2013 IEEE International Conference on Systems, Man, and Cybernetics

A Energy-Preserving Model for Wireless Sensors

Networks based on Heuristic Self-Organized Routing
Aurelio La Corte, Alessandro Di Stefano, Marialisa Scata, Marco Leotta
Department of Electrical, Electronics and Computer Science Engineering
DIEEI, University of Catania
Catania, Italy
lacorte@dieei.unict.it, alessandro.distefano@dieei.unict.it, lisa.scata@dieei.unict.it, marco.leotta@dieei.unict.it

AbstractOne of the main targets related to Wireless Sensor

Networks (WSNs) is to reduce power consumption of nodes and
of the whole network. An ideal WSN should be networked, scalable, fault-tolerant, energy-aware, and also smart and efcient.
Unfortunately, however, this is not always true. The basic idea of
the proposal is that senders use a heuristic approach to select the
sub-optimal next hop in order to reach just one sink, considering
some key requirements such as general performance (QoS and
security), efciency, trustability, high computational power and
energy-aware behaviour. Our model tries to satisfy the need for
reaching the nearest sink node, considering a trade-off between
the shortest path and heuristic decisions, in a top-level strategy
based on a heuristic approach in order to reduce the overall
power consumption of nodes of the network.
Index TermsWSNs, Heuristics, Self-Organization, EnergyAware Routing

I. I NTRODUCTION
A WSN consists of spatially distributed autonomous sensors
to monitor physical or environmental conditions. Nodes need
to cooperate using their processing abilities to carry out locally
simple computations and transmit the required data [1]. Each
node is an autonomous agent able to decide how to route information. There are many types of routing and forwarding algorithms for WSNs, characterized by different kinds of actions
about how to reach the destinations. In general, an efcient
routing protocol should perform aggregation, clustering, selforganization and cooperation for power saving and to increase
network lifetime; it should also consider a threshold for sensor
nodes in data transmission in order to get energy-saving, and
a multi-path dissemination to improve fault-tolerance. In this
paper, we propose an energy-preserving model for wireless
sensors networks based on a heuristic and self-organized
routing. We consider a network with specic nodes which
need to send information to just one of a set of sinks. In
order to make simpler and reduce the communication ow
that would produce a large amount of energy consumption,
each node routes information exploiting a cognitive behaviour
of its own neighbourhood, without using traditional routing
protocols that could produce large routing tables, for a huge
amount of sensors. The hierarchy follows a trend ranging from
the sink to the probable sources. From the latter, following the

978-1-4799-0652-9/13 $31.00 2013 IEEE

DOI 10.1109/SMC.2013.545

hierarchy created previously, each node will choose the next

hop according to an heuristic mechanism, based on trustability,
goodness perception of the path and battery level of node.
The heuristic approach considered within the model will allow
us to explain how the nodes take decisions to solve complex
problems with incomplete information using trustability level
and perception, following a top-level strategy that leads to
solve problems exploiting this underlying heuristic, so that it
results in a heuristic approach.
The paper is organized as follows: in the rst section we give
a brief overview of wireless sensors networks, also describing
the main features and the design factors. The second section
deals with some energy-aware protocols, also introducing
what is required to achieve an efcient energy-aware routing
protocol. In the third section we analyze the heuristic theory
underlying the proposed model, and how we use and propose
this strategy, along with trustability and perception, in multiagent systems. In the fourth section we present our model and,
nally, in the last section we close our work with conclusions
and future works.
II. A BRIEF OVERVIEW OF W IRELESS S ENSORS
N ETWORKS
A. Introduction
Wireless Sensors Networks (WSNs) are large networks that
consist of many autonomous small-sized sensors nodes, characterized by low-power and low-cost. Sensors are generally
equipped with data processing and communication capabilities
to collect data and route information back to a sink. Recent
advances in WSNs show an increased interest in the potential
use of WSNs in elds such as military and environmental
applications, health (scanning), space exploration, vehicular
movement, mechanical stress levels, disaster recovery management, and so on.
In general, sensors are expected to be remotely deployed in
unattached environments [11], in order to measure physical or
environmental changes. Sensors commonly measure preassure,
chemical reactions, sound intensity ad temperature, etc. [1],
and are tted with an on-board processor to locally carry out
simple computations and transmit only the required processed
data [1] [8] [9]. Sensors nodes are scattered in a special domain
called sensor eld and nodes typically have the capability to
3204
3198

collect data, analyze them, and route them to a designed sink

point, through communication interfaces (e.g. wireless links).
Some of the main design factors of WSNs, in terms of communication architecture, protocols and algorithms are listed below
[9] [10]: reliability, density and network size/scalability, sensor
network topology, energy consumption, hardware constraints,
data aggregation/data diffusion and self-conguration. These
nodes, spatially distributed, communicate with each other,
sharing data collected or other information to monitor specic
context and environmental conditions, such as temperature,
sound, vibration and pressure, and cooperatively pass their
data through the network to a main location. Cooperation
among nodes is one of the main issues of this type of
network, because sensor nodes use their processing capabilities
to locally carry out simple computations and transmit only
the required data [1]. The ideal WSN should be networked,
scalable, fault tolerant, energy aware, software programmable,
efcient, capable of fast data acquisition, reliable and accurate
over long terms and low cost maintanance [8]. At the same
time the network must possess self-organizing capabilities,
since positions of nodes are not predetermined. Topology
absence is counter-balanced by node intelligence, trying always
to keep the key requirements related to quality of service,
efciency, security, trustability, high computational power and
energy-aware behaviour.
B. Energy-Aware Routing Protocols
Routing or Forwarding of data packets in WSNs can be divided into three categories [2]: at-based routing, hierarchicalbased routing, location-based routing. In at-based routing
nodes have the same role and responsibility in forwarding/routing data, while in hierarchical-based routing, decisions
are inuenced by the hierarchical rank of sensor nodes. Furthermore, another classication of routing protocols is based on
how routes are created [12]. In a proactive approach all routes
are created in advance and updated regularly; in a reactive
approach routes are computed only when they are required,
and a hybrid approach is a combination of these two ideas.
The most well-known routing protocols for WSNs are [9]:
ooding, gossiping, SPIN (Sensor Protocols for Information
via Negotiation), directed diffusion, LEACH (Low Energy
Adaptive Clustering Hierarchy), PEGASIS (Power-Efcient
Gathering in Sensor Information), GEAR (Geographycal and
Energy Aware Routing). In general an efcient routing protocol
should perform data aggregation for power saving, dynamic
clustering to increase network lifetime, a threshold for sensor
nodes on data transmission and dissemination, in order to help
energy-saving, multi-path selection dissemination to improve
fault-tolerance, self-conguration and adaptation of the sensors nodes to changes in network topology and nally time
synchronization.
In this paper we propose an energy-efcient model based on a
heuristic approach. The key idea, applied to a WSNs scenario,
is that of giving to sinks the power to create a dynamic
hierarchy. The root of the hierarchical tree is represented by
the sink and leafs follow iteratively the hierarchy in order to

cover the entire network spatially. Using the previously created

hierarchical topology, senders will be able to reach one of
the sinks available, exploiting only the knowledge of their
neighborhood. Heuristics will help single nodes to choose the
next hop until to the sink.
III. I NFERENCE AND H EURISTICS
In this section we are going to explain inference mechanisms and focus on the overall heuristic approach adopted in
our model. How nodes deployed in a topology make inferences
about unknown aspect of the context? The possible approaches
could be [3]: Laplacean demon theory, fully heuristics and
controlled heuristics [4]. Many decision problems may be considered as optimization problems. These problems are tipically
too difcult to be solved exactly within a reasonable amount
of time and heuristics become the best methods of solve
them. Furthermore, when the quality of solution is critical
and an optimal solution does not exist, it becomes important
to investigate efcient procedures to obtain the best possible
solutions, considering key factors to be minimized. Heuristics
are criteria, methods, or principles for deciding which, among
several alternative courses of action, promises to be the most
effective, in order to achieve some goal [5].
A. Trustability and Perception in Multi-Agent Systems
Our model follows principles of multi-agent systems such
as cooperation and self-organization. Cooperation is related to
the interactions among agents. It is a fundamental feature of
multi-agent system where the overall system exhibits signicantly greater functionality than the individual component [6].
Exploiting hyerarchical methods within an heuristic approach
to detect a good path, we achieve an autonomous, multi-agent
and self-organized system. Therefore nodes become smart
objects which have different communication, information and
processing capabilities to make decisions. Our design choice
consists of the implementation of a heuristic model. The
heuristic approach inside the model will allow us to explain
how the nodes make decisions, come to judgments and solve
complex problems with incomplete information [13] [14]. The
main advantage is that we use a heuristic model with a dynamic
hierarchical topology setup phase, from sink to source, and a
heuristic approach, from source to sink: this approach allows
us to reduce the complexity of tasks, obtaining a self-organized
energy-aware system. The trustability and perception are used
when a node/agent interacts with other agents to decide the
next-hop in order to reach the sink. We refer to the model
presented in [15], adding some mechanisms to maximize the
overall lifetime of the network.
IV. P ROPOSAL
A. Introduction
Sink nodes in WSNs are able to receive messages from
other nodes and collect various kind of data. Usage of multiple
sinks is related to power consumption reduction in WSNs,
as a mechanism to increase the system lifetime, as shown
in [16], for this reason our topology follows a hierarchical

3199
3205

TABLE I
T OPOLOGY S ETUP F RAME
TSF

Fig. 1.

SINK ID

ADDRESS

NODES LEVEL

Fig. 2.

HN3 routes on the right

Fig. 3.

HN3 routes on the left

Each node could reach multiple higher level nodes

organization in which the highest role is delegated to sink

nodes. We identify three phases: Topology Setup, Data Sending
and Heuristic Approach. Our model tries to satisfy the need of
reaching the nearest sink node, considering heuristic decisions
to reduce the overall power consumption of the nodes in the
network and networks errors. The aim of the proposal is to use
simple network signaling and light logic to maximize network
lifetime.
1) Topology setup phase: Topology setup phase is initialized by sink nodes, which represent 0-level hierarchical nodes
(HN0 ), sending a topology setup frame to their neighbors, at
a certain time interval, Ttopologysetup . Each HN0 s neighbour
will become HN1 , storing node IDs of previously identied
HN0 s and, in turn, they will send the topology setup frame to
their neighborhood. Topology setup frame (TSF) is dened as
follows:
Networks nodes, without a specic role, will listen to
topology setup frames (TSF) and they will become L-level
hierarchical node, choosing own level L as the minimum of
the levels of nodes which have sent it a topology setup frame:
L = min (nodes levels) + 1

(1)

Consequently the HNL will store IDs of the HNL1 nodes

so that it will be able to use one of them to reach a sink as
shown in Fig. 1.
The HNL1 black circled node is the delegated node,
chosen in a heuristic way described in the next section, by
HNL node to send data to the sink.
2) Data Sending: When a HNL node needs to send data
to a sink, it will use the delegated node, using a Data Frame,
dened as follows:

TABLE II
DATA F RAME
DF

Se

Re

DFC

DATA

At each Data Frame (DF), the node will store, the Sender
(Se), the Receiver (Re), the last Data Frame Counter (DFC),
the payload of Data packet, and in the check phase (explained
below) it will consider the values of link quality indicators
such as RSSI (Received Signal Strenght Indicator) and LQI
(Link Quality Indicator) related to the received frame. Fig. 2
and Fig. 3 represent a network with two sinks and two different
paths.
3) Heuristic Approach: If a source node needs to send
data to a sink, it will refer to one of the hierarchical upper
level neighbours. It will choose the specic neighbour, using
a heuristic approach, similar to [15], taking in account both
battery level and communication quality.
A generic node i stores a risk perception level, Ai , that
intuitively represents a mean trouble level with neighbour
nodes. Each node will store a trustability level, ij , related
to each higher level hierarchical node and they will use
the most trustable node in the neighborhood to reach the
sink. The choosen node will be called delegated node.
Considering a generic node i, and its delegated node j, if
the trustability level, ij , related to j is greater than the risk
perception perceived by i, Ai , it will send data to node j
without considering js battery level, or connection quality

3200
3206

TABLE III
T RUSTABILITY R EQUEST F RAME
TREQF

Se

Re

TABLE IV
T RUSTABILITY R ESPONSE F RAME
TRESF

Se

Re

Battery Level

Link quality

between them.
Condition : if ij > Ai => i sends data to j

(2)

The model considers nodes with a nite memory, so they

forget their history, following an OblivionM echanism; as
a consequence risk perception and trustability will be updated
at certain time steps, as follows:
ij = ij (1 r ) ; Ai = Ai (1 rA )

(3)

In this way the model will force nodes to check models

parameters against neighbors and refresh knowledge perceived.
The r parameter represents the rate of decrease of the
Trustability Level, likewise rA is the rate of decrease of Risk
Perception, and are the time steps before the present time.
If node i does not trust nobody for the Oblivion Mechanism
or Negative Checks (explained below) it will start a check
phase. The check phase consists in asking to delegated node
the link quality indicators (RSSI, LQI) of the last Data Frame
and the battery levels to each higher level node, using
a Trustability Request Frame (TREQF) and a Trustability
Response frame (TRESF), as follows:
Battery levels and connection quality parameters will be
used to recalculate trustability levels related to each higher
level neighbour. We propose to evaluate the correction of
trustability level related to the delegated node, as follows:


Cij + Bweight
bj )
ij next = ij previous + v (Cweight



bnc + Cweight
Cnc
Bweight
) (5)
totalchecks
where bnc represent the total number of negative checks
related to battery levels and Cnc represents the total amount of
negative checks on network connection quality. T otalchecks is
the amount of total checks, both for battery and communication
rightness. vA is a another key parameter because it rules how
the Risk Perception should grow up and it represents a design
parameter as v .
We are going to simulate in future works this proposal model
to nd out the best values of design parameters. We will
consider different topologies and different sinks distribution,
aiming to minimize nodes energy consumption, to identify
bad data frames sent to sinks and to evaluate the performance
of the proposed model.

Ainext = Aiprevious + vA (

(4)

where Cij is the parameter related to connection quality


is the weight related to the imbetween i and j, Cweight
portance of considering the connection quality. bj is equal
to 1 if the battery level of node is lower than the quantity:
M eanbattery levels Standard Deviationbattery levels , otherwise it is equal to zero. We choose such kind of threshold
(Standard Deviationbattery levels ), to consider the global
amount of energy in the neighborhood and its distribution

is the weight related to the imporamong nodes. Bweight

and
tance of considering battery levels in our model. Bweight

Cweight are design values that will be choosed according to the
specic target to obtain. The Ai factor is a key parameter in
our model because it guides the behaviour of a node, according
to the amount of ij , following a heuristic approach. v is a
key parameter because it rules how the trustability should grow
up, so it is a design parameter. For this reason, periodically, at
a certain time step Tperception , nodes will recalculate the risk
perception perceived, using the following relation:

V. C ONCLUSION AND F UTURE W ORKS

In a WSNs context, nodes are able to collect data and send
only the required information to the sink. In a scenario with
multiple sinks, an ideal model should satisfy the need to reach
the nearest sink node. In order to reduce power consumption
of the network and, at the same time, maintain the same
communication performance, we have proposed, in this paper,
a new strategy based on a heuristic approach that rules the
underlying heuristic decision process under certain parameters
such as battery level, communication quality, trustability and
risk perception. In this way, we propose a model based
on heuristic assumptions, dynamic hierarchy and decision
processes to obtain a cognitive energy-preserving behaviour
for the sensor nodes, exploiting a self-organized routing. In
the future works we plan to extend the proposed model,
considering different scenarios, and to generalize, prove and
test the algorithm in order to realise and demonstrate how
the nodes can decide the best path following a top-level
strategy that considers the cognitive knowledge both of the
neighborhood and of the entire network.
R EFERENCES
[1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci: Wireless sensor
networks: a survey. Computer Networks. 38, 393422 (March 2002).
[2] Al-Karaki, J. N. and A. E. Kamal: Routing techniques in wireless sensor
networks: A survey. Wireless Communications, IEEE 11(6), 628 (2004).
[3] T. J. Glynn: Psychological sense of community: Measurement and application. Human Relations, 34, 780818 (1981).
[4] G. Gigerenzer and D.G.Goldstein: Reasoning the Fast and Frugal Way.
Psychological Review, 103(4), 650659 (1996).
[5] J. Pearl: Heuristics: Intelligent Search Strategies for Computer Problem
Solving. Addison-Wesley, Reading, 1984.
[6] G. Gigerenzer, P.M. Todd and the ABC Research Group: Simple Heuristics that make us smart. Evolution and Cognition Series, Oxford University
Press, 1999.
[7] Deepak Goyal, Malay Ranjan Tripathy: Routing Protocols in Wireless
Sensor Networks: A Survey. In: Second International Conference on Advanced Computing & Communication Technologies, IEEE Press, Rohtak,
India (2012).
[8] A. Gandomi, V. A. Tafti: On Wireless Sensor Networks: Architectures,
Information Dissemination Protocols, Applications. In: International Conference on Innovation, Management and Service IPEDR, (2001).
[9] Y. Al-Obaisat, R. Braun: On Wireless Sensor Networks: Architectures,
Protocols, Applications, and Management. In: Auswireless Conference,
(2006).

3201
3207

[10] A. Di Stefano, A. La Corte, M. Leotta, P. Lio, M. Scata: It

measures like me: An IoTs algorithm in WSNs based on heuristic behavior and clustering methods, Ad Hoc Networks (2013).
http://dx.doi.org/10.1016/j.adhoc.2013.04.011.
[11] P. Gajbhiye, A. Mahajan: A Survey of Architecture and Node deployment
in Wireless Sensor Network.In: International Conference on Applications
of Digital Information and Web Technologies (ICADIWT) (2008).
[12] Hasbullah H. , Nazir B. : Region-based Energy-aware cluster (REC)
for efcient packet forwarding in WSN. In: International Symposium in
Information Technology (ITSim), Volume: 3, pp. 16, IEEE (2010).
[13] D. Miorandi, S. Sicari, F. De Pellegrini, I. Chlamtac: Internet of Things:
Vision, Applications and Research Challenges. Ad Hoc Networks, (2012).
[14] G. Kortuem, F. Kawsar, D. Fitton and V. Sundramoorthy.: Smart objects
as building blocks for the Internet of things. In: Internet Computing,
Volume 14(1), pp.44-51. pp. 16, IEEE (2010).
[15] Bagnoli F., Guazzini A., Lio P.: Human Heuristics for Autonomous
Agents. BIOWIRE 2007, LNCS 5151, SpringerVerlag Berlin Heidelberg,
pp. 340351 (2008).
[16] Soyturk, M. : A Routing Algorithm for Mobile Multiple Sinks in
Large-Scale Wireless Sensor Networks. 2nd International Symposium on
Wireless Pervasive Computing, (2007).

3202
3208