USF Libraries
USF Digital Collections

Scalable energy-efficient location aided routing (SELAR) protocol for wireless sensor networks

MISSING IMAGE

Material Information

Title:
Scalable energy-efficient location aided routing (SELAR) protocol for wireless sensor networks
Physical Description:
Book
Language:
English
Creator:
Lukachan, George
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Large scale
Sensor node
Multihop routing
Network lifetime
Power constraint
Dissertations, Academic -- Computer Science -- Masters -- USF
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: Large-scale wireless sensor networks consist of thousands of tiny and low cost nodes with very limited energy, computing power and communication capabilities. They have a myriad of possible applications. They can be used in hazardous and hostile environments to sense for deadly gases and high temperatures, in personal area networks to monitor vital signs, in military and civilian environments for intrusion detection and tracking, emergency operations, etc. In large scale wireless sensor networks the protocols need to be scalable and energy-efficient. Further, new strategies are needed to address the well-known energy depletion problem that nodes close to the sink node face. In this thesis the Scalable Energy-efficient Location-Aided Routing (SELAR) protocol for wireless sensor networks is proposed to solve the above mentioned problems. In SELAR, nodes use location and energy information of the neighboring nodes to perform the routing function. Further, the sink node is moved during the network operation to increase the network lifetime. By means of simulations, the SELAR protocol is evaluated and compared with two very well-known protocols - LEACH (Low-Energy Adaptive-Clustering Hierarchy) and MTE (Minimum Transmission Energy). The results indicate that in realistic senarios,SELAR delivers up to 12 times more and up to 1.4 times more data packets to the base station than LEACH and MTE respectively. It was also seen from the results that for realistic scenarios, SELAR with moving base station has up to 5 times and up to 27 times more lifetime duration compared to MTE and LEACH respectively.
Thesis:
Thesis (M.S.)--University of South Florida, 2005.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by George Lukachan.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 56 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001912727
oclc - 174042330
usfldc doi - E14-SFE0001366
usfldc handle - e14.1366
System ID:
SFS0025686:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Scalable Ener gy-ef cient Location-Aided Routing (SELAR) Protocol for W ireless Sensor Netw orks by Geor ge Lukachan A thesis submitted in partial fulllment of the requirements for the de gree of Master of Science in Computer Science Department of Computer Science and Engineering Colle ge of Engineering Uni v ersity of South Florida Major Professor: Miguel Labrador Ph.D. K en Christensen, Ph.D. Adriana Iamnitchi, Ph.D. Date of Appro v al: No v ember 1, 2005 K e yw ords: lar ge scale, sensor node, multihop routing, netw ork lifetime, po wer constraint c r Cop yright 2005, Geor ge Lukachan

PAGE 2

A CKNO WLEDGEMENTS I tak e this opportunity to e xpress my sincere thanks to Dr Miguel Labrador for gi ving me this w onderful opportunity of w orking on this project. I am also grateful to him for his e xtended support and guidance throughout the course of this w ork, and for making my study at USF a pleasant and e xciting educational e xperience. My sincere thanks to Dr Christensen and Dr Iamnitchi, for being in my committee and for their v aluable comments and suggestions. It tak es more than w ords to e xpress my thanks to my f amily for their constant moti v ation and support, without which this w ork w ould not ha v e been possible. I thank all my friends for their continuous encouragement and support.

PAGE 3

T ABLE OF CONTENTS LIST OF T ABLES iii LIST OF FIGURES i v ABSTRA CT vi CHAPTER 1 INTR ODUCTION 1 1.1 Routing in wireless sensor netw orks 5 1.1.1 Design challenges for routing 5 1.1.2 Solution pro vided by SELAR 6 1.2 Contrib utions of this thesis 6 1.3 Document structure 6 CHAPTER 2 RELA TED W ORK 8 2.1 Design of wireless sensor netw ork routing protocols 8 2.1.1 Ov ervie w of routing protocols 8 2.2 Re vie w of wireless sensor netw ork routing protocols 9 2.2.1 Lo w-Ener gy Adapti v e Clustering Hierarchy (LEA CH) 10 2.2.2 Minimum T ransmission Ener gy (MTE) 11 2.2.3 Small Minimum Ener gy Communication Netw ork (SMECN) 12 2.2.4 Flooding and gossiping 12 2.2.5 Sensor Protocols for Information via Ne gotiation (SPIN) 13 2.2.6 Sequential Assignment Routing (SAR) 14 2.2.7 Directed dif fusion 14 2.2.8 SPEED 15 2.3 Distance Routing Ef fect Algorithm for Mobility (DREAM) 15 CHAPTER 3 THE SCALABLE ENERGY -EFFICIENT LOCA TION AIDED R OUTING (SELAR) PR O T OCOL 17 3.1 Design considerations 18 3.2 W orking of the protocol 19 3.2.1 Control pack et dissemination phase 20 3.2.2 Data pack et dissemination phase 21 3.2.3 Algorithm for determining neighbor nodes within the for w arding zone 23 3.3 Mo ving the base station 24 i

PAGE 4

CHAPTER 4 SIMULA TION AND RESUL TS 25 4.1 CMU wireless and mobility platform 25 4.2 MIT uAMPS ns-2 code e xtensions 26 4.2.1 Resource-adapti v e node 26 4.2.2 Radio model 28 4.2.3 Netw ork interf ace 28 4.2.4 MA C implementation 30 4.2.5 Base station application 31 4.2.6 Implementation of LEA CH and MTE 31 4.3 Implementing SELAR using the MIT uAMPS ns-2 code e xtensions 32 4.3.1 SELAR with mo ving base station 32 4.4 Restriction applied to radio range 32 4.5 Simulation scenario 33 4.6 Results 34 4.6.1 Data pack ets 35 4.6.2 Netw ork lifetime 38 4.6.3 Ener gy distrib ution for MTE 41 4.6.4 Ener gy distrib ution for LEA CH 44 4.6.5 Ener gy distrib ution for SELAR with stationary base station 47 4.6.6 Ener gy distrib ution for SELAR with mo ving base station 49 CHAPTER 5 CONCLUSION AND FUTURE W ORK 52 REFERENCES 54 ii

PAGE 5

LIST OF T ABLES T able 4.1 P arameters used for radio model [1 ] 28 T able 4.2 A v erage number of data pack ets recei v ed at the base station 36 T able 4.3 A v erage number of data pack ets recei v ed at the base station after restricting the maximum radio range to cross-o v er distance (86.2 m) 37 T able 4.4 A v erage netw ork lifetime in simulation seconds 39 T able 4.5 A v erage netw ork lifetime in simulation seconds after restricting the maximum radio range to cross-o v er distance (86.2 m) 40 iii

PAGE 6

LIST OF FIGURES Figure 1.1 En vironmental monitoring using sensor netw orks 2 Figure 1.2 Components of a sensor node [2 ] 3 Figure 1.3 Sensor netw ork protocol stack 4 Figure 2.1 Classication of routing protocols [3 ][4 ] 10 Figure 2.2 W orking of SPIN protocol 14 Figure 3.1 Control pack et dissemination 21 Figure 3.2 F ormat of the state table maintained at e v ery sensor node 21 Figure 3.3 Data pack et dissemination 22 Figure 3.4 Determining whether a neighbor node is within the forw arding zone 23 Figure 4.1 Block diagram of an ns-2 mobile node [5 ] 26 Figure 4.2 Block diagram of a resource-adapti v e node [5 ] 27 Figure 4.3 Radio ener gy dissipation model [1] 28 Figure 4.4 Mo ving the base station in SELAR 33 Figure 4.5 A v erage number of data pack ets recei v ed at the base station 36 Figure 4.6 A v erage number of data pack ets recei v ed after restricting radio range to cross-o v er distance (86.2 m) 37 Figure 4.7 A v erage lifetime of the netw ork in simulation seconds 39 Figure 4.8 A v erage lifetime of the netw ork in simulation seconds after restricting the maximum radio range to cross-o v er distance (86.2 m) 40 Figure 4.9 Ener gy distrib ution for MTE with around 90 nodes ali v e 42 Figure 4.10 Ener gy distrib ution for MTE with around 50 nodes ali v e 42 Figure 4.11 Ener gy distrib ution for MTE with around 10 nodes ali v e 42 i v

PAGE 7

Figure 4.12 Ener gy distrib ution for MTE (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 90 nodes ali v e 43 Figure 4.13 Ener gy distrib ution for MTE (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 50 nodes ali v e 43 Figure 4.14 Ener gy distrib ution for MTE (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 10 nodes ali v e 43 Figure 4.15 Ener gy distrib ution for LEA CH with around 90 nodes ali v e 45 Figure 4.16 Ener gy distrib ution for LEA CH with around 50 nodes ali v e 45 Figure 4.17 Ener gy distrib ution for LEA CH with around 10 nodes ali v e 45 Figure 4.18 Ener gy distrib ution for LEA CH (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 90 nodes ali v e 46 Figure 4.19 Ener gy distrib ution for LEA CH (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 50 nodes ali v e 46 Figure 4.20 Ener gy distrib ution for LEA CH (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 10 nodes ali v e 46 Figure 4.21 Ener gy distrib ution for SELAR (stationary base station) with around 90 nodes ali v e 48 Figure 4.22 Ener gy distrib ution for SELAR (stationary base station) with around 50 nodes ali v e 48 Figure 4.23 Ener gy distrib ution for SELAR (stationary base station) with around 10 nodes ali v e 48 Figure 4.24 Ener gy distrib ution for SELAR (mo ving base station) with around 90 nodes ali v e 50 Figure 4.25 Ener gy distrib ution for SELAR (mo ving base station) with around 50 nodes ali v e 50 Figure 4.26 Ener gy distrib ution for SELAR (mo ving base station) with around 10 nodes ali v e 50 v

PAGE 8

SCALABLE ENERGY -EFFICIENT LOCA TION-AIDED R OUTING (SELAR) PR O T OCOL FOR WIRELESS SENSOR NETW ORKS Geor ge Lukachan ABSTRA CT Lar ge-scale wireless sensor netw orks consist of thousands of tin y and lo w cost nodes with v ery limited ener gy computing po wer and communication capabilities. The y ha v e a myriad of possible applications. The y can be used in hazardous and hostile en vironments to sense for deadly gases and high temperatures, in personal area netw orks to monitor vital signs, in military and ci vilian en vironments for intrusion detection and tracking, emer genc y operations, etc. In lar ge scale wireless sensor netw orks the protocols need to be scalable and ener gy-ef cient. Further ne w strate gies are needed to address the well-kno wn ener gy depletion problem that nodes close to the sink node f ace. In this thesis the Scalable Ener gy-ef cient Location-Aided Routing (SELAR) protocol for wireless sensor netw orks is proposed to solv e the abo v e mentioned problems. In SELAR, nodes use location and ener gy information of the neighboring nodes to perform the routing function. Further the sink node is mo v ed during the netw ork operation to increase the netw ork lifetime. By means of simulations, the SELAR protocol is e v aluated and compared with tw o v ery well-kno wn protocols LEA CH (Lo w-Ener gy Adapti v e-Clustering Hierarchy) and MTE (Minimum T ransmission Ener gy). The results indicate that in realistic senarios, SELAR deli v ers upto 12 times more and upto 1.4 times more data pack ets to the base station than LEA CH and MTE respecti v ely It w as also seen from the results that for realistic scenarios, SELAR with mo ving base station has upto 5 times and upto 27 times more lifetime duration compared to MTE and LEA CH respecti v ely vi

PAGE 9

CHAPTER 1 INTR ODUCTION The adv ances in semiconductor technology has led to a gi v en computing capacity becoming smaller and less e xpensi v e with each passing year This has led to the creation of miniature radios and sensors which can sense forces in the physical w orld. These ine xpensi v e radios and sensors are combined to create what is kno wn as a sensor node. W ireless sensor netw orks consist of hundreds to thousands of tin y sensor nodes which are constrained in terms of ener gy po wer and communication capabilities. Sensor nodes transmit the information the y sense to a special node kno wn as the base station. Base station, also kno wn as a sink node, has signicantly higher ener gy and computational capabilities compared to a sensor node. Base station can be rechar ged using e xternal sources or be pro vided with, for e.g., a solar panel to rechar ge itself. W ireless sensor netw orks ha v e a myriad of possible applications. The applications of wireless sensor netw orks can be roughly classied into three cate gories [6]: monitoring space : This cate gory of applications include en vironmental and habitat monitor ing, precision agriculture, indoor climate control, surv eillance, treaty v erication and intelligent alarms. monitoring things : Structural monitoring, ecophysiology condition-based equipment maintenance, medical diagnostics and urban terrain mapping are the applications of wireless sensor netw orks f alling under this cate gory monitoring the inter action of things with other things and the encompassing space : Most of the dramatic applications of wireless sensor netw orks f all under this cate gory which include applications lik e en vironmental monitoring, disaster management, emer genc y response, asset tracking and manuf acturing process o w 1

PAGE 10

Figure 1.1 En vironmental monitoring using sensor netw orks Figure 1.1 depicts en vironmental monitoring using wireless sensor netw orks. Nodes scattered in dif ferent areas of the forest can be used to track v arious physical conditions lik e temperature, sunlight incidence, and so on. One e xample of en vironmental monitoring is monitoring the microclimate throughout the v olume of redw ood trees to form a sample of entire forests [6 ]. Redw ood trees are lar ge enough to en v elope an entire ecosystem. By placing wireless sensor nodes at v arious ele v ations of the tree we can measure incident light, radiant light, relati v e humidity barometric pressure and temperature. Using the data recorded it can be seen ho w the weather front mo v es up and do wn the tree. Another application of wireless sensor netw orks, namely motion monitoring can be used for condition-based maintenance [6 ]. Physical structures such as motors, airplane wings, and bridges ha v e typical modes of vibration, acoustic emissions and response to stimuli. Mechanical changes to these physical structures will be reected in their vibration modes, acoustic emissions, and response to stimuli. T in y wireless sensor nodes can be placed on the physical structure to sense the vibrations, acoustic emissions and can transmit them to the monitoring station. Alternati v ely the sensor nodes can continuously process the information it senses, if it nds an y aberrations it can transmit the necessary data to the monitoring station. Figure 1.2 graphically depicts the components of a sensor node. The four basic units of a sensor node are the sensing unit pr ocessing unit tr ansceiver and power unit A Sensing unit consists 2

PAGE 11

Figure 1.2 Components of a sensor node [2] of sensors and analog-to-digital con v erters (ADCs). Sensors produce analog signals based on the physical phenomenon the y observ e. The analog signals produced by the sensors are con v erted to digital signals by the ADC. The processing unit recei v es signal from the ADC. The processing unit consists of a processor as well as storage. The processing unit contains procedures to collaborate with other nodes in the netw ork. Ev ery node is connected to the netw ork via the transcei v er The po wer unit po wers all other units in the sensor node to perform sensing, processing, transmission and reception of data. The sensor node can ha v e additional units lik e po wer generators to supply po wer to the po wer unit. Sensor node can ha v e a location nding system to calculate relati v e or absolute location of itself as well as that of other nodes. F or certain applications the sensor node might require to be mobile and will ha v e a mobilizer unit attached to it. The protocol stack used in wireless sensor netw orks is sho wn in Figure 1.3. A brief description of each layer is as follo ws: Physical layer : It addresses the needs of rob ust modulation, transmission and recei ving techniques. 3

PAGE 12

Figure 1.3 Sensor netw ork protocol stack Data link layer : This layer should be po wer -a w are and minimize collisions with neighboring broadcasts. It should also pro vide f air access to the media and high netw ork utilization. Network layer : Ener gy-ef cient routing is performed at this layer Netw ork layer should be simple with no high requirements in terms of storage, computations and communication o v erhead. T r ansport layer : This layer maintains the o w of data. Most of the time UDP protocol is utilized to send small amounts of data. Application layer : This layer can run application softw are depending on the type of sensing task. This thesis focuses on the netw ork layer proposing a scalable and ener gy-ef cient routing protocol for lar ge-scale wireless sensor netw orks. 4

PAGE 13

1.1 Routing in wir eless sensor netw orks The routing function in wireless sensor netw orks is challenging because of the po wer storage and computational constraints of the nodes. In this section the design challenges posed by routing in wireless sensor netw orks and the solution pro vided by SELAR [7] are re vie wed. 1.1.1 Design challenges f or r outing The main challenges in designing a routing protocol for wireless sensor netw orks are as follo ws: F ault T oler ance : The ability to sustain sensor netw ork functionalities without an y interruption due to sensor node f ailures [8 ] [9 ]. Sensor nodes may f ail due to physical damage or en vironmental interference. The f ailure of fe w nodes should not af fect the o v erall producti vity or functionality of the netw ork. The routing function should be able to route around f ailures. Scalability : A sensor netw ork is said to be scalable if an increase in sensor nodes increase the functionality of the netw ork. The routing protocol should be designed in such a w ay that it is able to w ork with lar ge number of nodes spread throughout lar ge area. Pr oduction Costs : The production cost of a single sensor node should be such that the o v erall cost of deplo ying a wireless sensor netw ork is signicantly cheaper than deplo ying traditional sensors. F or a lar ge-scale sensor netw ork to be feasible, the cost of a single sensor node should be much less than US$1 according to [2]. The routing function should be as simple as possible so that no high po wer CPU and storage capabilities are needed. P ower/Ener gy Constr aints : Lar ge scale wireless sensor nodes ha v e e xtremely lo w amount of po wer and ener gy at their disposal. The routing protocol should be designed in such a w ay that each and e v ery transmission and reception performed is justied. Rotuing o v erhead must be k ept to a minimum. 5

PAGE 14

1.1.2 Solution pr o vided by SELAR Scalable Ener gy-ef cient Location Aided Routing protocol is meant to be a f ault-tolerant, scalable and ener gy-ef cient protocol. Ev ery sensor node in a sensor netw ork using SELAR maintains a state table containing the location and ener gy information of all its neighbor nodes. In SELAR, a sensor node forw ards data pack ets to its neighbor node based on the ener gy left in the neighbor node. Hence, the ener gy of the netw ork is used in a uniform manner Further the random f ailure of nodes in the netw ork do not af fect its o v erall functioning. The SELAR protocol is capable of handling a lar ge number of sensor nodes as routing decisions are localized. Therefore, the addition of more nodes into the netw ork do not de grade the performance of the netw ork. SELAR does not use comple x computations at the sensor nodes and routes pack ets in the netw ork by being po wer a w are leading to ef cient utilization of node ener gies. 1.2 Contrib utions of this thesis The contrib utions of this thesis are the follo wing: Pro vides a comprehensi v e re vie w of wireless sensor netw ork routing protocols. Proposes an ener gy-ef cient, f ault-tolerant and scalable routing protocol for wireless sensor netw orks. Ev aluates the proposed protocol by comparing it with e xisting protocols using standard per formance metrics. Implements the protocol in the ns-2 simulator 1.3 Document structur e This document be gins with a discussion on wireless sensor netw orks and their routing protocols in particular Chapter tw o pro vides an o v ervie w of the routing protocols in wireless sensor netw orks, well kno wn routing protocols and w ork related to the design of the proposed protocol, SELAR. Chapter 3 describes the proposed protocol in detail. In Chapter 4, SELAR is implemented 6

PAGE 15

using ns-2 and compared with tw o other well kno wn routing protocols LEA CH (Lo w-Ener gy Adapti v e-Clustering Hierarchy) [10 ] [1 ] and MTE (Minimum T ransmission Ener gy) [11 ] [12 ]. Finally Chapter 5 pro vides conclusions and future w orks related to SELAR. 7

PAGE 16

CHAPTER 2 RELA TED W ORK In this chapter the design considerations for wireless sensor netw ork routing protocols are discussed. Then some of the well kno wn wireless sensor netw ork routing protocols are re vie wed. The chapter concludes with a description of DREAM (Distance Routing Ef fect Algorithm for Mobility), ad hoc netw ork routing protocol, which inuenced the design of SELAR. 2.1 Design of wir eless sensor netw ork r outing pr otocols Routing is v ery important to the ef cient performance of wireless sensor netw orks. A signicant amount of research is being done to design ef cient routing protocols for wireless sensor netw orks. Some of the desired features of wireless sensor netw ork routing protocols are ener gy-ef cienc y scalability and f ault-tolerance. In the subsections which follo w the design considerations for wireless sensor netw ork routing protocols and their classication are discussed. 2.1.1 Ov er view of r outing pr otocols Routing protocols in wireless sensor netw orks can be broadly classied as follo ws [3] [4 ]: Flat/Data-centric r outing : In this type of routing protocols, each node typically plays the same role and collaborates together to perform the sensing task. The y use attrib ute based addressing because of the infeasibility of assigning global identiers to e v ery node in the netw ork. In at/data-centric routing, the sink node queries sensor nodes in a particular re gion and w aits for data from the sensors located in that particular re gion. Examples of at/datacentric routing protocols are SPIN (Sensor Protocols for Information via Ne gotiation)[13 ], Directed Dif fusion [14 ], and Rumor Routing [15 ]. 8

PAGE 17

Hier ar c hical : Hierarchical routing protocols are cluster -based. These type of routing protocols typically ha v e good scalability and ef cient communication. In hierarchical routing the nodes with higher ener gy can tak e care of aggre gating/processin g data and sending it to the base station while sensor nodes with lo wer ener gy can sense data and send it to the nearby higher ener gy node. Hierarchical routing tries to impro v e the o v erall ener gy-ef cienc y lifetime, and scalability of the sensor netw ork by creating clusters, clusterheads with special tasks assigned to them and by performing data fusion within the cluster The main principle on which hierarchical routing is based on is that it tak es more ener gy to send tw o pack ets than to send one pack et with more data. Examples of hierarchical routing protocols are LEA CH (Lo w-Ener gy Adapti v e Clustering Hierarchy) [10 ], PEGASIS (Po wer -Ef cient Gathering in Sensor Information Systems) [16 ] and SMECN (Small Minimum Ener gy Communication Netw ork) [17 ]. Location-based : In location based routing protocols, location of the sensor node is used to address the node. Sensor nodes can use incoming signal strength to estimate the distance as well as the relati v e coordinates of the neighboring nodes. In some instances, GPS (Global Positioning System) may be used to nd the location of the sensor nodes. Examples of locationbased routing protocols are GAF (Geographic Adapti v e Fidelity) [18 ], GEAR (Geographic and Ener gy A w are Routing) [19 ] and SP AN [20 ]. Figure 2.1 graphically depicts the classication of routing protocols for wireless sensor netw orks. Based on the protocol operation, the wireless sensor netw ork routing protocols can be classied as Multipath-based Query-based Ne gotiation-based QoS-based and Coher ent-based More information about this classication can be found in [3 ]. 2.2 Re view of wir eless sensor netw ork r outing pr otocols This section contains re vie ws of the follo wing wireless sensor netw ork routing protocols: Lo wEner gy Adapti v e Clustering Hierarchy (LEA CH)[10], Minimum T ransmission Ener gy (MTE) [11 ] [12 ], Small Minimum Ener gy Communication Netw ork (SMECN) [17 ], Flooding and Gossiping 9

PAGE 18

Figure 2.1 Classication of routing protocols [3 ][4 ] [21 ], SPIN (Sensor Protocols for Information via Ne gotiation)[13 ], SAR (Sequential Assignment Routing) [22 ], Directed Dif fusion [14 ], SPEED [23 ]. 2.2.1 Lo w-Ener gy Adapti v e Clustering Hierar ch y (LEA CH) Lo w-Ener gy Adapti v e Clustering Hierarchy (LEA CH), a clustering based protocol for wireless sensor netw orks, is discussed in [10 ]. In LEA CH, a set of nodes act as clusterheads and the rest of the nodes perform the sensing function. The nodes which act as clusterheads are changed randomly at re gular interv als of time so that there is uniform dissipation of ener gy throughout the wireless sensor netw ork. The operation of LEA CH is di vided into the setup phase and the steady phase. The duration of the steady phase is set to be longer than that of the setup phase so as to minimize o v erhead. During the setup phase e v ery sensor node chooses a random number between 0 and 1. If the random number chosen by the clusterhead is less than the threshold T(n) then the sensor node 10

PAGE 19

becomes a clusterhead. The Threshold T(n) is determined using Equations 2.1 and 2.2. T ( n ) = P 1 P [ r mod (1 =P ) ] ; if n 2 G ; (2.1) T ( n ) = 0; if n = 2 G (2.2) where P is the desired percentage to become a clusterhead, r is the current round and G is the set of nodes that ha v e not been clusterheads for the past 1 P rounds. Once the clusterheads are chosen, the ne w clusterheads broadcast the ne ws throughout the whole wireless sensor netw ork. Each sensor node recei ving the adv ertisement from clusterheads, chooses the clusterhead it w ants to belong to based on the signal strength of the adv ertisement. Each sensor node informs the clusterhead from which it recei v es the strongest signal that it will become a member of that clusterhead. F ollo wing this each clusterhead assigns a time slot to e v ery sensor node in its cluster during which the sensor node can send data to its respecti v e clusterhead. A TDMA approach is used by the clusterheads to assign time slots to each sensor node in its cluster Once the steady phase be gins, the sensor nodes sense and transmit data to their respecti v e clusterheads. The clusterheads perform data fusion on the data recei v ed from the sensor nodes and sends the aggre gated data to the base station. After a certain amount of time, the steady phase ends and a ne w setup phase be gins follo wed by another steady phase. This process continues until all the sensor nodes in the wireless sensor netw ork ha v e no ener gy left. Data pack et collisions in LEA CH are minimal since it uses a TDMA mechanism within each cluster and a CDMA mechanism when clusterheads transmit data to the base station. 2.2.2 Minimum T ransmission Ener gy (MTE) Minimum T ransmission Ener gy (MTE) protocol is a multihop routing protocol in which data pack ets are forw arded to intermediate nodes in the path to the base station such that transmission ener gy is minimized [11 ] [12 ]. In MTE, if a node needs to send a data pack et to the base station, it selects intermediate nodes on its path to the base station such that the sum of the transmission ener gies dissipated is less than that required to send the data pack et directly to the base station. If a 11

PAGE 20

sensor node A has a node B on its path to the base station C, then node A will transmit to the base station C through node B if and only if d AB 2 + d B C 2 < d AC 2 : (2.3) 2.2.3 Small Minimum Ener gy Communication Netw ork (SMECN) The Small Minimum Ener gy Communication Netw ork (SMECN) is proposed in [17 ]. SMECN is an e xtension of Minimum Ener gy Communication Netw ork (MECN) [24 ]. Gi v en a netw ork topology MECN constructs ener gy-ef cient subnetw orks for e v ery sensor node in the netw ork. MECN nds global minimum po wer paths using localized search for each node without considering all the sensor nodes in the wireless sensor netw ork. MECN assumes that e v ery node can transmit to e v ery other node in the netw ork. SMECN considers the netw ork to be fully connected b ut does not assume e v ery node to be capable of transmitting to e v ery other node in the netw ork. The subnetw ork constructed by SMECN is smaller than the one constructed by MECN if the broadcast re gion is circular around a broadcaster for a gi v en po wer setting. SMECN creates subnetw orks that assist in sending messages on minimum-ener gy paths by constructing a subnetw ork where the minimumener gy path is guaranteed to e xist. By creating smaller subnetw orks than MECN, SMECN increases the probability that the path used is the one requiring minimum ener gy 2.2.4 Flooding and gossiping Flooding is one of the oldest routing techniques. In ooding, each node recei ving a pack et broadcasts that pack et to e v ery other node unless the pack et has reached its destination or the maximum number of hops assigned to the pack et has been reached. Flooding is a simple routing protocol which requires no topology maintenance or route disco v ery algorithms. Flooding protocol has the follo wing dra wbacks: Implosion : Duplicated pack ets are recei v ed by a sensor node. 12

PAGE 21

Overlap : T w o nodes sharing the same observ ation re gion may sense a stimuli at the same time resulting in duplicated messages. Resour ce blindness : Flooding w orks without taking into account the a v ailability of resources lik e ener gy Gossiping [21 ] is a deri v ation of ooding. In gossiping, a sensor node randomly selects a neighbor to forw ard a pack et instead of broadcasting to all its neighbors. The neighbor node, on recei ving the data pack et, selects another sensor node randomly and forw ards the pack et. Gossiping a v oids the implosion problem b ut it may tak e a long time to send the pack et to its destination. 2.2.5 Sensor Pr otocols f or Inf ormation via Negotiation (SPIN) Sensor Protocols for Information via Ne gotiations (SPIN) [13 ] is designed to address the deciencies of classic ooding. SPIN uses ne gotiation and resource adaptation as each node disseminates information to e v ery other node in the netw ork considering them to be potential base stations. SPIN is designed based on the idea that sensor nodes operate more ef ciently and conserv e ener gy by sending data that describe the sensor data instead of sending all the data. SPIN uses three types of messages: AD V : Data adv ertisement messages. REQ : Request for data messages. D A T A : The data pack et/message. The w orking of SPIN is sho wn in Figure 2.2. Initially a sensor node broadcasts an AD V message containing a description of the data it has. The neighbor nodes who are interested in the data sends a REQ message to the rst sensor node. The sensor node holding the data then sends the D A T A message to the nodes from which it recei v ed the REQ message. The neighbor nodes then repeat this process. Finally all nodes in the netw ork interested in the data will ha v e a cop y of the data. In SPIN topological changes are localized since each node needs to kno w only its single-hop neighbors. 13

PAGE 22

Figure 2.2 W orking of SPIN protocol 2.2.6 Sequential Assignment Routing (SAR) In Sequential Assignment Routing (SAR) [22 ], multiple trees are created with the root of each tree being a single-hop neighbor of the base station. Each tree gro ws a w ay from the base station by including nodes with high QoS and ener gy reserv es. Most of the sensor nodes in the netw ork belong to dif ferent trees, enabling the sensor nodes to choose a tree to send data to the base station. The tw o parameters associated with each path in a tree are: Ener gy resources. Additi v e QoS metric. Each node selects the path to the base station based on the ener gy resources, QoS metrics of that path and the pack et' s priority le v el. Single W inner Election (SWE) and Multi W inner Election (MWE) handle the necessary signaling and data transfer tasks during local cooperati v e information processing.2.2.7 Dir ected diffusion Directed dif fusion is a data-centric protocol proposed in [14 ]. In directed dif fusion the base station sends out a task description, which is also kno wn as interest to all the sensor nodes in the netw ork. The naming of task descriptors are done by assigning attrib ute-v alue pairs that describe the task. Each sensor node sa v es the interest in its memory The interest entry consists of se v eral 14

PAGE 23

gradient elds and a timestamp eld. When a sensor node recei ving the interest has data to send to the base station, it becomes the source node. The gradients from the source node back to the sink are setup as the interest is propagated throughout the sensor netw ork. When the base station starts recei ving data from the source node it should reinforce the interest. 2.2.8 SPEED SPEED [23 ] uses geographic forw arding to nd paths to the base station. In SPEED each sensor node is required to maintain information about its neighbors. SPEED uses a routing module called Stateless Non-deterministic Geographic F orw arding (SNGF). The SNGF module w orks with four other modules which are as follo ws: Beacon Exc hang e : This module collects information about nodes and their location. Delay Estimation : This module estimates the delay to each neighbor by calculating the elapsed time when an ackno wledgement is recei v ed from a neighbor in response to a pack et transmitted to it. Neighborhood F eedbac k Loop : This module pro vides the relay ratio which in turn is calculated by considering the miss ratios of the neighbors of a node. Bac kpr essur e r er outing : This module tries to w ork around congestion in the netw ork. When a node f ails to nd out a ne xt hop neighbor it sends the pack et back to the node it recei v ed the pack et from so that a ne w route can be follo wed. 2.3 Distance Routing Effect Algorithm f or Mobility (DREAM) Distance Routing Ef fect Algorithm for Mobility (DREAM) is a routing protocol designed for wireless Ad hoc netw orks [25 ]. DREAM is included here because SELAR' s design is partially inuenced by this protocol. In DREAM each mobile node maintains a location table for all other nodes in the netw ork. Each location table consists of the coordinates of the source node based on some reference system, the source node' s speed and the time the location pack et w as transmitted by 15

PAGE 24

the source node. Each mobile node transmits location pack ets to nearby mobile nodes in the netw ork at a gi v en frequenc y and to f ara w ay mobile nodes in the ad hoc netw ork at a lo wer frequenc y When a node S needs to send information to node D, it calculates a circle around the most recent location information of D using R = V max ( t 1 t 0 ) (2.4) where R is the radius of the circle, V max is the kno wn maximum speed of node D, t 1 is the current time and t 0 is the timestamp for this node' s information from the location table. Then node S denes a forw arding zone whose v erte x is at S and whose sides are tangent to the circle around D. S then sends a data pack et to all its neighbors in the forw arding zone. The neighbor nodes in turn calculate their o wn forw arding zones to D and repeat the process. Mobile node D on recei ving the data pack et sends an ackno wledgement pack et (A CK) to S. The A CK pack et is send to S in the same manner as the data pack et w as send to D. If node S does not recei v e an A CK pack et it uses a reco v ery procedure. Flooding is a reco v ery procedure suggested in [25 ]. The forw arding zone concept in SELAR w as inuenced by DREAM. In SELAR, the destination D is al w ays the base station. Furthermore, SELAR increases its forw arding angle as and when needed so that at least one sensor node is present in the forw arding zone. A detailed description of the SELAR protocol is gi v en in Chapter 3. 16

PAGE 25

CHAPTER 3 THE SCALABLE ENERGY -EFFICIENT LOCA TION AIDED R OUTING (SELAR) PR O T OCOL The Scalable Ener gy-ef cient Location-Aided Routing (SELAR) protocol for wireless sensor netw orks is a protocol that assumes that a location mechanism e xists to pro vide the location to all nodes and therefore perform the routing function. Ev ery node can kno w its location using GPS (Global Positioning System) or some distrib uted localization protocol [26 ][27 ][28 ]. One of the main considerations during the design of SELAR w as to create a simple and scalable protocol with minimal computational o v erhead at the po wer -constrained sensor nodes and one that is capable of consuming the ener gy of the netw ork e v enly SELAR needs e v ery sensor node to kno w its o wn location as well as that of its neighbors and the base station. F or this, SELAR can use an y e xisting location mechanism. Once this is achie v ed, e v ery sensor node which needs to send data to the base station, selects from amongst its neighbors in the forw arding zone, the node with the maximum ener gy This process continues until the base station recei v es the data pack et. Ev ery sensor node has to be concerned only about forw arding the data pack et to its immediate neighbor in the direction of the base station. This leads to sensor nodes nearer to the base station dissipating their ener gy f aster Since sensor nodes that ha v e the base station nearer to them dissipate ener gy f aster this thesis proposes to mo v e the base station to better utilize the remaining ali v e nodes. The base station could be attached to a robot, thus enabling it to mo v e. The rest of this chapter discusses in detail the design considerations, w orking of the protocol and mo ving of the base station. 17

PAGE 26

3.1 Design considerations The main considerations during the design of the protocol were as follo ws: Simplicity : SELAR is designed to be a simple protocol which is easy to implement since it requires v ery lo w amount of computation at indi vidual nodes. SELAR attempts to optimize the amount of calculations to be done by the indi vidual sensor node. Each sensor node kno ws the ener gy and location information of itself as well as that of the nodes within its radio range. In addition, e v ery node kno ws the location information of the base station. In SELAR, the nodes within the radio range of a particular sensor node form its neighbors. Each sensor node selects the neighbor node to which it forw ards its pack et, based on the location information of the base station as well as the location and ener gy information of its neighbors. As such, nodes in SELAR need to maintain a v ery small routing table and mak e just a fe w computations to forw ard the pack ets. Scalability : Scalability is an important issue in wireless sensor netw orks. SELAR is a scalable routing protocol because the routing function does not depend on the number of nodes in the netw ork. Ener gy-ef ciency : SELAR is an ener gy-ef cient protocol. Ev ery sensor node forw ards data pack et to its most ener gy-rich neighbor at that particular point in time. Ev ery node periodically e xchanges its a v ailable ener gy with its neighbors. Thus sensor nodes forw ard data pack ets based on most recent ener gy information about its neighbors. This ensures that sensor nodes ha v e their ener gy drained uniformly Ho we v er it is e xpected that sensor nodes closer to the base station will be used more and therefore will die rst. Therefore, SELAR by itself does not guarantee a global e v en ener gy distrib ution. In order to solv e this problem, this thesis proposes to mo v e the base station. 18

PAGE 27

3.2 W orking of the pr otocol In SELAR, e v ery wireless sensor node has to kno w the follo wing: The ener gy and location information about itself. The ener gy and location information about all its neighbors. The location information of the base station. The location of e v ery wireless sensor node can be found, as suggested in [29 ], by using GPS (Global Positioning System) or some distrib uted localization protocol. The base station can broadcast its location information to all the wireless sensor nodes. This should be possible for the base station since it is a high po wer node with superior capabilities than the sensor nodes. The wireless sensor netw ork is considered to be relati v ely static. In SELAR sensor nodes transmit tw o types of pack ets control and data. Initially during the control pack et dissemination phase, e v ery sensor node broadcasts its location and ener gy information to all its neighbors. Ev ery sensor node maintains a table containing the location and ener gy of all its neighbors. This table is updated based on the information contained in the control pack ets recei v ed from its neighbors. In the simulations run in chapter four data packets are transmitted by each sensor node at re gular time interv als. The control pack et dissemination phase can be triggered at re gular interv als when signicant amount of ener gy has been dissipated in the netw ork. During the control pack et dissemination phase, e v ery sensor node broadcasts its ener gy information to all its neighbors. During the data pack et dissemination phase, e v ery sensor node forw ards one pack et each to the base station. Each sensor node forw ards its pack et to the neighbor with maximum ener gy in the direction of the base station. The subsections belo w discuss in detail the control pack et dissemination model, the data pack et dissemination model and the algorithm used to determine the neighbor nodes within a gi v en forw arding zone. 19

PAGE 28

3.2.1 Contr ol pack et dissemination phase SELAR be gins with a control pack et dissemination phase. During the control pack et dissemination phase, e v ery sensor node broadcasts the ener gy and location information about itself to all its neighbors. Future control pack et dissemination phases require sensor nodes to broadcast their a v ailable ener gy information alone. The control pack et dissemination model is graphically depicted by Figure 3.1, where sensor node S broadcasts control information to its neighbors N1, N2 and N3. All other sensor nodes sho wn in Figure 3.1, being outside the radio range of node S, do not recei v e the control pack et broadcasted by node S. Similar to node S, all other sensor nodes broadcast their ener gy information to their respecti v e neighbors. The main purpose of the control pack et dissemination phase is to increase the producti v e lifetime of the wireless sensor netw ork. Each sensor node by kno wing the ener gy and location information of its neighboring sensor nodes, can forw ard pack ets to the base station, by making use of sensor nodes with maximum amount of ener gy Alternati v ely each sensor node decide when to send control pack ets to neighbors. F or e xample, each time a sensor node' s ener gy le v el decreases by a certain amount, it can broadcast a control pack et with its current ener gy le v el to all its neighbors. Each sensor node, on recei ving a control pack et from its neighbor node, updates the corresponding entry in its state table. The format of the state table maintained at each node is sho wn in Figure 3.2. The state table at each sensor node has the follo wing elds: Neighbor ID : This eld contains the identication for the corresponding neighbor node. This identication can either be one that is global in nature or generated locally Location : This eld contains the location information for the corresponding node. The location information can be absolute or relati v e. Ener gy : This eld contains the a v ailable ener gy of the corresponding node. This eld is updated whene v er a control pack et is recei v ed from the respecti v e neighbor Sensor nodes can be pro vided with the hardw are and softw are capability to kno w its remaining ener gy T imestamp : This eld contains the time at which the last control pack et w as recei v ed from the corresponding node. If a certain amount of time has passed since the reception of control 20

PAGE 29

Figure 3.1 Control pack et dissemination Figure 3.2 F ormat of the state table maintained at e v ery sensor node pack et from a particular neighbor node, then the sensor node sets the ener gy eld in the state table corresponding to that neighbor node to 0. This is done so as to a v oid sending data pack ets to dead sensor nodes. The ne xt time a control pack et is recei v ed from that particular sensor node, the state table is updated accordingly 3.2.2 Data pack et dissemination phase The data pack et dissemination model in SELAR is v ery much similar to that used in [25 ]. Whene v er a sensor node needs to send a data pack et to the base station, it considers all the sensor nodes in the forw arding zone. F orw arding zone for a particular sensor node is the sector formed within the radio range of that sensor node, in the direction of the base station. The sensor node then forw ards the data pack et to the maximum ener gy sensor node in the forw arding zone. This process continues until a sensor node which has the base station in its radio range recei v es the forw arded pack et. This sensor node then sends the data pack et directly to the base station. 21

PAGE 30

Figure 3.3 Data pack et dissemination A fe w important aspects of the data pack et dissemination phase in SELAR are discussed belo w: Ev ery sensor node maintains a state table. The format of the state table is sho wn in Figure 3.2. The state table contains the location and ener gy information of all the neighbors of the particular sensor node. Ev ery node mak es its data pack et forw arding decision based on the information contained in the state stable. Initially the forw arding zone angle is set to 15 The forw arding zone angle is graphically depicted in 3.3. The angle subtended by the forw arding zone at the corresponding node will al w ays be twice that of (hence, initially 30 ). If a sensor node does not nd an y ali v e neighbor node in its forw arding zone, it increases by steps of 15 until at least one ali v e sensor node is present in its forw arding zone. The maximum v alue which can be tak en by is 180 At this point, the forw arding zone of the particular sensor node is its whole radio range. If the sensor node is not able to nd ali v e neighbor nodes with set to 180 it broadcasts the data pack et within its radio range. This is done as a last resort, taking into account the possibility that a fe w neighbor nodes could still be ali v e b ut are not able to successfully send control pack ets. Alternati v ely the sensor node can stop sending data pack ets when it is not able to nd an ymore ali v e neighbor nodes. In some applications, more sensor nodes are 22

PAGE 31

Figure 3.4 Determining whether a neighbor node is within the forw arding zone dropped into the nearly dead netw ork. When this happens, the nodes can resume sending data pack ets. The data pack et dissemination model is depicted graphically in Figure 3.3 where the diameter of the nodes represent the amount of ener gy the y contain. Sensor node S has tw o neighbor nodes in its forw arding zone, nodes N2 and N4. Sensor node S forw ards its data pack et to node N4 since it has more ener gy than N2. In Figure 3.3 the shaded re gion within the sector is the forw arding zone of sensor node S. During the data pack et dissemination phase e v ery sensor node sends a single data pack et to the base station. Similar to control pack et dissemination, during data pack et dissemination a random f actor is introduced to the transmission start time of the pack ets so that all sensor nodes do not transmit at the same time. 3.2.3 Algorithm f or determining neighbor nodes within the f orwarding zone The algorithm for determining the neighbor nodes within the gi v en forw arding zone of a sensor node has been borro wed from the DREAM ns-2 code recei v ed from the authors of [30 ]. The algorithm is e xplained with the help of Figure 3.4. Consider the situation where sensor node S needs to determine whether its neighbor node N f alls within the forw arding zone. This can be determined by performing the follo wing: 23

PAGE 32

The angle ( 1 ) which the line connecting node S (x0, y0) and node N (x1, y1) mak es with respect to the x-coordinate is determined by atan( y 1 y 0 x 1 x 0 ). The angle ( 2 ) which the line connecting node S (x0, y0) and base station (x2, y2) mak es with respect to the x-coordinate is determined by atan( y 2 y 0 x 2 x 0 ). The angle which the base station and node N mak es at node S can be found out as: = 1 2 If is less than or equal to the forw arding angle then node N lies in the forw arding zone of node S else node N lies outside the forw arding zone of node S. 3.3 Mo ving the base station In this thesis, the follo wing method to deal with the ener gy depletion problem of the nodes closer to the base station is proposed. In order to increase the lifetime and producti vity of the wireless sensor netw ork, the base station is mo v ed around the area. In SELAR, after a certain amount of time, it is found that the nodes f arther a w ay from the base station ha v e more ener gy than those nearby T o tak e care of this, the base station is mo v ed to an appropriate point, such that the nodes which ha v e the base station in their radio range ha v e suf cient amount of ener gy The base station can be mo v ed by attaching it to a robot which can be teleoperated remotely An implementation of mo ving the base station is gi v en in Section 4.3.1. Consider monitoring a habitat spread o v er a 400 m 400 m area. W e can predetermine points to which it is feasible to send the mo v able base station. W e can mo v e the base station to these points as and when required. Mo ving the base station is proposed in [31 ] too. 24

PAGE 33

CHAPTER 4 SIMULA TION AND RESUL TS This chapter describes the implementation of SELAR in the netw ork simulator 2 (ns-2) [32 ]. SELAR is compared with tw o other protocols Lo w-Ener gy Adapti v e Clustering Hierarchy (LEA CH) [10 ] and Minimum T ransmission Ener gy (MTE) [11 ] [12 ]. Simulations were performed using the ns-2 code e xtensions obtained from MIT uAMPS project [5 ] which in turn is b uilt on the CMU W ireless and Mobility platform [33 ] included in ns-2.1b5. 4.1 CMU wir eless and mobility platf orm The ns-2.1b5 release has the CMU W ireless and Mobility platform included with it. The follo wing are the CMU additions to the baseline simulator: Mobility Model. MA C protocols. Channel propagation models [34 ]. Figure 4.1 sho ws the implementation of a mobile node. Data pack ets are created by the Application and then send to the Agent. The netw ork and transport layer functions are performed by the Agent. The Agent then sends pack ets to CMUT race which writes statistics about the pack ets to trace les. The Connector then recei v es these pack ets and sends them to the Link-Layer for data-link processing. After a small delay the Queue recei v es the pack ets from the Link-Layer The pack ets are queued at the Queue if there are pack ets w aiting to be transmitted. The MA C runs media access protocols on the pack ets once it recei v es them from the Queue. The MA C sends the pack ets to the Netw ork Interf ace, where the pack ets are sent through the Channel after adding the correct 25

PAGE 34

Figure 4.1 Block diagram of an ns-2 mobile node [5 ] transmit po wer A cop y of the data pack et is sent to each node connected to the Channel. Each node' s Netw ork Interf ace, on recei ving the data pack et, sends them up through the same functions b ut in the re v erse model. The Agent recei v es the data and sends a notication to the Application. 4.2 MIT uAMPS ns-2 code extensions The ns-2 code e xtensions from MIT uAMPS project add support for lar ge-scale wireless sensor netw orks. MIT uAMPS ns-2 code e xtensions were done in ns-2.1b5 release atop the CMU W ireless and Mobility model. 4.2.1 Resour ce-adapti v e node The MIT uAMPS e xtensions to ns-2 added a Resource-Adapti v e node [13 ]. The block diagram of the Resource-Adapti v e node is sho wn in Figure 4.2. Resource-Adapti v e nodes help in implementing Resource-Adapti v e protocols. The tw o ne w features of Resource-Adapti v e nodes are: 26

PAGE 35

Figure 4.2 Block diagram of a resource-adapti v e node [5 ] Resour ce Mana g er : Pro vides a common interf ace between Application and the indi vidual resources. Resour ces : An ything that needs to be monitored (ener gy node neighbors). The follo wing functions are used by Application to update the status of the node' s resources through the Resource Manager: add : T o add more of a resource to the node' s supply r emo ve : T o remo v e the specied amount of resource from the node' s supply query : T o enquire about the amount of resource the node currently has. These functions are used in the simulations to assign, decrement and adv ertise ener gy v alues of the nodes. 27

PAGE 36

Figure 4.3 Radio ener gy dissipation model [1 ] T able 4.1 P arameters used for radio model [1 ] P arameter V alue Electronics ener gy ( E el ec ) 50 nJ/bit Recei v er po wer threshold ( P r thr esh ) 6 nW Radio ener gy parameter for Friss model ( f r iss amp ) 100 pJ/bit/m 2 Radio ener gy parameter for T w o-ray model ( tw o r ay amp ) 0.013 pJ/bit/m 4 Radio bitrate ( R b ) 1 Mbps 4.2.2 Radio model The MIT uAMPS ns-2 code e xtension assume a simple radio ener gy model which is sho wn in Figure 4.3. The transmitter dissipates ener gy to run the radio electronics as well as the amplier whereas the recei v er dissipates ener gy to run the radio electronics only T able 4.1 lists the important parameters used by the radio model. These are typical v alues used in se v eral in v estigations. Using the v alues in 4.1 and gi v en a pack et of size k bits, the ener gy consumed in the entire transmission and reception processes can be calculated and therefore adjusted. F or more information about the radio ener gy model, please refer to Section 4.1.2 in [1 ]. The radio parameters in MIT uAMPS ns-2 code e xtensions are based on that of the a v ailable transcei v er baseband chips [35 ]. 4.2.3 Netw ork interface The physical layer functions are performed by the Netw ork Interf ace. When the Netw ork Inter f ace is ready to transmit a pack et, it remo v es the appropriate amount of ener gy to send the pack et, based on the distance to the recei v er Once a node has used up all its ener gy it is remo v ed from the channel. P ack ets sent to dead nodes, nodes that are remo v ed from the channel, are thro wn a w ay 28

PAGE 37

A node can either be in the sleep state or a w ak e state. If the node is in the sleep state while its Netw ork Interf ace recei v es a pack et, it drops the pack et. Else, if the node is a w ak e, the Netw ork Interf ace calculates the recei v ed po wer of the pack et. The follo wing three scenarios could happen: If the recei v ed po wer of the pack et is belo w a certain detection threshold, the pack et is dropped. If the recei v ed po wer of the pack et at the node is abo v e the detection threshold b ut belo w the successful reception threshold, the pack et is mark ed as erroneous and passed up the stack. If the recei v ed po wer of the pack et is abo v e the successful reception threshold, the pack et is considered to ha v e been recei v ed successfully and is passed up the stack to the MA C layer Depending on the distance between the transmitter and the recei v er the free space model or multipath f ading model is used, as dened by the channel propagation model in ns-2 [34 ] [36 ]. The Friss free space model is used if the distance between the transmitter and recei v er is less than the cross-o v er distance. Cross-o v er distance is the maximum distance upto which a pack et can be transmitted successfully using the Friss free space model. The cross-o v er ( d cr ossov er ) distance is calculated [1 ] using d cr ossov er = 4 p L H T H R (4.1) where L is the system loss f actor not related to propagation, H T and H R are the heights of the transmitter and recei v er antennae respecti v ely The Friss equation to estimate the recei v ed signal po wer [1] is gi v en by P R = G T G R P T (4 ) 2 d 2 (4.2) where, is the w a v elength, P R is the recei v ed signal po wer in W atts (or dBm), G T is the transmitter gain, G R is the gain at the recei v er P T is the transmitted signal po wer in W atts (or dBm) and d is the distance between the recei v er and transmitter antenna measured in meters. F or a tw o-ray ground reection model [36 ], the recei v ed signal po wer [1 ] is in v ersely propartional to d 4 and is estimated using P R = G T G R P T ( H T 2 H R 2 ) d 4 L : (4.3) 29

PAGE 38

F or the e xperiments, an omnidirectional antenna ha ving the follo wing parameters is used: G T = G R = 1, H T = H R = 1.5 m, L = 1 (no system loss), 914 MHz radio, = 3 10 8 914 10 6 = 0.328 m. From these v alues the cross-o v er distance is calculated as 86.2 m. These v alues ha v e been tak en from [5 ]. 4.2.4 MA C implementation The ns-2 code e xtensions of MIT uAMPS project has a ne w protocol called MacSensor The MacSensor protocol consists of the follo wing: CSMA (Carrier -Sense Multiple Access) : Implemented in the MacSensor class. The CSMA implemented is non-persistent. In non-persistent CSMA a node on detecting the channel to be b usy sets a random time interv al and tries to transmit again after that time interv al. TDMA (T ime-Division Multiple Access) : This is implemented within the Application, by setting the Application to send data to the Agent during the specied TDMA time-slot. This implementation assumes the clocks of all nodes to be synchronized. DS-SS (Dir ect-Sequence Spr ead Spectrum) : A simple model of DS-SS is implemented jointly within the Application and the MacSensor class. A node running the MacSensor protocol cannot recei v e a pack et while transmitting. If a node recei v es a pack et while it is transmitting, it marks it as erroneous and passes it up to the link layer where the erroneous pack et is dropped. If a node running the MacSensor protocol recei v es a pack et while it is already recei ving another pack et, the follo wing tw o scenarios can occur: If the pack et which is already being recei v ed, has more than a certain amount of signal strength than the ne w pack et which has arri v ed, the ne w pack et is dropped. Otherwise if the signal strength of the rst pack et being recei v ed is not signicantly more than that of the ne w pack et which has arri v ed, both pack ets collide and are dropped. 30

PAGE 39

4.2.5 Base station application The base station is a v ery po werful node and has virtually no ener gy constraints. Ev ery data pack et generated has the base station as its ultimate destination. The base station application needs to k eep track of e v ery data pack et it recei v es. The base station application helps us in determining the follo wing: Estimate of the latenc y of dif ferent protocols. Quality information about the dif ferent protocols. 4.2.6 Implementation of LEA CH and MTE The implementation of LEA CH [10 ] is done according to the description in Chapter 3 of [1 ]. LEA CH is implemented as a subclass of ns-2' s Application class. MTE routing is implemented as described in [10 ] and [1 ]. In MTE routing, each node chooses the closest node that is in the direction of the base station as its ne xt-hop neighbor Initially all nodes e xpend a certain amount of ener gy to nd their ne xt-hop neighbors. A node N' s upstream neighbors are the set of nodes which ha v e N as their ne xt-hop neighbor Consider node N1 is node N' s ne xt-hop neighbor Whene v er node N dies, its upstream neighbors become part of node N1' s upstream neighbors. The follo wing action tak es place in MTE routing: Each node' s transmit po wer is set to the minimum required to reach its ne xt-hop neighbor MTE routing uses non-persistent CSMA MA C protocol. Ev ery node forw ards data pack ets to the ne xt-hop neighbor until the base station recei v es it. T e data pack et transmission rate in MTE is similar to that for SELAR (4.3). F or LEA CH the data pack et transmission rates are much higher The data pack et transmission rate for LEA CH can be obtained from [5 ]. 31

PAGE 40

4.3 Implementing SELAR using the MIT uAMPS ns-2 code extensions SELAR has been implemented in ns-2.1b5 using MIT uAMPS ns-2 code e xtensions, as described in Chapter 3. Initially it is assumed that all nodes kno w their o wn locations as well as that of the base station. The protocol starts with a control pack et dissemination phase during which all sensor nodes broadcast their location and ener gy information to their neighbor nodes. Ev ery control pack et dissemination phase is follo wed by tw o data pack et dissemination phases. The algorithm to nd the neighbor nodes within the forw arding angle w as obtained from the ns-2 code for DREAM gi v en by the authors of [30 ]. SELAR sends data pack ets continuously The approximate data pack et transmission rates used by SELAR are 25 pack ets e v ery 0.6 s for the 200 m 200 m scenario, 50 pack ets e v ery 1.7 s for the 283 m 283 m scenario, 75 pack ets e v ery 3.1 s for the 347 m 347 m scenario, 100 pack ets e v ery 4.8 s for the 400 m 400 m scenario. 4.3.1 SELAR with mo ving base station As mentioned in Section 3.3, SELAR with mo ving base station is implemented. Figure 4.4 graphically represents the implementation of SELAR with mo ving base station. Initially the base station is placed at position 1. When the base station stops recei ving data pack ets, it mo v es to position 2 and broadcasts to all sensor nodes about its ne w position. The sensor nodes change the information in their respecti v e state tables accordingly and resets their forw arding angle to 15 in the direction of position 2. Again, when the base station stops recei ving data pack ets at position 2, it mo v es to position 3 and the same sequence of steps as abo v e are follo wed. Finally when the base station stops recei ving pack ets at position 3, it mo v es to position 4 and stays there till the end of the simulation. Cartesian reference system is used in the simulations. Positions 1, 2, 3 and 4 are midpoints along the four edges of the sensor netw ork topology 4.4 Restriction applied to radio range The maximum radio range of sensor nodes used in simulating SELAR is restricted to the crosso v er distance (refer Section 4.2.3). F or LEA CH and MTE, as implemented by the MIT uAMPS ns-2 32

PAGE 41

Figure 4.4 Mo ving the base station in SELAR code e xtensions, no po wer restriction has been placed. This results in them being able to transmit to as much distance as the y w ant, as long as the y ha v e the ener gy to perform the necessary po wer amplication. This unfortunately is not possible in reality Most of the lo w cost sensor nodes w ork at radio ranges of fe w tens of meters indoors and nearly and order of magnitude higher outdoors [37 ] [38 ]. LEA CH and MTE were run the w ay the y were obtained from [5 ] i.e., without an y maximum radio range restrictions. Then, the maximum range of sensor nodes running LEA CH and MTE w as restricted to the cross-o v er distance and simulations were rerun. The results obtained are described and discussed in the sections which follo w 4.5 Simulation scenario Fi v e dif ferent sensor node topologies each, for four dif ferent scenarios were created and LEA CH, MTE, SELAR and their v ariations were run on these topologies. The topologies were created randomly A random function which uses uniform distrib ution w as used to place nodes within each 33

PAGE 42

topology F or e xample, if 100 nodes are to be placed in a 400 m 400 m area, tw o random numbers are generated using uniform distrib ution such that the numbers lie between 0 and 400. The tw o numbers generated form the (x,y) coordinate of the sensor node. This is done 100 times to generate the 100 node positions. The scenarios were created in such a w ay that the node density is same for all. The performance of the v arious protocols with increasing number of sensor nodes had to be measured. The simulation tool, ns-2, restricts the maximum number of sensor nodes which can be simulated to 128. Based on these considerations, the four scenarios created are as follo ws: 200 m 200 m scenario with 25 nodes. 283 m 283 m scenario with 50 nodes. 347 m 347 m scenario with 75 nodes. 400 m 400 m scenario with 100 nodes. The v arious protocols and their v ariations are run on the v arious topologies of the four scenarios are as follo ws: MTE. MTE with maximum radio range restricted to cross-o v er distance (86.2 m). LEA CH. LEA CH with maximum radio range restricted to cross-o v er distance (86.2 m). SELAR with stationary base station. SELAR with mo ving base station. 4.6 Results This section discusses and analyzes the performance of SELAR routing protocol compared to that of LEA CH and MTE. 34

PAGE 43

4.6.1 Data pack ets The a v erage number of data pack ets recei v ed after running MTE, LEA CH, SELAR with stationary base station (SST) and SELAR with mo ving base station (SMV) are sho wn in T able 4.2 and Figure 4.5. Figure 4.5 sho ws the 95% Condence Interv al for the mean v alues found using the t-distrib ution with 4 de grees of freedom. From T able 4.2 and Figure 4.5, it is observ ed that LEA CH deli v ers the highest number of pack ets follo wed by MTE. SELAR with mo ving base station (SMV) is seen to deli v er more data pack ets as the size of the netw ork increases. SELAR with stationary base station is seen to deli v er the least amount of data pack ets. This is because both the SELAR protocols ha v e their maximum radio range restricted while MTE and LEA CH operate under no such restrictions (which is not practical for lar ge scale wireless sensor netw orks). Again the huge difference between the number of data pack ets deli v ered by LEA CH compared to other protocols is because LEA CH uses data fusion. The a v erage number of data pack ets recei v ed after running MTE and LEA CH with restricted radio range, SELAR with stationary base station (SST) and SELAR with mo ving base station (SMV) are sho wn in T able 4.3 and Figure 4.6. In these set of simulations all the protocols ha v e their maximum radio range set to cross-o v er distance (86.2 m). Figure 4.6 sho ws the 95% Condence Interv al for the mean v alues found using t-distrib ution with 4 de grees of freedom. From T able 4.3 and Figure 4.6, it is observ ed that SELAR with mo ving base station (SMV) deli v ers the highest number of pack ets follo wed by SELAR with stationary base station (SST). MTE and LEA CH are sho wn to deli v er signicantly less number of data pack ets, when their data ranges are restricted lik e that for the SELAR protocols. F or MTE this happens because as each node dies, the distance between the corresponding upstream neighbors and ne xt-hop neighbor increases, e v entually leading them to be out of each others range. F or LEA CH to w ork ef ciently e v ery node should ha v e e v ery other node within its radio range. This is so that during the set-up phase in LEA CH, e v ery cluster -head can broadcast adv ertisement pack ets to all the sensor nodes in the wireless sensor netw ork. W ith restricted radio range this is not possible, which results in some nodes forming part of non-optimal clusters and some other nodes not participating in data pack et deli v ery Also, in LEA CH, cluster heads transmit directly to the base station, so if the y are out of range the y cannot deli v er data. It is 35

PAGE 44

T able 4.2 A v erage number of data pack ets recei v ed at the base station Protocol Scenarios 200 200 283 283 347 347 400 400 MTE 7889 8981.4 9164.6 8371.4 95% CI (6979.73 to (8198.29 to (8523.75 to (7937.01 to for MTE 8798.27) 9764.51) 9805.45) 8805.79) LEA CH 8172.8 14392.2 17723.8 18865.2 95% CI (5727.49 to (10956.74 to (13823.16 to (12257.93 to for LEA CH 10618.11) 17827.66) 21624.44) 25472.47) SST 7044.6 7226.2 6494.6 6228.8 95% CI (6245.18 to (6090.97 to (5643.41 to (5453.73 to for SST 7844.02) 8361.43) 7345.79) 7003.87) SMV 7650.4 8335.6 8491.8 9524.4 95% CI (7032.49 to (7496.33 to (7395.97 to (7835.15 to for SMV 8268.31) 9174.87) 9587.63) 11213.65) 200x200 283x283 347x347 400x400 0 0.5 1 1.5 2 2.5 x 10 4 scenario (m x m) # of data packets delivered MTE LEA SST SMV Figure 4.5 A v erage number of data pack ets recei v ed at the base station observ ed that LEA CH deli v ers the least amount of pack ets and the number of pack ets it deli v ers to the base station decreases with increase in scenario size. This e xperiment pro v es that the LEA CH and MTE protocols are not scalable. On the other hand, SELAR is scalable b ut suf fers from the 36

PAGE 45

T able 4.3 A v erage number of data pack ets recei v ed at the base station after restricting the maximum radio range to cross-o v er distance (86.2 m) Routing Protocol Scenarios 200 200 283 283 347 347 400 400 MTE 4975.4 5920.2 5067 4997 95% CI (4474.57 to (4475.82 to (4114.7 to (4225.01 to for MTE 5476.23) 7364.58) 6019.3) 5768.99) LEA CH 1915 951.2 809.8 493 95% CI (1035.07 to (465.76 to (622.98 to (301.34 to for LEA CH 2794.93) 1436.64) 966.62) 684.66) SST 7044.6 7226.2 6494.6 6228.8 95% CI (6245.18 to (6090.97 to (5643.41 to (5453.73 to for SST 7844.02) 8361.43) 7345.79) 7003.87) SMV 7650.4 8335.6 8491.8 9524.4 95%CI (7032.49 to (7496.33 to (7395.97 to (7835.15 to SMV 8268.31) 9174.87) 9587.63) 11213.65) 200x200 283x283 347x347 400x400 0 0.5 1 1.5 2 2.5 x 10 4 scenario (m x m) # of data packets delivered MTE LEA SST SMV Figure 4.6 A v erage number of data pack ets recei v ed after restricting radio range to cross-o v er distance (86.2 m) problem of ener gy depletion of nodes close to the base station. Ho we v er it w as sho wn that the strate gy of mo ving the sink around addresses this problem. 37

PAGE 46

4.6.2 Netw ork lifetime F or our simulations, netw ork lifetime is dened as the time at which the last data pack et is recei v ed at the base station. The a v erage netw ork lifetime after running MTE, LEA CH, SELAR with stationary base station (SST) and SELAR with mo ving base station (SMV) are sho wn in T able 4.4 and Figure 4.7. Figure 4.7 sho ws the 95% Condence Interv al for the mean v alues in T able 4.4 found using t-distrib ution with 4 de grees of freedom. The a v erage netw ork lifetime after running MTE with maximum radio range restricted to cross-o v er distance (86.2 m), LEA CH with maximum radio range restricted to cross-o v er distance (86.2 m), SELAR with stationary base station (SST) and SELAR with mo ving base station (SMV) are sho wn in T able 4.5 and Figure 4.8. Figure 4.8 sho ws the 95% Condence Interv al for the mean v alues in T able 4.5 found using t-distrib ution with 4 de grees of freedom. The netw ork lifetime for sensor nodes has been dened as the time at which the base station recei v es the last data pack et from an y sensor node. From Figures 4.7 and 4.8, it is observ ed that all protocols ha v e some what similar lifetimes with restricted as well as unrestricted radio ranges. The SELAR protocols, in all cases, ha v e their maximum radio ranges restricted to the cross-o v er distance (86.2 m). The lifetime of sensor netw ork running MTE protocol drops when the maximum radio range is restricted. This can be seen from T ables 4.4 and 4.5. But, in both cases the netw ork lifetime for MTE protocol is greater than that for SELAR with stationary base station (SST). This can be e xplained based on the w orking of both protocols and the denition of netw ork lifetime. In SELAR, e v ery node forw ards data pack ets to its neighbor in the direction of the base station with maximum ener gy while in MTE the ne xt hop neighbor is x ed. Because of x ed ne xt hop neighbors in MTE, certain nodes ha v e fe wer data to forw ard to the base station than others and hence, li v e longer This can happen when the position of the node is such that initially all its neighbors ha v e better options than choosing the sensor node. This results in the base station recei ving pack ets for longer amount of time (though fe w data pack ets are recei v ed) from these sensor nodes with lesser load, and this in turn increases the netw ork lifetime. On the other hand, the lifetime of LEA CH protocol is much lesser than other protocols because it sends data at a f aster rate. 38

PAGE 47

T able 4.4 A v erage netw ork lifetime in simulation seconds Routing Protocol Scenarios 200 200 283 283 347 347 400 400 MTE 839.26 1414.04 2021.58 2087.14 95% CI (563.07 to (937.19 to (1414.66 to (1452.88 to for MTE 1115.45) 1890.89) 2628.5) 2721.4) LEA CH 150.34 218.28 288.06 351.28 95% CI (117.91 to (183.49 to (252.29 to (277.81 to for LEA CH 182.77) 253.07) 323.83) 424.75) SST 601.43 514 476 566 95% CI (222.67 to (151.88 to (223.61 to (98.13 to for SST 980.19) 876.12) 728.39) 1033.87) SMV 809.57 1343.83 2469.05 7151.31 95% CI (557.16 to (795.12 to (1469.41 to (2418.37 to for SMV 1061.98) 1892.54) 3468.69) 11884.25) 200x200 283x283 347x347 400x400 0 5000 10000 15000 scenario (m x m) # of simulation seconds MTE LEA SST SMV Figure 4.7 A v erage lifetime of the netw ork in simulation seconds 39

PAGE 48

T able 4.5 A v erage netw ork lifetime in simulation seconds after restricting the maximum radio range to cross-o v er distance (86.2 m) Routing Protocol Scenarios 200 200 283 283 347 347 400 400 MTE 494 1280 1542 1404 95% CI (231.14 to (534.55 to (285.59 to (408.13 to for MTE 756.86) 2025.45) 2798.41) 2399.87) LEA CH 153.2 188.78 170 265.56 95% CI (118.95 to (120.63 to (121.10 to (138.99 to for LEA CH 187.45) 256.93) 218.90) 392.13) SST 601.43 514 476 566 95% CI (222.67 to (151.88 to (223.61 to (98.13 to for SST 980.19) 876.12) 728.39) 1033.87) SMV 809.57 1343.83 2469.05 7151.31 95% CI (557.16 to (795.12 to (1469.41 to (2418.37 to for SMV 1061.98) 1892.54) 3468.69) 11884.25) 200x200 283x283 347x347 400x400 0 5000 10000 15000 scenario (m x m) }# of simulation seconds MTE LEA SST SMV Figure 4.8 A v erage lifetime of the netw ork in simulation seconds after restricting the maximum radio range to cross-o v er distance (86.2 m) 40

PAGE 49

4.6.3 Ener gy distrib ution f or MTE Figures 4.9, 4.10 and 4.11 sho w the ener gy distrib ution of sensor nodes running MTE with around 90, 50 and 10 sensor nodes ali v e respecti v ely using the 400 400 scenario with 100 nodes. Again, the diameter of the nodes represent their amount of ener gy From the gures the follo wing can be observ ed about MTE: Sensor nodes which are on the path of other sensor nodes w ay to the base station die f aster than the rest of the nodes. Node ener gy dissipation is une v en in such a w ay that certain parts of the sensor netw ork die f aster than others. Figures 4.12, 4.13 and 4.14 sho w the ener gy distrib ution of sensor nodes running MTE (with maximum radio range restricted to cross-o v er distance, 86.2 m) with around 90, 50 and 10 sensor nodes ali v e respecti v ely The dissipation of node ener gy is similar to that by nodes with no radio range restrictions. One dif ference being that the nodes f ar a w ay from the base station dissipate ener gy at a slo wer rate when compared to the ener gy dissipation of the same nodes running MTE with no radio range restrictions. This can be observ ed comparing 4.10 and 4.13. 41

PAGE 50

0 100 200 300 400 0 50 100 150 200 250 300 350 400 O x-axis (m)y-axis (m) Figure 4.9 Ener gy distrib ution for MTE with around 90 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.10 Ener gy distrib ution for MTE with around 50 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.11 Ener gy distrib ution for MTE with around 10 nodes ali v e 42

PAGE 51

0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.12 Ener gy distrib ution for MTE (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 90 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.13 Ener gy distrib ution for MTE (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 50 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.14 Ener gy distrib ution for MTE (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 10 nodes ali v e 43

PAGE 52

4.6.4 Ener gy distrib ution f or LEA CH Figures 4.15, 4.16 and 4.17 sho w the ener gy distrib ution of sensor nodes running LEA CH with around 90, 50 and 10 sensor nodes ali v e, respecti v ely From the gures the follo wing can be observ ed about LEA CH: Sensor nodes dissipate node ener gy randomly When around 50 nodes are dead, most of the dead nodes are found to be the ones f ar a w ay from the base station. This happens because the nodes f arther a w ay from the base station has to spend signicantly more ener gy on becoming clusterheads than that spend by the nodes nearer to the base station on becoming clusterheads. Hence, nodes f arther a w ay from the base station dissipate their ener gy f aster than the nodes nearer to the base station. Figures 4.18, 4.19 and 4.20 sho w the ener gy distrib ution of sensor nodes running LEA CH (with maximum radio range restricted to cross-o v er distance, 86.2 m) with around 90, 50 and 10 sensor nodes ali v e respecti v ely The follo wing can be observ ed about LEA CH with restricted radio range: Sensor nodes to w ard the center of the netw ork topology dissipate ener gy f aster than other nodes Sensor nodes near the edges and especially the ones at the corners seem to ha v e more ener gy than other nodes. This could be because the y do not recei v e adv ertisement pack ets e v ery set-up phase and subsequently do not send data pack ets during those rounds. 44

PAGE 53

0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.15 Ener gy distrib ution for LEA CH with around 90 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.16 Ener gy distrib ution for LEA CH with around 50 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.17 Ener gy distrib ution for LEA CH with around 10 nodes ali v e 45

PAGE 54

0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.18 Ener gy distrib ution for LEA CH (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 90 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.19 Ener gy distrib ution for LEA CH (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 50 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.20 Ener gy distrib ution for LEA CH (maximum radio range restricted to cross-o v er distance, 86.2 m) with around 10 nodes ali v e 46

PAGE 55

4.6.5 Ener gy distrib ution f or SELAR with stationary base station Figures 4.21, 4.22 and 4.23 sho w the ener gy distrib ution of sensor nodes running SELAR (stationary base station) with around 90, 50 and 10 sensor nodes ali v e respecti v ely From the gures the follo wing can be observ ed about SELAR with stationary base station: Ener gy dissipation follo ws a pattern. As nodes are f arther a w ay from the base station, the y consume less ener gy This is e xpected because nodes closer to the base station will route pack ets on behalf of f arther nodes. Sensor nodes with similar distance to the base station dissipate ener gy similarly This pro v es that pack et forw arding in SELAR is ener gy a w are. 47

PAGE 56

0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.21 Ener gy distrib ution for SELAR (stationary base station) with around 90 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.22 Ener gy distrib ution for SELAR (stationary base station) with around 50 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.23 Ener gy distrib ution for SELAR (stationary base station) with around 10 nodes ali v e 48

PAGE 57

4.6.6 Ener gy distrib ution f or SELAR with mo ving base station Figures 4.24, 4.25 and 4.26 sho w the ener gy distrib ution of sensor nodes running SELAR (mo ving base station) with around 90, 50 and 10 sensor nodes ali v e respecti v ely From the gures the follo wing can be observ ed about SELAR with mo ving base station: Sensor nodes with similar distance to the base station dissipate ener gy similarly The ali v e sensor nodes which were pre viously isolated are made use of. The producti vity of the sensor netw ork has been increased. This can be e xplained by taking into account the f act that more data pack ets reach the base station as well as that the netw ork lifetime of the sensor netw ork increases by mo ving the base station By mo ving the base station at the right time to the right spot, it is possible to uniformly dissipate ener gy of the sensor nodes. 49

PAGE 58

0 100 200 300 400 0 50 100 150 200 250 300 350 400 O x-axis (m)y-axis (m) Figure 4.24 Ener gy distrib ution for SELAR (mo ving base station) with around 90 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.25 Ener gy distrib ution for SELAR (mo ving base station) with around 50 nodes ali v e 0 100 200 300 400 0 50 100 150 200 250 300 350 400 O xaxis (m)yaxis (m) Figure 4.26 Ener gy distrib ution for SELAR (mo ving base station) with around 10 nodes ali v e 50

PAGE 59

By considering all the abo v e results it can be seen that SELAR with a mo ving base station is the best suited protocol for wireless sensor netw orks which ha v e a lar ge number of sensor nodes with limited ener gy and po wer By mo ving the base station in SELAR at the appropriate time to the right position, it is ensured that the ener gy dissipation within the netw ork is uniform. Ev en with a stationary base station, we see that SELAR performs better than MTE and LEA CH, as it deli v ers more data pack ets than both the protocols. LEA CH with restricted radio range deli v ers v ery fe w number of data pack ets because man y of the cluster heads chosen in a particular round do not ha v e the base station in their radio range. By performing the routing functions locally SELAR ensures that it is f ault-tolerant and simple. Again, e v ery sensor node tak es into account the ener gy left in the neighbor node, before forw arding data pack ets. This ensures that po wer is dissipated in a uniform manner SELAR is implemented assuming that the maximum radio range of each sensor node does not co v er the entire sensor netw ork. The LEA CH and MTE protocols obtained from the MIT uAMPS project do not place an y restriction on the maximum radio range of the sensor node. LEA CH when run without an y restrictions deli v er upto 3 times more data pack ets to the base station than SELAR. But this gain is hypothetical, since in reality the radio range of wireless sensor netw orks is restricted [37 ] [38 ]. The results indicate that in realistic senarios, SELAR deli v ers upto 12 times more and upto 1.4 times more data pack ets to the base station than LEA CH and MTE respecti v ely It w as also seen from the results that for realistic scenarios, SELAR with mo ving base station has upto 5 times and upto 27 times more lifetime duration compared to MTE and LEA CH respecti v ely W ith radio restrictions applied on LEA CH and MTE, it is seen that SELAR performs signicantly better than both of them. Further by mo ving the base station during the operation of the sensor netw ork, the performance of the wireless sensor netw ork is seen to impro v e e v en more. 51

PAGE 60

CHAPTER 5 CONCLUSION AND FUTURE W ORK W ireless sensor netw orks consist of hundreds or e v en thousands of po wer ener gy contrained sensor nodes. The y ha v e a myriad of possible applications. W ireless sensor netw ork can be used in habitat monitoring, surv eillance in b uildings, measuring pressure, temperature in hazardous areas, in military applications and so on. Routing in wireless sensor netw orks is important. In lar ge scale wireless sensor netw orks designing an appropriate routing protocol can be challenging due to the contraints in po wer ener gy and computational capabilities for indi vidual sensor nodes. The main design considerations for routing protocols in lar ge-scale wireless sensor netw orks are: fault toler ance scalability pr oduction costs power/ener gy constr aints SELAR has been designed taking these design considerations into account. Using simulations, SELAR has been e v aluated and compared with LEA CH (Lo w-Ener gy Adapti v eClustering Hierarchy) [10 ] [1 ] and MTE (Minimum T ransmission Ener gy) [11 ] [12 ], tw o v ery well kno wn routing protocols. The results sho w that SELAR deli v ers upto 12 times more and upto 1.4 times more data pack ets to the base station than LEA CH and MTE respecti v ely It w as also seen from the results that for realistic scenarios, SELAR with mo ving base station has upto 5 times and upto 27 times more lifetime duration compared to MTE and LEA CH respecti v ely The results indicate that SELAR is able to send more data, e xtend the netw ork lifetime and distrib ute the ener gy more uniformly than LEA CH and MTE. Further the scalability of LEA CH and MTE w as sho wn to be poor while SELAR will w ork the same re gardless of the number of nodes and size of the netw ork. From the results, it can be concluded that SELAR is an ener gy-ef cient, f ault-tolerant and scalable routing protocol for wireless sensor netw orks. The results obtained sho w that SELAR performs better than MTE and LEA CH. Further by mo ving the base station at appropriate times, it is sho wn that the performance of SELAR can be increased. 52

PAGE 61

Se v eral aspects related to SELAR need further in v estigation. More w ork need to be done to determine the optimal time interv als or time at which to broadcst control pack ets to neighbor nodes so that control o v erhead is minimized without losing ener gy a w areness. Also, a detailed study needs to be conducted on mo ving the base station during the operation of the wireless sensor netw orks. T w o of the main questions to be answered about mo ving the base station are: When to mo ve the base station? and Wher e to mo ve the base station? 53

PAGE 62

REFERENCES [1] W Heinzelman, “Application-Specic Protocol Architectures for W ireless Netw orks, ” PhD thesis, Massachusetts Institute of T echnology 2000. [2] I. Ak yildiz, W Su, Y Sankarasubramaniam, and E. Cayirci, “A Surv e y on Sensor Netw orks, ” in IEEE Communications Ma gazine v ol. 40, August 2002, pp. 102–114. [3] J. N. Al-Karaki and A. E. Kamal, “Routing T echniques in W ireless Sensor Netw orks: A Surv e y ” Department of Electrical and Computer Engineering, Io w a State Uni v ersity Ames, 50011. [4] K. Akkaya and M. Y ounis, “A Surv e y on Routing Protocols for W ireless Sensor Netw orks, ” Department of Computer Science and Electrical Engineering, Uni v ersity of Maryland, Baltimore County Baltimore, MD 21250. [5] “The mit uamps ns code e xtensions. ” [Online]. A v ailable: www-mtl.mit.edu/researchgroups /icsystems/uamps/resear ch/ lea ch /lea ch code.shtml [6] D. Culler D. Estrin, and M. Sri v asta v a, “Ov ervie w of Sensor Netw orks, ” in IEEE Computer Ma gazine v ol. V ol. 37, No. 8, August 2004, pp. 41–49. [7] G. Lukachan and M. A. Labrador “SELAR: Scalable Ener gy-Ef cient Location Aided Routing Protocol for W ireless Sensor Netw orks, ” in Pr oceedings of the 29th Annual International Confer ence on Local Computer Networks (LCN'04) No v ember 2004, pp. 694–695. [8] C. Shen, C. Srisathapornphat, and C. Jaikaeo, “Sensor Information Netw orking Architecture and Applications, ” in IEEE P er s. Commun. August 2001, pp. 52–59. [9] G. Hoblos, M. Saroswiecki, and A. Aitouche, “Optimal Design of F ault T olerant Sensor Netw orks, ” in IEEE Int'l Conf Cont. Apps. September 2000, pp. 467–472. [10] W R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Ener gy-Ef cient Communication Protocol for W ireless Microsensor Netw orks, ” in Pr oceedings of 33r d Hawaii International Confer ence on System Sciences (HICSS '00) January 2000. [11] M. Ettus, “System Capacity Latenc y and Po wer Consumption in Multihop-Routed SS-CDMA W ireless Netw orks, ” August, 1998, pp. 55–58. [12] T Shepard, “A Channel Access Scheme for Lar ge Dense P ack et Radio Netw orks, ” in Pr oc. A CM SIGCOMM August 1996, pp. 219–230. 54

PAGE 63

[13] W Heinzelman, J. K ulik, and H. Balakrishnan, “ Adapti v e Protocols for Information Dissemination in W ireless Sensor Netw orks, ” in Pr oc. F ifth Annual A CM/IEEE International Confer ence on Mobile Computing and Networking (MobiCom '99) August 1999, pp. 174–185. [14] C. Intanagonwiw at, R. Goo vindan, and D. Estrin, “Directed Dif fusion: A Scalable and Rob ust Communication P aradigm for Sensor Netw orks, ” in Pr oceedings of the 6th Annual A CM/IEEE International Confer ence on Mobile Computing and Networking (MobiCom '00) August 2000. [15] D. Braginsk y and D. Estrin, “Rumor Routing Algorithm for Sensor Netw orks, ” in Pr oceedings of the F ir st W orkshop on Sensor Networks and Applications (WSN A) October 2002. [16] S. Lindse y and C. Ragha v endra, “PEGASIS: Po wer -Ef cient Gathering in Sensor Information Systems, ” in Pr oc. IEEE Aer ospace Confer ence v ol. V ol. 3, 9-16, October 2002, pp. 1125– 1130. [17] L. Li and J. Y Halpern, “Minimum Enr gy Mobile W ireless Netw orks Re visited, ” in IEEE International Confer ence on Communications (ICC) v ol. 1, October 2001, pp. 278–283. [18] Y Xu, J. Heidemann, and D. Estrin, “Geography-informed Ener gy Conserv ation for Ad-Hoc Routing, ” in Pr oc. 7th Annual A CM/IEEE International /Confer ence on Mobile Computing and Networking 2001, pp. 70–84. [19] Y Y u, D. Estrin, and R. Go vindan, “Geographical and Ener gy-A w are Routing: A Recursi v e Data Dissemination Protocol for W ireless Sensor Netw orks, ” UCLA Computer Science Department T echnical Report, UCLA-CSD TR-01-0023, May 2001. [20] B. Chen, K. Jamieson, H. Balakrishnan, and R. Morris, “SP AN: An Ener gy-Ef cient Coor dination Algorithm for T opology Maintenance in Ad Hoc W ireless Netw orks, ” in W ir eless Networks v ol. V ol. 8, No. 5, September 2002, pp. 481–494. [21] S. Hedetniemi, S. Hedetniemi, and A. Liestman, “A Surv e y of Gossiping and Broadcasting in Communication Netw orks, ” in Networks v ol. V ol. 18, 1998, pp. 70–84. [22] K. Sohrabi et al., “Protocols for Self-Or ganization of a W ireless Sensor Netw ork, ” in IEEE P er s. Commun. October 2000, pp. 16–27. [23] T He et al., “SPEED: A Stateless Protocol for Real-T ime Communication in Sensor Netw orks, ” in Pr oc. International Confer ence on Distrib uted Computing Systems Pro vidence, RI, May 2003. [24] V Rodoplu and T H. Meng, “Minimum Ener gy Mobile W ireless Netw orks, ” in IEEE J ournal Selected Ar eas in Communications v ol. V ol. 17, No. 8, August 1999, pp. 1333–1344. [25] S. Basagni, I. Chlamtac, V R. Syrotiuk, and B. A. W oodard, “A Distance Routing Ef fect Algorithm for Mobility (DREAM), ” in Pr oceedings of the F ourth A CM/IEEE International Confer ence on Mobile Computing and Networking (MobiCom'98) 1998, pp. 76–84. [26] N. Bulusu, J. Heidemann, and D. Estrin, “GPS-less Lo w Cost Outdoor Localization for V ery Small De vices, ” in IEEE P er sonal Communications Ma gazine v ol. V ol. 7, No. 5, October 2000, pp. 28–34. 55

PAGE 64

[27] N. Bulusu, J. Heidemann, and D. Estrin, “Adapti v e Beacon Placement, ” in Pr oceedings of the 21st International Confer ence on Distrib uted Computing Systems (ICDCS-21) April 2001. [28] N. Bulusu, J. Heidemann, V Bychk o vskiy and D. Estrin, “Density-adapti v e Beacon Placement Algorithms for Localization in Ad Hoc W ireless Netw orks, ” T echnical Report UCLA-CS-TR010013, Computer Science Department, Uni v ersity of California at Los Angeles, Los Angeles, California, USA, May 2001. [29] B. Blum, T He, S. Son, and J. Stank o vic, “IGF: A State-Free Rob ust Communication Protocol for W ireless Sensor Netw orks, ” Department of Computer Science, Uni v ersity of V ir ginia, USA, T ech. Rep. CS-2003-11, 2003. [30] T Camp, J. Boleng, B. W illiams, L. W ilcox, and W Na vidi, “Performance Comparison of T w o Location Based Routing Protocols for Ad Hoc Netw orks, ” Department of Math and Computer Sciences, Colorado School of Mines, Golden, CO 80401. [31] S. R. Gandham, M. Da w ande, R. Prakash, and S. V enkatesan, “Ener gy Ef cient Schemes for W ireless Sensor Netw orks with Mobile Base Station, ” Department of Computer Science, School of Management, Uni v ersity of T e xas at Dallas, Richardson, TX 75080. [32] “The Netw ork Simulator ns-2. ” [Online]. A v ailable: www .isi.edu/nsnam/ns/ [33] “CMU W ireless and Mobility Platform. ” [Online]. A v ailable: www .monarch.cs.rice.edu/cmuns.html [34] J. Broch, D. Maltz, D. Johnson, Y Hu, and J. Jetche v a, “A Performance Comparison of MultiHop W ireless Ad Hoc Netw ork Routing Protocols, ” in Pr oc. 4th A CM International Confer ence on Mobile Computing and Networking (Mobicom '98) October 1998. [35] S.Sheng, L. L ynn, J. Peroulas, K.Stone, I. O'Donnell, and R. Brodersen, “A Lo w Po wer CMOS Chipset for Spread Spectrum Communications, ” in Pr oceedings of the 42nd SolidState Cir cuits Confer ence (ISSCC'96) February 1996. [36] T Rappaport, W ir eless Communications: Principles & Pr actice Prentice-Hall Inc., Ne w Jerse y 1996. [37] “The uAMPS project. ” [Online]. A v ailable: www-mtl.mit.edu/researchgroups/ ics yst ems/u am ps/ [38] “Xbo w product details. ” [Online]. A v ailable: www .xbo w .com/Products/productsdet ail s.asp x? s id=62 56


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001912727
003 fts
005 20071011114503.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 071011s2005 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001366
040
FHM
c FHM
035
(OCoLC)174042330
049
FHMM
090
QA76 (ONLINE)
1 100
Lukachan, George.
0 245
Scalable energy-efficient location aided routing (SELAR) protocol for wireless sensor networks
h [electronic resource] /
by George Lukachan.
260
[Tampa, Fla] :
b University of South Florida,
2005.
3 520
ABSTRACT: Large-scale wireless sensor networks consist of thousands of tiny and low cost nodes with very limited energy, computing power and communication capabilities. They have a myriad of possible applications. They can be used in hazardous and hostile environments to sense for deadly gases and high temperatures, in personal area networks to monitor vital signs, in military and civilian environments for intrusion detection and tracking, emergency operations, etc. In large scale wireless sensor networks the protocols need to be scalable and energy-efficient. Further, new strategies are needed to address the well-known energy depletion problem that nodes close to the sink node face. In this thesis the Scalable Energy-efficient Location-Aided Routing (SELAR) protocol for wireless sensor networks is proposed to solve the above mentioned problems. In SELAR, nodes use location and energy information of the neighboring nodes to perform the routing function. Further, the sink node is moved during the network operation to increase the network lifetime. By means of simulations, the SELAR protocol is evaluated and compared with two very well-known protocols LEACH (Low-Energy Adaptive-Clustering Hierarchy) and MTE (Minimum Transmission Energy). The results indicate that in realistic senarios,SELAR delivers up to 12 times more and up to 1.4 times more data packets to the base station than LEACH and MTE respectively. It was also seen from the results that for realistic scenarios, SELAR with moving base station has up to 5 times and up to 27 times more lifetime duration compared to MTE and LEACH respectively.
502
Thesis (M.S.)--University of South Florida, 2005.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 56 pages.
590
Adviser: Miguel Labrador, Ph.D.
653
Large scale.
Sensor node.
Multihop routing.
Network lifetime.
Power constraint.
690
Dissertations, Academic
z USF
x Computer Science
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1366