Wireless Sensor Network, 2010, 2, 123-128
doi:10.4236/wsn.2010.22017 y 2010 (http://www.SciRP.org/journal/wsn/). Published Online Februar
Recharging Sensor Nodes Using Implicit Actor
Coordination in Wireless Sensor Actor Networks
Mohsen Sharifi, Saeed Sedighian, Maryam Kamali
School of Computer Engineering, Iran University of Science and Technology, Tehran, Iran
Received November 25, 2009 ; revised December 9, 2009; accepted December 14, 2009
Abstract
Wireless sensor actor networks are composed of sensor and actor nodes wherein sensor nodes outnumber
resource-rich actor nodes. Sensor nodes gather information and send them to a central node (sink) and/or to
actors for proper actions. The short lifetime of energy-constrained sensor nodes can endanger the proper op-
eration of the whole network when they run out of power and partition the network. Energy harvesting as
well as minimizing sensor energy consumption had already been studied. We propose a different approach
for recharging sensor nodes by mobile actor nodes that use only local information. Sensor nodes send their
energy status along with their sensed information to actors in their coverage. Based on this energy informa-
tion, actors coordinate implicitly to decide on the timings and the ordering of recharges of low energy sensor
nodes. Coordination between actors is achieved by swarm intelligence and the replenishment continues dur-
ing local learning of actor nodes. The number of actors required to keep up such networks is identified
through simulation using VisualSense. It is shown that defining the appropriate number of actor nodes is
critical to the success of recharging strategies in prolonging the network lifetime.
Keywords: Wireless Sensor Actor Networks, Coordination, Energy Harvesting, Swarm Intelligence
1. Introduction
Wireless sensor and actor networks (WSANs) are made
of two types of nodes called sensor nodes and actors.
Sensor nodes are tiny, low-cost, low-power devices with
limited sensing, computation, and wireless communica-
tion capabilities. Actors are usually resource-rich nodes
with higher processing capabilities, higher transmission
powers and longer life. In WSANs a large number of
sensor nodes are randomly deployed in a target area, may
be in the order of hundreds or thousands, to perform a
coordinated sensing task. Such a dense deployment is
usually not necessary for actor nodes because actors are
sophisticated devices with higher capabilities that can act
on wider areas. Actors collect and process sensor data
and perform actions on the environment based on their
gathered information.
Common applications of WSANs include all types of
surveillance, target tracking, attack detection, medical
sensing, and environment monitoring. Actors in many
applications should coordinate together to maximize
their overall task performance by sharing and processing
the sensors’ data, making decisions and taking appropri-
ate actions [1].
One of the greatest concerns in WSANs is energy espe-
cially for sensor nodes. A sensor node without energy
cannot do its duties unless the source of energy is re-
charged or changed. Since such networks are usually de-
ployed in large scales and for a long period of time, human
intervention for replenishment of energy is not feasible.
The rest of paper is organized as follows. Section 2
presents notable related work. Section 3 describes our
energy harvesting approach in detail. Section 4 argues
the validity of the approach and presents simulation re-
sults. Section 5 concludes the paper.
2. Related Work
Many researches on extending the lifetime of WSANs
have focused on minimizing energy usage in different
layers of WSANs like in data aggregation to decrease
data traffic [2], in energy-efficient networking protocols
[3], in managing sleep cycles [4] and in using low power
MAC [5].
There are other variants to tackle energy problem that
consider different ways of recharging or changing the
energy sources (mostly, batteries) of sensor nodes. A
method to improve the battery lifetime of sensor nodes is
to supplement the battery supply with environmental

M. SHARIFI ET AL. 124
energy. Environmental energy that is usually harvested
to generate electricity includes solar, wind, water and
thermal energies.
The main sources of energy for use by sensor net-
works are solar, mechanical and thermal energies. Solar
power is the current matured energy among other forms
of energy harvesting and is more attractive in outdoor
applications. A solar energy harvesting module for use
by Crossbow/Berkeley motes has been developed by Lin
et al. [6] in the Helimote project.
Vibration and mechanical energy are generated by
movements of objects. Traffic sensors [7] are powered
by the short duration vibration when a vehicle passes
over the sensor.
Thermal energy harvesting uses temperature differ-
ences to generate electricity. Thermo generators (TEGs)
[8] harvest energy from the human body and are used by
devices with direct contact to the human body.
Moreover, some other works have focused on energy
replenishment in sensor networks by robots [9]. Robots
based on solar cell are used to recharge sensor nodes and
to water plants. Another energy harvesting study [10]
with focus on mobile nodes has also researched the fea-
sibility of a system wherein mobile nodes have the abil-
ity to move in search of energy and delivery of energy to
static nodes.
3. Energy Harvesting Through Actor Nodes
In this paper we propose a new approach for recharging
sensor nodes by actor nodes to prolong the lifetime of the
network. We assume that sensor nodes are static and ac-
tors, which are charged by solar cells, are mobile nodes
that charge sensor nodes and do the actuating tasks too.
To replenish the network energy, actor nodes learn the
usage energy model of the network. They then coordi-
nate to select appropriate sensors to service and move
towards their selected sensor nodes to charge them.
Whenever actors request energy related information of
nearby sensor nodes, these sensor nodes send their power
capacity, their current power, their power usage ratio,
their location and their interest to the requesting actor.
This information helps actors to find their service area
for recharging sensors. Power capacity shows the maxi-
mum energy that a node can have, and the current power
shows the remaining energy. Interest is a parameter that
shows the tendency of sensor nodes to actor nodes.
Whenever a sensor node receives a message from an
actor node, the interest of the sensor node is increased to
that actor node. In addition, when an actor charges a
sensor node, the sensor node’s interest to that actor will
be highest. When power decreases in sensor nodes, the
interests to actor nodes also decrease. In fact, actor nodes
learn the usage energy model of the network by power
information that is gathered from sensor nodes. When an
actor learns the usage energy model of sensor nodes, it
figures out the sensor nodes to target for recharge. Ac-
cording to the achieved model and by considering the
sensors’ parameters, the actor selects an appropriate
sensor node for recharging and then moves towards that
sensor and charges it.
Figure 1 shows the pseudo code for recharging sensors
by actors. When an actor receives a message from a sen-
sor node, the actor selects the part of message that relates
to energy information of the sensor. EnergyList is a list
that holds energy information of sensor nodes. Then Se-
lectSensor method chooses the best sensor node that
needs to be charged. A sensor is selected for charge by
//Preposition: actor has gathered energy information of nearby sensor nodes, implying that the state of actor is getting energy info
DoRechare(ActorID)
1 EnergyList  GetEnergyInformationOfSensors();
2 //It returns a List of <SensorID, CurrentPower, PowerRatio, InterestToActor, Location>
3 SelectedSensor = SelectSensorForCharge(EnegyList);
4 MoveToSensor(SelectedSensor.Location);
5 //actor state: moving
6 ChargeSensor(SelectedSensor);
7 //actor state: charging
8 SetFullSensorInterestToActor (SelectedSensor, ActorID);
9 Return;
10 //actor state: getting energy info
SelectSensor(EnergyList)
11 For each node Є EnergyList
12 If (node.InterestToActor == ActorId)
13 node.compareValue 
14 (node.currentpower / node.powerRatio +0.1))
15 end if
16 end for
17 return FindMinValue(EnergyList.compareValue)
Figure 1. A pseudo code for recharging sensor nodes by actor nodes.

M. SHARIFI ET AL. 125
this method based on the energy information of sensor
nodes in the EnergyList. Afterwards, the location of the
selected sensor node is passed to MoveToSensor and the
actor moves towards the sensor and charges the selected
sensor node.
During the time actors are learning the energy usage
model of sensors, the network is divided into parts. This
division is formed by energy information that is sent by
sensors.
If the network architecture is automated, there is no
central controller (sink) and actor nodes process all in-
coming data and initiate appropriate actions [11]. In such
a network, events might happen everywhere uniformly
and the usage of power for each sensor node is considered
approximately the same as for other nodes. In this sce-
nario, the sensor node deployment has a random uniform
distribution; the divided parts are about the same size.
However, if the network architecture is considered
semi-automated, the sink in a WSAN coordinates net-
work activities [11] and nodes that are closer to the sink
lose more power because they are involved more in
communication than nodes farther from the sink. So in
semi-automated networks, the usage of power in a sensor
differs from power usage in other sensor nodes, resulting
in different sizes of network partitions. Because sensor
nodes closer to the sink lose their power faster, actors
have to recharge them more frequently. Therefore, the
service area of actors near the sink is smaller in size.
4. Experimental Validation
The proposed approach was simulated by VisualSense
[12] with different number of actor nodes and 100 sensor
nodes. Actors and sensor nodes were randomly placed in
an area of 1000m × 1000m. The transmission ranges for
each actor node and for each sensor node were 100m and
50m, respectively. The model was used in different sce-
narios to evaluate the performance of the model. We
used a short range for actor nodes because our algorithm
is really localized.
In Figure 2, there are 3 actor nodes and 80 sensor
nodes scattered randomly in the area. Also there is a sink
at the bottom right corner of the area. Actually, the net-
work is assumed semi-automated. As it is shown, the
area that actor #1 has created is smaller than the area of
actor #2 and actor #3. In this scenario, since energy of
sensor nodes near the sink decreases by higher ratio, ac-
tor #1 has to act in a smaller area to be able to supply
enough energy to sensor nodes. As in this scenario, there
are limited actor nodes with a lot of sensor nodes, so
some sensor nodes die out. But because there are a lot of
sensor nodes, sensing coverage does not decrease quickly.
The learning steps of actors for the above scenario are
shown in Figure 4. As it is depicted in the third part of
Figure 4, sensor nodes at the left top corner need not to
Figure 2. Sensor nodes charged by actor nodes in a network
wherein the sink is placed at the right bottom corner (cir-
cles represent sensor nodes and triangles represent actor
nodes).
Figure 3. Charging sensors by actor nodes in a network
with the sink placed at the centre.
be charged because they are not involved in interface
communication to other nodes; this means that they only
use power when they need to send their information.
Figure 3 shows the same scenario as the previous sce-
nario except that the sink is placed in the middle of the
area. As it is shown in Figure 3, network service areas
are different from those shown in Figure 2. Furthermore,
Figure 5 shows the evolutionary progress of the service
area formation as a result of application of learning
process.
Figure 6 shows a scenario for an automated architec-
ture containing the same number of sensor and actor
nodes as in the last two scenarios. Figure 7 shows the
learning process in this scenario and illustrates how some
sensor nodes die out because of limited number of actor
nodes.
In our proposed approach, we have an important chal-
lenge for the number of actors. Some important parame-
ters that affect the number of actors needed for the net-
work are the number of sensor nodes, the area, average
recharge ratio, local recharge ratio, speed of actors,
charge time and the network architecture. If the number
of actors does not match the network requirements, the
network sensing coverage will decrease, leading to a
dead network.
C
opyright © 2010 SciRes. WSN

M. SHARIFI ET AL. 126
Figure 4. Learning process in a semi-automated network with the sink placed at the right bottom corner.
Figure 5. Learning process in a semi-automated network with the sink placed at the centre.
Figure 6. Charging sensors by actor nodes in a network
without sink.
Figure 8 shows the result of the number of actors on
the sensing network coverage in a network with an
Figure 7. Learning process in an automated network.
C
opyright © 2010 SciRes. WSN

M. SHARIFI ET AL.127
Figure 8. Sensing network coverage in an automated net-
work based on the number of actors (76 sensor nodes).
automated architecture. If we consider the sensing cov-
erage under 50% as a dead network, the network will die
with less than 2 actor nodes; otherwise the network will
stay alive although some sensor nodes might have died.
Figure 9 and Figure 10 show the impact of the number
of actors and semi-automated architecture of the network
on network coverage. Moreover, these figures show that
the placement of the sink in the centre changes the sens-
ing coverage. It is shown that in a network with
semi-automated architecture, wherein the sink is placed
in the centre of the network, even 2 actor nodes cannot
provide enough sensing coverage; the sensing coverage
decreases to less than 50% and the network is considered
as a dead network.
5. Conclusions and Future Work
In this paper we introduced a new approach for energy
replenishment of sensor nodes by mobile actor nodes in
wireless sensor and actor networks. The replenishment
starts when local learning of actor nodes is finished. To
show the significance of the number of actor nodes
withcharging duty on the lifetime of the network, some
Figure 9. Sensing network coverage in a semi-automated
network with the sink placed at the bottom right corner (76
sensor nodes).
Figure 10. Sensing network coverage in a semi-automated
network with the sink placed at the centre (76 sensor
nodes).
scenarios were simulated. Simulations showed that a
network with 76 sensor nodes scattered randomly in a
1000m×1000m area needs at least 3 actor nodes to stay
alive. Although some sensor nodes may die because ac-
tor nodes cannot charge them before they are completely
discharged, the network still stays alive with a more lim-
ited coverage.
We are currently studying the case where actor nodes
share their learned energy model in order to make more
optimized decisions for recharging sensor nodes that
prolong the network lifetime (i.e., explicit coordination).
When actor nodes learn the energy model of the whole
network in addition to their own local learning, they can
cooperate to service sensor nodes. This way, the number
of sensor nodes recharged can be optimized and the
number of dead sensor nodes will decrease especially in
networks with semi-automated architecture. We also
know that explicit coordination may need more energy
consumption in actor nodes, so our desire is a trade-off
between implicit and explicit coordination.
Another future work that can increase the performance
is to recharge sensor nodes based on their required en-
ergy and not fully. In our approach, actor nodes charge
sensor nodes completely while they can charge them just
according to their real needs that can be extracted from
their ratio power, power capacity, current power and
time. This way, the time actor nodes spend on each sen-
sor node decreases and they can charge more sensor
nodes. The number of movements of actor nodes in-
creases though because sensor nodes require more fre-
quent recharges.
5. References
[1] B. P. Gerkey and M. J. Mataric, “Sold!: auction methods
for multirobot coordination,” IEEE Transactions on
Robotics and Automation, Vol. 18, No. 5, pp. 758–768,
2002.

M. SHARIFI ET AL. 128
[2] C. Q. Hua and T. S. P. Yum, “Optimal routing and data
aggregation for maximizing lifetime of wireless sensor
networks,” IEEE-ACM Transactions on Networking, Vol.
16, No. 4, pp. 892–903, 2008.
[3] S. Zhou, R. P. Liu and Y. J. Guo, “Energy efficient
networking protocols for wireless sensor networks,”
Proceedings of IEEE International Conference on
Industrial Informatics, Singapore, pp. 1006–1011, 2006.
[4] M. I. Shukur, L. S. Chyan, and V. V. Yap, “Wireless
sensor networks: delay guarantee and energy efficient
MAC protocols,” World Academy of Science, Engi-
neering and Technology, Vol. 50, 2009.
[5] M. H. F. Ghazvini, M. Vahabi, M. F. A. Rasid, R. S. A. R.
Abdullah, and W. M. N. M. W. Musa, “Low energy
consumption MAC protocol for wireless sensor
networks,” Proceedings of Second International Confe-
rence on Sensor Technologies and Applications, France,
pp. 49–54, 2008.
[6] K. Lin, J. Yu, J. Hsu, S. Zahedi, D. Lee, J. Friedman, A.
Kansal, V. Raghunathan and M. Srivastava, “Heliomote:
enabling long-lived sensor networks through solar energy
harvesting,” Proceedings of 3rd International Confe-
rence on Embedded Networked Sensor Systems,
California, USA, pp. 309, 2005.
[7] K. Vijayaraghavan and R. Rajamani, “Active control
based energy harvesting for battery-less wireless traffic
sensors: theory and experiments,” Proceedings of
American Control Conference, Washington, USA, pp.
4579–4584, 2008.
[8] L. Mateu, C. Codrea, N. Lucas, M. Pollak, and P. Spies,
“Energy harvesting for wireless communication systems
using thermogenerators,” Proceeding of the XXI
Conference on Design of Circuits and Integrated Systems
(DCIS), Barcelona, Spain, 2006.
[9] A. LaMarca, D. Koizumi, M. Lease, S. Sigurdsson, G.
Borriello, W. Brunette, K. Sikorski, and D. Fox,
“Plantcare: An investigation in practical ubiquitous sys-
tems,” Lecture Notes in Computer Science, Vol. 2498, pp.
316–332, 2002.
[10] M. Rahimi, H. Shah, G. S. Sukhatme, J. Heideman, and D.
Estrin, “Studying the feasibility of energy harvesting in a
mobile sensor network,” Proceeding of IEEE Inter-
national Conference on Robotics and Automation,
Taiwan, pp. 19–24, 2003.
[11] I. F. Akyildiz and I. H. Kasimoglu, “Wireless sensor and
actor networks: research challenges,” Ad Hoc Networks,
Vol. 2, No. 4, pp. 351–367, 2004.
[12] P. Baldwin, S. Kohli, E. A. Lee, X. Liu, Y. Zhao, C.
Brooks, N. V. Krishnan, S. Neuendorffer, C. Zhong, and
R. Zhou, “Visualsense: visual modeling for wireless and
sensor network systems,” Technical Memorandum UCB/
ERL M04/08, Electronics Research Laboratory, College
of Engineering, University of California, 2004.