Transcript
Page 1: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

A Topology Aware

Routing Algorithm for

Mobile Ad-hoc Networks

Tyng Wey, Leong

This report is submitted as partial fulfilmentof the requirements for the Honours Programme of theSchool of Computer Science and Software Engineering,

The University of Western Australia,2005

Page 2: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Abstract

A mobile ad-hoc network (MANET) is a collection of wireless mobile hosts form-ing a temporary network without any centralised administration or standard sup-port services. In the MANET, cooperation between nodes is required to routemulti-hop messages from the source to destination due to the limited transmissionrange of each node.

Traditional routing protocols such as DSDV, AODV, DSR and TORA do notconsider sufficient topology information in their routing decision, thus they areslow to react to drastic topology changes. We explore some methods of obtainingnetwork topology information and propose that a topology aware routing protocolwill have better performance as it will be able to adapt to the changes caused bynode mobility.

Based on the belief that a more stable route will outlast its more dynamiccounterparts, we propose a new metric, the Route Stability Metric (RSM), tobe used for route selection in routing protocols. RSM measures the rate atwhich a route changes based on the relative distances and speeds of node pairsin the route without the use of the Global Positioning System (GPS) or otherspecialised hardware. We implement the new metric in DSR-RSM and compareits performance with DSR under the cnet simulation environment. Simulationresults demonstrate that including topology information into the routing decisionimproves throughput, reduces overheads and route failures in dynamic mobilenetworks.

Keywords: MANET, protocol, topology, metricCR Categories: C.2.1, C.2.2

ii

Page 3: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Acknowledgements

It has been a long four years since I came to Western Australia from Singaporein July 2001 to pursue a degree in Computer Science at UWA. This last one yearhas been especially enriching and rewarding for me during my honours researchand it would not be possible without the love and support of many people.

I am fortunate to be supervised in my project by Dr Chris McDonald. Hisguidance and support kept me on my toes throughout the year and allowed me tocomplete my project on the cnet network simulator. He also devoted his precioustime to read through and correct my many drafts and reports.

Help also came from Amer Filipovic, an outstanding researcher and cnet ex-pert, who was willing to help me understand much of his DSR code on whichmy project was based on. He was always eager to share and willing to help inwhatever way he could, even on those Sunday afternoons.

Much care and support during this hectic period came from my girlfriend andsoul mate Jocelyn who was always there for me when the stress builds up. She’salways there with a smile to cheer me up. Her love kept me going when the goinggets tough.

Huichoo and Celina made sure that I always have a hot dinner waiting when Icome back late on Monday evenings after meeting with Dr Chris. Their friendshipprovided me with the warmth and energy to work on my project throughout thechilly winter.

Finally there was Skippy, my adorable chihuahua, who never fails to lift myspirit with her love and affection. All those long nights in front of the computer,she was always there on my lap.

I am forever grateful and indebted to all of you.

iii

Page 4: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Contents

Abstract ii

Acknowledgements iii

1 Introduction 1

2 The DSR Routing Protocol 4

2.1 DSR Route Discovery . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 DSR Route Maintenance . . . . . . . . . . . . . . . . . . . . . . . 6

3 Basic 802.11 MAC Layer Operations 7

3.1 Hidden and Exposed Node Problems . . . . . . . . . . . . . . . . 7

3.2 Medium Access Basics . . . . . . . . . . . . . . . . . . . . . . . . 8

3.3 Medium Contention in DCF Mode . . . . . . . . . . . . . . . . . 8

3.3.1 Back Off Mechanism . . . . . . . . . . . . . . . . . . . . . 9

3.4 Positive Acknowledgement . . . . . . . . . . . . . . . . . . . . . . 10

4 Range, Position and Orientation in Ad-hoc Networks 11

4.1 Ranging and Localisation Methods . . . . . . . . . . . . . . . . . 11

4.1.1 Received Signal Strength Estimation . . . . . . . . . . . . 11

4.1.2 Time Difference of Arrival(TDOA) . . . . . . . . . . . . . 13

4.1.3 Triangulation and Trilateration . . . . . . . . . . . . . . . 14

5 Related Work 17

5.1 Topology Information in Routing . . . . . . . . . . . . . . . . . . 17

5.2 Topology Information as A Route Selection Metric . . . . . . . . . 18

5.3 Topology Information for Faster Route Repairs . . . . . . . . . . 19

5.4 Our Approach and Hypothesis . . . . . . . . . . . . . . . . . . . . 19

iv

Page 5: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

6 Route Stability Metric 21

6.1 Free Space Loss (FSL) model . . . . . . . . . . . . . . . . . . . . 21

6.2 Relative Speed Between Adjacent Nodes . . . . . . . . . . . . . . 22

6.3 Measuring Link and Route Stability . . . . . . . . . . . . . . . . . 22

7 Implementation of RSM 25

7.1 Implementation of RSM in DSR . . . . . . . . . . . . . . . . . . . 25

7.2 DSR Modifications . . . . . . . . . . . . . . . . . . . . . . . . . . 25

7.3 802.11 MAC Layer Modifications . . . . . . . . . . . . . . . . . . 27

8 Methodology 28

8.1 Simulation Environment . . . . . . . . . . . . . . . . . . . . . . . 28

8.2 Test Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

8.2.1 Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

8.2.2 Traffic Load . . . . . . . . . . . . . . . . . . . . . . . . . . 30

8.3 Wireless LAN link Characteristics . . . . . . . . . . . . . . . . . . 31

8.4 Signal Propagation Model . . . . . . . . . . . . . . . . . . . . . . 31

8.5 MAC layer and Routing Protocol Setup . . . . . . . . . . . . . . . 32

9 DSR-RSM Performance Analysis 34

9.1 Protocol Verification . . . . . . . . . . . . . . . . . . . . . . . . . 34

9.2 Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

9.3 Routing Overhead . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

9.4 Route Failure Rate . . . . . . . . . . . . . . . . . . . . . . . . . . 41

10 Discussion & Conclusions 45

A Original Honours Proposal 47

A.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

A.2 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

A.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

A.4 Software and Hardware Requirements . . . . . . . . . . . . . . . . 50

v

Page 6: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

B DSR packet formats 51

C Basic IEEE 802.11 Frame formats 52

D Example of a cnet topology file 54

E Example of a cnet statistics file 55

vi

Page 7: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

List of Tables

8.1 Default characteristics of cnet WLAN links . . . . . . . . . . . . . 32

8.2 802.11 MAC layer configuration used in the simulations. . . . . . 33

8.3 DSR-RSM setup used in the simulations. . . . . . . . . . . . . . . 33

vii

Page 8: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

List of Figures

1.1 A simple ad-hoc network diagram. . . . . . . . . . . . . . . . . . . 2

2.1 DSR route discovery diagram. . . . . . . . . . . . . . . . . . . . . 5

2.2 DSR route error message diagram. . . . . . . . . . . . . . . . . . . 6

3.1 802.11 Hidden node diagram. . . . . . . . . . . . . . . . . . . . . 8

3.2 NAV diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.3 802.11 Exponential backoff. . . . . . . . . . . . . . . . . . . . . . 10

4.1 RSSI ranging diagram. . . . . . . . . . . . . . . . . . . . . . . . . 12

4.2 Simple TDOA diagram. . . . . . . . . . . . . . . . . . . . . . . . 13

4.3 Triangulation diagram. . . . . . . . . . . . . . . . . . . . . . . . . 15

4.4 Trilateration diagram. . . . . . . . . . . . . . . . . . . . . . . . . 16

6.1 Distance and speed resolution diagram. . . . . . . . . . . . . . . . 23

8.1 Cnet simulation screenshot. . . . . . . . . . . . . . . . . . . . . . 29

9.1 Results from a static network simulation. . . . . . . . . . . . . . . 35

9.2 Packet Delivered Ratio Comparison. . . . . . . . . . . . . . . . . . 36

9.3 Combined Packet Delivered Percentage Comparison. . . . . . . . . 37

9.4 Routing Overhead Comparison. . . . . . . . . . . . . . . . . . . . 39

9.5 Combined Routing Overhead Comparison. . . . . . . . . . . . . . 40

9.6 Route Error Comparison. . . . . . . . . . . . . . . . . . . . . . . . 42

9.7 Combined Route Error Comparison. . . . . . . . . . . . . . . . . . 43

B.1 DSR packet format. . . . . . . . . . . . . . . . . . . . . . . . . . . 51

C.1 General 802.11 MAC frame format. . . . . . . . . . . . . . . . . . 52

C.2 RTS frame format. . . . . . . . . . . . . . . . . . . . . . . . . . . 52

viii

Page 9: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

C.3 CTS frame format. . . . . . . . . . . . . . . . . . . . . . . . . . . 52

C.4 DATA frame format. . . . . . . . . . . . . . . . . . . . . . . . . . 53

C.5 ACK frame format. . . . . . . . . . . . . . . . . . . . . . . . . . . 53

D.1 Cnet topology file format. . . . . . . . . . . . . . . . . . . . . . . 54

E.1 Example Cnet output statistics file. . . . . . . . . . . . . . . . . . 55

ix

Page 10: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 1

Introduction

Since the advent of the Intel Centrino mobile technology, wireless networkinghas exploded in popularity [28]. If you purchase a notebook computer nowadays,it will almost always come packaged with some form of wireless network card.Even palmtop devices like PDAs are commonly sold with wireless capabilitiesnowadays. As these devices become increasingly common, the need to share datawirelessly increases. From the military to the office, wireless data communicationplays an important role in modern computing. Consider the case of a group ofscientists collecting data in the field with PDAs without the use of a Wide AreaNetwork (WAN) or the Internet. In this case, a mobile ad-hoc network would beideal for sharing and synchronising information among them.

A mobile ad-hoc network (MANET) consists of mobile computing entitiessuch as laptop and palmtop computers which communicate with each otherthrough wireless links and without relying on a static infrastructure [32] suchas a base station or access point. Without centralised administration, a MANETis highly unpredictable due to its unstable links and resource-poor as most ofthe nodes have limited battery power. Due to these physical limitations, nodesrequire the cooperation of other nodes to successfully send a message to a desti-nation over multiple hops, as shown in Figure 1.1. In reality, this routing problemis much more complex. It is highly dependent on the environment and the topol-ogy of the network. The topology refers to the arrangement in which nodesare connected to each other. In their communications, mobile nodes also haveto contend with data losses from packet collisions, electromagnetic interference,node movement and failures therefore routing protocols are necessary for effectivemulti-hop communications. For example, the route in Figure 1.1 would easily bebroken if any of the involved nodes moved out of range or become unavailable.The routing protocol will then have to find an alternate route to the destination.

The task of finding and maintaining routes in MANETs is non-trivial dueto unpredictable topology changes caused by node mobility. The problem ariseswhen a source node discovers multiple routes to an intended destination and hasto make a route selection [27]. Most routing protocols choose their routes based

1

Page 11: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 1.1: A wireless ad-hoc network. Due to the limited transmission range ofa node, cooperation between nodes is required to send a message over multiplehops.

on a lowest hop-count measure (shortest route) or simply the first discoveredroute to the destination. In many cases, this metric does not yield the best net-work performance [16] as shorter routes generally imply that the spatial distancebetween intermediate nodes on the selected route will be larger than other al-ternate routes. As the nodes are further away from each other, this may lead tomore frequent route failures and degrade the performance of the network.

Many protocols have been proposed for MANETs with the goal of achievingefficient routing. Most of these protocols can be classified either as source-basedor table-based routing protocols with a few hybrid protocols emerging in the re-cent years. Some of the popular table-driven algorithms are the Destination Se-quenced Distance Vector (DSDV) routing protocol [24], the Temporally-OrderedRouting Algorithm (TORA) [23] while source-based algorithms include the Dy-namic Source Routing (DSR) protocol [14] and Ad-hoc On-demand DistanceVector (AODV) routing protocol [25]. Hybrid protocols like the Zone RoutingProtocol (ZRP) combines proactive and reactive approaches at different stages ofthe routing process. These algorithms differ in the approach used for searchinga new route and/or modifying a known route when nodes move [12]. They aresimilar in that insufficient network topology information is considered in theirrouting decisions. For example, the routing algorithms do not consider the phys-ical location of the destination node when choosing a route. They are also notconcerned about information like the density of the network, movement speedand direction of the nodes. Consequently, these routing algorithms are slow inreacting to dynamic changes in the topology of the network resulting in reduced

2

Page 12: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

throughput when they occur.

Many attempts have been made to incorporate topology information intorouting protocols. Ko and Vaidya [17] proposed the Location-Aided Routing al-gorithm (LAR) that uses the destination node’s location when a packet is sent toit. The basic idea is to perform restricted flooding in the region where the desti-nation node is predicted to be when the source node performs a route discovery.This reduces the number of route discovery packets broadcast in the network.Other algorithms based on observation of nature were also explored. Camaraand Loureiro [8] proposed the the GPS/Ant-Like routing algorithm (GPSAL)which also uses the physical location of a destination node in routing and mobilesoftware agents modeled on ants. In the algorithm, agent software is used tocollect and disseminate information about nodes’ locations in the MANET.

The aim of this research is to explore some methods of obtaining and usingtopology information in MANETs. More importantly, we aim to provide insightto some methods where we can incorporate obtained topology information intocurrent routing algorithms like DSR and AODV to improve their routing effi-ciency. Like many others, we believe that a topology-aware routing algorithm isessential to fully utilise and maximise the capacity of a MANET.

Rather than reviewing all the common routing protocols, we focus our researchon a source-based single on-demand routing protocol such as DSR because of itsproven efficiency. We propose that DSR can be further improved by includingnetwork topology information in its routing decision and present the Route Sta-bility Metric which describes the dynamic nature of a route. This metric is usedby the routing protocol to help select a more stable route. We give considerationto minimising the network overhead required for obtaining and disseminatingtopology information in order to preserve DSR’s efficiency.

Comprehensive testing and simulation will show the performance of a simu-lated network using network topology information. The results will provide uswith guidelines to design and improve topology-aware routing protocols allowingperformance maximisation of the MANET.

3

Page 13: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 2

The DSR Routing Protocol

Published in 1996, the Dynamic Source Routing (DSR) protocol was designedspecifically for MANET applications [32]. Unlike conventional routing protocols,DSR was designed for the dynamic topology changes present in MANETs. As anon-demand routing protocol, DSR only attempts route discovery when necessaryand there are no periodic control packets of any kind in the network. If thenetwork is static and all the routes needed for communications have been found,the number of control packet in the network drops to zero. This is one reasonwhy DSR is more efficient than the other routing protocols.

In DSR, multiple routes to any destination node are allowed and the sourcenode can select the route used to deliver its packets [15]. The main feature ofsource routing is that every data packet follows the route stored in its packetheader. This route provides the address of every node through which the packetmust travel in order to reach its destination. Therefore every node on the selectedroute has a routing role in forwarding the packet to the next hop in the route. Atthe same time, each node also maintains a route cache for storing every sourceroute that it has learnt.

When a node needs to send a data packet, it checks its route cache for a routeto the destination. If a route exists, it is used to send the data packet. If there isno route to the destination in the cache, the node attempts to find one using theroute discovery mechanism. While a node is using a source route, a monitoringmechanism known as route maintenance checks the validity of the route used.When route maintenance detects a problem with a route in use, route discoverymay be used again to discover a new and valid route to the destination [14].

2.1 DSR Route Discovery

The process of finding a route between a source node S and a destination nodeD when no pre-existing route is available is known as route discovery. When

4

Page 14: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 2.1: DSR route discovery between initiator node S and destination nodeD.

node S needs to communicate with node D, it finds a route on-demand usingroute discovery when it cannot find a route to D in its route cache. To initiatethe process, node S broadcasts a Route Request (RREQ) packet in the network.This RREQ packet contains the address of the initiating node S, the address ofthe target node D, a request id to uniquely identify the RREQ packet and a routerecord. The route record holds the sequence of hops accumulated by the RREQpacket as it transverses the network. Each node in the network also maintains atable of the 〈 initiator address, request id 〉 pairs in order to detect and discardduplicate RREQ packets received.

When a node received a RREQ packet, it propagates the packet if it is notthe intended destination of that packet and if it is the first time it receives thatpacket. The first node receiving this RREQ packet and has a valid route to thedestination D in its route cache will initiate a Route Reply (RREP) packet backto node S. This RREP packet contains the addresses of nodes along the pathfrom the source node S to target node D. (see Figure 2.1). The first part of thislist was gathered along the path of the RREQ packet while the rest comes fromthe route cache of the replying node.

In the case where the destination node D receives the RREQ packet, i.e. nonode along the way before node D has a valid route from node S to node D inits route cache, node D sends a RREP packet containing the path traversed bythe RREQ packet from node S to D. This method assumes bi-directional linksare used in the network. In his original DSR proposal, Johnson [14] propose thatthe destination node D piggyback its RREP packet on its own RREQ packettargeted at the source node D. This method assumes that some uni-directionallinks may be present in the network. When multiple paths exist between node Dand S, node D selects the path to S based on the shortest hop count from nodeS to node D. Node D then ignores all RREQ packets corresponding to the sameroute discovery process after the arrival of the first one.

5

Page 15: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 2.2: A broken link between node C and D triggers a DSR route errormessage back to the initiator node A.

2.2 DSR Route Maintenance

Route maintenance in DSR is used for two reasons:

1. When the mobility of a node causes a route to become invalid. For example,an intermediate node C moves away from node B and D causing the routefrom B to D through C to become invalid.

2. When the availability of a node prevents it from participating in the com-munication. For example, node C shuts down due to a low battery powercausing the path from B to D to fail.

In both cases, the route maintenance mechanism has to ensure that the pathsstored in the node route cache are valid. As DSR does not use periodic controlpackets, route failure can only be detected during data communications. Asshown in Figure 2.2, when forwarding a packet, node C detects a link failurebetween it and node D. Node C then initiates a Route Error (RE) packet backto the initiator of the packet, node A. This RE packet indicates which link isbroken, i.e., the node that detected the error and the node it was trying to reach.

When a node receives a RE packet, it removes the invalid link from its routecache and, for each route containing this link, truncates the route from the hopbefore the broken link. In order to obtain the status of each link, several acknowl-edgement schemes can be used, e.g. data link layer acknowledgement, explicitrouting layer acknowledgement from the next-hop receiver, or passive acknowl-edgement where a node overhears its neighbours forwarding its packet. If thesource node has an alternative route to the destination, it can then continuesending packets using this alternate route. Otherwise, a Route Discovery is re-quired to establish a new route.

6

Page 16: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 3

Basic 802.11 MAC Layer Operations

As part of our proposed routing protocol improvement requires the MAC layerproviding it with topology information, we will briefly discuss the basic 802.11MAC layer operations here. Specific interactions between the MAC layer androuting layer will be discussed in Chapter 7.

The 802.11 family specifies a common Medium Access Control (MAC) layerwhich provides a variety of functions to support the operation of fixed and movingstations within a local area. It is by far the most common standard adopted bymany wireless network card manufacturers [18]. In the standard, both radio andinfra-red physical layer specifications are covered but we will only cover the radiotransmission portion here.

As in wired Local Area Networks, the 802.11 MAC layer manages and main-tains communications between 802.11 stations (wireless network interface cards(NIC) and access points) by coordinating access to a shared radio medium. How-ever, unlike Ethernet where frame collision detection is possible on the medium,wireless nodes implementing 802.11 use Carrier Sense Multiple Access/CollisionAvoidance (CSMA/CA ) which attempts to avoid collisions rather than detectsthem. This mechanism is needed as not all nodes are within transmission rangeof each other. Consider the typical situation in Figure 3.1 :

3.1 Hidden and Exposed Node Problems

1. Node A wishes to send a frame to node B but cannot sense that node B isbusy receiving messages from node C. Detecting that the medium is free,node A transmits to B resulting in a frame collision at B. This phenomenonis known as the hidden node or hidden terminal problem. Node C and Aare hidden from each other but their communications with B can interfere.

2. Node B wishes to transmit to node C but senses that transmission is ongoingbetween node A and another node (possibly hidden to B). Node B then

7

Page 17: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 3.1: Node B is within transmission range of both node A and C. Node Aand C cannot receive each other’s transmission.

incorrectly assumes that it cannot transmit to node C for fear of causinga collision. This problem is described as the exposed terminal or exposednode problem [31].

3.2 Medium Access Basics

Before a station can start transmitting frames, it must first gain access to themedium (a shared radio channel in the network). In 802.11, two forms of mediumaccess, distributed coordination function (DCF) and point coordination function(PCF) are described. DCF is mandatory and is based on the CSMA/CA protocol.With DCF, 802.11 stations contend for access to the medium and attempt to sendframes when no other stations are transmitting. On the the hand, in PCF, theaccess point (AP) grants an individual station access to the medium by pollingthe station during a contention free period. This method supports the deliveryof time-bounded data frames. Both DCF and PCF can coexist because differentinter-frame timings are used for the different modes. As MANETs only implementDCF, PCF operations are beyond our discussion scope.

3.3 Medium Contention in DCF Mode

When a station is operating in the DCF mode, the MAC layer determines whetherthe medium is free or busy using a virtual and physical carrier sense. The virtual

8

Page 18: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 3.2: An example of how stations update their NAV counter from thereceived frame.

carrier sense uses a Network Allocation Vector (NAV), which is a resident counterin each station representing the amount of time a previous station has reservedthe channel for. The NAV value must be zero before a station can attempt tosend a frame. Prior to transmitting, the station calculates the time it needsto complete the frame transaction (data frame and its acknowledgement frame)based on the frame length and data rate. This time value is then placed in theduration field in the frame header. Stations receiving this frame will update theirNAV counter with the duration value if they are not the intended recipient of theframe [9].

In additional to the virtual carrier sense, control frame exchanges betweencommunicating stations are also required before data frames can be sent. 802.11uses a Request To Send (RTS)/Clear To Send (CTS) control frame sequencebetween communicating stations to reserve the medium. Stations overhearingthese RTS/CTS frames shall update their NAV counter with the duration valuescarried in the frame headers (see Figure 3.2) if the value in the frame header islarger than their NAV counter.

3.3.1 Back Off Mechanism

Another important aspect of the DCF is the random backoff mechanism thata station uses if it detects a busy medium when it wants to transmit a frame.The station shall then wait until the medium become free, further defer for anadditional random period before attempting to access the medium again. Ifthe channel is again determined to be busy, the station increases its ContentionWindow (CW) value exponentially (see Figure 3.3) and uses this value to generate

9

Page 19: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 3.3: An example of the exponential increase of CW. (Figure obtained fromthe IEEE 802.11 standards.)

another random backoff period.

The value of CW is bounded by predetermined CWmin and CWmax values.When CW reaches CWmax, it remains at that value until a successful transmissionresets it to CWmin. This backoff mechanism reduces the probability that multiplestations wanting to send data will transmit at exactly the same time after sensinga free channel. This leads to improved stability of the protocol under high trafficload by reducing the number of collisions and their corresponding retransmissions.However it also increases the latency of the network.

3.4 Positive Acknowledgement

As wireless stations cannot switch both their receiver and transmitter on at thesame time (half-duplex transmissions), transmitting stations cannot listen forcollisions while they are sending data. As a result, the receiving station needs tosend an Acknowledgement (ACK) frame to the sender if no errors were detectedin the received frame. If the sending station does not receive an ACK framewithin a specified period, it will assume that the frame was lost due to a collisionand reinitialise the frame transmission sequence, i.e., RTS, CTS, DATA, ACK.

10

Page 20: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 4

Range, Position and Orientation inAd-hoc Networks

In many MANET applications, knowledge of the physical locations of mobilenodes is beneficial if not essential. An example is the relaying of tactical in-formation in the modern day battlefield between unit commanders and soldiers.Vital location information provided by the wireless devices carried by individualsoldiers allow their commanders to communicate with their troops in real timeand track their movement [1]. This information facilitates decision making duringthe course of battle.

In wireless sensor network applications, position and orientation informationof individual nodes allow target tracking, coverage management and labelling ofdata messages propagating in the network [20]. In some ad-hoc network routingprotocols, knowledge of the geographical position of nodes is required and loca-tion discovery capabilities are assumed [8][17][30][32]. Using a variety of nodecapabilities, different methods of obtaining range, position and orientation infor-mation are possible.

4.1 Ranging and Localisation Methods

4.1.1 Received Signal Strength Estimation

Most wireless networks operate on the principle of Radio Frequency (RF) trans-mission and there are four primary units of measurement that can be used torepresent RF signal strength. They are mW (milliwatts), dBm (dB milliwattswhere 0db is one mW), RSSI (Received Signal Strength Indicator) and a per-centage measurement. All of the above measurements are related and can beconverted from one form to another with varying degrees of accuracy [3].

Most wireless network interface cards implement the popular IEEE 802.11

11

Page 21: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 4.1: Received signal strength ranging provides a circle on which the trans-mitting node S must lie, absolute positioning is not possible with just one signal.

standard and measure the RF energy received by the circuitry using the RSSI.This numeric value is an integer with an allowable range of 0-255(a 1 byte value).However, no NIC manufacturer measures all 256 signal levels thus different vendorcards may have different scales for measuring the RSSI. In the standard statedin IEEE 802.11, it was also noted that the RSSI is intended to be used only ina relative manner and absolute accuracy of the RSSI reading was not specified.This implies that individual card vendors can choose to provide their own levelsof accuracy, granularity and range for the actual power and their RSSI values.

Using the received signal strength at a node, if the transmission power ofthe transmitting node is also known, it is possible to estimate range informationto the transmitting station. Using a suitable signal loss model that describesthe signal attenuation through the environment, we obtain a distance givinga circle on which the transmitting station must lie (see Figure 4.1). In manysensor networks operations, received signal strength ranging is commonly usedto estimate distances to neighbouring nodes.

However, received signal strength measurements are not always accurate andcan result in errors up to 50% of the measured distance especially if the signal

12

Page 22: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 4.2: A simple application of TDOA in everyday life. Estimating thedistance of a storm from the time difference between seeing a lightning flash andthen hearing the thunder [13].

loss model does not reflect the environment accurately. It is also known to bevery noisy under certain circumstances. Nevertheless, in many wireless networkswith limited node capabilities, it is the only method available to approximaterange information. To reduce the effects of this error, nodes trilaterate (seesection 4.1.3) their position by measuring distances to as many reference pointsas possible. In some cases, this has been shown to reduce the node location errorsto less than 5% [20].

4.1.2 Time Difference of Arrival(TDOA)

As the name suggests, time difference of arrival (TDOA) is based on the sixorders of magnitude difference between the speeds of sound and light [22]. Asa requirement, a positioning system based on this principle needs to transmitsignals in both RF and ultrasound. The receiver also needs to be able to receiveboth signals, therefore TDOA systems require some specialised hardware.

When a node transmits the two signals, a receiving node infers the range tothe originating node based on the difference in time of arrival of the two signals.For example, node S and D are x metres apart. Node D receives a RF signal attime t and an ultrasound signal at time t + 1 from node S. Based on 343 ms−1

for the speed of sound under average circumstances and 3×108ms−1 for the speedof light (c), the distance x between the nodes S and D would be approximately343 metres. An everyday example of how we use this principle is in the distanceestimation of a storm [13]. As shown in Figure 4.2, when we see a lightning flash,we count the number of seconds before hearing the thunder crack. Multiplyingthe speed of sound by the time delay, we can estimate our distance to the storm.

Other than obtaining range information, if a receiving node uses two ultra-

13

Page 23: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

sound receivers placed at a known distance d apart from each other, it is alsopossible to obtain an Angle of Arrival (AoA) measurement from the received ul-trasound signal [21]. However, this method imposes significantly more hardwarecapabilities and is only seen in specialised wireless sensor networks such as theCricket Compass project [26] developed at the Massachusetts Institute of Tech-nology (MIT). Similar to received signal strength estimations, TDOA requiresbeacons in the network in order to position nodes absolutely.

In digital mobile phone networks, a variation of TDOA is used to estimate thelocation of a mobile phone user. It works by calculating the difference in time asignal from the mobile phone takes to reach different access points in the mobilephone network. Using the propagation delay of the signal to the different accesspoints, cooperation between access points enables them to position the mobilephone user[7].

4.1.3 Triangulation and Trilateration

Triangulation is a positioning procedure that relies on angular measurementsto three known landmarks while trilateration uses the distances to these threelandmarks to determine position. Historically, triangulation has been more com-monly used in navigation than trilateration as angles are more easily measuredwith simple methods and equipment than distances [22]. For example, it is easierto measure the angle of a ship to a known island with a magnetic compass than itis to estimate the distance between them without modern equipment like a laserrangefinder or radar.

When a wireless node uses triangulation to determines its position in thenetwork, it requires the positions of at least three landmarks from which it isable to receive RF signals as well as the angle of arrival of these signals.

In Figure 4.3, node D infers its position by finding the point of intersectionof the three bearings from the landmarks. The bearings represent the directionfrom which the RF signals are coming. In triangulation, nodes are assumed tohave at least AoA capabilities before they can infer their locations.

If the distances (obtained using TDOA or received signal strength) and po-sitions of the landmarks are known to the node, location information can alsobe inferred using trilateration. As shown in Figure 4.4, we draw arcs from thelandmarks using their distance to node D as the radii of these arcs. The intersec-tion of two arcs from landmarks A and B will produce two possible positions fornode D and the 3rd arc from landmark C will confirm its position. In this case,Possible location 2 is the accurate position. In order to use trilateration, nodesonly need ranging capability (see section 4.1.1 and 4.1.2 above). Trilateration is

14

Page 24: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 4.3: Triangulation using angles to three known points A, B and C.

also the working principle behind the popular Global Positioning System (GPS)where instead of terrestrial landmarks being used, satellites orbiting the Earthact as landmarks. As the working principle of GPS is very similar to trilateration,we will not further explain it in this paper.

In order for wireless nodes to obtain absolute positioning from triangulationand trilateration, there must at least be some nodes in the network which haveself-positioning capabilities. These nodes must act as beacons and advertisetheir positions to less capable nodes in the network. In many wireless sensornetworks, trilateration or its extension, multilateration, is commonly used fornode localisation [5]. A sufficient number of beacons is necessary or nodes whichare not aware of their positions will not be able to calculate their positions.At the same time too many beacons will be too expensive and also cause selfinterference [6].

15

Page 25: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 4.4: Distance to the 3rd landmark C confirms the node’s location.

16

Page 26: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 5

Related Work

In this chapter, we discuss some of strategies taken by others to incorporatetopology information into routing protocols. It also outlines the ways their ap-proaches differ from our research. These approaches taken are as diversed as thetype of topology information used.

5.1 Topology Information in Routing

Drawing inspiration from nature, Camara and Loureiro [8] propose the algorithmGPS/Ant-Like (GPSAL) which is based on mobile software agents modeled onAnt algorithms. Based on a population approach, Ant algorithms are inspired bythe foraging behaviour of ant colonies. They have only been developed recentlyand applied successfully to solve some Non-Polynomial (NP) hard problems. Intheir algorithm, Camara and Loureiro used ant-like software agents to collect anddisseminate node position information throughout the MANET. Their approachis unique because there is no flooding of packets during route discovery.

Similar to LAR [17] mentioned in Chapter 1, GPSAL relies on GPS to pro-vide accurate location and movement information to the individual nodes. Thisimplies that GPSAL cannot be easily implemented in many current wireless net-works. However, unlike other pure ad-hoc network approaches, GPSAL alsoconsiders that fixed nodes may be present in the wireless network to act as land-marks.

In GPSAL, nodes update their routing tables locally by periodically receiv-ing topology change information from their neighbours. This process adds someoverhead to the network as periodic control packets are required. Globally, antagents are sent randomly to nodes on the network. During their propagation,these agents collect and scatter node position information throughout the net-work. Upon reaching a node, they update the routing tables in the node andhelp accelerate route discovery in the network. At the same time, these “ants”

17

Page 27: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

also add large overheads and increase the traffic in the network. As Camara andLoureiro did not elaborate on the structure and operations of these ant agentprograms, the exact load these agents place on the network is unclear and thegains cannot be easily verified.

5.2 Topology Information as A Route Selection Metric

Exploring another way to use topology information, Kim, Kwon and Suh [16]propose the Reliable Route Selection (RRS) algorithm using location and mobilityinformation to select the most reliable route impervious to failure caused by nodemobility. Similar to LAR and GPSAL, RRS assumes that a node acquires itsown location and speed from GPS.

The implementation of RRS was in AODV and it introduced the conceptof a stable/caution zone around a node. A stable zone is the area in whichtransmission can be safely undertaken between two nodes while a caution zone isthe region in which the transmissions carried out are likely to fail. The descriptionof the zones is based on the direction and speed information of a node.

During a route discovery process, when a node receives a RREQ packet, itinserts its location and stable zone information into the RREQ packet after ver-ifying that it is in the stable zone of the previous node. The RREQ packet isthen re-propagated in the network. This process ensures that during the propa-gation of the RREQ packet, only nodes in each other’s stable zone will forwardthe packet. When a node receives a RREQ packet and discovers that it is inthe caution zone of the previous node, it ignores the RREQ packet. This pro-cess eliminates flooding by nodes that will not be used in the route and reducesoverhead packets in the network.

The final selection of the routes is based on its perceived reliability rather thanthe smallest hop count. From this aspect, RRS sacrifices shorter packet deliverytime for route longevity thus ensuring more stable performance. In cases wherenode mobility is very high, network overheads will increase as the stable zonebecomes smaller due to increased node speed. This reduction in stable zone sizemay also lead to more frequent route discovery failures if all of the forwardingnodes are not in each other’s stable zones.

18

Page 28: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

5.3 Topology Information for Faster Route Repairs

Rather than selecting a more reliable route, Quintero, Pierre and Macabeo [27]chose to improve the route repair capabilities of routing protocols with node den-sity and availability information from a route’s neighbourhood. The availabilityof a node indicates its willingness to participate in wireless communication andcan be affected by its battery level and state.

Using the route discovery mechanism in AODV, provisions for the availabilityand density parameters were added to the RREQ packet. When a RREQ packet ispropagated through the network during route discovery, receiving nodes evaluatethe density and availability of their neighbourhood and update the fields in theRREQ packet. The destination node then chooses a route based on its perceivedrobustness and ease of repair. This procedure requires some additional overheadpackets to be broadcast periodically so that each node can maintain a validdensity and availability table.

The route selection is biased towards routes with higher node density andavailability so that local route repair can be carried out should the route breaks.Local route repair refers to repairing of the route at the point where the routewas determined to be broken. It is favoured over global route repair which startsfrom the source as it is faster and less detrimental to the overall performanceof the network. Similar to the RRS algorithm, this modified AODV protocolmay produce longer routes and have higher network overheads. This protocoldistinguishes itself by enabling and promoting local route repairs at the brokenlink which the other protocols do not consider.

5.4 Our Approach and Hypothesis

Using a similar approach as the RRS algorithm, our research focuses on improv-ing routing protocol performance with a topology aware metric. Our primaryobjective is to provide a new metric to describe the stability of a route and al-low the selection of more static and longer lasting routes. We believe that amore informed routing decision will significantly improve the performance of anetwork through higher throughput, lower overheads and less frequent route fail-ures. Unlike the approaches mentioned, we also aim to implement our protocolwithout the the use of specialised positioning hardware such as GPS and withoutimposing additional overhead packets on the network. We verify the utility ofour algorithm by implementing it in a well established routing protocol, DSR,and analysing the following performance by comparing it with the standard DSR

19

Page 29: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

protocol:

1. Packet delivered ratio - the number of messages delivered compared to thenumber of generated messages. A higher ratio achieved indicates betternetwork performance.

2. Packet overhead ratio - the number of overhead packet per message deliv-ered. A lower ratio indicates a more efficient routing protocol.

3. Route error ratio - the number of route failures per message delivered. Ahigher ratio indicates poorer route selection leading to more route failures.

The selection of these common performance metric allows us to compare ourresults with other similar research. Our secondary objective is to develop astandard compliant 802.11 MAC layer required for realistic routing protocol sim-ulations so that our work can be verified and continued in the future. This willalso allow future enhancement to be undertaken at the MAC layer.

20

Page 30: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 6

Route Stability Metric

In this chapter, we derive the metric that we use to describe the stability of aroute. We name this route description the Route Stability Metric (RSM). In orderto compute this metric, we estimate the rate at which the separation betweenadjacent pairs of nodes in a route changes. The measure of such a rate is given bythe relative speed of the nodes and the distance between them. In Section 4.1.1,we discussed how we can obtain distance information between communicatingnodes from the received signal strength. We apply the method here with a FreeSpace Loss model to obtain the relative distance and speed of adjacent nodepairs.

6.1 Free Space Loss (FSL) model

This model assumes the ideal propagation condition that there is only one clearline of sight (without obstacles) between the transmitter and receiver. As thetransmitted signal traverses the medium, its power level decreases at a rate in-versely proportional to the square of the distance travelled and proportional tothe wavelength of the signal [12]. In this model, the received signal strength isgiven by Equation 6.1.

RXpower = TXpower − TXcableloss + TXgain − FSL + RXgain −RXcableloss (6.1)

where FSL is the free space signal loss given by

FSL = 20log10[4πd

λ]{dB} (6.2)

where d is the distance travelled by the signal and λ is its wavelength with bothin the same units. If the RXpower, TXpower and the transmitting node’s NICcharacteristics are known to the receiver, we can estimate the distance d using the

21

Page 31: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

above equations. In our simulations, we assume a homogeneous network whereall nodes have the same characteristics. However, if a non-uniform network isused, distance calculation is still possible if transmitting nodes include their NICcharacteristics in the MAC layer frames so that receiving nodes can use thatinformation for distance calculations.

6.2 Relative Speed Between Adjacent Nodes

As shown in Figure 6.1, node A receives two signals from node B at time t1and t2. Resolving the received signals to distances d1 and d2 using the FSLmodel described above, node A is able to calculate the distance change duringthe duration t2 − t1. Using the Newtonian equation for velocity:

V =d2 − d1

t2 − t1(6.3)

we can find out how fast the node pair is converging or diverging. In the calcula-tion, we assume that the relative node motion is in a straight line. An importantthing to note here is that we are only concerned with the relative distance andspeed in this algorithm. If a node is moving around another node in a radial,their relative distance will remain constant and the speed between them will bezero.

This method does not require location knowledge of the nodes, the use ofGPS or other specialised hardware in the computation of the distances and speed.There are also no periodic beacon packets to be broadcasted and thus no addi-tional traffic is imposed on the network. However, this method does require thatthe MAC layer maintain a simple node signal strength table of its neighbours.

6.3 Measuring Link and Route Stability

After we obtain the relative distance and speed of a pair of nodes from Equa-tion 6.2 and Equation 6.3, we estimate the dynamics of the wireless link be-tween the two nodes using these two variables. We know that the received signalstrength falloff is relative to the distance travelled by the signal, therefore wewant our link metric to reflect this. If the node movement was nearer to themaximum transmission range of the node, the link metric value should be high.Similarly, if the movement occurs when the nodes are close to each other, thelink should have a lower metric value. Based on this, we establish the link metricm using the following equation.

22

Page 32: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

d1 d2

Node A

Node B

Node B

Figure 6.1: Node A calculates the relative speed between itself and node B afterobtaining two distances d1 and d2. Node A obtains these distances by resolvingthe received signal strength of packets from node B.

mi =Di

Dmax

× (vi

vmax

)2 (6.4)

where mi is the link metric for the ith pair of nodes, Di is the relative distancebetween them and Dmax is the maximum transmission range of a node. vi repre-sents the relative velocity between the two nodes and vmax is a fixed value basedon the known maximum speed of the nodes. We use the square of the normalisedvelocity in the link metric calculation as we do not differentiate between converg-ing or diverging nodes. The square of the normalised velocity removes negativefigures from the link metric.

Equipped with a measure of the link dynamics between a pair of nodes, wecombine the individual link metric values into a single route metric using thefollowing equation,

RSM =N∑

i=1

mi (6.5)

where RSM is the Route Stability Metric for a route, N is the total numberof hops for the route and mi is the link metric for the ith hop obtained fromEquation 6.4 above.

Observe that RSM is a positive quantity and a higher value indicates that theroute is more dynamic and therefore less stable. If the length of two routes is thesame, we would simply select the route with the smaller RSM value because it

23

Page 33: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

is more stable. However, if the routes are of different length, choosing the routewith the smaller RSM value would always favour the shorter route.

In order to remove the bias toward shorter routes, we use the average linkmetric when comparing the stability of multiple routes. Consider the case oftwo routes R1 and R2 with route metric RSM1 and RSM2 respectively. Let N1

and N2 be the number of hops in the respective routes. When we compare theaverage link metric of the routes, we use the following condition to determine ifR1 is more stable than R2:

RSM1

N1

<RSM2

N2

(6.6)

With Equation 6.6, we now have a procedure to quantify and compare thestability of two routes. In summary, RSM has the following properties:

1. It considers a route with a RSM value of zero as the most stable route.These routes feature nodes that are moving with zero relative speed or, inthe rare case, nodes which have relative distances of zero from each other.

2. Both converging and diverging links are penalised equally as both will even-tually lead to route failures.

3. RSM encourages the use of more static routes and does not favour shorterroutes. Using a normalised link metric comparison (see Equation 6.6 above),RSM allows the selection of longer but more stable routes.

We discuss the details of implementing RSM in the next chapter.

24

Page 34: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 7

Implementation of RSM

7.1 Implementation of RSM in DSR

In the previous chapter, we discussed how to compute and compare the RSMvalues of different routes. In this chapter, we present our strategy to enhance theDSR routing protocol with RSM. We also outline the modifications required tothe DSR routing protocol and the 802.11 MAC layer in order to use RSM.

As RSM is intended to be used as a route selection metric, we deploy it in theDSR route discovery process (see Section 2.1) and it is also carried in a RREQpacket. As the route discovery process for DSR is very similar to that of AODV,RSM can also be implemented in AODV using the same strategy. Our strategyfor RSM deployment is as follows:

1. When node S needs to find a route to a destination, a RREQ packet isgenerated and its RSM field initialised to zero before the packet is broadcastonto the network.

2. Intermediate nodes receiving the RREQ packet updates the RSM field inthe RREQ packet based on their calculation and further propagates theRREQ packet until it reaches the intended destination.

3. After collecting all the possible RREQ packets from source node S, thedestination node compares all the RSM values and replies to the RREQpacket with the route described by the lowest RSM.

7.2 DSR Modifications

For RSM to operate in DSR, a few minor modifications to the standard DSRprotocol are required. Firstly, the RREQ packet needs to carry a metric payloadin order for the route discovery process to ascertain the stability of a route. This

25

Page 35: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

requires us to enhance the RREQ packet header to carry a RSM field and a linkcount field (see Appendix B for original DSR packet format). The link count fieldcontains the number of links in the route that did not contribute to the RSMvalue. For some links, the MAC layer may only have one received signal strengthmeasurement from a neighbouring node thus the relative speed of the node paircannot be ascertained. Under such circumstances, the link will not contribute tothe RSM in the RREQ packet header. It will increment the link count field andforward the RREQ packet to the next hop.

Other than modifying the RREQ packet header, a time delay is also requiredat the destination node so that it does not reply to the first RREQ packet received.Upon reception of the first RREQ packet from source S, destination node D storesthis RREQ packet into a RREQ packet buffer. We implement a Route ReplyDelay (RRD) to allow node D to review the other RREQ packets arriving fromdifferent routes, instead of ignoring them as prescribed by standard DSR. Thehigher the value of the RRD, the more routes node D will be able to receive.However, this delay also increases the latency of the node communication as thetransmission of the RREP packet is delayed by the same duration.

In order to select a suitable value for the RRD, we restrict the routes examinedto a fixed number of hops. Using the first RREQ packet as a benchmark torepresent a route of H hops, we set the maximum route length considered atH + n. In order to receive a RREQ packet that passes through H + n hops, weexpected to wait an additional time equivalent to the delay of the RREQ packetat n hops. To calculate the time delayed required, we assume the worse casescenario where the RREQ packet is queued and delayed at all n hops. Using Tas the time required to transmit a data frame at the MAC layer and S as the sizeof the MAC layer frame buffer, we estimate the RRD value with RRD = nST .This duration does not include the delays due to the ever present MAC layercontention and backoffs. Depending on whether the RTS/CTS frame sequence(see Chapter 3.3) is required for every data packet transfer at the MAC layer, thetime T will vary between the duration needed to transmit one data frame to theduration needed to send a whole frame exchange sequence (a RTS, CTS, DATAand ACK frame). In networks where congestion is not a big factor, this worsecase delay allows routes with lengths longer than H + n hops to be consideredby the destination node.

When the RRD expires at the destination node, the RSM and link countfield of all the available routes in the RREQ packet buffer are compared and thedestination node initiates a RREP packet back to the source using the route withthe lowest link count and average RSM.

26

Page 36: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

7.3 802.11 MAC Layer Modifications

In order to provide distance and speed information to the routing protocol, theMAC layer needs to store the information in a table. This table has very lowstorage requirements as the typical number of neighbours is in the order of 6-8for optimal network operations [29]. Therefore, we are not introducing significantcomputation or memory requirements in the nodes.

Whenever a station receives a RTS or DATA frame, it updates its neighbourinformation table using the received signal strength of that frame. We only usethese two types of frames as they contain the MAC address of the transmittingstation in their frame headers. With the CTS and ACK frames, the receivingstation needs to keep track of who the transmitter is and this adds additionalmemory overheads to the receiving stations.

The neighbour state information is requested by the routing protocol onlywhen it needs to update the RSM field in a RREQ packet. In order to avoidstale entries in the table, the contents of the table are periodically purged. Theaddition of this table does not affect normal MAC layer operations.

27

Page 37: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 8

Methodology

8.1 Simulation Environment

We select the simulation environment for our experiments based on the followingcriteria:

1. Support for wireless networking - How easy/difficult it is to implement ormodify a wireless network topology in that environment.

2. Simulation visualisation- What kind of visual feedback do we get duringthe simulation and whether the feedback is in real time or not.

3. Simulation time - how long it will take to perform the simulations.

4. Support for the simulator - how easy is it to get help for the simulator fromits creator or user community.

We considered several software packages including the popular ns-2 [10], OP-NET Modeller and cnet [19]. Among them, we find cnet to be the most appro-priate for our application based on the above criteria.

Cnet is a discrete event network simulator developed at The University ofWestern Australia, Perth. It supports experimentation with different data-linklayer, network layer, routing and transport layer networking protocols in wide-area-networking (WAN), local-area-networking (LAN) and wireless-local-area-networking (WLAN) environments. In cnet, the Application and Physical layersare provided and users are required to provide their own protocols for other nec-essary internal layers. Cnet ’s biggest advantage over the other simulators is itsease in building and installation in the Linux environment. In addition, cnet alsoprovides an implementation of the DSR routing protocol which we require.

Our simulation topology files were created with a text editor and adheres tocnet ’s topology file format. A typical cnet WLAN topology file is included in

28

Page 38: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 8.1: A screenshot of cnet in operation. Cnet provides real time feedbackto users and topology configurations are also changeable at run-time.

Appendix D for reference. For the experiments, a topology file of 50 nodes wascreated. All other protocol files were implemented in C and cnet handles thecompilation and linking of these files during its execution.

The default output of cnet simulations are in the form of plain text files.Users can also specify that an output file be written for each individual node inthe simulated network for debugging purposes. In addition, cnet provides realtime network animation and visual feedback through Tcl/Tk windowing (seeFigure 8.1). During the simulation, users can also make changes to the nodeand link attributes, reboot individual nodes or change other network parameters.While graphing functions are not provided in cnet, statistics can be easily collatedusing standard Unix/Linux commands like grep, awk and the output plotted withMATLAB, Gnuplot, Microsoft Excel or any other statistical program.

29

Page 39: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

8.2 Test Cases

As we are dealing with complex routing in a dynamic environment, attemptsto capture the behaviour of the routing protocol require extensive testing undervaried network conditions.

In order to allow comparison of our protocol with similar work by others, wechosed commonly used parameter values in our simulations. For all our experi-ments, we used a set of 50 mobile agents moving in a flat grid area of 1500m x300m. These 50 nodes were randomly positioned on the grid area at the begin-ning of each simulation. We fixed the length of the simulations at 300 secondsas also chosen by Kim, Kwon and Suh [16]. A rectangular area was chosen inorder to force more multi-hop communications between randomly chosen source-destination pairs. In our setup, we assumed that all links in the network arebi-directional for simplicity.

8.2.1 Mobility

The movement of the nodes in the area is based on the random waypoint model [33],a commonly used synthetic model for mobility in Ad Hoc networks. The mod-elled motion is a straight line linear trajectory with the nodes pausing for afixed amount of time before changing direction. In our first simulation, we setthe pause time to the duration of the simulation and verified the performanceof DSR against DSR-RSM in a static network. In all other simulations, thepause time was fixed at 1s. The velocities of the nodes were chosen randomlyand limited by a fixed upperbound. In order to test the effects of mobilityon route selection, we varied the maximum node speed at 5ms−1(18km/hr),10ms−1(36km/hr), 15ms−1(54km/hr) and 20ms−1(72km/hr). We set the up-perbound at 20ms−1(72km/hr) to account for mobiles nodes located in movingvehicles. Higher speed were not considered as ad-hoc networks are not applicablefor extreme high speed environments.

8.2.2 Traffic Load

Apart from varying the mobility of the network, different traffic loads were alsoemployed in our simulations. The traffic load refers to the amount of data thatis transfered onto the medium in a specific time period. It is a measure of how“busy” the network is. In our simulations, we varied the number of nodes activelytransmitting packets in the network. The type of packets generated are addressedand not broadcast packets. In all, three scenarios were constructed:

30

Page 40: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

1. Light traffic load - 10 randomly chosen nodes actively transmitting.

2. Medium traffic load - 20 randomly chosen nodes actively transmitting.

3. Heavy traffic load - 30 randomly chosen nodes actively transmitting.

In our simulations, all transmitting nodes are constant bit rate (CBR) sourceswith a period of 1

4or 0.25 seconds. As the traffic load is varied, we do not need to

vary the packet sizes or the rate of transmission. Fixing the size of Applicationlayer data generated by cnet at 512 bytes, we achieve a 4 packets/second packetrate from each transmitting node. These configurations were controlled by ourcnet topology file.

8.3 Wireless LAN link Characteristics

The characteristics of the nodes in our simulation is determined by the defaultvalues given in Table 8.1. These value were derived from the specification usedon the Orinocco PCMCIA Silver/Gold cards. With the default transmissionpower, the transmission range reaches a maximum of 173 metres. We increasedthe transmission range to 200 metres in our experiments in order to test theperformance difference of longer and shorter route selection.

With a higher transmission range, the performance difference between a pro-tocol that uses the shortest hop count metric and one that uses another metricwill become more apparent. If transmission ranges are short, the probability ofhaving multiple routes to a destination will be low and we are unlikely to seesignificant performance difference in the protocols as there is only one route fromthe source to the destination in most cases.

8.4 Signal Propagation Model

Radio communication deteriorates with the distance between source and des-tination node. In the real environment, it is difficult to predict the variationand propagation of signals without a radio map of the area. A radio map is alookup table that holds collections of packet signal strengths and the locationswhere these signals were measured. An example of such a system would be theRADAR system developed by Balachandran, Bahl and Padmanabhan [2]. How-ever, this methods fixes the deployment of a wireless network in a mapped areathus restricting its versatility. Cnet currently employs a simple Free Space Loss

31

Page 41: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Characteristic Default valuetransmission frequency 2.45GHztransmission power 14.771dBmcable loss in transmitter 0.0dBmantenna gain in transmitter 2.14dBifree-space-loss equation 92.467 + 20log10(distance) + 20log10(freq)antenna gain in receiver 2.14dBicable loss in receiver 0.0dBminterface card sensitivity -82.0dBmreceiver signal-to-noise ratio 16.0dBmcurrent drawn in sleep state 10.0mAcurrent drawn in idle state 156.0mAcurrent drawn in transmit state 285.0mAcurrent drawn in receive state 185.0mA

Table 8.1: Default characteristics of cnet WLAN links

model (see section 6.1 for details) of signal propagation therefore the same modelis also used for processing the received signal strength. In simulating a wirelessnetwork dispersed over a large open area, the free space loss model is adequatefor our purpose. During the writing of this paper, additional features allowingthe user to specify their own signal propagation model were added to the latestversion of cnet. However, as it was still undergoing testing, we did not use it inour experiments.

One observation from the use of this model in cnet simulations is that eventhough the transmission range may be restricted to approximately 170metres,the carrier sense range can be as much as 1000 metres away. This differencebetween carrier sense and transmission range will cause nodes to back off theirtransmission unnecessarily in many cases. We expect to see our results reflect thisin the form of consistently lower throughput compared to experiments conductedwith other network simulators.

8.5 MAC layer and Routing Protocol Setup

In this section, we describe the specific setup of our 802.11 MAC layer and DSR-RSM used in the simulations. As cnet only provides an Application and Physicallayer, we implemented our own MAC layer and modified the existing DSR routingprotocol for our DSR-RSM implementation.

32

Page 42: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

MAC layer type 802.11bFrame queue size 64Topology information table size 10CWmin 3CWmax 63Slot time 20µsecsSIFS 10µsecsDIFS 50µsecsFragmentation NILEncryption NILLink bandwidth 11Mbps

Table 8.2: 802.11 MAC layer configuration used in the simulations.

vmax 40ms−1

Dmax 200mWorse case delay Hop 1RRD duration 230ms

Table 8.3: DSR-RSM setup used in the simulations.

Allowing for future development on the MAC layer, we implemented a stan-dard compliant 802.11 MAC layer for our experiments. To this MAC layer, wethen added distance and speed reporting capabilities which the routing protocolscan use to obtain topology information. The contents of this topology informa-tion table is purged once every 60 seconds. Other MAC layer configurations usedin our simulations are listed in Table 8.2.

In order to fully evaluate the use of RSM in route discovery, we disabled routecaching in DSR-RSM such that new route discoveries are initiated when existingroutes become invalid. This also allows us to compare the overheads betweena route caching and a non route caching algorithm in dynamic networks. InTable 8.3, we show the Route Reply Delay and some RSM specific configurationsused in the simulations.

33

Page 43: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 9

DSR-RSM Performance Analysis

In this chapter, we present the results of enhancing DSR with RSM. The resultsfrom our simulations were collated into several statistics files and subsequentdata analysis was undertaken by analysing these files. Before simulating mobilenetworks, we also tested DSR and DSR-RSM in a static network to verify our newprotocol and establish some benchmarks. This result is detailed in Section 9.1below. For the simulation of mobile networks, the results are divided into threemain parts.

The first part compares the throughput achieved by DSR and DSR-RSM. Weused the ratio of packets delivered against packets generated by the Applicationlayer as the throughput measure. This is important as any claims to improvingthe routing algorithm must be backed up by an increase in performance in thisarea.

The second part deals with the efficiency of the protocols by analysing theratio of control messages generated by the two protocols. We compare the ratio ofcontrol packets generated per data packet delivered of each protocols to determinetheir efficiency. The efficiency of the protocol significantly affects the energyexpenditure of nodes in a MANET [11].

The last part shows the difference in route failure rate between the two proto-cols. We compare the number of Route Error packets generated per data packetdelivered in order to prove that DSR-RSM improves the longevity of data com-munication routes.

9.1 Protocol Verification

Before testing the protocols on mobile networks, we analyse DSR-RSM and DSRin a completely static network and establish some performance benchmarks. Insuch an environment, we do not expect DSR-RSM and DSR to display signif-icant performance difference. In fact, due to the Route Reply Delay used in

34

Page 44: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 9.1: Performance comparison of DSR and DSR-RSM in a static network.

35

Page 45: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 9.2: Ratio of Packets delivered with varying number of sources and speeds.

DSR-RSM slowing down route discoveries, we expect DSR-RSM to have slightlylower throughput than DSR. We used random topologies for both simulationsand carried out the experiment for a total of 10 times. The plots in FIgure 9.1shows the average values obtained. Surprisingly, the results we obtained in Fig-ure 9.1 showed DSR-RSM to have a slight edge over DSR under light and optimaltraffic loads. Under heavy traffic loads, DSR reversed the trend and obtained aslight edge over DSR-RSM. However, the results indicate that the difference inperformance between the two protocols is minimal and insignificant in a staticnetwork. We believe that the performance difference is due more to the initialdistribution of nodes in the simulated area then to the workings of the protocols.In terms of overheads and route failures, the results also do not show any cleardominance of one protocol over the other.

36

Page 46: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 9.3: Combined DSR and DSR-RSM packet delivered ratio comparison.

37

Page 47: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

9.2 Performance

The result of our throughput experiment is depicted in Figure 9.2. The plotsshow the ratio of packets delivered successfully against an increasing number ofactively transmitting nodes over networks of different maximum node speeds.

During the design of our protocol, one of our main concerns was balancing thelatency caused by the slower route discovery in DSR-RSM and the performancegain from longer-lived routes. Our belief was that the increase in throughputfrom the longer-lived routes would more than compensate for any latency in theroute discovery process. From the results shown in Figure 9.2, we observe thatDSR-RSM consistently outperforms DSR regardless of the number of activelytransmitting nodes and the maximum node speeds.

As the traffic load in the network increases, both the throughput for DSR-RSM and DSR decreases accordingly due to packet congestion, collision and MAClayer backoffs. In networks where nodes move at higher speeds, the performancegain for DSR-RSM is significantly higher than that of a slower moving network.For example, at the maximum node speed of 20ms−1 (see Figure 9.2 (d)), DSR-RSM showed an average performance improvement of 19% against DSR. Thisindicates that more stable routes were being selected and used, therefore leadingto the improved performance. At the other extreme, at the maximum node speedof 5ms−1 (see Figure 9.2 (a)), we achieved a 4% average performance improvementfrom DSR-RSM. This result is expected since at higher mobility, the probabilitythat the shortest hop route being the best route would be lower, therefore select-ing an alternate route will usually result in performance improvement. On theother hand, in a slow moving network, the probability of a shorter route failingis lower thus the performance difference between DSR and DSR-RSM would alsobe less significant. The extreme case would be a completely static network whichwe tested in the previous section. In Figure 9.3, we show the packet deliveredratio for the different node speeds in a single chart for comparison.

9.3 Routing Overhead

After comparing the throughput, we analyse the network overheads for the twoprotocols. This is important as routing protocol overheads contribute signifi-cantly to the power consumption and longevity of the MANET [4]. As a criteriafor improving the performance of a routing protocol, our aim was to ensure thatDSR-RSM does not generate more network overheads than DSR. In order tocompare the amount of overhead imposed on the network by the two routing

38

Page 48: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 9.4: Ratio of control packets generated and data packets delivered withvarying number of sources and speeds.

39

Page 49: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 9.5: Combined chart showing the ratio of control packets generated anddata packets delivered with varying number of sources and speeds.

40

Page 50: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

protocols, we use the ratio of control packets generated against data packetstransmitted. We compare the overhead ratio of networks with various maximumspeeds and traffic loads in Figure 9.4 and Figure 9.5.

One of the main reason why longer lasting routes are better than short-livedones is the reduced control overhead resulting from fewer route discovery pro-cesses. This reduction in route discovery processes comes as a direct result ofreduced route failures in longer lasting routes. As route selection in DSR-RSMis based on the stability of the routes, the average DSR-RSM source route willhave better longevity than the average DSR source route. Therefore, we expectto see fewer routing packets transmitted by DSR-RSM in our results.

From observation, we see that there is a large increase in overhead for DSRwhen the traffic load is over the network capacity (30 transmitting nodes) whilethe overhead increase for DSR-RSM is relatively linear to the increase in trafficload. This can be explained by the longer lifetime of the routes chosen by DSR-RSM leading to fewer route discovery processes being generated. With fewerRREQ packets generated, a DSR-RSM network encounters less congestion andfewer dropped packets in a heavy traffic load environment compared to DSR.In DSR, the extra control packets generated for reporting route failures anddiscovering new routes results in a larger number of packets lost or dropped whentraffic load is heavy. The dropping of these control packets will be interpreted asroute failures by the routing protocol and result in more control packets beinggenerated by the communicating nodes.

9.4 Route Failure Rate

Other than using the packet delivered ratio to compare the performance of DSR-RSM and DSR, we are also interested to know if our protocol is able to reduce thenumber of route failures and RE packets generated in a network by choosing morestable routes. The number of RE packets generated will be directly proportionalto the number of overheads in the network, therefore the results in Figure 9.6and Figure 9.4 are closely related.

As shown in Figure 9.6 and Figure 9.7, we collate the number of RE packetsgenerated by the routing protocol and obtain its ratio against the number ofsuccessfully transmitted data packets. This ratio gives us an indication of howoften a route breaks during simulation and the average longevity of the routesunder each of the two protocols. The higher the RE ratio, the more often routefailure occurs in the network. From the results shown in Figure 9.6, we observethat, once again, DSR-RSM outperforms DSR consistently in all the simulations.

41

Page 51: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 9.6: Ratio of RE packets generated and data packets delivered with vary-ing number of sources and speeds.

42

Page 52: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure 9.7: Combined chart showing the ratio of RE packets generated and datapackets delivered with varying number of sources and speeds.

43

Page 53: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

This can be explained by several reasons. Firstly, as DSR-RSM chooses routesthat tend to more static, this selection reduces the number of route failures androute error packets generated. Secondly, in networks with higher mobility, DSRtends to accumulate more stale routes in its cache which eventually lead to moreroute failures and route error packets generated. As the route discovery processis actively used in DSR-RSM, it does not suffer from this stale route cachingproblem. To a certain extent, this problem demonstrates that a topology awareroute selection is superior to plain route caching as a route failure reductionstrategy.

44

Page 54: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

CHAPTER 10

Discussion & Conclusions

In ad-hoc networks where the topology is dynamic and unpredictable due to amobile node’s mobility, routing protocols that do not consider these changingtopology information in their routing decisions are incomplete. Relying on theshortest hop count as its route selection metric, DSR has the problem of a fragileroute. Consequently, the selected routes tend to have short lifespans which leadsto additional network overheads when new routes need to be established.

In order to overcome this problem, we propose a new route selection metric,RSM, to improve the route selection of routing protocols. Without relying onspecialised hardware and node capabilities such as GPS and directional antennas,we use the received signal strength from transmitted packets to obtain the relativenode distance and speed between node-pairs in a route. We use this informationto compute our metric using a distributed algorithm. This metric is then usedto reflect the stability of a route and is implemented in our DSR-RSM routingprotocol. We simulate our protocol and the standard DSR in a wide range ofnetwork conditions to compare their performance difference.

From our experiments with DSR-RSM and DSR, we confirmed that in highmobility networks, significant performance improvements can achieved with atopology aware routing algorithm. On average, we achieved an increase in through-put of 19% in high mobility networks and a 4% performance improvement in lowermobility networks. In these mobile networks, choosing more static routes allowedcommunications to last much longer than using the shortest routes. As a result,the number of route failures and the corresponding network overheads were alsosubstantially reduced. In terms of reliability, we have proven that RSM signifi-cantly improves the routing efficiency of the network. With the achieved results,we have also shown that reliable topology information for routing protocols canbe obtained and used without having specialised hardware and equipment. Thisachievement will allow topology information to be more easily incorporated intocurrent routing protocols and networks.

Future extension to this work can be taken in many directions. Other than the

45

Page 55: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

reliability of the network, we also want to examine the effects of a topology-awarerouting protocol on its energy efficiency. This is important as the lifetime of awireless network operating on batteries is crucial to its usefulness. The energyconsumption pattern in the network may also reveal if a topology aware protocolis practical in real life. For example, if a protocol keeps using certain nodes in thenetwork because they are more static, it may lead to an isolated network afterthese nodes deplete their power supply and go offline.

Expanding the use of a route stability metric beyond route discovery mayalso yield interesting results. If the stability of each link in a route is evaluatedduring data communications, preemptive route failure predictions may be issuedby nodes sensing an imminent link failure. This will improve protocol perfor-mance on two fronts: first by reducing the latency of re-establishing a valid routeto the destination and secondly, improving the energy efficiency by reducing thenumber of failed transmission.

46

Page 56: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

APPENDIX A

Original Honours Proposal

Title: Topology discovery and dissemination algorithms for Mobile Ad-hocNetworks

Author: Tyng Wey, Leong

Supervisor: Dr Chris McDonald

A.1 Background

A mobile ad-hoc network (MANET) is a collection of wireless mobile hosts form-ing a temporary network without the aid of any centralised administration orstandard support services. In such an environment, it is often necessary for onemobile host to enlist the aid of others in forwarding a packet to its destination, dueto the limited propagation range of each mobile hosts wireless transmissions[1].

This technology has existed for over 20 years and has been commerciallyavailable for the past 10 years[2]. It is the increasing popularity of the Internet,the use of laptops and mobile computing devices (palmtop computers, mobilephones, Bluetooth devices.) that led to the gain in popularity of MANETs. Theability to quickly establish connectivity and operate without the need of any wiredinfrastructure (e.g. base stations, routers, etc.) makes them promising candidatesfor military, disaster relief, geological fieldwork, law enforcement applications,commercial and personal use like conferencing and personal area networks.

The growing interest in sensor network applications has also created a need forprotocols and algorithms for large-scale self-organising ad-hoc networks, consist-ing of hundreds or thousands of nodes[3]. However, the potential widespread useof MANETs is currently restricted by their energy source, the battery. Until aninexpensive, continuous source of power can be readily obtained to power thesemobile nodes, we have to rely on network protocols, technological improvementsand prudent usage to prolong the lifetime of the network and their nodes.

47

Page 57: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Routing protocols such as DSDV, AODV, DSR, TORA and others attemptto be energy efficient by reducing the number of non-data packets in the networkand the size of these packets. At the same time, some of the protocols frequentlyupdate their routing tables in the attempt to maintain the most updated andaccurate routes to other nodes. However, Filipovic[4] has shown that much ofthe energy expended by MANETs is in fact wasted when these overhead packetsare resent after they were dropped by intermediate nodes due to poor packetsending and dropping selection management in many of the protocols.

We believe that if thorough and updated knowledge about the MANET topol-ogy could be acquired by the nodes (although initially costly), it can help makeoptimal routing and packet management decisions and reduce energy consump-tion in the network.

Possible network topology attributes that may assist in the calculation andselection of better routes include network density, shape and size. It is in thisfield that we are particularly interested, as it can potentially benefit existingMANET protocol performance. Some methods of location and topology discoveryinclude the usage of GPS, mobile Agents, the Location-Aided Routing (LAR)[5]protocol, Distance Routing Effect Algorithm for Mobility (DREAM)[6], etc. Asthe DREAM paper available to me was not published in English, I was unableto gather much information about this protocol but it will remain as a topic ofinterest in my research.

A.2 Aim

The aim of this project is to explore the various common topology gatheringand dissemination methods used in Mobile Ad-hoc Networks and devise newmethods for implementation to investigate their effects on the performance andenergy state of the nodes in a network.

Areas that I will attempt to cover will include the types of topology discoveryand dissemination methods, how the location and route discovery is carried out,the associated hardware requirements for the methods, how the gathered infor-mation is disseminated to other nodes, their associated costs and performanceenhancement to the network if any.

Some of the problems likely to be experienced include how the nodes shouldcommunicate with each other in a heterogeneous mobile network where nodeshave dissimilar hardware, software specifications and configurations, what kindof system information should be made available to data collecting agents and thepossible security risks posed by topology gathering programs to the individual

48

Page 58: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

nodes.

I believe that in solving some of the above listed problems, it would makethe implementation of topology gathering/dissemination into routing protocolsmore popular and effective, hence helping to achieve the ultimate goal of creatingenergy/cost efficient MANETs.

A.3 Method

The initial phase of the project will focus on researching and reviewing some ofthe available topology discovery/dissemination methods published before movingto the phase of devising and implementing our own method.

Of the research we have done so far, methods of exporting local node capa-bilities to global capabilities[7] and location information gathering/disseminationusing quorum[8] are of particular interest to us. The former describes how nodeswith different capabilities might be able to infer their location information andsubsequently propagate this information for the use of other nodes in the networkwhile the latter attempts to facilitate the tracking and propagation of node loca-tion information using selections of nodes(quorums) in the network. The mergingof these two methods is a possible candidate for our implementation.

The designing and coding phase would involve implementing an existing net-work protocol with our devised topology discovery/dissemination method incor-porated in the chosen network simulator and the testing phase will compare theperformance of our design with the original routing protocol. Some performancevariables that we would be interested in are the total throughput of the network,energy consumption by the nodes during the simulation, eventual energy stateof nodes at the end of the simulation and the number of live nodes at the end ofthe simulation and the experiments conducted will be similar to the performanceexperiments conducted by Filipovic[4].

If time permits, we would implement other existing protocols with our devisedmethod and compare the consistency of the performance gain from our method.

The following is a tentative schedule for the project which is subjected torevision:

49

Page 59: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Week Due Date Task DescriptionWeek 4 08/12/04 Research topology gathering methodsWeek 5 08/16/04 Project proposal presented to research groupWeek 7 09/01/04 Research on topology dissemination techniquesWeek 8 09/10/04 Selection and familiarisation on network simulatorsWeek 10 09/24/04 Review research and update project proposalWeek 12 10/21/04 Final Proposal on Project

End of Year Exams and Break(Continue coding and testing of topology software)

Week 2 03/10/05 Implement and run simulationsWeek 4 03/31/05 Analyse results and repeat testsWeek 5 04/06/05 Compile test results into draft dissertationWeek 7 04/21/05 Draft dissertation due to project supervisorWeek 9 05/07/05 Revise dissertation from collected draft,

finalise seminar requirementsWeek 10 05/12/05 Seminar title and abstract due to 4th Year CoordinatorWeek 12 05/26/05 Submit Final dissertation to 4th Year CoordinatorWeek 13 05/31/05 Seminar presented to seminar marking panelWeek 13 06/02/05 Poster submitted and End of Project

A.4 Software and Hardware Requirements

For this project, I will require the use of a standard desktop PC available in thelaboratories running on Linux with sufficient hard disk space to store the outputfrom simulations programs. Based on experience from other similar projectsattempted in the University, an estimated disk space of 2GB would be requiredto store the generated data.

The simulation software to be used is currently not confirmed and work mightbe done on cnet, ns-2 or other network simulators. The software packages willbe sourced from either the University or from other open source avenues. Allrequired software will need to be installed on the Linux PC.

50

Page 60: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

APPENDIX B

DSR packet formats

Figure B.1: DSR packet format.

51

Page 61: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

APPENDIX C

Basic IEEE 802.11 Frame formats

Figure C.1: General 802.11 MAC frame format.

Figure C.2: RTS frame format.

Figure C.3: CTS frame format.

52

Page 62: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Figure C.4: DATA frame format.

Figure C.5: ACK frame format.

53

Page 63: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

APPENDIX D

Example of a cnet topology file

Figure D.1: Cnet topology file format.

54

Page 64: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

APPENDIX E

Example of a cnet statistics file

Figure E.1: Example Cnet output statistics file.

55

Page 65: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

Bibliography

[1] Abrams, M. UC San Diego researchers win Pentagonfunding to develop ad-hoc wireless networking technologyfor battlefield environments. UCSD News (March 2004).http://ucsdnews.ucsd.edu/newsrel/science/20040322Zeidler.asp.

[2] Bahl, P., and Padmanabhan, V. N. A software system for locatingmobile users: Design, evaluation, and lessons. MOBIHOC 2001 (2001),106–116.

[3] Bardwell, J. Converting signal strength percentage to dBm values.Tech. rep., November 2002. www.wildpackets.com/elements/ whitepa-pers/Converting Signal Strength.pdf.

[4] Broch, J., Maltz, D. A., Johnson, D. B., Hu, Y.-C., andJetcheva, J. A performance comparison of multi-hop wireless ad hocnetwork routing protocols. In ACM/IEEE International Conference on Mo-bile Computing and Networking (MOBICOM) (Dallas, Texas, USA, October1998), vol. 4, ACM Press, pp. 85–97.

[5] Bulusu, N., Heidemann, J., and Estrin, D. Gps-less low cost out-door localization for very small devices. IEEE Personal CommunicationsMagazine 7, 5 (October 2000), 28–34.

[6] Bulusu, N., Heidemann, J., and Estrin, D. Adaptive beacon place-ment. In Twenty-first International Conference on Distributed ComputingSystems (ICDCS-21) (April 2001), University of California, Los Angeles,IEEE Computer Society.

[7] Caffery, J., and Stuber, G. Radio location in urban CDMA microcells.In Sixth IEEE International Symposium on Personal, Indoor and MobileRadio Communications (Toronto, Canada, September 1995), vol. 2, pp. 858–862. citeseer.ist.psu.edu/caffery95radio.html.

[8] Camara, D., and Loureiro, A. A. F. A novel routing algorithm for adhoc networks. In HICSS ’00: Proceedings of the 33rd Hawaii InternationalConference on System Sciences-Volume 8 (Washington, DC, USA, 2000),IEEE Computer Society, p. 8022.

56

Page 66: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

[9] Chayabejara, A., Zabir, S. M. S., and Shiratori, N. An enhance-ment of the IEEE 802.11 MAC for multihop ad hoc networks. In VehicularTechnology Conference, 2003. VTC 2003-Fall. 2003 IEEE 58th (October2003), vol. 5, IEEE, pp. 3020 – 3024.

[10] Fall, K., and Varadhan, K. The ns Manual. University of California,Berkley, February 2002. The VINT Project.

[11] Feeney, L. M. An energy-consumption model for performance analysisof routing protocols for mobile ad hoc networks. ACM Journal of MobileNetworks and Applications 6, 3 (June 2001), 239–249.

[12] Filipovic, A. Energy efficiency of ad hoc routing protocols, November2002. School of Computer Science & Software Engineering, The Universityof Western Australia.

[13] Heidorn, K. C. “Watching” Thunder. Online (May 1999).http://www.islandnet.com/ see/weather/eyes/thunder3.htm.

[14] Johnson, D. B., and Maltz, D. A. Dynamic source routing in ad hocwireless networks. In Mobile Computing, T. Imielinski and H. Korth, Eds.,vol. 353. Kluwer Academic Publishers, 1996, ch. 5, pp. 153–181.

[15] Johnson, D. B., Maltz, D. A., and Hu, Y.-C. The dy-namic source routing protocol for mobile ad hoc networks (dsr).http://www.ietf.org/internet-drafts/draft-ietf-manet-dsr-10.txt.

[16] Kim, W.-I., Kwon, D.-H., and Suh, Y.-J. A reliable route selectionalgorithm using global positioning systems in ad-hoc networks. In IEEE In-ternational Conference on Communications (ICC) (Helsinki, Finland, 2001),vol. 10, pp. 3191–3195.

[17] Ko, Y.-B., and Vaidya, N. H. Location-aided routing (lar) in mobilead hoc networks. In Mobile Computing and Networking (Dallas, TX, 1998),pp. 66–75.

[18] Lough, D. L., Blankenship, T. K., and Krizman, K. J. A shorttutorial on wireless LANs and IEEE 802.11, Summer 1997. The BradleyDepartment of Electrical and Computer Engineering, Virginia PolytechnicInstitute and State University, Blacksburg, Virginia.

[19] McDonald, C. A network specification language and execution environ-ment for undergraduate teaching. ACM Computer Science Education Sym-posium (March 1991), 25–34.

57

Page 67: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

[20] Meguerdichian, S., Slijepcevic, S., Karayan, V., and Potkonjak,M. Localized algorithms in wireless ad-hoc networks: Location discovery andsensor exposure. MOBIHOC 2001 (2001), 106–116.

[21] Nath, B., and Niculescu, D. Position and orientation in ad hoc net-works. Ad Hoc Networks, 2 (2004), 133–151.

[22] Niculescu, D., and Nath, B. Ad hoc positioning system (aps). InGLOBECOM (San Antonio, November 2001), pp. 2926–2931.

[23] Park, V. D., and Corson, M. S. A highly adaptive distributed routingalgorithm for mobile wireless networks. In IEEE Computer and Communi-cations Societies (INFOCOM) (Kobe, Japan, April 1997), vol. 3, pp. 1405–1413.

[24] Perkins, C. E., and Bhagwat, P. Highly dynamic destination-sequenced distance-vector routing (DSDV) for mobile computers. In ACMSIGCOMM Conference on Communications Architectures, Protocols andApplications (London, UK, August 1994), vol. 24, ACM Press, pp. 234–244.

[25] Perkins, C. E., and Royer, E. M. Ad-hoc on-demand distance vectorrouting. In IEEE Workshop on Mobile Computing Systems and Applications(WMCSA) (New Orleans, LA, USA, February 1999), vol. 3, pp. 90–100.

[26] Priyantha, N. B., Chakraborty, A., and Balakrishnan, H. Thecricket location-support system. In Mobile Computing and Networking(2000), pp. 32–43.

[27] Quintero, A., Pierre, S., and Macabeo, B. A routing protocol basedon node density for ad hoc networks. Ad Hoc Networks Ad Hoc Networks2(2004) (2004), 335–349.

[28] Roush, W. Intel’s centrino solution. An MIT En-terprise. Technology Review.com(online) (February 2005).http://www.technologyreview.com/articles/05/02/issue/brief intel.asp.

[29] Royer, E., Melliar-Smith, P., and Moser, L. An analysis of theoptimal node density for ad hoc mobile networks. IEEE ICCC 2001 3 (2001),857–861.

[30] Shah, S. H., and Nahrstedt, K. Predictive location-based qos routing inmobile ad hoc networks. IEEE International Conference on Communications2 (2002), 1022–1027.

58

Page 68: A Topology Aware Routing Algorithm for Mobile Ad-hoc Networks · 2017-11-17 · network topology information and propose that a topology aware routing protocol will have better performance

[31] Tanenbaum, A. S. Computer Networks, 3rd ed. Prentice Hall, Inc, UpperSaddle River, New Jersey, 1996. Printed in USA.

[32] Vaidya, N. H., Lee, H., and Welch, J. L. Location tracking usingquorums in mobile ad hoc networks. Ad Hoc Networks 1 (2003), 371–381.

[33] Yoon, J., Liu, M., and Noble, B. Random waypoint considered harm-ful. In Proceedings of Infocom ’03 (San Fransisco, California, USA, April2003), pp. 1312–1321.

59


Recommended