J. Parallel Distrib. Comput.

72 (2012) 856867
Contents lists available at SciVerse ScienceDirect
J. Parallel Distrib. Comput.
journal homepage: www.elsevier.com/locate/jpdc
Coordination in wireless sensoractuator networks: A survey
Hamidreza Salarian

, Kwan-Wu Chin, Fazel Naghdy

School of Electrical, Computer, and Telecommunications Engineering, University of Wollongong, Northfields Avenue, Australia 2522, Australia
a r t i c l e i n f o
Article history:
Received 21 March 2011
Received in revised form
16 August 2011
Accepted 19 February 2012
Available online 28 March 2012
Keywords:
Wireless sensor and actuator networks
Coordination
Energy efficiency
Actuators
a b s t r a c t
Wireless SensorActuator Networks (WSANs) have a myriad of applications, ranging frompacifying bulls
to controlling light intensity in homes automatically. An important aspect of WSANs is coordination.
Unlike conventional Wireless Sensor Networks (WSNs), sensor and actuator nodes must work hand-in-
hand to collect and forward data, and act on any sensed data collaboratively, promptly and reliably. To this
end, this paper reviews current state-of-the-art techniques that address this fundamental problem. More
specifically, we review techniques in the following areas: (i) sensoractuator coordination, (ii) routing
protocols, (iii) transport protocols, and (iv) actuator-to-actuator coordination protocols. We provide an
extensive qualitative comparison of their key features, advantages and disadvantages. Finally, we present
unresolved problems and future research directions.
2012 Elsevier Inc. All rights reserved.
1. Introduction
Advances in micro-electro-mechanical system (MEMS) tech-
nology and wireless communications have enabled the devel-
opment of small size, low-cost, low-power and multi-functional
devices that are typically equipped with sensors such as seis-
mic, magnetic, thermal, visual, infrared, acoustic and radar.
Consequently, they can be used to measure tremor, distance, di-
rection, speed, load and pressure, temperature, humidity, light,
vibration, motion and acoustic parameters. Fig. 1 shows a block di-
agram of a sensor node. Typically, a sensor node is equipped with
a power supply, transceiver, memory, and processing and sensing
unit. There is also an analog to digital converter (ADC) unit. The
processor collects and processes the signal taken from the sensor
unit and sends it to the transmission unit [3,6].
Sensor nodes are able to self-organize and self-configure into
a Wireless Sensor Network (WSN), which has wide ranging
applications, including precision agriculture [8,54,42], monitoring
of pests [20], and volcanoes [27] to name a few. Fig. 2 shows a
typical WSN, comprising of one sink (or base station) and a number
of sensor nodes scattered randomly or according to a predefined
methodology in a physical space. A key characteristic of WSNs is
that nodes help each other forward data to the sink node, which is
usually connected to the Internet via a gateway. Interested users
obtain information about the state of the field via the gateway and
base station.
Recently, a number of researchers have considered equipping
sensor nodes with an actuator. Briefly, an actuator is a transducer

Corresponding author.
E-mail addresses: hs528@uow.edu.au, hs528@uowmail.edu.au (H. Salarian),
kwanwu@uow.edu.au (K.-W. Chin), fazel@uow.edu.au (F. Naghdy).
that converts an electric signal to a physical movement. Example
actuators include valves that control the water or gas outflow
from a pipe, electrical motors that open/close doors and windows,
and switches that turn on/off heaters or lights. Fig. 3 shows the
key components of an actuator node, i.e., transmission, processor,
storage, controller (decision), Digital to Analog Converter (DAC)
and actuation. The decision unit functions as an entity that takes
sensor readings and generates action commands. An example
actuator node is the ME8300 Wireless Zone Valve actuator from
Spartan [48], which includes a TransCeiver Module (TCM) [16],
16 MHz 8051 CPU, 32 kB Flash and 2 kB RAM, 8bit DAC, and
868 MHz/315 MHz transceiver, Alternating Current (AC) motor as
an actuation unit and 24 V AC power supply.
The resulting Wireless Sensor/Actuator Network (WSAN) not
only consists of sensor nodes that measure specific environmental
parameters, but some of them have the capability to act on
the environment through actuators. A distinguishing feature of
actuator nodes is that they are usually resource-rich, i.e., they
have ample power supply, higher computational power and longer
communication range than sensor nodes. For these reasons, there
are fewer actuators deployed in a given environment [40].
Fig. 4 shows a typical WSAN, comprising of geographically
distributed network of wireless sensor and actuator nodes. Similar
to WSNs, sensor nodes gather information, but instead of a sink
node, they transmit collected data to actuator nodes via single-hop
or multi-hop transmissions. These actuator nodes then decide the
appropriate responses to effect the sensor field. Note that the sink
node is used to provide monitoring and management ability for
end users by communicating with sensor and actuator nodes.
Table 1 outlines the various applications of WSANs. For
example, home automation [37,18,25], animal control [55], mon-
itoring the health of an infrastructure [44,65,28,29,23], preci-
sion agriculture [32,42,17,61], intelligent buildings [35,11,26,39],
0743-7315/$ see front matter 2012 Elsevier Inc. All rights reserved.
doi:10.1016/j.jpdc.2012.02.013
H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867 857
Fig. 1. A block diagram of a typical sensor architecture.
reducing sewer overflow [52,31] and traffic light control [68,46].
The number of sensor and actuator nodes varies according to appli-
cationtypes. For example, insewer management systemandtraffic
light control, due to the large coverage area, the number of sensors
and actuators may be in the order of hundreds as compared to a
few nodes when they are used to monitor buildings. In home au-
tomation and precision agriculture, there are more sensor nodes
than actuator nodes due to the coverage area. In animal control and
infrastructure health monitoring applications, because sensor and
actuator nodes are integrated together, there are equal number of
sensors and actuators.
A key design parameter is the delay tolerance of applications.
Here, delay tolerance is defined as the allowed time delay between
sensing and acting. This parameter is important because it governs
the time frame before actuators respond to sensed data. For
example, whena WSANis usedtomonitor aninfrastructure, sensor
nodes have an extremely low duty cycle, and hence, are asleep
most of the time. They either wake up periodically to test the
structure, say once a day at midnight, or may be woken to test
the structure immediately after a catastrophic event such as an
earthquake, a collision, or a blast [23]. In precision agriculture [17]
and sewer management system [52], sampling rates of 10 and
14 min are sufficient to control and monitor key parameters.
In lighting system applications, a sampling rate of two seconds
is sufficient as light intensity follows an analogue pattern [18].
Lastly, animal control applications require a sampling interval of
500 ms [55].
In WSANs, guaranteeing that actions are carried out as per
the delay tolerance of an application is highly dependent on
three processes: distributed collection of data by sensor nodes,
forwarding said data to actuator nodes, and cooperative decision
by actuator nodes on how to perform the required action. These
three processes are referred to as coordination in WSANs. In
addition to delay tolerance, energy efficiency is important because
each sensor and actuator node has finite battery capacity.
This paper focuses on the coordination problem in WSANs.
Specifically, we review past studies that improve communications
between sensors and actuators, and also between actuators. The
critical problems include routing and transport of sensed data and
commandsboth of which play a crucial role in the operation of
a WSAN as they govern how actuators respond to one or more
events. Moreover, any solutions must ensure packet delays are
Fig. 3. A block diagram of a typical actuator architecture.
bounded, and packets are delivered reliably as they may contain
information used to locate actuator nodes, and also control the
movement and responses of actuator nodes. Given the importance
of these problems, this paper makes the following contributions:
It presents an extensive review of WSAN applications. We
categorize them according to their requirements in terms
of bounded delay, reliability, bounded task completion time,
service differentiation and energy efficiency.
Unlike a prior work [50], it reviews four newly published
sensor-to-actuator coordination protocols. Also, we compare
each protocol in terms of coordination model, localization
procedure, actuator mobility and energy efficiency. We like to
point out that [50] is only a categorization of papers as opposed
to a proper literature review. Hence, it lacks analysis on the
strengths and weaknesses of WSAN protocols.
We compare six routing protocols designed for WSANs in terms
of route discovery and maintenance, delay-bound and energy
consumption. In addition, we review five transport protocols
and analyze their ability to provide reliability and service
differentiation.
We highlight the advantages and drawbacks of six actuator-to-
actuator protocols. We outline their ability in terms of handling
multiple events and guaranteeing task completion time.
Lastly, our paper highlights key challenges and problems
associated with coordination in WSANs, and present key future
works.
In the next section, we give an overview of two main WSAN
architectures. The type of architecture will have a significant
impact on the coordination protocol used to meet application re-
quirements. With this in mind, we present an overview of the co-
ordination problem in Section 3, followed by prior solutions to
sensoractuator, actuator-to-actuator, routing and transport prob-
lems in Section 4. Then in Section 5, we present future trends
and open research problems. Finally, Section 6 presents our
conclusions.
2. Network architecture
The network architecture of WSANs can be categorized as
either fully-automated or semi-automated [3]; see Fig. 5. In the
fully automated architecture, actuator nodes coordinate amongst
themselves and decide on a plan of action based on sensed data.
Fig. 2. Typical WSN architecture.
858 H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867
Fig. 4. A wireless sensoractuator network.
Table 1
Summary of WSAN applications.
Applications Sensor types Actuator types Delay tolerance Number of sensors Number of actuators
Home automation [37,18,25] Temperature, humidity,
light, acoustic and
occupancy sensors.
Relays, light, switches,
motors, noise-masking
system adjusting and
valves.
2 s More than 100
sensor nodes
2050 actuator nodes
Animal control [55] Location and velocity
meter sensors
Stimuli board 500 ms 2030 sensor
nodes
2030 actuator nodes
Infrastructure health
monitoring [44,65,28,29,23]
Velocity meter and
piezoelectric sensors
Hydraulic or piezoelectric
shakers
1 Day 20100 sensor
nodes
20100 actuator
nodes
Precision agriculture [32,42,17,61] Temperature, humidity,
light and soil moisture
sensors
Valves, relays light
switches
10 min More than 100
sensor nodes
2030 actuator nodes
Intelligent Buildings [35,11,26,39] Smoke detectors, glass
break and motion sensors
Water sprinklers and
cameras
2 s 2050 sensor
nodes
2030 actuator nodes
Sewer management [52,31] Water level Valves or city sewers 14 min More than 100
sensor nodes
More than 100
actuator nodes
Traffic light control [68,46] Magnetic sensor Relays of traffic lamp 10 s More than 100
sensor nodes
More than 100
actuator nodes
In the semi-automated architecture, however, sensor nodes route
their data to actuators via a sink. Note, similar to WSNs, a central
entity or sink may be used to collect and process sensed data, and
send commands to actuator nodes. Alternatively, the sink node
may be usedjust for monitoring andmanaging the overall network.
Both architectures have their advantages and disadvantages.
A semi-automated architecture is similar to WSNs. In addition to
using existing WSNs protocols, due to its centralized property,
there is no need for a distributed communication and coordination
protocol between sensor and actuator nodes. However, since
sensed data is routed through a sink rather than a nearby actuator,
communication latency can be significant. Moreover, nodes near
the sink will deplete their energy quicker than those further
away [33]. On the other hand, in the fully-automated architecture,
sensed data is sent to different actuator nodes. Hence, the
communication load can be distributed evenly amongst actuator
nodes, and thereby, extend the lifetime of a WSAN [33].
We like to point out that WSNs are passive, where sensor
nodes simply record data and send them to one or more sinks
for processing, which may then issue new commands to change
the sample rate. However, WSANs are active in that sensed data
governs the behaviors of actuators, which in turn causes these
actuator nodes to effect the environment, and hence change data
collected by sensors. In short, there is a close coupling between
sensor and actuator nodes. As we will elaborate in Section 3, this
coupling yields a number of novel challenges.
3. Coordination
Coordination is a fundamental problem in WSANs. The exact
coordination used, however, is dependent on the network
a
b
Fig. 5. Two architectures of WSANs: (a) fully-automated and (b) semi-automated
architectures.
H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867 859
a
b
Fig. 6. Sensoractuator coordination mechanism: (a) single and (b) multiple actuators.
architecture. More specifically, coordination is carried out by
the sink in the semi-automated architecture. On the other
hand, in the fully-automated architecture, there are two modes
of communication: sensor-to-actuator and actuator-to-actuator
coordination. In the former, sensor nodes are required to find an
appropriate actuator node to send their data. In the latter, the
actuator nodes coordinate amongst themselves to deal withsensed
data [3,50,45]. In this study, the focus will be on fully-automated
architecture as existing WSNs protocols can be deployed in the
semi-automated architecture. Readers who are interested in these
WSNs protocols are referred to [56,25] and references therein.
In the fully-automated architecture, distributed protocols are
needed for sensor-to-actuator coordination. This is an important
function as a correct response from actuator nodes cannot be
achieved unless sensed data arrives in a timely manner and is
received correctly by the corresponding actuator nodes. This gives
rise to the following fundamental question: how do sensor nodes
determine the best actuator node(s) to send their data? This
is important as it enables them to determine the location and
intensity of an event before taking action.
Once an event occurs the data may be forwarded to an actuator
node (see Fig. 6(a)), or multiple actuator nodes (see Fig. 6(b)).
In the first case, sensor nodes in the event area communicate
with each other to find the nearest actuator node(s) that covers
the said area and has sufficient energy to carry out the required
task. The advantage here is that actuator nodes are excluded from
this coordination effort. Instead, coordination is carried out by
sensor nodes. As a result, sensor nodes around the event area
will deplete their energy quicker. In the second case, each sensor
node independently selects an actuator node [62]. However, there
is no sensor-to-sensor coordination, which may overload a given
actuator or cause sensor nodes around an actuator node to deplete
their energy quicker.
Once sensor nodes decide on one or more actuators, their next
task is tolocate andselect suitable actuator nodes. This task is made
difficult in some applications where actuator nodes are mobile.
Hence, there needs to be mechanisms that enable actuator nodes
to update sensor nodes of their location, and vice-versa. Moreover,
these mechanisms must be energy efficient as continuous tracking
of actuator nodes increases the duty cycle of resource constrained
sensor nodes.
The next issue is establishing one or more routing paths to
actuator nodes. The main challenge is constructing one or more
paths that meet the delay requirement of applications, and are
sufficiently reliable to carry data between sensors and actuator
nodes. A key consideration is congestion avoidance, especially in
areas surrounding an actuator node. Otherwise, a congestion path
will lead to packet loss and increased end-to-end delay. Another
issue of importance is route construction and maintenance to one
or more actuator nodes. Moreover any unicast or multicast routing
protocols must consider the low duty cycle of nodes, and the
mobility of actuator nodes.
Actuators must coordinate amongst themselves to ensure at
least one of them will respond to any event that arises. Fig. 7
shows two decision categories of actuatoractuator coordination.
In the centralized decision category, whenever an actuator node
receives an event from a sensor node, it sends the information to
a predetermined, central actuator node or decision center, which
then decides the best group of actuator nodes to perform the
required task; see Fig. 7(a). In the distributed decision category,
after receiving event information, actuator nodes communicate
with each other and send sensed data, their residual energy,
current position and action range to other actuator nodes in the
network. This information is then used by actuators to determine
whether to participate in an action; see Fig. 7(b) [62].
The advantage of centralized actuatoractuator coordination is
that the decision center is able to select the best actuator nodes
to carry out a task, especially when there are multiple events. This
advantage, however, is the increased end-to-end delay as actuator
nodes need to send their data to a central station. On the other
hand, distributed actuatoractuator coordination reduces actuator
response time in comparison to the centralized model because
each actuator node is able to make local decision based on received
data. Onthe other hand, the distributedmodel increases the energy
consumption of actuator nodes as they need to communicate
with each other after each event. Lastly, in both coordination
models, there needs to be a mechanism to handle the occurrence
of multiple events.
In summary, the close coupling between actuator and sensor
nodes poses a number of issues not found in conventional WSNs.
These issues include [3,57,33,49,58]:
Bounded packet delayactuators have to act on sensed data
quickly. Otherwise, it would be detrimental to the operation
of a WSAN. Specifically, any communication protocol must
ensure packets do not exceed an applications delay tolerance.
This means sensor nodes need to form low delay and reliable
paths to one or more actuator nodes, and actuator nodes must
collaboratively reach a consensus on the best response to one
or more events. For example, intelligent building applications
require the control of water sprinklers within two seconds; see
Table 1.
860 H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867
a
b
Fig. 7. Actuatoractuator coordination mechanism: (a) centralized, (b) distributed.
Reliabilityto have correct execution of actions, actuator nodes
need to receive sensed data within a pre-determined time
period in order to re-construct an event, understand its
intensity, location and coverage, and lastly determine the
appropriate number of actuators that are deployed in response
to the event. Unfortunately, sensed data and commands may be
lost due to congestion, bit error, or bad connectivity. Therefore,
protocols must be developed to address one or more of these
issues such that sensed data and control information are
communicated to sensor and actuator nodes reliably.
Bounded task completionapplications such as fire control
systems require bounded task completion time. Specifically,
they need a bound on the elapsed time from when sensor
nodes report the occurrence of a fire to when water sprinklers
completely extinguish the fire.
Service differentiationmultiple events with different urgency
level will need to be treated accordingly. For instance, in home
automation systems, there will be sensors that are used to sense
temperature and lighting, while some will be responsible for
tracking the whereabouts of a person. These two types of sensor
nodes generate traffic flows which require different delivery
and task completion time.
MobilityThe mobility nature in WSAN is completely different
to WSNs. Mobile elements in WSN are designed to save the
energy of sensor nodes by collecting data from sensor nodes
directly or via rendezvous or cache points [60,9,53]. On the
other hand, inWSANs, mobile actuator nodes are usedto reduce
end-to-end packet delay and guarantee task completion times;
e.g., in areas with high frequency of events.
We like to point out that protocols developed for WSNs are
generally not applicable to WSANs for a number of reasons. Sensors
nodes are power constrained, and have limited transmission
range and memory size, while actuator nodes are resource rich
and have better transmission capabilities as well as buffer size.
Hence, WSAN protocols must consider the resource constraints
of both types of nodes. The information flow in WSNs is from
sensor nodes to a sink, which forma many-to-one communication
pattern. In a WSAN, the communication is many to many as it
can take place between any nodes. Moreover, sensed data must
be routed to the corresponding actuators that are able to mount
the appropriate actions. Hence, localization, reliability and real-
time communications are crucial to the operation of WSANs. We
note that there are similar challenges andproblems betweenWSNs
and WSANs in the following areas: spectrum management [4,5],
MediumAccess Control (MAC) [15] and security [38,12]. Hence, we
do not consider them in this paper, and refer interested readers
to the previously cited papers, and references therein for more
information.
4. Literature review
We divide the literature into four categories: sensor-to-
actuator, routing, transport, and actuator-to-actuator coordination
each of which encapsulates the problems and challenges
discussed in Section 3.
4.1. Sensoractuator coordination
The fundamental problem addressed by the previous studies is
to determine energy efficient methods/protocols that allowsensor
nodes to select and locate an appropriate actuator node. The main
issues to consider include the coordination model used, i.e., single
or multiple actuators, and how actuators update sensor nodes of
their location.
Shah et al. [45] propose a Cluster-based Coordination and
Routing (CCR) protocol. Clustering is a standard approach used [1]
extensively in WSNs to decrease the energy consumption of sensor
nodes. The authors construct a number of clusters in a given
network region based on a formula that considers the dimension of
an area, number of deployed sensor nodes, and transmission range
of sensor nodes. Then for each cluster, a cluster head is selected
based on the residual energy of sensor nodes, data transmission
rate and node density. A new cluster head is selected once its
residual energy becomes the lowest in the cluster. The key roles
played by cluster heads are data aggregation, and forwarding
packets onpaths that meet the following criteria: end-to-end delay
that is within a given delay bound, and least energy expenditure.
If there are multiple paths, a cluster head selects the path with
nodes having the most residual energy. CCR distributes energy
consumption uniformly amongst nodes and prevents the energy
holes problem [59]. However, the most important issue is the
overheads required to form cluster. In [45], sensor nodes are
required to collect position information from all nodes in the
network before they are able to construct the optimal number of
clusters for a given WSAN.
Melodia et al. [40] proposed a distributed and adaptive event-
based partitioning method for sensoractuator coordination.
When an event happens, sensors in the event area independently
determine the nearest actuator node. More specifically, all sensor
H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867 861
Table 2
Comparison of sensor-to-actuator proposals.
Prior works Coordination Actuator localization Actuator mobility Energy conservation techniques
Melodia et al. [40] Multiple actuators Dynamic clustering No Rotate cluster head role and use alternate
paths
Shah et al. [45] Single actuator Static clustering No Event-based clustering
Zeng et al. [64] Multiple actuators Actuator nodes periodically broadcast their
position
Yes No consideration
Melodia et al. [39] Multiple actuators Voronoi diagram and Kalman filtering to
predict position.
Yes Exploit Voronoi diagrams to scope broadcast
messages
nodes in the event area that is sending data to the actuator form
a cluster and a delivery tree that is rooted at the actuator node.
The key advantage of this method is that sensors and actuator
within a given scope of the detected event are required to be
active. This means sensor nodes save energy when there are no
events in the network because there is no need to spend energy
for cluster maintenance. On the other hand, clusters are formed
independently by sensor nodes using only local information. Each
sensor node selects the nearest actuator node as a cluster head.
This means sensor nodes are not required to communicate with
other nodes to select cluster heads, as is the case in [45]. This
method, however, is not suitable for scenarios withfrequent events
as constructing clusters and locating the nearest actuator node will
consume a significant amount of energy.
Zeng et al. [64] developed a real-time sensoractuator coor-
dination protocol. Sensor and actuator nodes are aware of their
position. Inaddition, actuator nodes are mobile. Actuator nodes pe-
riodically broadcast their position, residual energy and load. When
sensor nodes receive the broadcast message, they select the near-
est actuator node with the maximum residual energy and mini-
mum load for event reporting. Interestingly, a sensor node is able
to request an actuator to move closer whenever the end-to-end de-
lay of its current path exceeds a given threshold. This model, how-
ever, is not energy efficient as it requires actuators to broadcast
messages periodically, which creates significant overheads. Also,
the movement of an actuator node in response to a sensor nodes
request may lead to higher end-to-end delays.
Melodia et al. [39] propose a sensor-to-actuator framework
that considers mobile actuator nodes, which comprises of a novel
location management scheme, where an actuator broadcasts a
message to inform all sensor nodes of its new position. Actuator
nodes are aware of the location of all sensor nodes, and are
equipped with two radios tuned to a distinct channel; one for
communicating with sensor nodes, and other one to communicate
with other actuator nodes. In order to restrict the broadcast
scope of an actuators messages to only relevant sensor nodes,
the authors use Voronoi diagram to partition the network area
into a number of convex polygons or scope. Each sensor node is
then assigned to its nearest actuator node. Every time an actuator
node changes its position, it communicates with other actuator
nodes to determine the polygon of each actuator node. As a result,
energy consumption is reduced because there is no need for sensor
nodes to relay update messages. Melodia et al. [39] also proposed
a method to reduce energy consumption when actuator nodes
are mobile. Specifically, they use Kalman filtering to predict the
position of a mobile actuator at a given time instance. Actuators
send update messages periodically, which sensor nodes then use
to predict an actuators future location. This has the effect of
reducing location update messages, and as a result, there is less
communication overhead. In the Voronoi diagram and Kalman
filtering method [39], each actuator node is aware of the position
of all sensor nodes in the network. The location information is
either set up manually, which is impractical, or communicated
by actuators. The drawback, however, is that as the number
of actuators increases, the network will incur a non-negligible
amount of signaling overheads associated with localization.
Table 2 summarizes sensoractuator coordination protocols
according to their coordination, and actuator localization mecha-
nism as well as whether actuators are mobile. All proposed mod-
els are based on multiple actuator coordination except for the
CCR protocol [45]. As mentioned in Section 3, multiple actuator
coordination incurs communication overhead, in addition to in-
creasing the energy consumption of actuator nodes. The difference
between static [45] and dynamic [40] clustering is whether sensor
nodes communicate with each other to elect an actuator node as
the cluster head. More specifically, in dynamic clustering, sensor
nodes individually select an actuator node as a cluster head while
in static clustering, all sensors in a cluster communicate with each
other and choose an actuator node as their destination. This means
dynamic clustering can be categorized as a multiple actuator coor-
dination mechanism while static clustering is a single actuator co-
ordination mechanism. Actuator localization, mobility and energy
conservationare three key challenges insensoractuator coordina-
tion. As Table 2 shows, all proposed models, except Zeng et al. [64],
provide an energy efficient way for sensor nodes to locate actu-
ator nodes. However, Zeng et al. [64]s model is not energy effi-
cient as they only focus on actuator mobility. The use of Voronoi
diagram and Kalman filtering, as proposed in [39,22], is promis-
ing for static WSANs. However, it is unclear, how this methods
fare in WSANs with mobile actuators, especially when considering
the non-negligible amount of signaling overheads associated with
localization.
4.2. Routing protocols
Routing protocols play a critical role in WSANs as sensed
data or control messages have pre-determined Quality of Service
(QoS) requirements. The main issues are route discovery, route
maintenance, selecting a path that is within the required delay
tolerance, and energy efficiency. Many routing protocols have been
proposed for WSNs in recent years. However, to the best of our
knowledge, none of the existing papers to date consider research
challenges resulting from the coexistence of sensors and actuator
nodes. That is, the communication path is between sensors and
one or more actuator nodes. For example, a sensor node can send
data to multiple actuator nodes or select the best actuator node
according to predefined parameters. Moreover, as actuator nodes
are resource rich, they can bear the burden of the routing process.
In addition, actuator nodes play the role of mobile sinks, which has
a significant impact on how routing is carried out in a WSAN.
The authors of [51] analyzed the performance of three ad-hoc
network routing protocols in terms of their ability to decrease
end-to-end packet delays. They analyzed Destination Sequenced
Distance Vector (DSDV) [10], Dynamic Source Routing (DSR) [30]
and Ad Hoc On-Demand Distance Vector (AODV) [41]. The authors
use these protocols for actuator-to-actuator communication and
coordination. DSDV is a proactive routing protocol that requires
nodes to periodically exchange routing tables. DSR, a reactive
routing protocol, embeds the route to be taken in the header of
each packet. In addition, each node maintains a cache of routes,
which they compile from packets they have processed, and from
862 H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867
those transmitted by neighboring nodes. AODV, another reactive
routing protocol, works similarly to DSR where route requests are
sent in an on-demand manner. However, it does not maintain
caches nor require source routes in packets. The results in [47]
show that at startup time, packets routed using DSR and AODV
arrive quicker than those forwarded by DSDV. This is because
routers using DSDV require a non-negligible amount of time to
achieve route convergence. However, after the network becomes
stable, DSDV yields the lowest end-to-end delays as a source
does not need to establish a route in an on-demand manner. The
drawback of DSDV protocol is that it is not energy aware and
wastes network bandwidth. To guarantee low end-to-end packet
delays, sensor nodes are required to update their routing tables
periodically, which leads to an increased in power and bandwidth
consumption.
Hu et al. [22] propose to use anycast and mobile actuator nodes
to minimize energy consumption in WSANs. The proposed model
is a variant of the Directed Diffusion (DD) [19] communication
paradigm, which was proposed for WSNs. An anycast tree is built
froma sensor to actuator nodes, where actuators formthe leaves of
the resulting tree. Every time an actuator node joins the network, it
broadcasts a route discovery message. This causes all sensor nodes
to update their anycast tree. To decrease energy consumption and
communication overheads, an actuator node floods an explicit
leave message to remove itself from the anycast tree. This model
is suitable for applications that can tolerate minor packet loss.
The main drawback is that actuator nodes may generate excessive
signaling overhead when they change their location given the use
of flooding to update their location.
Boukerche et al. [7] developed a QoS aware routing protocol.
Actuator nodes broadcast subscription messages with the energy
level and hop count field set to zero. Each sensor node that receives
this message updates these fields and collects information about
its hop level from actuator nodes and the energy level of its
neighboring nodes. Whenever a sensor node wants to send a data
packet, it updates the packets delivery time based on the elapsed
time from when the packet was generated. Sensors uniformly
distribute packets to actuators based on packets delivery time
in order to normalize their energy usage. When paths have the
same length, sensors will select the one with the highest energy
level. The QoS-aware source routing protocol selects a path that
meets a packets delivery time. In the scenario where no paths
can be found, the authors did not propose any solution. One
mechanismto alleviate this problemis for sensor nodes to increase
their transmission power. However, this needs a coordination
mechanism that determines which nodes on a path are required
to increase their transmission power. Moreover, such mechanism
must balance energy consumption due to higher transmission
power and end-to-end delay.
Cayirci et al. [14] propose a Power Aware Many-to-many
Routing (PAMR) protocol. The basic idea is to create a multicast tree
rooted at a sensor node toward actuator nodes using a publish and
subscribe method. At start-up, actuators broadcast their interest
to sensor nodes. A sensor node with the required data records
the number of hops, minimum energy of nodes on the route, and
energy that will be consumed by nodes when forwarding packets.
A sensor node then selects the route with the smallest weight to
minimize the formation of energy holes, and also provides end-
to-end packet delay. In other words, each sensor node builds an
energy efficient delay aware multicast tree to actuator nodes that
are interested in its sensed data. The drawback of this model is
that PAMR cannot guarantee end-to-end packet delays. This is
because sensor nodes have a fixed transmission power, and they
do not have any option when the delay on the minimum weight
path is larger than the required packet delivery time. To overcome
this problem, the authors also propose a different PAMR version,
called Power Controlled PAMR (PCPAMR), where sensor nodes
increase their transmissionpower whenthe required delivery time
of a packet is less than the delay of an existing path. In other
words, PCPAMRbalances the requiredend-to-endpacket delay and
energy consumption of sensor nodes. The problemwith PCPAMR is
that when each sensor node individually changes its transmission
power, it causes non-uniform energy consumption and energy
holes to occur along the packet forwarding path.
The authors of [36] propose Delay-Energy Aware Routing Pro-
tocol (DEAP). Sensor nodes wake up once in every predetermined
time period, transmit data and then sleep. The time length of an
active period depends on the number of data packets in the trans-
mission queue. A large queue size causes a long active time. Hence,
sensor nodes make local decisions as to sleep or to be active based
on their queue length. Whenever a sensor node wants to send data
packets, it selects one that is closest to an actuator node. DEAP dis-
tributes energy consumption between a source nodes forwarding
set because each time a source node wants to send data packet,
its forwarding set may change based on the active and sleep pe-
riod of its neighbors. DEAP prolongs WSAN lifetime and reduces
energy consumption in sensor nodes but at the expense of end-to-
endpacket delay. This happens because innormal scenarios, source
node builds the shortest forwarding path to the nearest actuator
node but when sensor nodes sleep according to their load, other
nodes have to rebuild their routing path as nodes may decide to
enter sleep mode, or new nodes may be available.
In applications such as the water irrigation system [17],
actuators are in charge of opening or closing valves depending
on the need to water areas as determined by sensed data
provided by sensors. This means sensors may send data to multiple
actuators. Therefore, in order to save energy, Sanchez et al. [43]
proposed an energy-efficient multicast routing protocol. As the
problem of finding an energy-efficient multicast tree is NP-
complete, the authors used a heuristic that incorporates nodes
location to build energy-efficient multicast paths. In particular,
their heuristic used the destination and energy cost to each
destination during tree construction. Moreover, they consider
merging paths onto a common node, which then serves multiple
destinations. The multicast problem proposed by Sanchez et al.
however is not delay-aware. When the number of actuator nodes
increases, the computational power and storage requirement
associated with finding candidate sets and conducting a merge
become significant for sensor nodes.
Table 3 summarizes the aforementioned routing models
according to their route discovery and maintenance mechanism.
We also compare the approach used to ensure the end-to-
end packet delay of a path is bounded. A notable approach in
this respect involves actuator nodes issuing feedback to sensor
nodes on a given path to increase their transmission power
until the required end-to-end delay path is met. On the other
hand, protocols that place the responsibility of route discovery
and maintenance on sensor nodes have high communication
overheads and energy consumption.
4.3. Transport protocols
Reliability is a critical issue as actuators use sensed data to
reconstruct events before launching the corresponding action(s)
for a given event. Therefore, it is important for actuators to
ascertain the type, location and intensity of events reliably.
The main approaches taken in prior works include varying the
transmission power of nodes, and number of retransmissions.
Apart from that, service differentiation is a key concern in
scenarios with multiple events. In general, transport protocols
should provide both reliability and real-time data communication
between sensor and actuator nodes. In addition, they must also
guarantee event reliability; defined as the amount of data that is
H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867 863
Table 3
Comparison of routing protocols.
Prior works Route discovery Route maintenance Delay bound Energy conservation techniques
Dinh et al. [51] Sensor nodes Sensor nodes periodically exchange
routing tables
No None
Hu et al. [22] Actuator nodes Actuator nodes Select the nearest actuator node Anycast tree
Boukerche et al. [7] Actuator nodes Actuator nodes periodically broadcast
subscription messages
Select the path that meets packet
delivery time
Distributes packets between
existing paths
Cayirci et al. [14] Actuator nodes Actuator nodes periodically broadcast
subscription messages
Vary transmission range Multicast tree
Durresi et al. [36] Sensor nodes On demand No Sleep and wake-up
Sanchez et al. [43] Sensor nodes None. No Multicast tree
reliably delivered to an actuator node within a given time bound.
In addition, anycast communication is used frequently as sensed
data needs to be routed to the closest actuator node(s). For these
reasons, transport protocols developed for WSNs are generally not
suitable for use in WSANs.
Zhou et al. [34] propose a real-time data transport protocol
called POWER-SPEED. Sensor nodes calculate the hop-by-hop
delay that a packet will experience on a given path. They then use
the expiration time of each packet to determine their transmission
power. A packet that has a higher delay tolerance will traverse
more hops as sensor nodes will use a lower transmission power,
which increases the number of hops on a given path. Otherwise,
the packet will be forwarded over a route with fewer hops. The
limitation of POWER-SPEEDis that it does not consider congestion.
In other words, increasing or decreasing transmission power does
not necessarily alleviate or avoid congestion on a given route. In
addition, POWER-SPEED cannot handle multiple events.
Gungor et al. [2] propose RT2, an energy efficient, reliable and
real-time transport protocol. The authors define event reliability
as the percentage of event data received by actuator nodes within
a given time bound. To provide event reliability, RT2 uses the
Time-Critical Event First (TCEF) [2] scheduling algorithm, where
sensor nodes service packets according to their deadline. An
interesting feature is that actuator nodes will send feedback
messages to sensor nodes in order to decrease their sampling or
transmission rate if the observed event reliability is above a given
percentage. Sensor nodes are notified of impending congestion
when their buffer overflows or when the average packet delay
exceeds a threshold. Inthis situation, sensor nodes set a Congestion
Notification (CN) bit in their packets to inform actuators of the
impending congestion [21]. Upon receiving packets with the
CN bit set, and observing that the event reliability is below a
given threshold, actuators inform sensor nodes to decrease their
sampling or transmission rate. The advantage of RT2 is that it saves
the energy of sensor nodes when there is congestion. On the other
hand, its drawback is that there is no mechanism to overcome
congestion and guarantee event reliability.
Ngai et al. [67,66] propose a latency-oriented fault tolerant
(LOFT) transport protocol for WSANs. Sensors and actuators are
aware of their location. Each sensor node maintains and schedules
queues according to the urgency of events. When a sensor node
wants to send a data packet to a destination, it selects the path
that meets the required delivery time. In scenarios where there
are multiple paths, the source node uses the least congested path.
The authors also propose to use mobile actuators to service areas
with high frequency of events. The drawback of [67,66] is that
sensor nodes are required to send the status of their queue to
their neighbors, whichincreases signaling overhead. Moreover, the
authors assume nodes are aware of their location. For example,
they suggest the use of the Geographic Positioning System (GPS),
which consumes non-negligible amount of energy. Ngai et al.
[67,66] propose a novel replication method to increase reliability.
Nodes maintain the link loss rate to each of their respective
neighbor, and use this information to determine their transmission
strategy. If the loss rate is high, a sensor node sends its packets to
multiple neighbors that have a path to the actuator. If all neighbors
fail to meet the required loss rate, the sensor node sends a feedback
message to a packets previous hop. The process is repeated until
another path is found or the feedback message is received by
the source node, which then decides whether to terminate the
transmission.
Melodia et al. [40] propose a transport protocol that provides
event reliability to applications. Here, event reliability is defined
as the percentage of data that arrived within a given time bound.
Whenever a sensor node needs to transmit a data packet, it
builds a routing path to its nearest actuator using the minimum
transmission power. When an actuator receives a data packet from
a sensor node, it computes its event reliability and broadcasts the
result to sensor nodes. If an applications event reliability is not
met, sensor nodes adjust their transmission power according to a
given probability. For example, if the calculated event reliability
is 0.1 of the required event reliability, more sensor nodes in the
packet forwarding path will increase their transmission power. On
the other hand, if the calculated event reliability is 0.9, sensors
reduce their transmission power. This mechanism, however, does
not consider congestion at nodes. Hence, increasing transmission
power may not necessarily alleviate the low event reliability
reported by actuators. In addition, the authors assume that in each
network area, only one event will occur, and do not consider the
possibility of multiple events requiring different reliability bounds
within said area. For example, in some applications such as home
automation, light or vision sensors will generate different sensed
data with different event reliability requirements.
Melodia et al. [39] propose a transport protocol for sen-
soractuator network similar to the distributed heuristic model
in [40], which trade-offs the energy consumption of sensor nodes
and providing minimum event reliability. The only difference is
that whenever the reliability is low, even after an increase in trans-
mission power, the actuator node detects congestion occurrences.
In this scenario, a new actuator is chosen, and half of the traffic is
routed to the selected actuator. The limitation with their approach
is that the new actuator is chosen with respect to the old actu-
ator. Hence, the new actuator may not be close to sensor nodes.
As a result, the redirected traffic may unnecessarily consume more
energy.
Table 4 summarizes transport protocols developed for use in
WSANs. Most of them rely on the existence of multiple paths,
and the use of varying transmission range to shorten forwarding
paths. The former technique means more nodes are required to
participate inthe forwarding of packets, whereas the latter method
leads to increasedinterference. Moreover, the use of multiple paths
and increased transmission range is likely to degrade network
capacity as nodes on different paths are likely to interfere and
increase contention delay. Apart from that, transport protocols
that rely on retransmissions may cause unwanted responses. For
example, in a fire system, a temperature sensor node may send
864 H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867
Table 4
Comparison of transport protocols.
Prior works Delay bound Reliability Service priority? Energy efficiency
Zhou et al. [34] By changing transmission range of
sensor nodes to shorten or elongate a
forwarding path
No consideration Yes Forwarding paths are selected
based on required delivery time of
packets
Gungor et al. [2] Scheduling packets according to their
deadlines and dynamic adjustment of
sampling rate
TCEF scheduling algorithm and the
changing sampling rate
Yes During congestion, sampling rate of
sensor nodes is reduced
Ngai et al. [67,66] Forwards packets on the least
congested path. They also consider
the use of mobile actuators
Measuring links loss rate and sending
multiple copies of a packet along
many paths
Yes No
Melodia et al. [40] Probabilistic transmission schedule,
and increasing nodes transmission
range to reduce route length
Probabilistic recruitment of sensor
nodes
No Adjust sensor nodes transmission
range
Melodia et al. [39] Changing sensor nodes transmission
range and shifting load to less
congestion actuator nodes
Changing sensor nodes transmission
range and shifting packets onto less
congested paths or actuators
No Reduce sensor nodes transmission
range if event reliability is above a
given threshold
a data packet on two different paths, and as a result, they have
different arrival times. Upon receiving the first packet, the actuator
starts the water sprinklers. However, when the second packet
arrives, it may conclude that more water is required to extinguish
the fire. It thus increases the water flow to the sprinklers, which
unfortunately leads to flooding.
4.4. Actuator-to-actuator coordination
In actuator-to-actuator coordination, it is critical that tasks
completes on time, so bounding this time is one of the most im-
portant application requirement that must be met. Vassis et al. [24]
propose a multi-channel actuatoractuator communication proto-
col to decrease task completion time. Each actuator uses two inde-
pendent radio transceivers, with one used for communicating with
single-hop or neighboring actuators, and the other for multi-hop
or remote actuators. The transceiver used for multi-hop commu-
nications has a longer range, and lower data rate. As each node has
two transceivers, the network is less likely to be congested due to
the separate collision domain used for local and remote communi-
cations. As a result, nodes experience shorter medium access de-
lay. On the other hand, the multi-hop transceiver with its longer
transmission range reduces end-to-end delay as packets traverse
fewer hops to their destination actuator. The downside with this
approach is that it incurs additional cost and energy due to the use
of two radio transceivers.
The authors of [40] developed a localized auction algorithm to
minimize task completion time. Each actuator node is assigned
to a given area. As actuators may have overlapping areas, the
algorithm selects one that can complete the task with minimum
energy expenditure and within a given delay bound. The selection
is carried out according to actuator nodes residual energy and
minimum load. This approach can only be used when there are
multiple static actuators in an overlapping area. The proposed
algorithm is not capable of handling the occurrence of multiple
events with different task completion times because it does not
prioritize tasks according to their urgency.
Zeng et al. [64] proposed an approach that uses mobile actuator
nodes. The first actuator node that receives an event occurrence
report starts collecting information such as residual energy and
current position from other actuators. Actuators also indicate
whether they have receivedthe same report. Basedonthe collected
information, an actuator that has the highest residual energy and
located in range of the event area is then assigned to complete
the task. A key limitation of this work is that the authors have
not considered multiple events. In particular, an actuator may be
selectedtocarry out multiple tasks but its residual energy may only
be sufficient to complete one task.
Melodia et al. [39] propose a novel mechanismto control mobile
actuators in scenarios where events occur in partially overlapped
space and/or time. They formulated a Mixed Integer Non-Linear
Program (MINLP) that aims to minimize energy consumption
whilst ensuring tasks complete within the given time bound. In
the presence of multiple events, the proposed mechanism aims
to guarantee task completion time for high priority events. For
each task, the proposed model selects an ideal number of actuators
and determines the actuator velocity required to move into an
event area using minimal energy whilst adhering to a given delay
bound. The key limitation of this work is that the MINLP model
is a centralized method, and the authors have not proposed any
heuristics or distributed models for the problem.
Ngai et al. [67] developed a relocation method where actuators
are mobile and visit areas that have a high frequency of events.
The network area is divided into cells, and a selected actuator
periodically collects the event frequency of all cells in the
network. The selected actuator node runs a relocation algorithmto
balance load and move more actuators to areas with higher event
frequencies. This method requires actuator nodes to periodically
communicate with selected actuator nodes, which increases
energy consumption. To reduce power consumption, the authors
assume fixed event frequency. This means after an actuator node
is assigned to a given area, there is no longer any need for
communication between actuator nodes. Unfortunately, in some
applications, different areas will have varying number of events
occurring within a given time period.
Shah et al. [45] also developed an actuator relocation method.
The network area is divided into clusters, where cluster heads are
responsible for collecting sensed data fromtheir respective cluster,
and for sending them to actuator nodes. An actuatoractuator
coordinating procedure is triggered when a cluster head is not
in the action range of any actuator nodes. This cluster head then
causes a relocation message to be broadcasted to other actuator
nodes, which then determine the possibility of covering said
cluster head whilst remaining in range to cover their existing
cluster heads. If an actuator node can move toward the cluster
head that issued the relocation message, it sends back a reply
message that contains the residual energy andnumber of clusters it
is responsible for. The actuator node with the maximumamount of
residual energy and minimumnumber of attending cluster head(s)
is then selected to help the cluster head. The disadvantage of this
actuator relocation model is that while actuator nodes are mobile,
the authors have not proposed any solution when there are no
actuators that could move toward uncovered cluster heads.
Table 5 summarizes actuator-to-actuator coordination mod-
els. As we can see, only a handful of works exist in this cate-
gory. The main theme of these works is developing protocols to
H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867 865
Table 5
Comparison of actuator-to-actuator protocols.
Prior works Coordination model Multiple events? Bounded task completion time? Minimize actuators energy?
Vassis et al. [24] Distributed No Yes No
Melodia et al. [40] Distributed No Yes Yes
Zeng et al. [40] Distributed No Yes Yes
Melodia et al. [40] Centralized Yes Yes Yes
Ngai et al. [40] Centralized Yes Yes No
Shah et al. [40] Distributed No No Yes
control mobile actuator nodes. A key observation is that guaran-
teeing task completion time when there are multiple events is
currently not possible with distributed actuator-to-actuator coor-
dination mechanisms. Besides that, coordinating multiple actua-
tors, especially in scenarios with varying event frequencies, incurs
a significant amount of communication overheads. On the other
hand, centralized coordination models have high delays and as a
consequence, they are not suitable for tasks that require fast com-
pletion time.
5. Future research directions
The coordination problem cuts across different layers of
the protocol stack. Moreover, unlike WSNs, protocols must be
developed according to a set of requirements that is distinct to
WSANs. To this end, there are a number of open research issues.
They include,
Developing techniques to distribute energy consumption and
preventing energy holes from forming around actuator nodes.
In this regard, multiple actuator nodes can be distributed
uniformly to balance power consumption. However, this
assumes uniform events occurrence and sampling rate. One
approach to alleviate these problems is to periodically change
the target actuator(s) which sensor nodes send their data to, or
change the mobile actuator nodes that visit areas with frequent
events.
Designing energy efficient methods for delivering sensed data
to actuator nodes in a timely manner. Current techniques
rely primarily on wireless communications. Unfortunately, the
wireless channel is inherently lossy and shared by multiple
contending nodes. As a result, multiple techniques are required
to overcome packet loss, ensure fair channel access, and
minimize delay all of which require nodes to budget a
significant portion of their energy on addressing the vagaries
of the wireless channel. To this end, a promising approach is
to use mobile sinks [63,13,47]. As a result, communications can
be kept to a minimum as sensors are able to forward their data
directly to a mobile sink or to a node, via a limited number of
hops, that is on a mobile sinks path. A key challenge is meeting
applications deadline and ensuring reliability.
Designing actuator-to-actuator protocols that guarantee task
completion time in scenarios with multiple events. Centralized
coordination models are well equipped to handle multiple
events but are unable to guarantee the task completion time
of each event. On the other hand, distributed coordination
models incur a non-negligible amount of overheads when used
to guarantee the task completion time of multiple events. A
promising approach is to utilize clustering, where an actuator
node is designated as the cluster head. When multiple events
occur in a cluster, the cluster head is then responsible for
assigning each task to different actuator nodes and ensuring all
tasks are completed on time.
6. Conclusion
The potential of WSANs has attracted the attention of
researchers from diverse disciplines. The key driving factor is
their wide range of applications as sensor nodes are now able
to interact with the environment they are in by effecting key
parameters. In this context, this paper has provided an extensive
reviewof solutions pertaining to the coordinationprocess between
sensor and actuator nodes, as well as between actuator nodes. It
is evident that addressing the coordination problem will have to
involve energy efficient protocols froma number of protocol layers.
Moreover, incomparisonto WSNs, a significant amount of research
remains. In particular, those related to developing distributed
protocols to control multiple actuator nodes, and providing an
upper bound on task completion time. To this end, we have
outlined a number of open research problems and issues to guide
future researchers.
References
[1] A.A. Abbasi, M. Younis, A survey on clustering algorithms for wireless sensor
networks, Computer Communications 30 (2007) 28262841.
[2] O. Akan, I. Akyildiz, V.C. Gungor, A real time and reliable transport protocol for
wireless sensor and actor networks, IEEE/ACMTransactions on Networking 16
(2) (2008) 359370.
[3] I.F. Akyildiz, I. Kasimoglu, Wireless sensor and actor networks: research
challenges, Ad Hoc Networks 2 (4) (2004) 351367.
[4] I.F. Akyildiz, W.-Y. Lee, M.C. Vuran, S. Mohanty, Next generation/dynamic
spectrum access/cognitive radio wirelessnetworks: a survey, Computer
Networks 50 (13) (2006) 21272159.
[5] I. Akyildiz, W.-Y. Lee, M. Vuran, S. Mohanty, A survey on spectrum
management in cognitive radio networks, IEEE Communications Magazine 46
(4) (2008) 4048.
[6] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, Wireless sensor
networks: a survey, Computer Networks 38 (4) (2002) 393422.
[7] R. Araujo, L. Villas, A. Boukerche, A wireless actor and sensor networks QoS
aware routing protocol for emergency prepareness class of applications, in:
IEEE Local Computer Networks, Tampa, FL, USA, January 2006.
[8] A. Baggio, Wireless sensor networks in precision agriculture, in: ACM
WorkshoponReal-WorldWireless Sensor Networks, Stockholm, Sweden, June
2005.
[9] S. Basagni, A. Carosi, E. Melachrinoudis, C. Petrioli, Z.M. Wang, Controlled
sink mobility for prolonging wireless sensor networks lifetime, ACM/Springer
Wireless Networks 14 (2008) 831858.
[10] P. Bhagwat, C.E. Perkins, Highly dynamic destination-sequenced distance-
vector routing (DSDV) for mobile computers, in: ACM SIGCOMM, London, UK,
August 1994.
[11] J. Capella, A. Bonastre, J. Serrano, R. Ors, A new robust, energy-efficient
and scalable wireless sensor networks architecture applied to a wireless
fire detection system, in: Intl. Conf. on Wireless Networks and Information
Systems, Shanghai, China, December 2009.
[12] H. Chan, A. Perrig, Security and privacy in sensor networks, Computer 36 (10)
(2003) 103105.
[13] I. Chatzigiannaki, A. Kinalis, S. Nikoletseas, Sink mobility protocols for data
collection in wireless sensor networks, in: ACM International Workshop on
Mobility Management and Wireless Acces, Torremolinos, Spain, 2006.
[14] T. Coplu, O. Emiroglu, E. Cayirc, Power aware many to many routing in wireless
sensor and actuator networks, in: 2nd European Workshop on Wireless Sensor
Networks, Istanbul, Turkey, February 2005.
[15] I. Demirkol, C. Ersoy, F. Alagoz, MAC protocols for wireless sensor networks: a
survey, IEEE Communications Magazine 44 (4) (2006) 115121.
[16] enocean. Tcm320: transceiver module for programmable systemcomponents.
http://www.enocean.com/en/enocean_modules/tcm-320/.
866 H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867
[17] R. Evans, W. Iversen, Y. Kim, Remote sensing and control of an irrigation
system using a distributed wireless sensor network, IEEE Transactions on
Instrumentation and Measurement 57 (7) (2008) 13791387.
[18] M. Gauger, D. Minder, P.J. Marron, A. Wacker, A. Lachenmann, Prototyping
sensoractuator networks for home automation, in: Proceedings of the
Workshop on Real-World Wireless Sensor Networks, Glasgow, Scotland,
March 2008.
[19] R. Govindan, D. Estrin, J. Heidemann, F. Silva, C. Intagonwiwat, Directed difus-
sion for wireless sensor networking, IEEE/ACMTransactions on Networking 11
(1) (2003) 216.
[20] W. Hu, V.N. Tran, N. Bulusu, C.-T. Chou, S. Jha, A. Taylor, The design
and evaluation of a hybrid sensor network for cane toad monitoring, in:
Information Processing in Sensor Networks, IPSN, Los Angeles, CA, April 2005.
[21] IETF. Internet engineering task force. http://www.ietf.org.
[22] S. Jha, N. Bulusu, W. Hu, Acommunicationparadigmfor hybridsensor/actuator
networks, in: IEEE PIMRC, Barcelona, Spain, September 2004.
[23] E. Johnson, R. Govindan, K. Chntalapudi, Structural damage detection using
wireless sensoractuator networks, in: IEEE International Symposium on,
Mediterrean Conference on Control and Automation, Budapest, Hungary, June
2005.
[24] G. Kormentzas, C. Skianis, D. Vassis, Performance evaluationof single andmulti
channel actor to actor communication for wireless sensor actor networks, Ad
Hoc Networks 4 (4) (2006) 487498.
[25] S.-F. Li, Wireless sensor actuator network for light monitoring and control
application, in: IEEE CCNC, Las Vegas, January 2006.
[26] S. Liu, D. Tu, Y. Zhang, Multiparameter fire detection based on wireless sensor
network, in: Intl. Conf. on Intelligent Computing and Intelligent Systems,
Shanghai, China, November 2009.
[27] K. Lorincz, M. Welsh, O. Marcillo, J. Johnson, M. Ruiz, J. Werner-Allen,
Deploying a wireless sensor network on an active volcano, IEEE Internet
Computing 10 (2) (2006) 1825.
[28] J.P. Lynch, K.J. Loh, Asummary reviewof wireless sensors and sensor networks
for structural health monitoring, Shock and Vibration Digest 38 (2) (2005)
91128.
[29] J.P. Lynch, R. Swartz, Strategic network utilization in a wireless structural
control system for seismically excited structures, Journal of Structural
Engineering 135 (5) (2009) 597608.
[30] D. Maltz, D. Johnson, Dynamic Source Routing in Ad Hoc Wireless Networks,
Kluwer Academic Publishers, 1996, pp. 153181 (Chapter 5).
[31] M. Montestruque, L. Lemmon, Csonet: a metropolitan scale wireless sen-
soractuator network, in: International Workshop on Mobile Device and Ur-
ban Sensing, MODUS, St. Louis, MO, USA, April 2008.
[32] V. Narasimhan, A. Arvind, K. Bever, Greenhouse asset management using
wireless sensoractor networks, in: IEEE Mobile Ubiquitous Computing,
Systems, Services and Technologies, Papeete, French Polynesia, Tahiti,
November 2007.
[33] I.S.A. Nayak (Ed.), Wireless Sensor and Actuator Networks: Algorithms and
Protocols for Scalable Coordination and Data Communication, John Wiley and
Sons Inc., 2009.
[34] E. Ngai, M. Lyu, J. Liu, Y. Zhou, POWER-SPEED: a power controlled real time
data transport protocol for wireless sensor actuator networks, in: IEEE WCNC,
Hong Kong, China, March 2007.
[35] J. Paradells, C. Gomez, Wireless home automation networks; a survey of
architectures and technologies, IEEE Communications Magazine 48 (6) (2010)
92101.
[36] V. Paruchuri, L. Barolli, A. Durresi, Delay-energy aware routing protocol for
sensor and actor networks, In: Proc. of 11th Intl. Conf. on Parallel and
Distributed Systems, Fuduoka, Japan, July 2005.
[37] S. Wu, P. Noy, A conceptual design of a wireless sensor actuator system
for optimizing energy and well-being in buildings, Intelligent Buildings
International 2 (2010) 4156.
[38] A. Perrig, J. Stankovic, D. Wagner, Security in wireless sensor networks,
Communications of the ACM 47 (2004) 5357.
[39] D. Pompili, I. Akyildiz, T. Melodia, Handling mobility in wireless sensor and
actor networks, IEEE Transactions on Mobile Computing 9 (2) (2010) 100120.
[40] D. Pompili, V. Gungor, I. Akyildiz, T. Melodia, Communication and coordination
inwireless sensor andactor networks, IEEE Transactions onMobile Computing
6 (10) (2007) 11161129.
[41] E. Royer, C. Perkins, Ad-hoc on-demand distance vector routing, in: IEEE
Workshop on Mobile Computing Systems and Applications, NewOrleans, USA,
February 1999.
[42] L. Ruiz-Garcia, L. Lunadei, P. Barreiro, I. Robla, A review of wireless sensor
technologies and applications in agriculture and food industry: State of the
art and current trends, Sensors 9 (6) (2009) 47284750.
[43] J. Sanchez, P.M. Ruiz, I. Stojmenovic, Energy-efficient geographic multicast
routing for Sensor and Actuator Networks, Computer Communications 30 (13)
(2007) 25192531.
[44] E. Sazonov, R. Jha, K. Janoyan, V. Kidshnamurthy, Wireless intelligent sensor
and actuator network (WISAN): a scalable ultra-low-power platform for
structural health monitoring, in: Health Monitoring and Smart Nondestructive
Evaluation of Structural and Biological Systems, 2006.
[45] G.A. Shah, M. Bozyigit, F.B. Hussain, Cluster-based coordination and routing
framework for wireless sensor and actor networks, Wireless Communications
and Mobile Computing (2009) 11401154.
[46] Y. Shang, H. Shi, M. Tubaishat, Adaptive traffic light control with wireless
sensor networks, in: IEEE CCNC, Las Vegas, USA, January 2007.
[47] A. Somasundara, M. Srivastava, D. Jea, Multiple controlled mobile elements
(data mules) for data collection in sensor networks, in: First IEEE Intl. Conf.
on Distributed Computing in Sensor Systems, DCOSS, Marina del Rey, CA, USA,
June 2005.
[48] Spartan. Me8300 wireless actuator based on the enocean chipset.
http://www.spartan-pd.com.
[49] G.M.R. Verdone, D. Dardari, A. Conti, Wireless Sensor and Actuator Networks:
Technologies, Analysis and Design, Elsevier, 2008.
[50] L. Villasenor-Gonzalez, E. Ruiz-Ibarra, Cooperation mechanism taxonomy for
wireless sensor and actor networks, in: IFIP Conference on Wireless Sensor
and Actor Networks, Ontario, Canada, July 2008.
[51] M. Vuong, H. Nguyen, H. Ngyen, D. Dinh, Wireless sensor actor networks
and routing performance analysis, in: Intl. Workshop on Wireless Ad-Hoc
Networks, London, UK, May 2005.
[52] M. Wan, P. Lemmon, Distributed flow control using embedded sen-
soractuator networks for the reduction of combined sewer overflow (CSO)
events, in: IEEE Conference on Decision and Control, New Orleans, USA, De-
cember 2007.
[53] Z. Wang, S. Basagni, E. Melachrinoudis, C. Petrioli, Exploiting sink mobility
for maximizing sensor networks lifetime, in: Proceedings of the 38th Annual
Hawaii International Conference on System Sciences, January 2005, p. 287a.
[54] N. Wang, N. Zhang, M. Wang, Wireless sensors in agriculture and food
industryrecent development and future perspective, Computers and Elec-
tronics in Agriculture 50 (2005) 1.
[55] T. Wark, C. Crossman, W. Hu, Y. Guo, P. Valencia, P. Sikka, P. Corke, C. Lee, J.
Henshall, K. Prayaga, J. OGrady, M. Reed, A. Fisher, The design and evalaution
of a mobile sensor/actuator network for autonomous animal control, in:
Proceedings of the 6th International Conference on Information Processing in
Sensor Networks, Cambridge, MA, USA, April 2007.
[56] H. Wattar, G. Scholl, H. Korber, Modular wireless real-time sensor/actuator
network for factory automation applications, IEEE Transactions on Industrial
Informatics 3 (2) (2007) 111119.
[57] F. Xia, QoS challenges and opportunities in wireless sensor/actuator networks,
Sensors 8 (2) (2008) 10991110.
[58] F. Xia, Y. Tian, Y. Li, Y. Sung, Wireless sensor/actuator network design for
mobile control applications, Sensors 7 (10) (2007) 21572173.
[59] D. Xia, N. Vlajic, Wireless sensor networks: to cluster or not to cluster?, in: IEEE
WoWMoM, Buffalo, NY, USA, June 2006.
[60] G. Xing, T. Wang, Z. Xie, W. Jia, Rendezvous planning in wireless sensor
networks with mobile elements, IEEE Transactions on Mobile Computing 7
(2008) 14301443.
[61] X. Yang, X. Guo, M. Zhou, L. Wang, Y. Zhou, A design of greenhouse monitoring
& control systembased on ZigBee wireless sensor network, in: ACMMobicom,
Montreal, CA, September 2007.
[62] H. Yuan, H. Ma, H. Liao, Coordination mechanism in wireless sensor and
actor networks, in: Proceeding of the First International Multi-Symposiums
on Computer and Computational Sciences, Hangzhou, China, June 2006.
[63] Y. Yun, Y. Xia, Maximizing the lifetime of wireless sensor networks withmobile
sink in delay-tolerant applications, IEEE Transactions on Mobile Computing 9
(9) (2010) 13081318.
[64] Y. Zeng, C. Sreenan, G. Zheng, A real-time architecture for automated wireless
sensor and actuator networks, in: 5th Intl. Conf. on Wireless and Mobile
Computing, France, August 2009.
[65] X. Zhao, H. Gao, G. Zhang, B. Ayhan, F. Yan, C. Kwan, J. Rose, Active health
monitoring of an aircraft wing with embedded piezoelectric sensor/actuator
network: I. Defect detection, localization and growth monitoring, Smart
Materials and Structures 16 (4) (2007) 1208.
[66] E.C.H. Ngai, Yangfan Zhou, M.R. Lyu, Jiangchuan Liu, LOFT: a latency-oriented
fault tolerant transport protocol for wireless sensoractuator networks, in:
IEEE Globecom, Washington, DC, USA, November 2007.
[67] Y. Zhou, M. Lyu, J. Liu, E. Ngai, A delay-aware reliable event reporting
framework for wireless sensoractuator networks, Ad Hoc Networks 8 (7)
(2010) 694707.
[68] F. Zou, B. Yang, Y. Cao, Traffic light control for a single intersection based
on wireless sensor network, in: Intl. Conf. on Electronic Measurement and
Instruments, Beijing, China, August 2009.
Hamidreza Salarian received the B.S. degree in computer
engineering (hardware major) from Ferdowsi University
Mashhad and the M.S. degree in computer architecture
from Isfahan University of Technology, Iran in 2005
and 2008. He is currently a Ph.D. student with the
School of Electrical, Computer and Telecommunications
Engineering (SECTE) at University of Wollongong. His
research focused on energy consumption in wireless
sensor networks and wireless sensoractuator networks,
in particularly, the effect of mobility in increasing system
lifetime.
H. Salarian et al. / J. Parallel Distrib. Comput. 72 (2012) 856867 867
Kwan-WuChinobtained his Bachelor of Science with First
Class Honours from the Curtin University of Technology,
Australia. He thenpursuedhis Ph.D. at the same university,
where he graduated with distinction and the vice-
chancellors commendation. He then joined Motorola
Research Lab as a senior research engineer, where he
developed zero configuration home networking protocols
and designed new medium access control protocols
for wireless sensor networks. In 2004, he joined the
University of Wollongong as a Senior Lecturer. His current
research areas include medium access control protocols
for wireless networks, routing protocols for delay tolerant networks, RFID anti-
collision protocols, and discrete optimization problems related to computer
communications. To date, he holds four US patents, and has published more than
50 articles in numerous conferences and journals.
Fazel Naghdy has a demonstrated track record and leader-
ship in research, teaching, and management. His research
has had its focus on machine intelligence and control par-
ticularly in embedded mechatronics and robotics systems.
He has more than 250 publications in international jour-
nals and conferences and as book chapters. He is also
contributing editor to IEEE Transactions on Mechatronics
Engineering, and International Journal of Intelligent Au-
tomation and Soft Computing. He has served on a large
number of International scientific committee of various in-
ternational conferences. He is the Director of Centre for
Intelligent Mechatronics Research. His current research interests include hap-
tic rendered virtual manipulation of clinical and mechanical systems, intelligent
control and learning in non-linear and non-structured systems. He is currently a
Professor at University of Wollongong, School of Electrical, Computer & Telecom-
munication Engineering, Australia. Fazel Naghdy was born in Tehran, Iran. He re-
ceived his first degree from Tehran University in 1976, M.Sc. and Ph.D. from the
Postgraduate School of Control Engineering, University of Bradford, England, in
1979, 1982, respectively.