JOURNAL OF COMMUNICATIONS AND NETWORKS, VOL. 17, NO. 2, APRIL 2015
203
An Energy-Efficient Mobility-Supporting MAC
Protocol in Wireless Sensor Networks
Fei Peng and Meng Cui
Abstract: Although mobile applications are an essential characteristic of wireless sensor networks, most existing media access control (MAC) protocols focus primarily on static networks. In these
protocols, fixed periodic neighbor discovery and schedule updating
are used to connect and synchronize neighbors to provide successful data transmission; however, they cannot adapt to mobile speed
variation and degrade the network performance dramatically. In
this paper, we propose a mobile-supporting mechanism for MAC
protocols, in which the decision to update the neighbors of a mobile node is made adaptively according to the mobile speed. Analysis and simulation results demonstrate that the mechanism efficiently avoids the disconnection of a mobile node from its neighbors
and achieves a better performance as compared with fixed periodic
neighbor discovery.
Index Terms: Media access control (MAC), mobility, performance
evaluation, wireless sensor networks.
I. INTRODUCTION
OST traditional applications in wireless sensor networks
(WSNs) focus mainly on static networks, in which the
nodes are, in general, fixed from the time of their deployment.
Recently, mobility emerged and has become a typical characteristic in the application of WSNs. Mobility in WSN applications refers to the fact that some network elements, such as sensor nodes, sinks, monitored targets, or actuators, may be mobile for reasons such as environmental influences, e.g., wind
or water, or because the sensors are attached to or carried by
mobile entities or possess automotive capabilities to quickly react to emergency situations. As the number of mobile entities
monitored by sensors has gradually increased, the role played
by mobility has become increasingly important in WSNs. However, most existing media access control (MAC) protocols are
customized for stationary networks, where the topology is considered fixed and the neighbors of each node remain unchanged
for a long period. According to our survey, only several MAC
protocols consider the support of mobility. These MAC protocols are generally classified into three types; contention-based,
time division multiple access (TDMA)-based, and hybrid. We
concentrate on contention-based MAC protocols, since they do
M
Manuscript received February 26, 2014; approved for publication by Niyato,
Dusit (Tao), Division III Editor, September 19, 2014.
This work is jointly supported by the National Nature Science Foundation under Grant 61102009 and the Natural Sciences and Engineering Research Council
of Canada.
Fei Peng is with University of British Columbia, Vancouver, BC, Canada V6T
1Z4. (Institute of Information on Traditional Chinese Medicine, China Academy
of Chinese Medical Sciences), email: fpengca@yahoo.ca.
Meng Cui is with Institute of Information on Traditional Chinese Medicine,
China Academy of Chinese Medical Sciences, email: cm@ail.cintcm.ac.cn.
Digital object identifier 10.1109/JCN.2015.000034
not use a complicated scheduling algorithm for channel division. Furthermore, contention-based MAC protocols provide an
equal probability for mobile and static nodes to access the channel and they do not incur the overhead produced by hybrid protocols. To handle the schedule updating problem of the mobile
node when it crosses between different virtual clusters, in [1] the
border nodes add all the schedules’ information that they follow
to the SYNC packets, so that the mobile node performs schedule
adaptation according to the schedules it receives from the border nodes. However, since SYNC packets broadcast at a fixed
interval of 10 listen-sleep cycles, it is possible that the mobile
node will miss them if they pass through the border nodes at a
speed faster than 10 cycles. In [1], the authors also proposed
adjusting the establishing period of the active/sleep schedule according to the estimation of the change in the number of onehop neighbors; however, their method does not consider making the schedule adaptable to the variation in mobile speed. The
mobility-aware delay-sensitive sensor MAC (MD-SMAC) protocol [2] inherited the basic idea of the mobility-aware sensor
MAC (MS-MAC) protocol: The mobile node increases the frequency of the neighbor discovery and leaves the other nodes to
continue their normal operations. When the mobile node has discovered the new schedule, it returns to its normal neighbor discovery frequency (every 5 m) and removes its old schedule. In
[3], an enhanced MS-MAC protocol named EMS-MAC was introduced to predict the node’s movement more accurately by
using both received signal strength indication (RSSI) and link
quality indication (LQI), since MS-MAC causes energy wastage
because of the inappropriate mobility prediction produced by
using only RSSI. However, EMS-MAC has the same periodic
procedure for connection setup as MS-MAC. In existing variations of contention-based MAC to support mobility, the fixed
period of neighbor discovery and schedule updating does not allow the mobile node with its varying speed to connect to new
neighbors in a sufficiently short time. In the contention-based
MAC protocols mentioned above, a mobile node synchronizes
with its neighbors after hearing a SYNC frame. We propose a
mobility support mechanism based on the exchange of SYNC to
allow the mobile nodes to set up connections with their neighbors adaptively, while a low power level is maintained. The rest
of this paper is organized as follows. In Section II, we present
our mobility-supporting mechanism in detail. In Section III, we
describe the performance evaluation of our mechanism. Section
IV concludes the paper.
II. THE MOBILITY - SUPPORTING MECHANISM
A. Brief Review of S-MAC
S-MAC [4] is one of the most popular contention-based MAC
protocols designed to ensure energy efficiency in WSNs. It
1229-2370/15/$10.00 c 2015 KICS
204
JOURNAL OF COMMUNICATIONS AND NETWORKS, VOL. 17, NO. 2, APRIL 2015
SYNC
Synchronization period=10 frames ≈ 10 s
Type
One frame ≈ 1 s
Length
srcAddr
synNode sleepTime
state
Length
srcAddr
synNode sleepTime
state
RSSI
crc
MSYNC
SYNC
(broadcast) DATA
Sleep
SYNC
DATA
Sleep
SYNC
DATA
…
SYNC
(broadcast)
DATA
…
Type
speed
angle
crc
MACK
Type
Neighbor
discovery
…
srcAddr
synNode1 sleepTime1 synNode2(opt) sleepTime2(opt) NextAnnounceInterval
crc
Neighbor
discovery
5 m or 30 s
Fig. 2. Formats of SYNC, MSYNC, and MACK.
Fig. 1. Description of S-MAC synchronization period.
adopts a periodic listen-sleep cycle (called a frame, in general,
constituting 1.0 s, with a listen period of about 100 ms and a
sleep period) as the schedule for the node to follow. Nodes that
have the same schedule form a virtual cluster to wake up for
listening at the same time. The listen period is further divided
into a SYNC period and a Data period for broadcasting or receiving SYNC packets (nearly 30 ms) and sending or receiving DATA packets, respectively. As shown in Fig. 1, each node
broadcasts a SYNC packet every synchronization period, which
is generally 10 frames, about 10 s. Furthermore, to avoid loss
of SYNC among neighbors, S-MAC is defined to repeatedly execute neighbor discovery every 5 m, when the node has at least
one neighbor, and more frequently, for instance every 30 s, when
the node has no neighbors. During the neighbor discovery period, the node continues to listen for 10 listen-sleep cycles without sleeping to achieve more chances of hearing a new neighbor
node. MS-MAC has the same packet format as S-MAC, while
DE-ASS [1] adds multiple schedule information in the SYNC
packet at the border node. The packet structures of other MAC
protocols designed to support mobility [5], [6] are more complicated than those of S-MAC. The design of our mechanism is
based on the S-MAC format.
B. Design of Control Packet Format
There are three types of SYNC packets in our mobilitysupporting mechanism; the original SYNC, MSYNC, and
MACK. MSYNC is used by the mobile node and MACK is
broadcast by the nodes that receive the MSYNC. The original
SYNC is still used by the ordinary static nodes and border nodes.
The two fields sleepTime and srcAddr in the SYNC packet,
shown in Fig. 2, represent the sender’s schedule information to
determine when the sender will enter the next sleep period after
it sends the SYNC packet and the sender address through which
the receiver can update its neighbor list, respectively. The mobile nodes broadcast only MSYNC with two fields of speed and
angle added in the original SYNC packets. In addition to the usages similar to those of the original SYNC, MSYNC can inform
the receivers that the mobile node has entered their communication range.
We built a reply packet of MACK specifically corresponding
to MSYNC, as shown in Fig. 2. The field called NextAnnounceInterval in the MACK packet helps the mobile node to determine when to send the MSYNC packet. sleepTime1 in the
MACK packet represents the schedule of the sender’s virtual
cluster. If the MACK packet is sent by border nodes, sleepTime2
represents the schedule of another virtual cluster; else, sleepTime2 is null. Here, we assume the border nodes follow only
two different schedules. To ensure that the MSYNC/MACK
handshake can be completed in a frame time, we reserve half
of the duration in the SYNC period for the mobile to receive the
MACK packet in the same frame as MSYNC.
C. Location Estimation from RSSI
When a mobile node initiates movement in the networks, it
sends MSYNC, through which it becomes synchronized with its
neighbors. One of the neighbor nodes is chosen to deduce the
appropriate announcement time of MSYNC and broadcasts a
MACK packet to the mobile node. The mobile node then adaptively broadcasts MSYNC to its one-hop neighbors during its
movement. Other neighbor nodes that are not chosen to send the
MACK packet compete to broadcast original SYNC packets at
the following frames.
In the proposed MS-MAC protocol [5], a change in the radio
signal level of received SYNC messages is used as an indication
of a mobile’s presence. Although this method does not provide
a high level of accuracy, it is effective in our mechanism, which
is described in more detail in [5]. An incorrect detection of mobility may unnecessarily trigger a neighbor search, but it does
not degrade the network performance. Unlike the IR, acoustic,
or ultrasound signals used in other methods, the radio signal of
SYNC messages is available at no extra cost in terms of energy
consumption.
To expedite the connection set up process, we enable each
node to discover the presence as well as the level of mobility within its neighborhood, based on the RSSI values obtained
from the SYNC messages transmitted by its neighbors. Each
node maintains a one-hop neighbor list to record the relative
position of its neighbors. The relative position can be easily
obtained from the measurable location. Please refer to [6] for
more details. If the RSSI value from one and the same neighbor changes during a time interval, it realizes that either this
neighbor, the node itself, or both are moving, since a one-to-one
mapping between the distance and the RSSI values is assumed.
According to the change in the RSSI values, the relative moving
speed of the mobile individual can be deduced. Based on this information, a mobile node broadcasts an MSYNC message containing its own schedule and additional mobility information,
i.e., the estimated speed.
D. The Announcement Time of MSYNC
We assume each node has the same communication range R.
For simplicity without losing generality, first only one mobile
node is considered. The proposed approach can then be extended
to multiple mobile nodes.
Suppose the topology around the mobile node shown in
PENG AND CUI: AN ENERGY-EFFICIENT MOBILE-SUPPORTING MAC PROTOCOL...
205
node g, and node m constitute a triangle, as shown in Fig. 3.
Assuming the distance between point O and m is d(vm , g), according to the cosine theorem we obtain the relation
N
d2 (vm , g)−2d (v, g) Lm→g cos(β−γ)+L2m→g −R2 = 0. (4)
g
i
From (4), we obtain d(vm , g) if (1) is satisfied,
k
R
m
q
d (vm , g) = Lm→g cos(β − γ) ± R2 − L2m→g sin2 (β − γ)
 q
 L2m→g − R2 , if it takes − ;
q
=
2
 L2
m→g COS (2 (β − γ)) + R , if it takes + .
O
h
j
(5)
Fig. 3. Topology and abstraction around the mobile node.
Fig. 3. The mobile node m broadcasts MSYNC. Sensor nodes i
and j are the new one-hop neighbors, while nodes k and h are
the previous one-hop neighbors of the mobile node. Assuming
both nodes i and j can receive MSYNC, we propose a limitation to select one. Node i is selected to send MACK if the angle
ωlvi→j between the directed edge li→j and the movement direction of the mobile node meets the condition:
ωlvi→j ≥ 90◦ .
(1)
From all the neighbor nodes that satisfy (1), we select that
which is farthest from the mobile node and has largest ωlvi→j ∈
[90◦ , 180◦ ]. In Fig. 3, node i is selected from among the onehop neighbors of the mobile node, and is entitled to compute
the next MSYNC announcement time and send the results to
the mobile node through MACK. We then take the next step
to select node g, which is the one-hop neighbor of node i, and
the two-hop neighbor of the mobile node. The distance that the
mobile node moves from its one-hop communication range to its
two-hop range is denoted by dg . Fig. 3 shows the relationships
among the mobile node, node i, and node g. Li→g , Lm→i and
Lm→g represent the distance from node i to g, from the mobile
node to node i and from the mobile node to node g respectively.
α, β, and γ represents the angle between directed line m → i
and i → g, the angle between directed line m → i and the
movement direction, and the angle between directed line m →
i and m → g respectively. We determine the condition under
which the mobile node can move into the communication range
of its two-hop neighbor g through its one hop neighbor i:
R ≥ Lm→g sin |β − γ| .
(2)
We can see in Fig. 3 that nodes i and g and the mobile node
m constitute a triangle, so that a cosine theorem can be applied
to obtain the relation among Lm→g , Lm→i , and Li→g :
q
L2i→g + L2m→i − 2Li→g Lm→i cos(180 − α).
(3)
Further, the intersection point O (between the movement direction and the edge of the communication range of node g),
Lm→g =
q
Since Lm→g sin(β − γ) ≤ R in (2),
L2m→g − R2 ≤
q
L2m→g COS (2 (β − γ)) + R2 , we take
q
d (vm , g) = L2m→g − R2 .
(6)
Lm→g expressed in (3) is related to the speed direction of
mobile node m. Suppose Vm is the set of speed of mobile node
m and Vm := IR+ .
D = minv,g {d(vm , g)}, g ∈ Ni and vm ∈ Vm
(7)
Considering that node i has several neighbors, the neighbor set
of node i is denoted by Ni , and node g is in Ni as assumed. The
general equation we take as the NextAnnounceInterval value in
MACK is the minimum d(vm ,g) in (7). Node i is selected to
v
<
reply to the mobile node if the condition (2), (4), and ωL
i→g
◦
90 are satisfied.
E. The Active Cycle of Neighbors
In S-MAC, each node maintains a one-hop neighbor list in
which the active time for each neighbor node is not included.
To record the duration for which the neighbor node stays active in the communication range of the mobile node, we add
an active cycle field to the neighbor list in the unit of number
of frames. The value of the active cycle is initialized and updated when the mobile node receives the first SYNC packet and
the MACK packet from any of its new neighbors, respectively.
The neighbor node should be deleted from the neighbor list and
disconnected from the mobile node when the time of the active
cycle has elapsed. As shown in Fig. 4, assume r is the passby distance of the mobile node when it first receives the SYNC
packet of node c until it moves out its communication range, q
represents the intersection point between the movement direction of the mobile node and the communication edge of node c,
θ represents the angle between Lm→c and the movement direction of the mobile node.
As assumed, we can determine the values of the Lm→c , R,
and θ in triangle ∆mcq. According to the triangle geometry theorem - cosine theorem, we obtain,
r2 − 2rLm→c cos θ + L2m→c − R2 = 0.
The solution to (8) is
(8)
206
JOURNAL OF COMMUNICATIONS AND NETWORKS, VOL. 17, NO. 2, APRIL 2015
c
R
m c
q
V
r
m
Fig. 4. Topology for a mobile node to calculate active cycle of node c.
r = Lm→c cos θ +
q
R2 − L2m→c sin2 θ
(9)
only if R ≥ Lm→c sin θ.
Therefore, the active cycle of node c in the mobile node’s
neighbor list is equal to:
q


2 
j r k 
 Lm→c cos θ + R2 − L2m→c sin θ 

=
activeCycle=
vT
vT
only if R ≥ Lm→c sin θ.
(10)
Similarly, the static node, i.e., node c, also records the active
cycle frames in its neighbor list when it receives the MSYNC
from the mobile node, which means the mobile node is just coming into its communication range. It can then keep account of
the frames of the active cycle. This avoids redundant transmission to the mobile node caused by the static node not knowing
when the mobile node has left its communication range.
F. The Process of the Mobility-Supporting Mechanism
For the sake of clarity, we divide our mobility-supporting
mechanism into intra-cluster and inter-cluster processes. In
both, the same neighbor updating process is applied. For example, the static node checks whether it meets the limitation (3) and
computes the NextAnnounceInterval field in the MACK packet
to inform the mobile node when to send MSYNC again. The
mobile node and the static node add the sender to the neighbor
list and set the active cycle field based on (11) if they receive
the control packets from the sender the first time; else, they
decrease the active cycles in the neighbor list by 1 per frame
to delete the disconnected neighbors. Our scheme employs an
adaptive MSYNC announcement and ensures that the mobile
node broadcasts MSYNC when it is approaching the border of
a virtual cluster. If two schedules are included in the MACK
packet, this means it is broadcast by the border node; the mobile
node can follow these two schedules and determine when it can
broadcast MSYNC again according to the NextAnnounceInterval field in the MACK packet.
G. Communication Overhead Analysis
The communication overhead in our analysis is considered to
be the impact of the control packets on the throughput, defined
as the ratio between the number of control packets and the data
packets transmitted. In our mechanism, the control packets refer primarily to MSYNC, MACK, and original SYNC. Since
MSYNC is proposed to insert small fields in SYNC, no additional SYNC control packet is included. In our mechanism,
since only one static node is selected to communicate with the
mobile node adaptively instead of periodically, the data collision is not significant. We ignore sync and data collisions for
simplicity.
According to the synchronization period defined in the original S-MAC and MS-MAC, when the mobile node moves into
the communication range of a new neighbor, it has to wait for
9 listen-sleep cycles to receive a SYNC from the new neighbor
and become synchronized with the new neighbor, after which
only one data packet can be sent. Since only one data and one
SYNC packet are sent during the synchronization period, we
obtain the communication overhead C = 1:1.
In our mechanism, when the mobile node enters the communication range of a new neighbor, it synchronizes with its new
neighbor through adaptively broadcasting MSYNC and receives
a MACK reply from the new neighbor during the same SYNC
period at the first frame; the remaining time of the first frame and
the following nine frames of the synchronization period can be
used to transmit data. In this way, 2 control packets and 10 data
packets can be sent during the synchronization period. Therefore, we obtain the communication overhead Cm in our mechanism as Cm = 2:10 = 1:5.
Through the comparison between C and Cm , we can draw the
conclusion that the communication overhead is much smaller in
our mechanism than in the original S-MAC.
III. PERFORMANCE ANALYSIS
In a network consisting of a large number of sensors placed
in a vast two-dimensional geographical region A, the sensor locations can be modeled by a stationary two-dimensional Poisson point process. The number of sensors located in a region
A, N (A), follows a Poisson distribution of parameter. Since the
mobile sensor covers a disk of radius R, the number of sensors
located in a virtual cluster region R, N (R), can be similarly
obtained by using the Poisson point process.
To study the effect of the size of the network, i.e., the number of nodes, on the performance of the protocol, we define the
following coverage measures.
Definition. Detection time, consider a mobile target located
at a random position of virtual cluster k at time t = tk , the
detection time of the target X, is defined to be time at which the
target first enters the sensing area of a sensor in the next virtual
cluster k+1 along the moving direction of the target, i.e., the
target is first detected by the sensor.
Theorem. Consider a sensor network at time t = k, let T be
the detection time of a randomly located mobile target moving
towards a next virtual cluster with speed vs and moving direction θi refer to the stationary node i, we have T ∼ exp(2λRvs ).
Here, X ∼ exp(u) stands for P (X < x) = 1 − exp(−ux),
namely, the random variable X is exponentially distributed with
parameter u.
The expected detection time of a mobile target with randomly
PENG AND CUI: AN ENERGY-EFFICIENT MOBILE-SUPPORTING MAC PROTOCOL...
207
Table 1. Parameters used in the simulations.
Parameter
Bandwidth(kbps)
Initial energy/node(J)
txPower (mw)
rxPower (mw)
idlePower(mw)
sleepPower (mw)
Duty Cycle (%)
Traffic
Length of data packet(bytes)
Packet generation interval(s)
Value
20
1,000
386.0
368.2
344.2
0.05
10
CBR
50
5
distributed direction is E(T) = 1 / 2λRvs . Therefore, in order to
quickly detect a target, one can add more sensors, use sensors
with larger sensing ranges, or increase the speed of the mobile
node. The detection time would directly reflect the performance
such as delay, packet loss and throughput, for example, short
detection time would enable short connection setup time and
high throughput, low packet loss and delay.
kZ
kY
kW
kX
Fig. 5. Intra-cluster topology.
kW
IV. SIMULATION RESULTS
In this section, we present our evaluation of the performance
of our mobility-supporting mechanism (named MS-SMAC) in
TOSSIM [7], which is a discrete time event simulator for
TinyOS. We compared MS-SMAC with MS-MAC, where a
shorter period of schedule synchronization and neighbor updating is adopted. Similar to MS-MAC, we use RSSI to detect
mobility. Although MS-MAC has a faster schedule frequency
than S-MAC, when the moving speed of the mobile node exceeds their updating frequencies, they obtain the same results.
The simulation is divided into two parts: intra-cluster and intercluster. The distance between adjacent nodes was 20 m. The
transmission range of a node was 30 m. Table 1 shows the main
parameters for the simulation.
A. Intra-Cluster and Inter-Cluster Topologies
The network intra-cluster topology for the simulation is
shown in Fig. 5. The mobile node is located at D0. The moving
track is shown as D1→D2→D3→D0. A series of simulations
were run for 3000 s. For demonstration, we assume that the mobile node works as a sink to collect data. The source node sends
data packets to the mobile node every 50 s and is located at the
bottom of the leftmost corner.
The inter-cluster topology for simulation is shown in
Fig. 6. Nodes wake up at different times and form two virtual
clusters with manual configuration. The mobile node is located
at D0. The trajectory is formed as the mobile node moves from
left to the right along the solid line and returns to D0 from right
to the left along the dashed line.
B. Energy Consumption Efficiency
We define the energy consumption efficiency (J/byte) as the
energy consumption on each valid byte received by the sink.
Fig. 6. Inter-cluster topology.
Fig. 7 shows that MS-SMAC achieves better energy consumption efficiency than S-MAC and MS-MAC for the intra-cluster
case, since the SYNC packet introduced in our MS-SMAC helps
the mobile node set up connections with its new neighbors in a
sufficiently short time to allow more data packets to be transmitted to the mobile node, which improves energy efficiency per
received packet. In comparison, S-MAC and MS-MAC send the
SYNC packet at fixed intervals so that it is difficult to update
the connection with its neighbors adaptively, the packet bytes
are not successfully received, and the energy efficiency per received packet is influenced.
Fig. 8 shows MS-SMAC achieves a 20% and 15% lower
energy consumption efficiency (J/byte) than S-MAC and MSMAC, respectively, for the inter-cluster case under different mobile speeds. In addition to synchronization with the neighbors,
the mobile node needs to change the schedule from one cluster to another. However, S-MAC and MS-MAC send the SYNC
packet at fixed time intervals that do not change the schedule
of the mobile node in a sufficiently short time when the mobile
node travels through different clusters. The overhead of ACK
for each SYNC in our mechanism is compensated by the large
number of data packets received at the sink.
C. Packet Loss
Fig. 9 shows that when the mobile speed is high, the difference between MS-SMAC, MS-MAC, and S-MAC is clear. As
shown in Fig. 9, the packet loss due to mobility in S-MAC and
MS-MAC is approximately 20% and 15%, respectively, more
than that in MS-SMAC, with the increase in the mobile speed.
The origin of this significant difference is the fact that the neigh-
208
JOURNAL OF COMMUNICATIONS AND NETWORKS, VOL. 17, NO. 2, APRIL 2015
Fig. 9. Packet loss comparison in intra-cluster case.
Fig. 7. Energy consumption efficiency comparison in intra-cluster case.
Fig. 10. Packet loss comparison in inter-cluster case.
and MS-MAC, respectively, which would not update the schedule as fast as the mobile speed and produce more packet losses.
Fig. 8. Energy consumption efficiency comparison in inter-cluster case.
bors of the mobile node change very fast with the increase in
mobile speed; however, S-MAC updates the neighbors at a slow
fixed period such that the connection between the mobile node
and its neighbor is easily broken and large packet losses are introduced; MS-MAC uses faster neighbor updating but still at
a fixed interval that cannot adapt to the mobile speed; in our
MS-SMAC, the neighbors help the mobile node to adjust the
MSYNC frequency according to the mobile speed, and therefore, the connection between the mobile node and its new neighbors can be set up in a sufficiently short time to avoid packet
losses.
Fig. 10 shows the packet loss ratio for the inter-cluster topology. The results suggest that MS-SMAC can reduce packet
losses by at least 10% and 20% as compared with S-MAC and
MS-MAC, respectively. The reason is that our schedule adaptation mechanism can keep pace with the mobile to adaptively
change its schedule across virtual clusters, whereas the mobile
node has to wait for about 5 m and 30 s to complete the change
in the schedules from one cluster to another in original S-MAC
D. Average End-to-End Delay
Figs. 11 and 12 shows the end-to-end delay difference between MS-SMAC, MS-MAC, and S-MAC. Since the mobile
node operates as the sink in the simulation, the connection to
its neighbors significantly influences the end-to-end delay. From
Fig. 11 with the intra-cluster configuration, we can see that
the end-to-end delay in MS-SMAC is 25% and 20% less than
that in S-MAC and MS-MAC, respectively, since our mobilitysupporting mechanism builds the connection for the mobile
node with its new neighbors adaptively so that the data can be
transmitted to the mobile node with a shorter delay.
In Fig. 12 with inter-cluster configuration, it can be seen that
MS-SMAC improves the average end-to-end delay by at least
20% and 15% as compared to S-MAC and MS-MAC, respectively. Our mechanism achieves the best end-to-end delay performance, since it can change the schedule of the mobile node
as quickly as it moves through the clusters.
In addition, it can be seen in Figs. 11 and 12 that the average
end-to-end delays for MS-SMAC, MS-MAC, and S-MAC are
greatly influenced by the hops between the source and the mobile node, since the data have to be transmitted through several
more hops to the mobile node.
PENG AND CUI: AN ENERGY-EFFICIENT MOBILE-SUPPORTING MAC PROTOCOL...
209
comments.
REFERENCES
[1]
[2]
[3]
[4]
[5]
Fig. 11. Average end-to-end delay comparison in intra-cluster case.
[6]
[7]
[8]
[9]
[10]
[11]
[12]
Fig. 12. Average end-to-end delay comparison in inter-cluster case.
[13]
Dalong Zhang et al., “DE-ASS: An adaptive MAC algorithm based on
mobility evaluation for wireless sensor networks,” in Proc. WiCOM,
Chengdu, China, Sept. 2010, pp. 1–5.
S. A. Hameed et al., “Mobility-aware MAC protocol for delay-sensitive
wireless sensor networks,” in Proc. ICUMT, St. Petersburg, Russia, Oct.
2009, pp. 1–8.
Mahdi Zareei et al., “EMS-MAC: Energy efficient contention-based MAC
protocol for mobile sensor networks,” Comput. J., vol. 54, no. 12,
pp. 1963–1972, 2011.
Wei Ye, John Herdemann, and Deborah Estin, “An energy-efficient MAC
protocol for wireless sensor networks,” in Proc. INFOCOM, NY, USA,
June 2002, pp. 1567–1576.
Sung-Chan Choi and Jang-Won Lee, “An energy-efficient mobilitysupporting MAC protocol for mobile sensor networks,” IEICE Trans.
Commun., vol. E91.B, pp. 2720–2723, 2010.
Stuart MacLean and Suprakash Datta, “Improving the accuracy of
connectivity-based positioning for mobile sensor networks,” in Proc. IEEE
PIMRC, Toronto, ON, Canada, Sept. 2011, pp. 1197–1202.
Philip Levis et al., “TOSSIM: Accurate and scalable simulation of entire tinyOS applications,” in Proc. ACM SenSys, Los Angeles, California,
USA, Nov. 2003, pp. 126–137.
Huan Pham and Sanjay Jha, “An adaptive mobility-aware MAC protocol
for sensor networks (MS-MAC),” IEEE MASS, Fort Lauderdale, Florida,
USA Oct. 2004, pp. 558–560.
Dusit Niyato, Ekram Hossain, and Afshin Fallahi, “Sleep and wakeup
strategies in solar-powered wireless sensor/mesh networks: Performance
analysis and optimization,” IEEE Trans. Mobile Comput., vol. 6, no. 2,
pp. 221–236, Feb. 2007.
Fei Peng et al., “An application oriented power saving MAC protocol
for wireless sensor networks,” Wireless Pers. Commun., vol. 67, no. 2,
pp. 279–293, Nov. 2012.
Fei Peng et al., “Theoretical performance evaluation of EDCA in IEEE
802.11e wireless LANs,” European Trans. Telecommun., vol. 21, no. 5,
pp. 478, Aug. 2010.
Fei Peng et al., “Performance analysis of IEEE 802.11e enhanced distributed channel access,” IET Telecommun., vol. 4, no. 6, pp 728–738,
Apr. 16, 2010.
Bharat Shrestha et al., “An optimization-based GTS allocation scheme for
IEEE 802.15.4 MAC with application to wireless body area sensor networks,” in Proc. IEEE ICC, Cape Town, South Africa, May 2010.
V. CONCLUSIONS
This paper focused on the key criteria of energy-efficient
MAC to support and promote the performances of mobile applications in WSNs. The most important feature of our mechanism is that the mobile node adaptively builds connections with
its new neighbors in an active manner based on the two-way
handshake for the speed and schedule information exchange in
a frame time of the SYNC period and minimizes the overhead.
Simulation results demonstrated that our mechanism can improve energy efficiency performances by approximately 20% as
compared with S-MAC under mobility situations. In the future,
we will conduct more experiments with random topologies and a
varying number of mobile nodes to investigate the effectiveness
of our method. Future work will also consider the loss of control
packets, packet collisions, and errors that affect the performance
of our method. Additionally, we will integrate the MAC layer
mechanism with the route protocols using a cross-layer method
to further improve the mobile performances in WSNs.
ACKNOWLEDGMENTS
This work is jointly supported by the National Nature Science Foundation under Grant 61102009, and the Natural Sciences and Engineering Research Council of Canada (NSERC).
The authors appreciate anonymous reviewers for their helpful
Fei Peng is a ssociated with UBC. She participated in
drafting of IETF and IEEE standard; undertook Program Committee for ICC, WCNC, IPCCC and etc,
reviewed for major international journals in the computer and networking areas; published over 40 papers in the world conferences and journals and held
national and international patents. Main research areas are in wireless/mobile communication networks
and information security, wireless sensor networks,
biomedical informatics, data (text) mining and distributed computing, and artificial intelligence.
Meng Cui is an Academic Leader of Traditional Chinese Medicine (TCM) and is the creator for the subject
of TCM informatics.