Optimization Techniques for Wireless Sensor Networks

Dharini Ganesh, Lalitha M Veeramachaneni, Linda Wong

INFS 612 – Summer 2009, PGN # 4

George Mason University dganesh@gmu.edu, lveerama@gmu.edu, lwong5@gmu.edu

Abstract:

Wireless sensor networks (WSNs) deal with

the major issue of energy limitation in their

deployments. In this paper, we will discuss the

quality of service (QoS) parameters considered

when designing an infrastructure for optimal energy

efficiency in a wireless sensor network. There are

several issues that constrain the WSNs and

challenges posed by the environment of handling

traffic and the lifetime of the battery in the nodes.

We discuss the different techniques proposed by

current studies used to deal with these challenges.

We briefly illustrate different factors and aspects to

evaluate the value each optimization technique

provides from scalability to reliability. Finally, we

designed an algorithm that reduces the amount of

data traffic and satisfies the QoS requirements and

further propose a technique that alleviates the

unbalanced energy usage, data redundancy and

prolong the effective system lifetime.

Introduction:

The wireless sensor network consists of

numerous applications for monitoring different

environments. In comparison to traditional

networks, wireless sensor networks offer improved

functionalities to monitor larger scaled and

changing topology with limited power and

computational abilities in dense deployments.

Wireless sensor networks have been deployed in

application areas for military, environment, health,

traffic control and other areas of interest for

gathering data. Due to the limited supply of energy

that resides on the sensor nodes, it is critical that

the lifetime of the nodes span over a long period

of months or years. Therefore, in the design

phase of the network infrastructure, energy

conservation plays a key role in creating a

network infrastructure that utilizes optimal energy

for their respected system and application. [1]

A distributed sensor network primarily

consists of one or more nodes, wireless

communications device and a sink node, also

known as the gateway node. Sensor node’s energy

source is provided by battery power. The lifetime

of a sensor node is expected to be months to

years, because replacing or recharging a node is

complicated and unfavorable. Due to this

limitation, efficiently using energy from the nodes

has become a crucial challenge. The main

problems that trouble architects are the power and

communication of data in the network. Because

communication is transferred wirelessly, it is

important to mimic the function of transferring

uninterrupted and reliable data that is normally

carried out by a cable. [2] In order to increase

energy efficiency, WSNs also need to reduce

energy consumption as to not completely drain the

life from the nodes. Our algorithm tries to

increase the lifetime of the sensor by

incorporating different levels in the network

hierarchy.

Sensor networks incorporated technologies

from three different research areas: sensing,

communication and computing. [3] Our paper will

focus on the communication issue of the sensor

network. There is a trade-off between reducing

energy consumption and system performance.

Decreasing the node’s energy use means a

downgrade in the transmission of data. For

example, if the packets of data that were sent

contained an error, in order to retransmit that same

data, additional power is necessary. To conserve

energy, applications can coordinate with the system

to places nodes in an inactive state if they are not

gathering data. However, as previously mentioned,

there is a trade-off in this situation as well. As the

nodes are inactive, it increases latency, or delay, of

transmission of packets because the nodes have to

be switched to the active state when called upon. [2]

Another point of concern in communication

networks is the transmission of messages to meet

the requirements set by Quality of Service (QoS)

parameters. Quality of service controls include

message delay, message due dates, bit error rates,

packet loss, economic cost of transmission,

transmission power, etc. [4] There can not be a set

protocol design for QoS standard networks due to

the different capabilities of the applications in its

particular setting. Therefore, there QoS parameters

are subjected to the constraints set by the hardware,

purpose of communication and energy.

Packet merging, data compression, and

aggregation and diffusion, duplicate suppression

can be employed to decrease the size of data to be

transferred in the network, and therefore save

energy of sensor nodes. Clustering has been a

common approach to meet the QoS requirements, in

addition to meeting the constraints of energy in the

sensor nodes. In clustering the wireless sensor

networks are organized into two tiers – cluster head

and cluster members. Cluster heads are the

managers in the cluster structure. They are

responsible for collecting data from the member

nodes and transferring them to the destination node

after the data has been reduced and fused into a

well-organized broadcast. The cluster models help

reduce the amount of data flowing from several

nodes to one sink node, in turn reducing and

extending the life of the member sensor nodes.[5]

This project concentrates on the network

architecture issues and QoS modeling in wireless

sensor networks. We study the various

optimization techniques of wireless sensor

networks in various aspects and focus on meeting

the requirements for efficiently utilizing network

resources with QoS mechanisms. We designed an

algorithm that reduces the amount of data traffic

and satisfies the QoS requirements and further

propose a technique that alleviate the unbalanced

energy usage, data redundancy and prolong the

effective system lifetime.

Research problem:

Saving energy and the communication of

the data sensed by the nodes are two major

issues in the wireless networks. Each node must

consume little power and should work on low

operating and system cost to maintain a large

scale deployment of wireless sensor networks.

Antenna and radio frequency transceiver are

used for communication of sensor nodes with

other nodes. Sensor nodes contain a memory

unit, a CPU, the sensor unit, and the power

source which is usually supplied by batteries.

Many applications need sensor nodes to be

designed as tiny as possible to create a small

network suitable for any location. The small size

of sensor nodes is beneficial for many situations

but the small space also means the availability

for battery capacity is small. A small network

structure also provides another benefit of a

reduced transmission range between nodes.

The main focus for wireless sensor

networks is mostly on energy conservation

through various optimization techniques. These

techniques should concentrate on

communication and operation management as

the power consumption needed for

communication typically dominates a node’s

power budget. In some applications there is a

minimum need for sensing activity for most of

the time and sometimes there is a need for

strong sensor processing at particular instants. In

such cases, there will be large variation of

workloads. Energy awareness must be included

into groups of communicating sensor nodes of the

entire network as well as into the individual nodes

in such applications. [13]

Wireless sensor networks have two

important challenges in the deployment. These

challenges include involvement of large number

of devices and the need to embed them in a

dynamic physical environment. The wireless

sensor network capabilities involve mechanisms

of distributing information over many nodes and

collecting data to a sink node. These mechanisms

should be low energy consuming, scalable with

the number of nodes, and fault tolerant. Some

sensor nodes may fail or paused due to lack of

power, physical damage, or environmental

obstruction. If there are multiple instances of

failure occurring at a time, these mechanisms

need to have the ability to form new links and

routes to the sinks. This may include rerouting

the packets or regulating the power to redirect the

path with ample energy for processing. [13]

Based on the above factors considered, we

classified our problems into two aspects: the first

research problem is network architecture design

issues and the second one is issues in QoS

support.

Design issues:

Network architectures require all nodes to

be able to perform routing, processing, etc. all of

which increase node cost. This node cost is in

terms of energy usage, weight, size, and

reliability. For large scale applications, it costs

the application a significant amount to handle the

robustness of the application. Different

architectural design issues have been considered

depending on various applications and

architectures. Some issues include network

dynamics, node deployment, node

communications, data delivery models, node

capabilities, and data fusion. [6]

Design Issue

Primary Factors

Network Dynamics Mobility of node,

target, and sink

Node Deployment Deterministic or Ad

Hoc

Node

Communications

Single-hop or multi-

hop

Data Delivery

Models

Continuous, event-

driven, query-driven,

or hybrid

Node Capabilities Multi- or single

function;

homogeneous or

heterogeneous

capabilities

Table 1: Architectural design issues [6]

Network dynamics challenges may occur

due to the use of power management or energy

efficient schemes. Network dynamics involve

determining whether the network consists of

static or dynamic nodes. Most sensor nodes are

stationary and monitoring the events in the

vicinity on an as-needed basis. For dynamic

nodes, the mobile nodes are more challenging

and require periodic reporting. Energy is a

primary concern because it may not be possible

to replace or recharge the battery life of the

nodes. Therefore, computing algorithms,

signaling protocols and network states must be

at a minimum to conserve energy spent

processing data. [6]

Node deployment concerns self

organizing or deterministic placement of nodes.

Self organizing nodes are scattered which

requires the system to interact in an impromptu

style because there are no fixed paths and

schedules created. Because the nodes are

randomly scattered, it is important to create a

cluster of nodes that are energy efficient.

Deterministic placement of nodes is manually

placed and the data routes are predetermined and

prescheduled that helps alleviate data collusion.

[6]

Communication between nodes is based

on multi-hop routing paths because it consumes

less energy. However, this technique requires

more management and MAC mechanisms. QoS

mechanisms also have to be designed to handle an

unbalanced traffic flow, as data merges from large

nodes to smaller nodes and finally to a sink node.

[6]

Delivery of the data falls into four

categories: continuous, event-driven, query-driven

and hybrid. Depending on the responsibility of

the node network operation, data could be

constantly sent back to the sink, triggered by an

event on the node, queried from the sink for

information in the vicinity of the node or a

combination of the categories. [6]

Sensor nodes also have different capabilities

based on the application. The nodes are

responsible for relaying, sensing, or aggregating

data collected. Some nodes may only specifically

have one function while certain nodes are

responsible for all three functions. Depending on

the purpose of the sensor, some networks may

only use a subset of the sensors. To avoid wasting

the energy, some of the sensors can be turned on

or off. Because the nodes deployed handle

various monitoring responsibilities, the network

infrastructure contains an assortment of node

models that generate data at different rates, have

different service constraints, and follow different

delivery methods. The infrastructure may also

include different sink nodes that request certain

types of data. WSNs need to support the different

needs of nodes as well. [6]

Quality of Service:

Quality of service (QoS) describes the

techniques used to measure elements of the

network performance including availability,

bandwidth, latency and error rate. QoS, as

defined by RFC 2385, is described as set of

service requirements that needs to be met when

transporting a packet stream from the source to

the destination. [6][7]

The infrastructure design of the QoS

network faces challenges in handling traffic or

routing data. Bandwidth limitations due to

energy constraints, limited computational

resources and collisions with transmissions

make it difficult to secure bandwidth needed to

meet QoS standards. Another challenge is

reducing and aggregating QoS traffic. Buffer

size limitations increase the delay the packets

incur during transmission. Packet queuing

conserves energy by sending a number of

packets at once instead of wasting energy

constantly sending packets but adds delay to the

transfer. The more hops the data has to make to

reach the destination the longer the delivery of

the data. QoS mechanisms also have to

differentiate packet priority based on the

application further adding delay to certain

packets. QoS requirement for delivery faces the

challenges of energy and delay trade-offs. [6][7]

The data redundancy in the wireless

sensor networks helps to loosen the

reliability/robustness requirement of data

delivery, but it unnecessarily spends much

precious energy. In order to achieve a more

network lifetime, energy load must be evenly

distributed among all sensor nodes so that the

energy at a single sensor node or a small set of

sensor nodes will not be drained out very soon.

QoS support should take this factor into

account. Scalability is another major factor to be

considered. QoS support designed for WSNs

should be able to scale up to a large number of

sensor nodes, i.e. it should not degrade quickly

when the number of nodes or their density

increases. WSNs should be able to support

different QoS levels associated with multiple

sinks. QoS mechanisms may be required to

differentiate packet importance and set up a

priority structure. [6][7]

We outlined below research efforts

currently in use to tackle the design and QoS

issues mentioned above. Different techniques like

duplicate suppression and directed diffusion under

data aggregation, address centric and data centric

approaches, in network processing and clustering

will be addressed.

Related Research Work:

Considering the research problems

discussed in the previous section, we have done a

lot of research on different techniques that are

developed to provide an efficient wireless

network. We did the research on various

optimization techniques of wireless sensor

networks in various aspects like network life time,

data loss, scalability, concurrency, reliability and

latency.

Routing in wireless sensor networks faces

many challenges due to the limited energy,

processing and storage capacity, the ad-hoc,

random deployment of the sensor nodes, and the

need for multiple sources to route their traffic to a

single destination. In this paper we will focus on

three routing techniques: data aggregation, in-

network processing, and clustering.

Background:

Event driven applications within WSNs

are characterized by interactive, real-time, mission

critical and require non-end-to-end performance,

meaning a cluster of sensors is connected to the

sink and not just as single sensor node. The

events within the vicinity are quickly detected and

relayed back to the sink. Query driven data

delivery is also interactive, non-end-to-end

applications, mission critical and is query specific.

Data and queries are pushed from the sink to the

sensor nodes versus data pulled from the nodes in

event driven data delivery strategy. Query driven

method is also used to manage and update the

nodes with software upgrades. In both methods,

data is highly redundant because the cluster of

nodes will gather similar data from the event in

the area. A continuous data delivery is a

constant pre-specified transfer rate of data from

the sensors to the sink. A hybrid model

introduces coexistence of all the models

mentioned earlier.

Data centric is a query process where the

routing algorithm searches for the route to the

destination node and then the data is transmitted

along that route. Data centric routing focuses

on the content of the data packets versus the

end-to-end node manner in address-centric

routing. Data centric concepts lay the

foundation for most of the techniques. Further

discussion of how they work together will be

reviewed later in the paper.

Routing Technique: Data aggregation

Data aggregation is the process of

combining data from different sources based on

certain aggregation function like duplicate

suppression, the max, minimum and average of

the data. Sensor nodes within the same vicinity

usually generate redundant data involving the

events that occurred. Based on how the

aggregation function is set up, similar packets

can be aggregated to reduce the number of

transmissions to the sink node. This technique

has proved effective in achieving optimal data

transfer along with achieving energy efficiency

for numerous routing protocols. [8]

Figure 1, Data Aggregation Issues [8]

Data aggregation’s technique reduces the

volume and congestion of traffic carried by the

network, as well as improving bandwidth, energy

and the lifetime of the network as a whole. Data

aggregation protocols can reduce the

communication cost, thereby extending the

lifetime of the sensor.

The idea of combining data from different

sources while enroute, has become a prototype for

wireless routing in WSNs, to eliminate

redundancy along with minimizing the number of

transmissions while reducing energy cost for the

transactions. This technique is a data-centric

approach, which means locating routes from

multiple sources and merging them into a single

destination. Its purpose is to consolidate

redundant data within the network. This approach

is more appropriate for data aggregation technique

compared to the traditional address-centric

approach that requires finding the shortest routes

between end-nodes.

The two key problem areas to focus on for data

aggregation are:

1. Maximize data rate possible from a group

of nodes to a sink under power constraint

and ensure the data is reliable.

2. Explain a technique to achieve optimal

data limit of aggregation

Features of Data Aggregation Techniques:

• Nodes are aware of each other and able

to communicate in order to form a path

to the sink.

• Focus on a data-centric approach, a

many-to-one information flow.

• Functions to only transmit relevant

information.

Examples:

Duplicate Suppression, Directed Diffusion

Directed Diffusion:

Directed diffusion is a communication

scheme where all the nodes are application

aware. Its purpose is to save energy by

diffusion of data through sensor nodes, which

will get rid of unnecessary operations of the

network layer routing. It allows data retrieval in

forms of node requests for named data using

attribute-value pairs and queries on an on-

demand basis.

Diffusion involves two phases of

operation. In the first phase, a "sink" node

requests and broadcasts exploratory interest

messages that flood and are geographically

routed towards nodes in the region throughout

the sensor network. Exploratory interest

messages are messages that contain the

interests, data messages, gradients and

reinforcements to interrogate the nodes on

specifically what the user needs [14]. The set of

characteristics, which specify constraints of the

data that the sink expects, is named using

attribute-value pairs. The attributes contain

information such as name of objects, interval

duration, geographical area, etc. When the

interest messages are broadcasted to the network

from the sink node, each node receiving the

interest can do caching for later use. The

interests in the caches are then used to compare

the received data with the values in the interests.

The interest contains gradients that are setup to

direct the interest to the events that match the

attributes. Gradients act as road signs along the

path of nodes to the source node of interest and

are characterized by the data rate, duration and

expiration time from the received interests. Paths

are established between the sink and the source

using the gradients and interest. [14] Sources

with data matching these constraints send

exploratory data messages back along the

gradients with the exploratory interests received

(which could involve several neighbors). In the

second phase, the sink, upon receiving interest

messages, sends reinforcement interest messages

to specific neighbors who delivered useful data.

These neighbors reinforce the paths established

that provided useful data and this process

continues. [10] The sink can resend the original

interest message along the path established with

smaller energy interval because the path is

reinforced with frequently sent data. The diagram

below depicts the operation of "original"

diffusion.

Figure 2: Direct Diffusion [9]

An important feature of directed diffusion is that

localized interactions, exchanging of messages

between neighboring nodes, governs how data is

aggregated and propagated. Data diffusion shows

that multi-path delivery can save energy when

neighboring nodes aggregate responses to

queries versus end-to-end communication in

traditional data networks. [14]

Pros

1. Data centric: All communications are

neighbor to neighbor with no need for a

node addressing mechanism. There are

no “routers” because each node can

interpret interest messages.

2. Each node can do aggregation &

caching.

3. Energy efficiency of the network

improves as reinforcement maintains an

adequate number of high quality paths.

Cons

1. On-demand, query-driven: Inappropriate

for applications requiring continuous

data delivery, e.g., environmental

monitoring.

2. Attribute-based naming scheme is

application dependent.

• For each application it should be

defined a priority.

• Extra processing overhead at

sensor nodes.

Duplicate Suppression:

The simplest data aggregation function

is duplicate suppression. Duplicate suppression

is restraining repeated notifications of the same

event from nodes in the nearby groups. In

figure 3 below, events sent from source 1 and 2

contain the same data that is sent to node B.

Based on time synchronization, where events

are time stamped with the frequency precision,

duplicates are easier to recognize and prevents

redundant notification of nearby nodes to be

recognized.

Figure 3: Duplication suppression [9]

Routing technique: In-network data processing

The basic idea behind in-network

processing is to reduce the amount of data to be

transmitted and equal amount energy usage of

entire network.

Figure 4: In-network data aggregation example

[12]

A typical in-network data aggregation scheme is

shown in the above figure. All sensor devices

inside the region detect the event. Each sensor

transmits its signal strength to its neighbors and if

the neighbor has a higher strength, the sender

becomes inactive and stops transmitting packets.

Otherwise, it waits for packets from other

sensors and after receiving packets from all its

neighbors, the sender with the highest signal

strength, becomes the data aggregator and all

other sensor devices stop detecting the event

and helps only in routing the packet to the sink

nodes.

In-network data processing is required

because transmitting data requires more energy

than processing data. In-network data

processing helps lengthen the network lifetime

and minimize the amount of data that needs to

be transmitted. It consolidates the information

acquired by different sensors at specific nodes

within the network versus each individual node

transmitting data to the sink. Other benefits of

in-network processing include equal energy

usage, suppression of duplicate messages and

prevention of bottleneck at the gateway.

Figure 5: In-network example [24]

Pros

1. It has the ability to perform cluster

communications

2. Less storage and computational charges.

3. This scheme is highly scalable.

Cons

1. Compromised nodes represent a big

threat to the security of in-network data

processing.

2. It assumes that sink node is never

compromised

3. Improper design of in network processing

may cause time delays.

There are several factors that need to be

considered during the design phase of the in-

network processing models. Such factors include

density of the network, degree of separation,

physical location where the aggregation will

occur, the separation between source and sink,

and the time-delay trade-off associated with each

of these factors should be taken into account.

Routing Technique: Clustering

In order to support data aggregation

through efficient organization of the network,

sensor nodes can be partitioned into a number of

small groups called clusters. Building the

hierarchy levels among the network nodes is

known as clustering. In a clustering mechanism,

the nodes that are adjacent geographically are

grouped to form a cluster. Every node in each

cluster is assigned a job of either being the cluster

head or the member nodes. The member node

takes care of the transmission and arrangement of

nodes within the cluster. The cluster head takes

care of transferring the data to other clusters

within the network by maintaining the routing

information.

Cluster heads (CHs) form the higher level

while member nodes form the lower level. Figure

6 illustrates data flow in a clustered network. The

member nodes report their data to the respective

CHs. The CHs aggregate the data and send them

to the central base through other CHs. Because

CHs often transmit data over longer distances,

they lose more energy compared to member

nodes. The network may be re-clustered

periodically in order to select energy-abundant

nodes to serve as CHs, thus distributing the load

uniformly on all the nodes. The cluster formation

helps reduce energy consumption, communication

latency, traffic load and routing overhead.

Figure 6: Clustering

Pros 1. Has high energy efficiency

2. Reduces channel contention and packet

collisions, resulting in better network

throughput under high load.

Cons

1. Overload to cluster heads results in

reduced reliability on them

Analysis

Based on our analysis of the existing

techniques, we summarized and compared these

techniques against different metrics like

scalability, flexibility, concurrency, energy

efficiency etc.

The following table classifies and provides a comparison of the various optimization techniques:

Solution to research problem:

In this paper, we have thoroughly

researched and reviewed some of the popular

optimization approaches for wireless sensor

networks. The objective of this paper is to provide

a comparison of these different techniques and

propose a feasible hybrid solution. For this

purpose, we derive a conclusion that a

combination of in-networking processing,

aggregation and clustering can be an efficient

solution which can be implemented to all types of

sensor nodes (homogenous or heterogeneous) as

well as applications involving mobility of nodes.

Our algorithm is depicted in the flowchart shown

below.

Figure 8: Flowchart of analysis

Figure 10: Level 1

Figure 9: Level 2

Figures 9 and 10 depict the two levels of

hierarchy in our solution. Level 1 is clustering

of nodes and level 2 is aggregation of the cluster

head nodes to form an aggregator node, which

transmits data to the destination sink.

The various sensor nodes placed in the

target environment search for the data from

surrounding events. When they detect the data,

the sensor nodes calculate the node coefficient

Nc for their nodes which is determined by the

following three factors, namely

• Energy in node’s battery = existing

energy/capacity of battery

• Success rate of the node = No of

successful transmissions/Total No of

transmissions

• Traffic load of each node = free space of

buffer/total space of buffer

Figure 9: Traffic load for node

The values for the three factors are

between 0 and 1. The threshold value for each of

the factors has been set as 0.5. The node

coefficient is given by the summation of all these

three values. The data in the sensor nodes which

have Nc below the threshold value are discarded

as the transmission of these data with very less Nc

through a very large distance will only cause

much loss to the energy in the nodes. The node

with the highest node coefficient becomes the

cluster head for this instance. Thus all the clusters

are assigned the cluster heads and the cluster

heads collect the data to pass it them to the next

higher level of aggregator nodes. The aggregator

nodes which are part of the next level have high

energy and are able to successfully transmit the

data to the source. Depending upon the

application, the data delivery can be implemented

as an event-driven, query-driven or a hybrid

model of continuous, event-driven and query-

driven data delivery. Our approach of integrating

hierarchical clustering wit in-network data

aggregation provides substantial energy

optimization with the limited tradeoff of data loss.

Thus, our proposed algorithm using this technique

achieves 1) flexibility for adding new sensor

network devices 2) scalability 3) high

concurrency and 4) low communication overhead.

However, depending upon the type of situation

and application, different techniques provide

different benefits.

Summary:

Wireless sensor networks propose

several new requirements for QoS parameters

that are different from the traditional models for

applications. The power and communication of

data are the main problems of a wireless sensor

network. With its limitations, it is important to

design a network that uses optimal energy

resources while transferring reliable data. We

discussed the major features with two of the

common techniques used today and lay out the

pros and cons for each technique.

Through this paper, we present an

algorithm that considers energy efficiency as the

important objective for routing and design of the

infrastructure. The algorithm also puts into

consideration the effectiveness of the data

gathered by combining aggregation and fusion

techniques. There are more open issues for

further research in the future that include

consideration of Bluetooth technologies and

encryption services that will stimulate more

requirements and extensive exploration for

techniques to suit the developing technical

world.

References:

[1] J. Sun, “Multi-threshold based data

gathering algorithms for Wireless Sensor

Nodes”,

http://www.academypublisher.com/jnw/vol04/n

o01/jnw04013041.pdf, 2009

[2] R. Johnstone, D. Caputo, U. Cella, A.

Gandelli, C. Alippi, F. Grimaccia, N. Haritos, R.

Zich, “Smart Environmental Measures &

Analysis Technologies (SEMAT): Wireless

sensor networks in marine environment”,

http://www.ict-

sensei.org/index2.php?option=com_docman&ta

sk=doc_view&gid=16&Itemid=49, 2009

[3] Chong, C.Y., Kumar, S.P., 2003. Sensor

networks: evolution, opportunities, and

challenges. Proceedings of the IEEE 91 (8),1247–

1256.

[4] F.L. Lewis, “Wireless Sensor Networks”,

http://arri.uta.edu/acs/networks/WirelessSensorNe

tChap04.pdf, 2004

[5] J. Luo, H. Liu, R. Li, L. Bao, “QoS-oriented

asynchronous clustering protocol in Wireless

Sensor Networks”,

http://www.ics.uci.edu/~lbao/publications/LuoWI

COM2008.pdf, 2008

[6] M. Younis, K. Akkaya, M. Eltoweissy, A.

Wadaa, “On Handling QoS Traffic in Wireless

Sensor Networks,”

www.cs.umbc.edu/~younis/Publications/HICSS20

04/QoS_final.pdf, 2004

[7] D. Chen, P. Varshney, “QoS Support in

Wireless Sensor Networks: A Survey”,

www.cs.binghamton.edu/~kang/teaching/cs580s/q

os-survey2.pdf, 2004

[8] D. Trigos, “Toward Optimal Data Aggregation

in Random Wireless Sensor Networks”,

http://z2.inf.tu-

dresden.de/dargie/wsn/slides/students/2008/002.p

pt, 2008

[9] J. Heidemann, F. Silva, C. Intanagonwiwat, R.

Govindan, D. Estrin, and D. Ganesan, “ Building

efficient wireless sensor networks with low-level

naming. In Proceedings of the Symposium on

Operating Systems and Principles, pages 146-159,

Banff, Alberta, Canada Oct. 2001

[10] K. Akkaya, M. Younis “A survey on routing

protocols for wireless sensor networks”,

http://www.adms.ethz.ch/teaching/lectures/ss2007

/hotdms/papers/routing_survey.pdf, 2003

[11] B. Krishnamachari, D. Estrin, S. Wicker

“Modelling data-centric routing in wireless sensor

networks”,

http://lecs.cs.ucla.edu/Publications/papers/Bhas

kar-DataCentric.pdf, 2002

[12] K. Vaidyanathan, S. Sur, S. Narravula, P.

Sinha, “Data Aggregation Techniques in Sensor

Networks”, ftp://ftp.cse.ohio-state.edu/pub/tech-

report/2004/TR60.pdf, 2004

[13] C. Mu-Huan, “Energy Optimization in

Sensor Networks”,

http://www.lib.ncsu.edu/theses/available/etd-

11022007-115419/unrestricted/etd.pdf, 2007

[14] C. Intanagonwiwat, R. Govindan, D. Estrin,

J. Heidemann, F. Silva, “Directed Diffusion for

Wireless Sensor Networking”,

http://www.isi.edu/~johnh/PAPERS/Intanagonw

iawat02b.pdf

[15] J. Elson, “Time Synchronization Services

for Wireless Sensor Networks”,

http://www.circlemud.org/~jelson/writings/qual

s.pdf, 2001

[16] Alessandro Sorniotti, Laurent Gomez,

Konrad Wrona and Lorenzo Odorico, “Secure

and Trusted in-network Data Processing in

Wireless Sensor Networks: a Survey” in Journal

of Information Assurance and Security 2 (2007)

189-199

[17] Farnaz Dargahi, Amir Masoud Rahmani,

Sam Jabehdari, “Nodes’ Credit based Directed

Diffusion for wireless sensor networks ” in

International Journal of Grid and Distributed

Computing, Dec 2008

[18] K. Akkaya and M. Younis, "A Survey of

Routing Protocols in Wireless Sensor Networks,

" in the Elsevier Ad Hoc Network Journal, Vol

3/3 pp. 325-349, 2005

[19] B. Krishnamachari, D. Estrin, and S.

Wicker, “Modelling data-centric routing in

wireless sensor networks” In Proceedings of

IEEE INFOCOM’02, June 002.

[20] K Vaidyanathan, S. Sur, S. Narravula, P.

Sinha. “Data-Aggregation Techniques in Sensor

Networks”, Technical Report OSU - 2004.

[21] E. Fasolo, M. Rossi, J. Widmer, and M.

Zorzi, “In-Network Aggregation Techniques for

Wireless Sensor Networks: A Survey”, IEEE

Wireless communication 2007.

[22] P. Vijayakumar, V. Ravichadran, “Wireless

Sensor Network Issues-An Overview”,

International Journal of Wireless Networks and

Communications Volume 1, Number 1 (2009), pp.

17–28

[23] P. Katopodis, “A Hybrid, large-scale wireless

sensor network for real-time acquisition and

tracking”, http://www.dtic.mil/cgi-

bin/GetTRDoc?AD=ADA470111&Location=U2

&doc=GetTRDoc.pdf, 1997

[24] S. Chatterjea, “In-Nework data aggregation

in Wireless Sensor networks”,

http://www.gleonrcn.org/pgm-

download_media.php?name=SensorNetworkSyste

ms.ppt,