spectrum and integrating various
the same unicast slot. To make the two nodes be able to networks into an advanced global network. Multichannel WSN
communicate, nodes are allowed to send in their assigned MAC protocols are helpful in this field.
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 HUANG et al.: THE EVOLUTION OF MAC PROTOCOLS IN WIRELESS SENSOR NETWORKS: A SURVEY 113

In multichannel MAC protocol designs, two questions need can be modeled as finding a minimum K-way cut that
to be answered: how to allocate channels and how to achieve minimizes inter-channel communication in the graph given by
cross-channel communication. The solution must also consider the network topology. To reduce the complexity of solving the
energy efficiency and cost. problem, a heuristic method is introduced in [60]. The idea
is to move some nodes to another channel when a channel
A. Address Cross-channel Communication becomes overloaded.
The first step is to define a metric that measures the effect of
MMSN [56] points out that multichannel MAC protocols
a crowded channel. All channels are organized as a ladder and
designed for general wireless networks are not suitable for
all nodes start at the lowest channel. Each node periodically
WSNs because of two reasons. First, they use some more
broadcasts the total number of times it successfully and
powerful radios such as radios that have multichannel sensing
unsuccessfully acquires the channel, s and f respectively.
ability. Second, the control overhead of channel negotiation
Node i estimates the probability that anyof its neighbors
 can
is acceptable considering that the data packet size is much
successfully access the channel as αi = ( j sj )/ j (sj +fj ).
larger. As an example, MMAC [57] assumes that time is
If the value is less than a threshold, node i considers a switch
synchronized and divided into fixed-length beacon intervals.
from its current channel to the next higher channel with a
Each beacon interval is further divided into an ATIM (Ad Hoc
probability that depends on the channel quality difference and
Traffic Indication Messages) window and a communication
how closely a node resembles a sink. A node that acts like
window. Nodes negotiate which channel to use for data
a sink has a low-cost outgoing link because it sends less
transmission on the same default channel in the ATIM window,
traffic and this makes it more appropriate to cut. The method
and then switch to the reserved channels for communication in
assumes that some intermediate nodes can do aggregation so
the communication window. Later, TMMAC [58] introduces
that they send less than they receive; otherwise the probability
dynamic ATIM window adjustment rules to increase the
of initiating channel switch will be zero for all intermediate
flexibility of the design. Because data packet size is usually
nodes. When a node switches, nodes that communicate heavily
small in WSNs, at most 128 bytes for CC2420 radio [59],
with it also switch. If a transmitter wants to send data to a
pairwise channel negotiation for each packet introduces high
receiver on a different channel, the transmitter simply switches
control overhead. Authors of MMSN thus proposed to assign
to the receiver’s channel. As a result, the transmitter may miss
channels to nodes in a static way. The idea is to let a node
some packets destined for it.
learn its neighbors’ channel choices and then select a channel
Similarly, TMCP [61] partitions a sensor network to K
that is not chosen by any of its neighbors within the two-hop
vertex-disjoint trees, each of which is assigned a channel.
communication neighborhood. The idea is similar to the time
Different from K-way cut problem that aims at minimizing
slot assignment for receivers, thereby the channel assignment
inter-channel communication, TMCP tries to minimize the
cannot guarantee collision-free data transmission.
maximum intra-tree interference value among all trees. TMCP
To allow two nodes of different channel choices to commu-
defines the interference value of a tree as the maximum inter-
nicate, MMSN introduces toggle snooping and toggle trans-
ference value among all non-leaf nodes and the interference
mission. If a node has data to send, it must listen on its own
value of a node is defined as the number of nodes within the
channel (fself ) for potential data reception and listen on its
node’s interference range. The definition is not very accurate
target receiver’s channel (fdest ) for clear channel assessment.
since a node’s neighboring nodes are potential but not actual
MMSN thus lets the node snoop on the two channels alter-
interferers. Two nodes with the same number of neighboring
nately. If the node does not sense any signal on both channels,
nodes can experience different levels of interference depending
the node transmits a preamble alternately on the two channels.
on their neighbors’ activities. Therefore, the metric does not
The toggle transmission not only prevents other nodes from
reflect the actual interference intensity.
sending data to the transmitter but also informs any node that
Comparing with node-based multichannel protocols such
intends to transmit on fdest to back off. The toggle snooping
as MMSN [56], link-based multichannel assignments require
period (TT S ) is twice as long as the toggle transmission period
fewer number of orthogonal channels and incur fewer channel
(TT T ). The setting ensures that when a node sends out a
switches. Currently, they only divide nodes into K groups
preamble using toggle transmission, any other node that is
given K available channels. To further increase parallel trans-
toggle snooping can detect the transmission within a maximum
missions, spatial reuse of channels can be considered.
delay of TT S if they have a shared channel. Considering that
GBCA [62] proves that finding a channel assignment that
the cost of channel switching is nontrivial [60], nodes that use
minimizes the total interference in a network is NP-hard.
the toggle snooping/transmission frequently may exhaust their
Therefore, it models a channel assignment game to solve the
energy earlier than other nodes.
problem with a suboptimal result. Since routing in WSNs
is generally static and of tree/forest structure, it models the
B. Channel Assignment based on Metric Optimization parent of each Parent-Children Set (PCS) as a player. These
Le et al. [60] claim that frequency synthesizer needs time to players negotiate channel usage base on the Best Response
stabilize and thus communication in the same channel incurs (BR) dynamic, trying to maximize their payoff in each itera-
less overhead compared with inter-channel communication. tion. It proves that there exists a Nash Equilibrium (NE) in the
Therefore, they propose to group nodes that communicate game and the solution converges to the NE in polynomial time.
frequently into the same channel and separate nodes that However, same as TMCP [61], the interference is measured
do not communicate much into different channels. The idea based on network topology without considering current traffic
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 114 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

CF slots contention
window
channel
F1 F2 F3 CM DATA F1 time slot

control CM DATA F2 f1
channel node 1 node 2 node 3
CM DATA F3
f2
node 1 node 2
Fig. 13. Time slot structure in MC-LMAC. f3
node 1
f4
pattern. As a result, it may not reach the best solution for time
current interference condition. To adapt to dynamic traffic, the
game theoretic approach may incur excessive overhead. channel data ACK Notify that a
polling reception (optional) node will wait
for messages

Fig. 14. Channel hopping mechanism in Y-MAC.

C. TDMA/FDMA for Sending

Toggle snooping and toggle transmission introduce addi-

tional overhead for each packet transmission. To avoid fre- D. TDMA/FDMA for Receiving
quent channel switching before data transmission, channel A node may have different amount of data to send in
assignment can be combined with time slot assignment. In different periods. To adapt to dynamic traffic, it is better to
MC-LMAC [63], time is slotted and several consecutive slots allocate network resources accordingly. Some protocols adopt
are organized as a frame. When a node joins the network, the frequency hopping technique.
it listens for neighbors’ occupied slots in frames, which use Y-MAC [64] is a TDMA-based dynamic channel selection
a bit 0/1 to indicate whether a slot is occupied. The slot scheme, where time slots are assigned to nodes for receiving.
occupancy is associated with frequencies. A node chooses a Recall that the benefit is fewer nodes wake up at each slot
slot/channel pair that is not occupied by any of its neighbors compared with scheduling senders, but multiple senders have
within its two-hop communication neighborhood. A node may to contend for sending. In Y-MAC, each time frame consists
choose the same slot that is used by its neighbors on a of a broadcast period and a unicast period. Both of them
different channel. To ensure that two direct neighbors can are further divided into small slots. To support broadcast, all
communicate, a node must exclude slots that are used by its nodes listen on the base channel during the broadcast period.
direct neighbors, but a node can share the same slot with its Nodes that have broadcast packets contend for sending in each
two-hop neighbors. To reduce the possibility that a node is broadcast slot. If there is no broadcast traffic, the radio can
addressed by multiple senders, MC-LMAC utilizes the data be turned off for energy saving.
gathering tree structure. It lets a node choose the slot that is The slots of the unicast period are assigned in the same
not used by any other children of the same parent with a higher graph-coloring way that each node owns an exclusive slot
priority. The control overhead of broadcasting slot occupancy within its two-hop communication neighborhood. To make use
on all channels is slightly high. Same as traditional TDMA, of slots owned by neighbors, Y-MAC lets a node hop to the
the channel utilization under low contention is low and this next channel to receive another packet in the next slot if it
leaves room for further improvement. receives a packet on the current channel. As shown in Fig. 14,
The aforementioned slot/channel selection allows parallel the procedure starts with the base channel in a node’s assigned
data transmission, which is achieved by letting all nodes listen slot. Any node must listen on the base channel in its assigned
on the control channel first in each slot. Each slot has a slot. If a node receives a packet in its assigned slot, it hops to
Common Frequency (CF) period for notifying receivers of data the next channel in the next slot for potential data reception.
reception. As shown in Fig. 13, a CF period is composed of k Although the next slot is owned by one of its neighbors,
small slots where k is equal to the total number of available its data reception will not interfere with its neighbor’s data
channels. Multiple nodes may control the current time slot, reception because they are on different channels. However,
but each of them is on a different channel (i.e., each chooses if two receivers own the same slot, they will always hop
an unique channel/slot pair). Therefore, each sender of the to the same channel and their senders may keep interfering
current slot can use the CF slot that is corresponding to its with each other. In addition, Y-MAC does not discuss when a
controlled channel number to broadcast its target receiver’s ID. node should stop receiving and change to sending, and which
After the CF period, each pair of sender and receiver switch channel can be used when all channels have been visited. The
to the sender’s controlled channel for data transmission. Note channel hopping method thus needs further investigation.
that the control channel can be reused for data transmission. MuChMAC [65] is another hybrid design of TDMA and
The Control Message (CM) period is reserved for control FDMA. Each node independently chooses its receiving chan-
information transmission (e.g., the acknowledgement to the nel for each slot. To make it easy to learn the hopping
sender and the occupied slots vectors). The protocol is suitable sequence of a neighbor, the hopping sequence is produced by
for scenarios of high demands on the medium where the a pseudo-random generator using the ID and the slot number
overhead is compensated by parallel data transmissions. as the input. To support broadcast, a broadcast slot is inserted
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 HUANG et al.: THE EVOLUTION OF MAC PROTOCOLS IN WIRELESS SENSOR NETWORKS: A SURVEY 115

Star Topology Peer-to-Peer Topology Cluster Tree Network

PAN ID 1 PAN ID 2

PAN Coordinator
Full Function Device
Coordinator
Reduced Function Device
Communication Flow

Fig. 15. IEEE 802.15.4 topologies and their use.

Beacon Interval

Superframe Duration (Active Portion) Inactive Portion

Contention-Free
Contention Access Period (CAP)
Period (CFP)

Beacon GTS 1 GTS 2 Beacon

Fig. 16. An example of the IEEE 802.15.4 superframe structure.

every u unicast slots. The channel numbers of these broadcast VI. S TANDARDS
slots also follow a pseudo-random hopping sequence and the
Along with academic researches, standards are developed
sequence is identical for all nodes.
to speed up utilization of WSNs. We briefly discuss some
Because nodes choose channels randomly for each slot, two
industrial progresses in this section.
nodes may choose the same channel for a particular slot.
If they are within the interference range of each other, the
amount of parallel traffic is reduced. To address the problem, A. IEEE 802.15.4 and ZigBee
MuChMAC divides a time slot into several subslots. A node The IEEE 802.15.4 standard [66] is proposed for Low-
will choose one of the subslots to wake up. Obviously, the Rate Wireless Personal Area Networks (LR-WPANs). The key
chance that two nodes choose the same channel/subslot pair features of low cost, low power consumption, and low data
is low, but many subslots are wasted. Therefore, how to assign rate typically fit the requirements of WSNs. Two different
channel/slot to nodes efficiently remains an open issue. device types are defined in IEEE 802.15.4: a full-function
device (FFD) and a reduced-function device (RFD). A FFD is
E. Summary a device that can serve as a PAN coordinator or a coordinator
Static channel assignments may be optimized for a certain while a RFD is intended for extremely simple applications and
topology, but considering that interference condition changes implemented with minimal resources and memory capacity.
in accordance with traffic, it is better to flexibly allocate The standard supports two topologies: the star topology
channels among nodes. Dynamic channel allocation, however, and the peer-to-peer topology as shown in Fig. 15. In the
is a tough task same as the adaptive time slot assignment. In star topology, all communications must go through the PAN
addition to channel assignment, cross-channel communication coordinator which is the central controller of the WPAN. In
is another challenge in multichannel MAC protocols. Cur- the peer-to-peer topology, any device can communicate with
rently, toggle snooping/transmission may draw energy quickly. any other devices if they are in the communication range of
An alternative way is to combine TDMA with FDMA. A one another, but still the PAN coordinator must be present to
good example of scheduling senders is introduced in MC- choose PAN identifier and manage the network. More complex
LMAC [63], but overhead is slightly high and channel/slot network structures can be constructed out of the peer-to-peer
utilization is low under low contention. If slots and channels topology. An example is the multicluster tree where the PAN
are assigned for receiving as in Y-MAC [64] and MuChMAC coordinator can instruct a device to become the coordinator of
[65], contentions among senders need to be addressed if two a new cluster adjacent to the first one. Other devices gradually
receivers happen to hop to the same channel. connect to the network and extend the coverage area.
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 116 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

Two modes of operation exist in IEEE 802.15.4, namely B. WirelessHART and ISA-100.11a
the beacon-enabled mode and the nonbeacon-enabled mode.
WirelessHART [75], released by the HART Communica-
In the nonbeacon-enabled star topology, the PAN coordinator
tion Foundation in September 2007, is the first open wireless
stays awake. Associated devices send to the PAN coordinator
communication standard for process control and related ap-
or request data from the PAN coordinator in a contention-
plications. Soon ISA-100.11a [76] was approved by the ISA
based way, applying the Carrier Sense Multiple Access with
Standards and Practices Board as an official ISA standard in
Collision Avoidance (CSMA/CA). In the nonbeacon-enabled
September 2009. In general, ISA-100.11a is more flexible by
peer-to-peer topology, the devices wishing to communicate
providing more configurable parameters than WirelessHART.
have to keep their radio on constantly or some synchronization
For example, the time slot size is fixed at 10 ms in Wire-
mechanisms have to be employed. How to achieve synchro-
lessHART while in ISA100.11a it is configurable on a per-
nization in the peer-to-peer topology, however, is beyond the
superframe base. Both standards use Time Division Multiple
scope of the standard. To facilitate low-power operation, a
Access (TDMA) with frequency hopping for channel access
superframe structure is defined in beacon-enabled mode. The
in the 2.4 GHz band. The combination of direct sequence
coordinator broadcasts beacons periodically and all associated
spread spectrum (DSSS) and frequency-hopping spread spec-
devices are synchronized by the beacons. As shown in Fig. 16,
trum (FHSS) makes WirelessHART and ISA-100.11a more
the superframe has an active and an inactive portion. The
robust to interference in harsh industrial environments. IEEE
active portion is divided into the contention access period
802.15.4 defines 16 channels (i.e., channels 11-26) in the
(CAP) and the contention-free period (CFP). All contention-
2.4 GHz band. WirelessHART uses channels 11-25 because
based transactions must be completed before the CFP begins,
channel 26 is not legal to use in some countries. In ISA-
which is reserved for low-latency applications or applications
100.11a, channel 26 is optional.
requiring specific data bandwidth. A device can request a
WirelessHART adopts a slotted hopping scheme as shown
guaranteed time slot (GTS) in the CAP. A coordinator can
in Fig. 18(a) where the channel is changed in each slot while
allocate up to seven GTSs, and a GTS is allowed to occupy
ISA100.11a defines three channel hopping schemes including
more than one slot period. The optional superframe structure
slotted hopping, slow hopping, and hybrid hopping. As shown
allows the coordinator to switch to the low-power sleep mode
in Fig. 18(b), a device stays on a channel for a collection of
periodically and provide the data with quality of service (QoS).
contiguous time slots in slow hopping. This supports devices
In multihop networks, the beacon-enabled mode is not
with imprecise timing settings or devices that want to join
that straightforward. Beacons of one coordinator may col-
the network. A node can mainly scan these channels for the
lide with beacons from other coordinators or data frames
advertisement of a target network. The slow hopping also
from nodes that are associated with other coordinators. The
serves as a way to improve support for event-based traffic.
beacon collisions are intolerable because beacons provide
Usually a group of devices share a slow hopping period in a
synchronization and define the superframe structure (i.e., CAP,
contention-based way, that is, transmissions in a slow hopping
CFP, and inactive period). In addition, if two coordinators’
period is CSMA/CA based. When an event triggers the need
superframes overlap with each other, the CFP is no longer
for a device to immediately transmit a data packet or an
collision-free. These problems motivate researchers to work
alarm, the device does not need to wait for the next time slot
on beacon scheduling [67] [68] [69] [70] [71]. Generally,
that is assigned to it, thereby reducing the latency. However,
there are two methods. The first approach is superframe
slow hopping increasing devices’ energy consumption as they
duration scheduling (SDS) where each coordinator transmits
have to listen to the channel for possible incoming packets.
its superframe during the inactive period of its neighbors in
Therefore, hybrid hopping combines slotted hopping and slow
the two-hop communication neighborhood as illustrated in
hopping (Fig. 18(c)), in which slotted hopping accommodates
Fig. 17(a). A long inactive period used in this approach may
periodical messages and slow hopping improves support for
introduce significant delay in data transmission. The second
less predictable messages. In WirelessHART, the channel hop
approach is to create a beacon-only-period (BOP) at the start
patterns is controlled by the network manager. In ISA100.11a,
of the superframe where each coordinator selects a free time-
five preprogrammed hopping patterns are defined as below.
slot to transmit its own beacon and thus avoid collisions
(Fig. 17(b)). Because the active periods overlap with each • Pattern 1: 19, 12, 20, 24, 16, 23, 18, 25, 14, 21, 11, 15,
other, how to allocate GTSs efficiently is still under research 22, 17, 13 (, 26)
[72]. • Pattern 2: pattern 1 in reverse
The IEEE 802.15.4 standard defines the physical layer • Pattern 3: 15, 20, 25 (intended for slow-hopping chan-
(PHY) and the medium access control (MAC) sublayer spec- nels)
ifications. ZigBee [73] builds up the network layer and the • Pattern 4: pattern 3 in reverse
application layer on top of the IEEE 802.15.4 physical and • Pattern 5: 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22,
data link layers, thus defining a full protocol stack for LR- 23, 24, 25 (, 26).
WPANs. In industrial applications, comparison [74] shows that Pattern 1 ensures that any retransmission will not be per-
Zigbee is not suitable because it cannot meet the stringent formed on the same IEEE 802.11 channel so as to avoid
industrial requirements, especially the deterministic delay and constant interference from IEEE 802.11 devices. In addition,
high reliability. Some industrial wireless standards thus high- both WirelessHART and ISA-100.11a support blacklisting that
light their features of strict latency and high reliability for eliminates channels of significant interference. The frequency
factory automation. diversity together with the time diversity make WirelessHART
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 HUANG et al.: THE EVOLUTION OF MAC PROTOCOLS IN WIRELESS SENSOR NETWORKS: A SURVEY 117

Beacon-Only Period (BOP)

Guard Interval

Beacon

Beacon
Wake-up PAN Active Inactive
Wake-up
PAN Coordinator
Coordinator Active Inactive Inactive Post Beacon Delay

Wake-up Wake-up Coordinator 1 Acctive Inactive

Coordinator 1 Inactive Active Inactive
Post Beacon Delay

Wake-up Wake-up Wake-up Coordinator 2 Active Inactive

Coordinator 2 Inactive Inactive Active
Post Beacon Delay

(a) Superframe Duration Scheduling (SDS) (b) Beacon-only-period Scheduling (BOP)

Fig. 17. Beacon scheduling approaches.

and ISA-100.11a favorable in industrial applications. A sys- data delivery from the common active period to the sleep
tematic comparison between WirelessHART and ISA-100.11a period. The common active period is used for arranging data
can be found in [77]. The IEEE 802.15.4e is also under devel- transmission in the sleep period. Current scheduling methods
opment for considering the real-time aspects [78]. Surveys on are good at delay reduction for individual packets but are very
WSN technologies for industrial automation are also available vulnerable to interference. It is highly demanded to design a
in [79], [80]. more robust scheduling scheme that could efficiently utilize
the sleep period for data transmission. Synchronous MAC
VII. C ONCLUSION AND FUTURE WORK protocols may gradually evolve to traffic-adaptive TDMA
In this article we have surveyed the recent developments in schemes, which utilize the time in the sleep period wisely by
the design of WSN MAC protocols. We classify WSN MAC updating traffic condition during the common active period. In
protocols into four categories: asynchronous, synchronous, fact, we have seen such a paradigm in TRAMA [50], but the
frame-slotted, and multichannel. The classification aims at channel utilization is low due to the inefficient slot assignment
identifying the research trend of WSN MAC protocols based and the delay is high because nodes that fail to claim slots
on techniques they employ. We show that throughput and delay in the random period cannot access the channel in the entire
metrics are also being considered besides energy efficiency cycle. Moreover, the downstream nodes are not notified in
in WSN MAC protocols when time synchronization and advance and thus they will not reserve time slots for use. The
multichannel are gradually supported. end-to-end delay is high. More future work can be done under
Energy saving by putting nodes into low-power sleep mode this framework.
periodically is a fundament mechanism in WSN MAC proto- In TDMA, slot stealing helps to improve the channel
cols. Asynchronous MAC protocols have integrated the predic- utilization under low contention, but it incurs high energy
tion mechanism to estimate the best wake-up time for sending. consumption as all nodes have to wake up to receive potential
Recently, the responsibility of establishing communication is data in a slot. Assigning time slots only to nodes that are
gradually shifted from the sender side to the receiver side. on active routes can increase channel utilization without
Although the transition is reasonable since only when the incurring too much overhead, but there are several issues that
receiver is ready to receive the sending is effective, quick and require further study. First, a mechanism is needed to rapidly
proper response to link breaks and target receivers’ leave is inform nodes of incoming traffic. Second, a lightweight time
a potential issue. It is also challenging to minimize the idle slot assignment algorithm is needed for dynamic time slot
listening of senders while guaranteeing that the sender will assignment among nodes on the active routing paths. Third,
not miss the beacon of the receiver. In other words, although how to utilize slots that are claimed by a node but left unused
LPP-based MAC protocols provide higher throughput, the is of interest.
energy efficiency is generally lower than that of LPL-based Scheduling senders in TDMA wakes all potential receivers
MAC protocols in light traffic load scenarios because channel up. Although scheduling receivers in TDMA leads to fewer
sampling duration is short in LPL-based MAC protocols operations of turning radio on, the strategy deviates away
but idle listening duration for a beacon is slightly longer. from the original goal of TDMA (i.e., collision-free data
Which one is more appropriate is application dependent. In transmission). It is thus necessary to address contentions
asynchronous MAC protocols, nodes’ random and independent among multiple senders efficiently.
schedules result in a long initial delay for event report. These To further boost network throughput, multichannel MAC
are the potential research directions in asynchronous MAC protocols became a hot topic. One challenge for future re-
protocols. search is the design of dynamic channel allocation algorithms
Synchronous MAC protocols synchronize a cluster of nodes that adapt to the dynamic traffic of sensor networks. The
to a common schedule. The trend in this branch is to move interference analysis of prior work is based on network
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 118 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

Time Slot
also challenging to design such a dynamic channel allocation
algorithm with low overhead.
11
12 Another challenge in multichannel MAC protocols is cross-
13 channel communication. For cost consideration, a sensor node
14 is usually equipped with a half-duplex radio transceiver and
15
the transceiver is unable to listen on multiple channels at the
16
same time. This implies that a node has to switch back and
Channel

17
18 forth between two channels to check whether it has data to
19 receive and whether the target receiver’s channel is clear. A
20 node is also responsible for letting its neighbors know that
21
it is going to send and thus is unable to receive. These
22
23
operations increase the overhead for a packet transmission
24 and frequent channel switch makes a node consume energy
25 faster than others. Therefore, how to perform cross-channel
communication efficiently is worth further study.
Time
The channel hopping used in industry standards is robust
(a) slotted hopping
to interference. However, the channel hopping strategies and
Time Slot
time slot assignment are under control of a central unit (i.e.,
network/system manager). To extend the network size, reliable
11
distributed TDMA/FDMA algorithms are highly desired.
12
13 Finally, most previous studies have only considered or-
14 thogonal channels. However, it is possible to utilize partially
15 overlapped channels. Some studies [81] [82] have shown
16
that the overall throughput can be improved by using non-
Channel

17
18
orthogonal channels even though the throughput on each in-
19 dividual channel may slightly drop. Although the introduction
20 of non-orthogonal channels has the potential to further boost
21 network capacity, how to utilize them wisely is an open issue.
22
As it becomes harder to model channel interference when
23
24
co-channel interference is introduced, it is more challenging
25 to design channel allocation algorithms for sensor networks,
especially when dynamic channel allocation is desired.
Time
(b) slow hopping
Time Slot R EFERENCES

11 [1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, “A survey

12 on sensor networks,” IEEE Commun. Mag., vol. 40, no. 8, pp. 102–114,
13 2002.
14 [2] “SENSEI (Integrating the Physical with the Digital World of the
15 Network of the Future),” http://www.sensei-project.eu.
16 [3] P. Naik and K. M. Sivalingam, A survey of MAC protocols for sensor
Channel

17 networks. Kluwer Academic Publishers, 2004, pp. 93–107.

18 [4] I. Demirkol, C. Ersoy, and F. Alagoz, “MAC protocols for wireless
sensor networks: a survey,” IEEE Commun. Mag., vol. 44, no. 4, pp.
19
115–121, 2006.
20
[5] K. G. Langendoen, Medium access control in wireless sensor networks.
21
Nova Science Publishers, Inc., 2008, pp. 535–560.
22
[6] A. Bachir, M. Dohler, T. Watteyne, and K. Leung, “MAC essentials for
23 wireless sensor networks,” IEEE Commun. Surveys Tutorials, vol. 12,
24 no. 2, pp. 222 –248, 2010.
25 [7] J. L. Hill and D. E. Culler, “Mica: A wireless platform for deeply
embedded networks,” IEEE Micro, vol. 22, no. 6, pp. 12–24, Novem-
Time ber/December 2002.
(c) hybrid hopping [8] A. El-Hoiydi, “Aloha with preamble sampling for sporadic traffic in ad
hoc wireless sensor networks,” in Proc. ICC, 2002.
Fig. 18. Frequency-hopping patterns. [9] , “Spatial TDMA and CSMA with preamble sampling for low
power ad hoc wireless sensor networks,” in Proc. 7th International
Symposium on Computers and Communications (ISCC), 2002, pp. 685–
692.
[10] J. Polastre, J. Hill, and D. Culler, “Versatile low power media access
topology, and thus the solutions may not be the best fit for for wireless sensor networks,” in Proc. SenSys, 2004.
[11] C. Schurgers, V. Tsiatsis, S. Ganeriwal, and M. Srivastava, “Optimizing
the current interference condition. It is desirable to allocate sensor networks in the energy-latency-density design space,” IEEE
network resources flexibility according to the traffic but it is Trans. Mobile Computing, vol. 1, no. 1, pp. 70–80, January-March 2002.
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 HUANG et al.: THE EVOLUTION OF MAC PROTOCOLS IN WIRELESS SENSOR NETWORKS: A SURVEY 119

[12] S. Lim, S. Kim, J. Cho, and S. An, “Medium access control with an [33] W. Ye, J. Heidemann, and D. Estrin, “An energy-efficient MAC protocol
energy-efficient algorithm for wireless sensor networks,” in Personal for wireless sensor networks,” in Proc. INFOCOM, 2002, pp. 1567–
Wireless Communications, ser. Lecture Notes in Computer Science, 1576.
P. Cuenca and L. Orozco-Barbosa, Eds. Springer Berlin / Heidelberg, [34] , “Medium access control with coordinated adaptive sleeping for
2006, vol. 4217, pp. 334–343. wireless sensor networks,” IEEE/ACM Transactions on Networking,
[13] A. Bachir, D. Barthel, M. Heusse, and A. Duda, “Micro-frame preamble vol. 12, no. 3, pp. 493–506, 2004.
MAC for multihop wireless sensor networks,” in Proc. IEEE Interna- [35] T. van Dam and K. Langendoen, “An adaptive energy-efficient MAC
tional Conference on Communications (ICC), vol. 7, June 2006, pp. protocol for wireless sensor networks,” in Proc. SenSys, 2003, pp. 171–
3365 –3370. 180.
[14] M. Awenuti, P. Corsini, P. Masci, and A. Vecchio, “Increasing the [36] S. Du, A. K. Saha, and D. B. Johnson, “RMAC: A routing-enhanced
efficiency of preamble sampling protocols for wireless sensor networks,” duty-cycle MAC protocol for wireless sensor networks,” in Proc. IN-
in Proc. First Mobile Computing and Wireless Communication Interna- FOCOM, 2007, pp. 1478–1486.
tional Conference (MCWC), September 2006, pp. 117 –122. [37] Y. Sun, S. Du, O. Gurewitz, and D. B. Johnson, “DW-MAC: A low
[15] K.-J. Wong and D. K. Arvind, “SpeckMAC: low-power decentralised latency, energy efficient demand-wakeup MAC protocol for wireless
MAC protocols for low data rate transmissions in specknets,” in Proc. sensor networks,” in Proc. MobiHoc, 2008, pp. 53–62.
2nd International Workshop on Multi-hop Ad Hoc Networks: from [38] G. Lu, B. Krishnamachari, and C. S. Raghavendra, “An adaptive energy-
Theory to Reality, 2006, pp. 71–78. efficient and low-latency MAC for data gathering in wireless sensor
[16] S. Lim, Y. Ji, J. Cho, and S. An, “An ultra low power medium access networks,” in Proc. IPDPS, 2004.
control protocol with the divided preamble sampling,” in Ubiquitous [39] N. Vasanthi and S. Annadurai, “Energy efficient sleep schedule for
Computing Systems, ser. Lecture Notes in Computer Science, H. Youn, achieving minimum latency in query based sensor networks,” in Proc.
M. Kim, and H. Morikawa, Eds. Springer Berlin / Heidelberg, 2006, IEEE International Conference on Sensor Networks, Ubiquitous, and
vol. 4239, pp. 210–224. Trustworthy Computing, vol. 2, June 2006, pp. 214–219.
[17] X. Shi and G. Stromberg, “SyncWUF: An ultra low-power MAC [40] W. Ye, F. Silva, and J. Heidemann, “Ultra-low duty cycle MAC with
protocol for wireless sensor networks,” IEEE Trans. Mobile Computing, scheduled channel polling,” in Proc. SenSys, 2006, pp. 321–334.
vol. 6, no. 1, pp. 115 –125, January 2007. [41] P. Lin, C. Qiao, and X. Wang, “Medium access control with a dynamic
[18] S. Mahlknecht and M. Bock, “CSMA-MPS: a minimum preamble duty cycle for sensor networks,” in Proc. IEEE Wireless Communica-
sampling MAC protocol for low power wireless sensor networks,” in tions and Networking Conference (WCNC), vol. 3, March 2004, pp.
Proc. IEEE International Workshop on Factory Communication Systems, 1534–1539 Vol.3.
September 2004, pp. 73 – 80. [42] Z. Chen, C. Hu, J. Liao, and S. Liu, “Protocol architecture for wireless
[19] E.-Y. Lin, J. Rabaey, and A. Wolisz, “Power-efficient rendez-vous body area network based on nRF24L01,” in Proc. IEEE International
schemes for dense wireless sensor networks,” in Proc. IEEE Interna- Conference on Automation and Logistics (ICAL’08). IEEE, 2008, pp.
tional Conference on Communications (ICC), vol. 7, June 2004, pp. 3050–3054.
3769 – 3776. [43] O. Omeni, A. Wong, A. Burdett, and C. Toumazou, “Energy efficient
[20] M. Buettner, G. V. Yee, E. Anderson, and R. Han, “X-MAC: a short medium access protocol for wireless medical body area sensor net-
preamble MAC protocol for duty-cycled wireless sensor networks,” in works,” IEEE Trans. Biomedical Circuits and Systems, vol. 2, no. 4,
Proc. SenSys, 2006, pp. 307–320. pp. 251–259, 2008.
[21] L. Bernardo, R. Oliveira, M. Pereira, M. Macedo, and P. Pinto, “A [44] A. Milenkovic, C. Otto, and E. Jovanov, “Wireless sensor networks for
wireless sensor MAC protocol for bursty data traffic,” in Proc. IEEE personal health monitoring: Issues and an implementation,” Computer
18th International Symposium on Personal, Indoor and Mobile Radio Communications, vol. 29, no. 13-14, pp. 2521–2533, 2006.
Communications (PIMRC), September 2007, pp. 1 –5. [45] B. Latre, B. Braem, I. Moerman, C. Blondia, E. Reusens, W. Joseph,
and P. Demeester, “A low-delay protocol for multihop wireless body area
[22] H. Wang, X. Zhang, F. Nait-Abdesselam, and A. A. Khokhar, “DPS-
networks,” in Proc. 4th Annual International Conference on Mobile and
MAC: An asynchronous MAC protocol for wireless sensor networks.” in
Ubiquitous Systems: Networking & Services (MobiQuitous), 2007, pp.
Proc. 14th International Conference on High Performance Computing
1–8.
(HiPC), 2007, pp. 393–404.
[46] S. Marinkovic, E. Popovici, C. Spagnol, S. Faul, and W. Marnane,
[23] S. Liu, K.-W. Fan, and P. Sinha, “CMAC: An energy efficient MAC
“Energy-efficient low duty cycle MAC protocol for wireless body area
layer protocol using convergent packet forwarding for wireless sensor
networks,” IEEE Trans. Inf. Technol. Biomed., vol. 13, no. 6, pp. 915–
networks,” in Proc. 4th IEEE Communications Society Conference on
925, 2009.
Sensor, Mesh and Ad Hoc Communications and Networks (SECON),
[47] I. Rhee, A. Warrier, M. Aia, and J. Min, “Z-MAC: a hybrid MAC for
June 2007, pp. 11–20.
wireless sensor networks,” in Proc. SenSys, 2005, pp. 90–101.
[24] C. J. Merlin and W. B. Heinzelman, “Network-aware adaptation of [48] I. Rhee, A. Warrier, J. Min, and L. Xu, “DRAND: Distributed ran-
MAC scheduling for wireless sensor networks,” in Proc. 3rd IEEE domized TDMA scheduling for wireless ad-hoc networks,” in Proc.
International Conference on Distributed Computing in Sensor Systems MobiHoc, 2006, pp. 190–201.
(DCOSS Poster Session), June 2007. [49] S. Gobriel, D. Mosse, and R. Cleric, “TDMA-ASAP: Sensor network
[25] A. El-Hoiydi and J.-D. Decotignie, “WiseMAC: An ultra low power TDMA scheduling with adaptive slot-stealing and parallelism,” in Proc.
MAC protocol for multi-hop wireless sensor networks,” in Proc. 9th ICDCS, 2009, pp. 458–465.
International Symposium on Computers and Communications (ISCC), [50] V. Rajendran, K. Obraczka, and J. J. Garcia-Luna-Aceves, “Energy-
vol. 1, June-July 2004, pp. 244–251. efficient collision-free medium access control for wireless sensor net-
[26] , “Low power downlink MAC protocols for infrastructure wireless works,” in Proc. SenSys, 2003, pp. 181–192.
sensor networks,” Mob. Netw. Appl., vol. 10, pp. 675–690, October 2005. [51] L. Bao and J. J. Garcia-Luna-Aceves, “A new approach to channel access
[27] R. Musaloiu-E, C.-J. Liang, and A. Terzis, “Koala: Ultra-low power scheduling for Ad Hoc networks,” in Proc. MobiCom, 2001, pp. 210–
data retrieval in wireless sensor networks,” in Proc. IPSN, 2008, pp. 221.
421–432. [52] S. Chatterjea, L. van Hoesel, and P. Havinga, “AI-LMAC: an adaptive,
[28] B. Jang, J. B. Lim, and M. L. Sichitiu, “AS-MAC: An asynchronous information-centric and lightweight MAC protocol for wireless sensor
scheduled MAC protocol for wireless sensor networks,” in Proc. MASS, networks,” in Proc. 2004 Intelligent Sensors, Sensor Networks and
2008. Information Processing Conference, 2004, pp. 381–388.
[29] Y. Sun, O. Gurewitz, and D. B. Johnson, “RI-MAC: A receiver-initiated [53] W.-Z. Song, R. Huang, B. Shirazi, and R. LaHusen, “TreeMAC:
asynchronous duty cycle MAC protocol for dynamic traffic loads in Localized TDMA MAC protocol for real-time high-data-rate sensor
wireless sensor networks,” in Proc. SenSys, 2008, pp. 1–14. networks,” in Proc. PerCom, 2009.
[30] P. Dutta, S. Dawson-Haggerty, Y. Chen, C.-J. M. Liang, and A. Terzis, [54] G. P. Halkes and K. G. Langendoen, “Crankshaft: an energy-efficient
“Design and evaluation of a versatile and efficient receiver-initiated link MAC-protocol for dense wireless sensor networks,” in Proc. EWSN,
layer for low-power wireless,” in Proc. SenSys, 2010, pp. 1–14. 2007, pp. 228–244.
[31] P. Huang, C. Wang, L. Xiao, and H. Chen, “RC-MAC: A receiver-centric [55] T. Zheng, S. Radhakrishnan, and V. Sarangan, “PMAC: An adaptive
medium access control protocol for wireless sensor networks,” in Proc. energy-efficient MAC protocol for wireless sensor networks,” in Proc.
IWQoS, 2010, pp. 1–9. IPDPS, 2005.
[32] L. Tang, Y. Sun, O. Gurewitz, and D. B. Johnson, “PW-MAC: An [56] G. Zhou, C. Huang, T. Yan, T. He, J. A. Stankovic, and T. F. Abdelzaher,
energy-efficient predictive-wakeup MAC protocol for wireless sensor “MMSN: Multi-frequency media access control for wireless sensor
networks,” in Proc. INFOCOM, 2011, pp. 1305–1313. networks,” in Proc. INFOCOM, 2006, pp. 1–13.
Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.
 120 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 15, NO. 1, FIRST QUARTER 2013

[57] J. So and N. H. Vaidya, “Multi-channel MAC for ad hoc networks: network technologies for industrial automation: The security and quality
handling multi-channel hidden terminals using a single transceiver,” in of service perspectives,” Future Internet, vol. 2, no. 2, pp. 96–125, 2010.
Proc. MobiHoc, 2004, pp. 222–233. [80] M. Paavola and K. Leiviska, “Wireless sensor networks in industrial
[58] J. Zhang, G. Zhou, C. Huang, S. H. Son, and J. A. Stankovic, “TMMAC: automation,” Factory Automation, 2010.
An energy efficient multi-channel MAC protocol for ad hoc networks.” [81] A. Mishra, V. Shrivastava, S. Banerjee, and W. Arbaugh, “Partially
in Proc. ICC, 2007, pp. 3554–3561. overlapped channels not considered harmful,” in Proc. SIGMETRICS,
[59] “CC2420 datasheet,” http://www.ti.com. 2006, pp. 63–74.
[60] H. K. Le, D. Henriksson, and T. Abdelzaher, “A practical multi-channel [82] X. Xu, J. Luo, and Q. Zhang, “Design of non-orthogonal multi-channel
media access control protocol for wireless sensor networks,” in Proc. sensor networks.” in Proc. ICDCS, 2010, pp. 358–367.
IPSN, 2008, pp. 70–81.
[61] Y. Wu, J. A. Stankovic, T. He, J. Lu, and S. Lin, “Realistic and efficient
multi-channel communications in wireless sensor networks,” in Proc.
INFOCOM, 2008, pp. 1193–1201.
[62] Q. Yu, J. Chen, Y. Fan, X. Shen, and Y. Sun, “Multi-channel assignment Pei Huang received the BS with distinction and MS degrees in electrical
in wireless sensor networks: a game theoretic approach,” in Proc. engineering from Southwest Jiaotong University, China, in 2005 and 2008,
INFOCOM, 2010, pp. 1127–1135. respectively. Currently, he is a research assistant with the Michigan State
[63] O. D. Incel, L. van Hoesel, P. Jansen, and P. Havinga, “MC-LMAC: University Computer Science and Engineering Department, where he is
A multi-channel MAC protocol for wireless sensor networks,” Ad Hoc pursuing his PhD degree. His research interests include the areas of wireless
Netw., vol. 9, pp. 73–94, January 2011. communication, wireless ad hoc and sensor networks, and cognitive radio
[64] Y. Kim, H. Shin, and H. Cha, “Y-MAC: An energy-efficient multi- networks.
channel MAC protocol for dense wireless sensor networks,” in Proc.
IPSN, 2008, pp. 53–63.
[65] J. Borms, K. Steenhaut, and B. Lemmens, “Low-overhead dynamic
multi-channel MAC for wireless sensor networks,” in Proc. EWSN,
2010, pp. 81–96. Li Xiao received the BS and MS degrees in computer science from North-
[66] IEEE Standard for Local and metropolitan area networks - Part 15.4: western Polytechnic University, China, and the PhD degree in computer
Low-Rate Wireless Personal Area Networks (LR-WPANs), IEEE Std. science from the College of William and Mary in 2002. She is an associate
802.15.4-2011, September 2011. professor of computer science and engineering at Michigan State University.
[67] B. Villaverde, R. De Paz Alberola, S. Rea, and D. Pesch, “Exper- Her research interests are in the areas of distributed and networking systems,
imental evaluation of beacon scheduling mechanisms for multihop overlay systems and applications, and wireless networks. She is a senior
IEEE 802.15.4 wireless sensor networks,” in Proc. 4th International member of the IEEE.
Conference on Sensor Technologies and Applications (SENSORCOMM),
July 2010, pp. 226–231.
[68] R. Burda and C. Wietfeld, “A distributed and autonomous beacon
scheduling algorithm for IEEE 802.15.4/ZigBee networks,” in Proc.
IEEE Internatonal Conference on Mobile Adhoc and Sensor Systems Soroor Soltani is a PhD candidate in the Electrical and Computer Engineering
(MASS), October 2007, pp. 1–6. Department at Michigan State University, East Lansing, MI. She graduated
[69] H.-I. Jeon and Y. Kim, “BOP (beacon-only period) and beacon schedul- with BE in Electrical and Communication Engineering with distinction from
ing for MEU (mesh-enabled USN) devices,” in Proc. 9th International Kuwait University, Kuwait in 2005. She received her MS in Electrical
Conference on Advanced Communication Technology, vol. 2, February Engineering from the Michigan State University, MI, in 2008. Her expertise
2007, pp. 1139 –1142. and research interests lie in wireless cognitive radio mesh networks operating
[70] E.-J. Kim and H.-H. Choi, “EBBS: Energy-efficient BOP-based beacon in urban areas, and multimedia communication over stochastic networks. She
transmission scheduling for WSNs,” in Proc. IEEE 19th International is a student member of the IEEE.
Symposium on Personal, Indoor and Mobile Radio Communications
(PIMRC), September 2008, pp. 1–6.
[71] A. Koubaa, A. Cunha, and M. Alves, “A time division beacon schedul-
ing mechanism for IEEE 802.15.4/Zigbee cluster-tree wireless sensor
Matt W. Mutka received the BS degree in electrical engineering from the
networks,” in Proc. 19th Euromicro Conference on Real-Time Systems
University of Missouri-Rolla, the MS degree in electrical engineering from
(ECRTS), July 2007, pp. 125–135.
Stanford University, and the PhD degree in Computer Sciences from the
[72] B. C. Villaverde, S. Rea, and D. Pesch, “Guaranteeing reliable com- University of Wisconsin-Madison. He is on the faculty of the Department
munications in mesh beacon-enabled IEEE802.15.4 WSN for industrial of Computer Science and Engineering at Michigan State University, where
monitoring applications,” in Proc. 2nd International ICST Conference he is currently professor and chairperson. He has been a visiting scholar at the
on Ad Hoc Networks (ADHOCNETS), 2010, pp. 359–370. University of Helsinki, Helsinki, Finland, and a member of technical staff at
[73] “ZigBee Alliance,” http://www.zigbee.org/. Bell Laboratories in Denver, Colorado. His current research interests include
[74] T. Lennvall, S. Svensson, and F. Hekland, “A comparison of Wire- mobile computing, sensor networking, wireless networking and multimedia
lessHART and ZigBee for industrial applications,” in Proc. IEEE networking.
International Workshop on Factory Communication Systems (WFCS),
May 2008, pp. 85 –88.
[75] HART Field Communication Protocol Specification, Revision 7.0, HART
Communication Foundation Std., September 2007.
[76] Wireless Systems for Industrial Automation: Process Control and Re- Ning Xi received his D.Sc. degree in Systems Science and Mathematics
lated Applications, Std. ISA-100.11a-2009, 2009. from Washington University in St. Louis, Missouri in December, 1993, and
[77] S. Petersen and S. Carlsen, “WirelessHART Versus ISA100.11a: The B.S. degree in electrical engineering from Beijing University of Aeronautics
format war hits the factory floor,” IEEE Industrial Electronics Mag., and Astronautics. He is the University Distinguished Professor and John D.
vol. 5, no. 4, pp. 23–34, December 2011. Ryder professor of Electrical and Computer Engineering at Michigan State
[78] F. Chen, R. German, and F. Dressler, “Towards IEEE 802.15.4e: A study University. Currently, he serves as the Head and Chair Professor of the
of performance aspects,” in Proc. 8th IEEE International Conference Department of Mechanical and Biomedical Engineering, the City University
on Pervasive Computing and Communications Workshops (PERCOM of Hong Kong. Dr. Xi is a fellow of IEEE. Currently, his research interests
Workshops), April 2010, pp. 68–73. include robotics, manufacturing automation, micro/nano manufacturing, nano
[79] D. Christin, P. S. Mogre, and M. Hollick, “Survey on wireless sensor sensors and devices, and intelligent control and systems.

Authorized licensed use limited to: LAKIREDDY BALI REDDY COLLEGE OF ENGINEERING. Downloaded on September 30,2024 at 15:36:58 UTC from IEEE Xplore. Restrictions apply.