ad hoc networks technologies and protocols

Markov chain model for the channel around a node Markov chain model for a node Illustration of “hidden” area influence Throughput comparison Network Model Illustration Example of collisi

AD HOC NETWORKS

Technologies and Protocols

Print ISBN: 0-387-22689-3

Print © 2005 Springer Science + Business Media, Inc.

All rights reserved

No part of this eBook may be reproduced or transmitted in any form or by any means, electronic, mechanical, recording, or otherwise, without written consent from the Publisher

Created in the United States of America

Boston

©200 5 Springer Science + Business Media, Inc.

Visit Springer's eBookstore at: http://www.ebooks.kluweronline.com

and the Springer Global Website Online at: http://www.springeronline.com

1 3 4 5 6 7

1.2.1

1.2.2

The BattlefieldThe Urban and Campus Grids: a case for opportunistic adhoc networking

1.3 Design Challenges

10 12 12 13 15 17 21 22

1.3.1

1.3.2

Cross Layer InteractionMobility and Scaling1.4

1.5

1.6

Evaluating Ad Hoc Network Protocols - the Case for a Testbed

Overview of the Chapters in this Book

Conclusions

References

2

J J Garcia-Luna-Aceves and Yu Wang

2.1 Performance of collision avoidance protocols

2.1.1

2.1.2

2.1.3

Approximate AnalysisNumerical ResultsSimulation Results

25 26 35 39 44 46 48 54 58

Conclusion

References 60 3

Routing in Mobile Ad Hoc Networks

Mahesh K Marina and Samir R Das

3.2.1 Efficient Flooding Techniques

References

4

Multicasting in Ad Hoc Networks

Prasant Mohapatra‚ Jian Li‚ and Chao Gui

4.7 Conclusions and Future Directions

References

119 119 5

Transport Layer Protocols in Ad Hoc Networks

Karthikeyan Sundaresan‚ Seung-Jong Park‚ Raghupathy Sivakumar

Loss-based Congestion Indication Linear Increase Multiplicative Decrease Dependence on ACKs and Retransmission Timeouts Absolute Impact of Losses

TCP-aware Cross-layered Solutions

Ad-hoc Transport Protocol

Summary

References

6

Energy Conservation

Robin Kravets and Cigdem Sengul

6.1 Energy Consumption in Ad Hoc Networks

Energy Conservation Approaches6.2 Communication-Time Energy Conservation

6.2.1

6.2.2

6.2.3

Power ControlTopology ControlEnergy-Aware Routing6.3 Idle-time Energy Conservation

Use of Smart Antennas in Ad Hoc Networks

Prashant Krishnamurthy and Srikanth Krishnamurthy

7.3.1 The IEEE 802.11 MAC Protocol in Brief

124 125 126 127 129 131 132 132 134 135 137 140 146 150 151 153 155 155 157 157 158 158 158 161 172 176 176 186 190 190 197 197 198 199 200 201 201 202

7.3.2 Directional Transmissions and the IEEE 802.11 MAC

QoS Support using DCF based Service Differentiation8.5 QoS Routing

QoS at other Networking Layers

Inter-Layer Design Approaches

8.7.1

8.7.2

INSIGNIACross-Layer Design for Data Accessibility8.8 Conclusion

References

9

Security in Mobile Ad-Hoc Networks

Yongguang Zhang‚ Wenke Lee

9.1 Vulnerabilities of Mobile Ad Hoc Networks

203 204 206 208 210 213 214 215 217 217 218 221 222 222 223 224 226 226 229 229 232 232 233 233 234 234 234 236 236 238 239 240 241 242 242 243 243 244 246 249 249

9.4 Intrusion Detection Techniques

9.5 Conclusion

References

Index

251 253 253 254 255 256 256 259 261 264 265 269

1.1 Internet in the sky architecture designed as part of the

ONR supported Minuteman project at UCLA

An example opportunistic ad hoc network

An example of LANMAR implementation

Markov chain model for the channel around a node

Markov chain model for a node

Illustration of “hidden” area

influence

Throughput comparison

Network Model Illustration

Example of collisions with data packets in the IEEE

802.11 MAC Protocol

Performance comparison of IEEE 802.11 with analytical resultsPerformance comparison of IEEE 802.11 with adjusted

analytical results

A simple network: node graph and flow contention graph

Network configurations with two competing flows

The adaptive backoff algorithm

The criteria to choose sender-initiated or receiver-initiated

handshake

Special tag processing for two-way flows

Multipoint Relay concept Two dotted circles around the

source S represent its logical 1-hop and 2-hop

neighbor-hood respectively

Comparison of search regions using expanding ring search

and query localization Dotted circles in each figure

in-dicate the search regions

Restricted directional flooding in LAR and DREAM

811142831323637394042454647515455

68

7682

3.4 Illustration of greedy forwarding failure and perimeter

routing in GPSR In this figure‚ S is the source and D is

the destination By greedy forwarding‚ S sends the packet

to node X But all neighbors of X are farther to D than

itself‚ so greedy forwarding fails at X X then switches to

perimeter mode and routes the packet along the perimeter

until it reaches Y (closer to D than itself) From Y‚ greedy

forwarding is used again until the packet reaches D For

simplicity‚ in this example we have assumed that actual

network graph is planar

4.1

4.2

4.3

A mobile ad hoc network

Multicast join operation of MAODV

Traffic flow from h (a) In a CAMP mesh (b) In the

equivalent shared tree

The forwarding group concept

Format of JOIN Query packet

Format of JOIN Reply packet

MCEDAR join procedure

Concept of virtual topology for overlay multicast

An example of area-based method: source node A sends

a broadcast packet‚ and intermediate node B‚ based on its

calculation of additional coverage area (shadowed in the

figure)‚ decides whether to rebroadcast the packet Note

that the additional coverage area of node B is a function

of transmission radius R and nodal distance d When d

= R‚ the maximum additional coverage area is reached‚

Number of route errors

Round-trip Time and Timeouts (1 Flow)

Slow-start and Loss-based Congestion Detection

Route Errors and Impact of Losses

Node power level is less than node and

communi-cation is not possible

Node CTS does not silence node and so node k can

interfere with node since node power level is higher

839298100100101102104107

111126128130133136139145149150160160

6.3 COMPOW computes a common power level of 100mW

for the network‚ which shows that a common power level

is not appropriate for non-homogeneous networks With

CLUSTERPOW‚ the network has three clusters

corre-sponding to 1mW‚ 10mW and 100mW The 100 mW

cluster is the whole network A

10mW-100mW-10mW-1mW route is used for node to reach node

6.4 Enclosure of node Node computes the relay regions

of nodes and Relay Regions 1‚ 2 and 3

(cor-responding to nodes and respectively) specify the

enclosure of node Node maintains links only to nodes

and Nodes and are not contained in node

enclosure‚ and therefore‚ are not its neighbors

6.5 Neighbor discovery in the cone-based algorithm‚

Node adjusts its power level to to reach all

neigh-bors in all cones Although‚ cone III (due to node being

outside the range )‚ node does not

unneces-sarily adjust to

6.6 An example of unidirectional links using LMST There

Neigh-borhood of node is and the neighbors of

node are nodes and The visible Neighborhood of

neighbor Therefore‚ but

6.7

6.8

6.9

Example topologies created by the LMST algorithm

IEEE 802.11 Power-save Mode

Mis-matched Beacon Intervals Node 2 can never hear

the ATIM from node 1

6.10 Alternating odd and even cycles ensure that all nodes can

hear each other’s notification messages

6.11

6.12

Using two notification windows guarantees overlap

Nodes remain awake once every T intervals (T = 4).

However‚ communication is delayed up to T times the

length of the beacon interval

6.13 Nodes remain awake once every intervals

Nodes each choose one row and one column (i.e.‚ node

chooses row and column and node chooses row

and column

6.14 Node chooses row 0 and column 1 and node chooses

row 2 and column 2 Both stay awake during intervals 2

and 9

166

168

170170178180181181

183

183

183163

6.15 Slot allocations determine when each node remains awake.

This figure shows an example slot allocation that

guar-antees at least one overlapping slot between any two nodes 1846.16 Nodes with offset slots are guaranteed to hear each other’s

beacon messages at least once per cycle 1846.17 Example Connected Dominating Set The black nodes

form the CDS Nodes 1-5 are all only one hop away from

6.18 GAF’s virtual grid One node in each grid location

re-mains awake to create a connected dominating set 1897.1 Footprint of (a) An Omni-directional Antenna and (b) A

7.2

7.3

The Cone and Sphere Radiation Pattern

The effect of omni-directional / directional transmissions

of control messages with the 802.11 MAC Protocol

2022047.4

A Scenario to Understand the Schemes Proposed in D-MAC

A Scenario to Understand the Problems with DMAC

The MMAC Protocol

The Circular RTS message

The Multi-Beam Antenna Array

Impact of omni-directional route requests

206209211214215219221222231231235236237

Packet formats for basic 802.11

Point Coordination Function (PCF)

Example of a 802.11 super-frame It relies on TXOPs

(Transmission opportunities) Polled TXOP may be

lo-cated in Contention Period or Contention-Free Period

8.7

8.8

8.9

Multiple backoff of streams with different priorities

Service Differentiation using different DIFS values

CEDAR: Core nodes in a network

8.10

8.11

INSIGNIA QoS Framework

Cross-Layer Design for Data Accessibility

237238239241243244

9.2

An IDS architecture for mobile ad-hoc network: IDS

agents run on monitoring nodes throughout the network

Each MANET node can be the monitoring node for itself

Alternatively‚ a cluster of neighboring nodes can share

one monitoring node

A conceptual model of an IDS agent

257258

IEEE 802.11 protocol configuration parameters

Equivalent configuration parameters for analytical model

Percentage of ACK timeout in BEB scheme

Notations used in the hybrid scheme

An example of two-way flow processing

IEEE 802.11 and TAFA specific configuration parameters

Throughput comparison for the IEEE 802.11‚ the hybrid

scheme and TAFA – two CBR flows (throughput

mea-sured in kbps)

2.8 Throughput comparison for the IEEE 802.11‚ the hybrid

scheme and TAFA – two FTP flows (throughput

mea-sured in kbps)

4.1

6.1

Classification

Transmit‚ receive and sleep mode energy costs for

se-lected wireless cards

6.2 Transmit power levels for selected wireless cards with

power control capabilities

6.3

9.1

9.2

Transmission rates for selected wireless card types

Feature Set I: Topology and route related features

Feature Set II: Traffic related features: dimensions and

57

5897155156156263263264264

Samir R Das is an Associate Professor in the Department of Computer Science

at the Stony Brook University‚ New York His email address is

samir@cs.sunysb.edu

J J Garcia-Luna-Aceves is the Baskin Professor of Computer

Engineer-ing at the University of California‚ Santa Cruz‚ CA His email address isjj@cse.ucsc.edu

Mario Gerla is a Professor in the Computer Science Department at the

Univer-sity of California‚ Los Angeles‚ CA His email address is gerla@cs.ucla.edu

Chao Gui is a doctoral candidate in the Department of Computer Science at the

University of California‚ Davis‚ CA His email address is guic@cs.ucdavis.edu

Robin Kravets is an Assistant Professor in the Department of Computer

Sci-ence at the University of Illinois in Urbana-Champaign Her email address isrhk@cs.uiuc.edu

Prashant Krishnamurthy is an Assistant Professor in the Telecommunications

Program at the University of Pittsburgh‚ PA His email address is

prashant@mail.sis.pitt.edu

Srikanth Krishnamurthy is an Assistant Professor in the Department of

Com-puter Science and Engineering at the University of California‚ Riverside‚ CA.His email address is krish@cs.ucr.edu

Wenke Lee is an Assistant Professor in the College of Computing at the Georgia

Institute of Technology‚ GA His email address is wenke@cc.gatech.edu

Jian Li is a doctoral candidate in the Department of Computer Science at the

University of California‚ Davis‚ CA His email address is lijian@cs.ucdavis.edu

Mahesh K Marina is a doctoral candidate in the Department of Computer

Science at the Stony Brook University‚ New York His email address is hesh @ cs sunysb edu

ma-Prasant Mohapatra is a Professor in the Department of Computer Science‚

University of California‚ Davis‚ CA His email address is prasant@cs.ucdavis.edu

Seung-Jong Park is a doctoral candidate in the School of Electrical and

Com-puter Engineering at the Georgia Institute of Technology‚ GA His email address

is sjpark@ece.gatech.edu

Cigdem Sengul is a graduate student in the Department of Computer Science

at the University of Illinois in Urbana-Champaign Her email address is gul@uiuc.edu

sen-Prasun Sinha is an Assistant Professor in the Department of Computer and

Information Science at Ohio State University‚ OH His email address is

prasun@cis.ohio-state.edu

Raghupathy Sivakumar is an Assistant Professor in the School of Electrical

and Computer Engineering at the Georgia Institute of Technology‚ GA Hisemail address is siva@ece.gatech.edu

Karthikeyan Sundaresan is a doctoral candidate in the School of Electrical

and Computer Engineering at the Georgia Institute of Technology‚ GA Hisemail address is sk@ece.gatech.edu

Yu Wang is a doctoral candidate in the Department of Computer

Engineer-ing at the University of California‚ Santa Cruz‚ CA His email address isywang@cse.ucsc.edu

Yongguang Zhang is a Senior Research Scientist at the HRL Laboratories‚ CA.

His email address is ygz@hrl.com

Wireless mobile networks and devices are becoming increasingly popular asthey provide users access to information and communication anytime and any-where Conventional wireless mobile communications are usually supported

by a wired fixed infrastructure A mobile device would use a single-hop less radio communication to access a base-station that connects it to the wiredinfrastructure In contrast‚ ad hoc networks does not use any fixed infrastruc-ture The nodes in a mobile ad hoc network intercommunicate via single-hopand multi-hop paths in a peer-to-peer fashion Intermediate nodes between apair of communicating nodes act as routers Thus the nodes operate both ashosts as well as routers The nodes in the ad hoc network could be potentiallymobile‚ and so the creation of routing paths is affected by the addition anddeletion of nodes The topology of the network may change randomly‚ rapidly‚and unexpectedly

wire-Ad hoc networks are useful in many application environments and do notneed any infrastructure support Collaborative computing and communications

in smaller areas (building organizations‚ conferences‚ etc.) can be set up using

ad hoc networking technologies Communications in battlefields and disasterrecovery areas are other examples of application environments Similarly com-munications using a network of sensors or using floats over water are otherapplications The increasing use of collaborative applications and wireless de-vices may further add to the need for and the usage of ad hoc networks.During the last few years‚ numerous papers and reports have been published

on various issues on mobile ad hoc networks Several tutorials and surveyreports have been also published on specific aspects of the mobile ad hoc net-works In fact‚ conferences and symposiums that are dedicated to ad hoc net-working have emerged However‚ a “one-stop” resource for overviewing orsummarizing the knowledge and progress on ad hoc networking technologies

is currently unavailable Our co-edited book is primarily motivated by theselines of thought

We have put together a set of interesting chapters that deal with various teresting focal aspects in ad hoc networks The first chapter is a forerunnerfor things to come It primarily motivates the need for ad hoc networks and

in-discusses the evolution of these networks and projects future directions andchallenges The second chapter primarily looks at contention based mediumaccess control in ad hoc networks Most of the research in ad hoc networksassume the use of either the IEEE MAC protocol or variants thereof and thischapter enuniciates by means of both discussion and analyses the nuances ofsuch MAC protocols The third chapter provides an in-depth discussion ofrouting in ad hoc networks Next‚ we provide a discussion of multicasting in

ad hoc networks‚ the issues that arise and the technologies that have emerged

We follow with a discussion of transport layer issues and the protocol designsthus far in the fifth chapter Since ad hoc networks consist of wireless bat-tery operated devices managing energy / power consumption is of paramountimportance The sixth chapter deals exclusively with issues related to powermanagement Lately‚ in order to increase the achievable capacity in ad hocnetworks there has been a lot of interest in the use of directional antennas and

we deliberate various protocols that have emerged for use with such antennas inChapter seven Various issues related to the provision of quality of service andmechanisms for dealing with these issues are presented in the eighth chapter.Finally‚ we have a chapter on security‚ a vital component that will determinethe successful deployment and emergence of ad hoc networks

We wish to express our heartfelt thanks to all of the authors for helping uswith this effort and for participating in this effort We also wish to thank ChaoGui for helping us with compiling and formating of the chapters Our thanks

to Alex Greene from Kluwer for his patience while we were in the process ofcompleting the book and for his support throughout Thanks are also due to ourfamilies for their support and patience during the entire process

Abstract This book covers the major design issues in ad hoc networks It will equip the

researchers in the field with the essential tools to attack the design of a complex ad hoc system, whether in standalone configurations like the 10,000 node battlefield networks or opportunistically connected to the Internet The introductory chapter that follows will provide a definition and characterization of ad hoc networks, followed by an overview of the main applications The design challenges at the various layers of the ad hoc network architecture are then reviewed, with particular emphasis on scalability and mobility An urban grid scenario that captures the complexities of both standalone and “opportunistically extended” ad hoc designs is introduced This scenario poses challenges at all the layers of the

ad hoc protocol stack It is thus the ideal framework to illustrate the impact and the key contributions of the various chapters in this book We conclude with a mention of problems that lie ahead, for further probing by future researchers.

Ad Hoc Network, MANET

Keywords:

1.1 Introduction and Definitions

Internet usage has skyrocketed in the last decade, propelled by web and timedia applications While the predominant way to access the Internet is still

mul-cable or fiber, an increasing number of users now demand mobile, ubiquitous

access whether they are at work, at home or on the move For instance, they

want to compare prices on the web while shopping at the local department store,access Internet “navigation” aids from their car, read e-mail while riding a bus

or hold a project review while at the local coffee shop or in the airport lounge

The concept of wireless, mobile Internet is not new When the packet ing technology, the fabric of the Internet, was introduced with the ARPANET

switch-in 1969, the Department of Defense immediately understood the potential of a

packet switched radio technology to interconnect mobile nodes in the

battle-field The DARPA Packet Radio project which began in the early 70’s helped

establish the notion of ad hoc wireless networking This is a technologythat enables untethered, wireless networking in environments where there is

no wired or cellular infrastructure (eg, battlefield, disaster recovery, etc); or, ifthere is an infrastructure, it is not adequate or cost effective

The term “ad hoc” implies that this network is a network established for

a special, often extemporaneous service customized to applications So, thetypical ad hoc network is set up for a limited period of time The protocolsare tuned to the particular application (e.g., send a video stream across the bat-tlefield; find out if a fire has started in the forest; establish a videoconferenceamong 3 teams engaged in a rescue effort) The application may be mobileand the environment may change dynamically Consequently, the ad hoc proto-cols must self-configure to adjust to environment, traffic and mission changes.What emerges from these characteristics if the vision of an extremely flexible,malleable and yet robust and formidable network architecture An architec-ture that can be used to monitor the habits of birds in their natural habitat, andwhich, in other circumstances, can be structured to launch deadly attacks ontounsuspecting enemies

Because of its mobile, non-infrastructure nature, the ad hoc network poses new design requirements The first is self-configuration (of addresses and

routing) in the face of mobility At the application level, ad hoc network userstypically communicate and collaborate as teams (for example, police, firefight-ers, medical personnel teams in a search and rescue mission).These applications

thus require efficient group communications (multicasting) for both data and real time traffic Moreover, mobility stimulates a host of location based services

non existent in the wired Internet

The complexity of mobile ad hoc network designs has challenged generations

of researchers since the 70’s, Thanks in part to the advances in radio technology,major success have been reported in military as well as civilian applications

on this front (eg, battlefield, disaster recovery, homeland defense, etc) At firstlook, these applications are mutually exclusive with the notion of “infrastructure

networks and the Internet” on which most commercial applications rely This

is in part the reason why the ad hoc network technology has had a hard timetransitioning to commercial scenarios and touching people’s everyday lives.This may soon change, however An emerging concept that will reverse thistrend is the notion of “opportunistic ad hoc networking” An opportunis-

tic ad hoc subnet connects to the Internet via “wireless infrastructure” linkslike 802.11 or 2.5/3G, extending the reach and flexibility of such links This

could be beneficial, for example, in indoor environments to interconnect out

of reach devices; in urban environments to establish public wireless mesheswhich include not only fixed access point but also vehicles and pedestrians,and; in Campus environments to interconnect groups of roaming students andresearchers via the Internet It appear thus that after more than 30 years of in-dependent evolution, ad hoc networking will get a new spin and wired Internetand ad hoc networks will finally come together to produce viable commercialapplications

1.1.1 Wireless Evolution

We begin by studying the evolution of wireless communications systems andnetworks The rapid advances of radio technology in the 70’s stimulated thedevelopment of mobile communications systems that would meet the needs ofyoung professionals on the move First, there came the need to communicatewhile on the move, or away from a fixed phone outlet or internet plug Thecellular phone explosion took the original developers by surprise, but it wasactually a very predictable phenomenon because telephony is by definition

a mobile application In fact, in our daily life we often use the phone justbecause we are on the move, for example, we call friends to obtain directions,

to coordinate our movements/schedules, etc

Next, on the heel of the success of cellular telephony came the interest toconnect to the Internet from mobile terminals The traditional Internet appli-cations are less “mobile” than telephony (most of us would prefer to read oure-mail from the convenience of a home than from the road) However, since

we are spending an increasing number of hours in cars, trains and planes, wewant to fully utilize the travel time with Internet work New emerging Internet

“location based” services (e.g., navigation assistance, store price comparisons,tourist/hotel/parking, etc.) will soon make the wireless connected PDA an in-dispensable companion The second wireless wave (mobile Internet access)

is supported by wireless LAN technology (predominantly, IEEE 802.11) and

by data cellular services (e.g., GPRS, 1xRTT and UMTS); again, a plethora ofstandards exist also for Data Both cellular and wireless networking servicesare supported by an infrastructure and address well established and understood

“commodity” needs of the users (e.g., conferencing, e-mail, web access etc).The third wave in this wireless revolution is the so called “Ad Hoc network-ing” This type of network was borne with goals very different from mobiletelephony and Internet access The primary goal was to set up communicationsfor specialized, customized, extemporaneous applications in areas where there

is no preexisting infrastructure (e.g., jungle explorations, battlefield), or wherethe infrastructure has failed (e.g., earthquake rescue), or it is not adequate forthe current needs (e.g., interconnection of low energy environmental sensors)

With the exception of environment sensor networks (where ad hoc networking

is motivated by lack of convenient, low cost infrastructure), most of the other adhoc applications are “mobile” In fact, they often reflect coordinated mobilitypatterns (e.g., group motion, swarming, etc.) They involve heterogeneous nodetypes (with different form, energy, transmission range and bandwidth factors);and heterogeneous traffic (voice, data and multimedia) They often pose criticaltime constraints (because of the multimedia traffic and the emergency nature

of the applications).In the following section we review the characteristics of adhoc networks in more detail

1.1.2 Ad hoc Networks Characteristics

Mobility: the fact that nodes can be rapidly repositioned and/or move is

the raison d’etre of ad hoc networks Rapid deployment in areas with no frastructure often implies that the users must explore an area and perhaps formteams/swarms that in turn coordinate among themselves to create a taskforce

in-or a mission We can have individual random mobility, group mobility, motionalong preplanned routes, etc The mobility model can have major impact onthe selection of a routing scheme and can thus influence performance

Multihopping: a multihop network is a network where the path from source

to destination traverses several other nodes Ad hoc nets often exhibit multiplehops for obstacle negotiation, spectrum reuse, and energy conservation Battle-field covert operations also favor a sequence of short hops to reduce detection

by the enemy

Self-organization: the ad hoc network must autonomously determine its

own configuration parameters including: addressing, routing, clustering, tion identification, power control, etc In some cases, special nodes (e.g., mobilebackbone nodes) can coordinate their motion and dynamically distribute in thegeographic area to provide coverage of disconnected islands

posi-Energy conservation: most ad hoc nodes (e.g., laptops, PDAs, sensors, etc.)

have limited power supply and no capability to generate their own power (e.g.,solar panels) Energy efficient protocol design (e.g., MAC, routing, resourcediscovery, etc) is critical for longevity of the mission

Scalability: in some applications (e.g., large environmental sensor fabrics,

battlefield deployments, urban vehicle grids, etc) the ad hoc network can grow

to several thousand nodes For wireless “infrastructure” networks scalability issimply handled by a hierarchical construction The limited mobility of infras-tructure networks can also be easily handled using Mobile IP or local handofftechniques In contrast, because of the more extensive mobility and the lack

of fixed references, pure ad hoc networks do not tolerate mobile IP or a fixedhierarchy structure Thus, mobility, jointly with large scale is one of the mostcritical challenges in ad hoc design

Security: the challenges of wireless security are well known - ability of the

intruders to eavesdrop and jam/spoof the channel A lot of the work done ingeneral wireless infrastructure networks extends to the ad hoc domain The

ad hoc networks, however, are even more vulnerable to attacks than the tructure counterparts Both active and passive attacks are possible An activeattacker tends to disrupt operations (say, an impostor posing as a legitimatenode intercepts control and data packets; reintroduces bogus control packets;damages the routing tables beyond repair; unleashes denial of service attacks,etc.) Due to the complexity of the ad hoc network protocols these active at-tacks are by far more difficult to detect/fold in ad hoc than infrastructure nets.Passive attacks are unique of ad hoc nets, and can be even more insidious thanthe active ones The active attacker is eventually discovered and physicallydisabled/eliminated The passive attacker is never discovered by the network.Like a “bug”, it is placed in a sensor field or at a street corner It monitors dataand control traffic patterns and thus infers the motion of rescue teams in anurban environment, the redeployment of troops in the field or the evolution of aparticular mission This information is relayed back to the enemy headquartersvia special communications channels (eg, satellites or UAVs) with low energyand low probability of detection Defense from passive attacks require powerfulnovel encryption techniques coupled with careful network protocol designs

infras-Unmanned, autonomous vehicles: some of the popular ad hoc network

applications require unmanned, robotic components All nodes in a genericnetwork are of course capable of autonomous networking When autonomousmobility is also added, there arise some very interesting opportunities for com-bined networking and motion For example, Unmanned Airborne Vehicles(UAVs) can cooperate in maintaining a large ground ad hoc network intercon-nected in spite of physical obstacles, propagation channel irregularities andenemy jamming Moreover, the UAVs can help meet tight performance con-straints “on demand” by proper positioning and antenna beaming

Connection to the Internet: as earlier discussed, there is merit in extending

the infrastructure wireless networks opportunistically with ad hoc appendices.For instance, the reach of a domestic wireless LAN can be extended as needed(to the garage, the car parked in the street, the neighbor’s home, etc) withportable routers These opportunistic extensions are becoming increasinglyimportant and in fact are the most promising evolution pathway to commercialapplications The integration of ad hoc protocols with infrastructure standards

is thus becoming a hot issue

1.1.3 Wireless Network Taxonomy

From the above, it is clear that ad hoc nets offer challenges (and opportunities)well beyond the reach of infrastructure networks So, where do these nets fit

in the overall wireless network classification? Most researchers will view adhoc wireless networks as a special subset of wireless networks In fact, the adhoc radio technology and most of the MAC technology will be driven by theadvancements in infrastructure wireless networks The unique design features

on ad hoc nets marking a departure from the former are in the network andtransport protocol areas (routing, multicast, ad hoc TCP and streaming, etc).Another important family of ad hoc networks, the sensor networks, can in turn

be viewed as a subset of ad hoc networks There are differences, however At thephysical, MAC and network layers, the major innovations and unique features

of sensor nets (which set them apart from conventional ad hoc networks) are theminiaturization, the embedding in the application contexts and the compliancewith extreme energy constraints At the application layer, the most uniqueand novel feature of sensor nets is undoubtedly the integration of transport andin-network processing of the sensed data

1.2 Ad Hoc Network Applications

Identifying the emerging commercial applications of the ad hoc networktechnology has always been an elusive proposition at best Of the three abovementioned wireless technologies - cellular telephony, wireless Internet and adhoc networks - it is indeed the ad hoc network technology that has been theslowest to materialize, at least in the commercial domain This is quite surpris-ing since the concept of ad hoc wireless networking was born in the early 70’s,just months after the successful deployment of the Arpanet, when the militarydiscover the potential of wireless packet switching Packet radio systems weredeployed much earlier than any cellular and wireless LAN technology Theold folks may still remember that when Bob Metcalf (Xerox Park) came upwith the Ethernet in 1976, the word spread that this was one ingenious way todemonstrate “packet radio” technology on a cable!

Why so slow a progress in the development and deployment of commercial

ad hoc applications? Main reason is that the original applications scenarioswere NOT directed to mass users In fact, until recently, the driving applica-tion was instant deployment in an unfriendly, remote infrastructure-less area

Battlefield, Mars explorations, disaster recovery etc have been an ideal match

for those features Early DARPA packet radio scenarios were consistently turing dismounted soldiers, tanks and ambulances A recent extension of thebattlefield is the homeland security scenario, where unmanned vehicles (UGVsand UAVs) are rapidly deployed in urban areas hostile to man, say, to establishcommunications before sending in the agents and medical emergency person-nel

fea-Recently an important new concept has emerged which may help extend

ad hoc networking to commercial applications, namely, the concept of

oppor-tunistic ad hoc networking This new trend has been in part prompted by thepopularity of wireless telephony and wireless LANs, and the recognition thatthese techniques have their limits The ad hoc network is used “opportunis-tically” to extend a home or Campus network to areas not easily reached bythe above; or, to tie together Internet islands when the infrastructure is cut intopieces - by natural forces or terrorists for examples).

Another important area that has propelled the ad hoc concept is sensor nets.Sensor nets combine transport and processing and amplify the need for lowenergy operation, low form factor and low cost - so, these are specialized adhoc solutions Nevertheless, they represent a very important growing market

In the sequel we elaborate on two applications, the battlefield and the the

urban and Campus grid

sup-of vast urban/suburban areas to track suspects; search and rescue operations inunfriendly areas (e.g., chemical spills, fires, etc), exploration of remote plan-ets, reconnaissance of enemy field in the battle theater, etc In those applica-tions, many different types of Unmanned Vehicles (UVs) will be required, eachequipped with different sensor, video reconnaissance, communications supportand weapon functions A UV team may be homogeneous (e.g., all sensor UVs)

or heterogeneous (i.e., weapon carrying UVs intermixed with reconnaissanceUVs etc) Moreover, some teams may be airborne, other ground, sea and pos-sibly underwater based As the mission evolves, teams are reconfigured andindividual UVs move from one team to another to meet dynamically changingrequirements In fact, missions will be empowered with an increasing degree

of autonomy For instance, multiple UV teams collectively will determine thebest way to sweep a mine field, or the best strategy to eliminate an air defensesystem The successful, distributed management of the mission will requireefficient, reliable, low latency communications within members of each team,across teams and to a manned command post In particular, future naval mis-sions at sea or shore will require effective and intelligent utilization of real-timeinformation and sensory data to assess unpredictable situations, identify andtrack hostile targets, make rapid decisions, and robustly influence, control, andmonitor various aspects of the theater of operation Littoral missions are ex-

pected to be highly dynamic and unpredictable Communication interruptionand delay are likely, and active deception and jamming are anticipated.The Office of Naval Research (ONR) is currently investigating efficient sys-tem solutions to address the above problems ONR envisions unmanned systems

of Intelligent, Autonomous Networked Agents (AINS) to have a profound fluence on future naval operations allowing continuous forward yet unobtrusivepresence and the capability to influence events ashore as required Unmannedvehicles have proven to be valuable in gathering tactical intelligence by surveil-lance of the battlefield For example, UAVs such as Predator and Global Hawkare rapidly becoming integral part of military surveillance and reconnaissanceoperations The goal is to expand the UAV operational capabilities to includenot only surveillance and reconnaissance, but also strike and support mission(e.g., command, control, and communications in the battle space) This newclass of autonomous vehicles is foreseen as being intelligent, collaborative,recoverable, and highly maneuverable in support of future naval operations

in-In a complex and large scale system of unmanned agents, such as designed

to handle a battlefield scenario, a terrorist attack situation or a nuclear disaster,there may be several missions going on simultaneously in the same theater Aparticular mission is “embedded” in a much larger “system of systems” In such

a large scale scenario the wireless, ad hoc communications among the teamsare supported by a global network infrastructure (the “Internet in the sky”).The global network is provisioned independently of the missions themselves,but it can opportunistically use several of the missions’ assets (ground, sea orairborne) to maintain multihop connectivity

Figure 1.1 Internet in the sky architecture designed as part of the ONR supported Minuteman

project at UCLA.

The development of the Internet in the Sky hinges on three essential nologies:

tech-Robust wireless connectivity and dynamic networking of autonomousunmanned vehicles and agents

1

2 Intelligent agents including: mobile codes, distributed databases andlibraries, robots, intelligent routers, control protocols, dynamic services,semantic brokers, message-passing entities

3 Decentralized hierarchical agent-based organization

As Figure 1.1 illustrates, the autonomous agents have varying domains ofresponsibility at different levels of the hierarchy For example, clusters ofUAVs operating at low altitude (1K-20K feet) may perform combat missionswith a focus on target identification, combat support, and close-in weaponsdeployment Mid-altitude clusters (20-50K feet) could execute knowledge ac-quisition, for example, surveillance and reconnaissance missions such as de-tecting objects of interest, performing sensor fusion/integration, coordinatinglow-altitude vehicle deployments, and medium-range weapons support Thehigh altitude cluster(s) (50K-80K feet) provides the connectivity At this layer,the cluster(s) has a wide view of the theater and would be positioned to providemaximum communications coverage and will support high-bandwidth robustconnectivity to command and control elements located over-the-horizon fromthe littoral/targeted areas

We use this example to focus on mission oriented communications and more

precisely on a particular aspect of it, team multicast In team multicast themulticast group does not consist of individual members, rather, of teams Forexample, a team may be a special task force that is part of a search and rescuemission The message then must be broadcast to the various teams that are part

of the multicast group, and, to all UVs within each team For example, a weaponcarrying airborne UV may broadcast an image of the target (say, a poison gasplant) to the reconnaissance and sensor teams in front of the formation, inorder to get a more precise fix on the location of the target The sensor UVteam(s) that has acquired such information will return the precise location Asanother example, suppose N teams with chemical sensors are assessing the

“plume” of a chemical spill from different directions It will be important foreach team to broadcast its findings step by step to the other teams using teammulticast In general, team multicast will be common place in ad hoc networksdesigned to support collective tasks, such as occur in emergency recovery or inthe battlefield

1.2.2 The Urban and Campus Grids: a case for

opportunistic ad hoc networking

In this section we describe two sample applications that illustrate the researchchallenges and the potential power of ad hoc as opportunistic extension of thewireless infrastructure

Two emerging wireless network scenarios that will soon become part of

our daily routines are vehicle communications in an urban environment, and

Campus nomadic networking These environments are ripe for benefitingfrom the technologies discussed in this report Today, cars connect to thecellular system, mostly for telephony services The emerging technologies

however, will soon stimulate an explosion of new applications Within the car, short range wireless communications (e.g., PAN technology) will be used for

monitoring and controlling the vehicle’s mechanical components as well as forconnecting the driver’s headset to the cellular phone Another set of innovative

applications stems from communications with other cars on the road The

potential applications include road safety messages, coordinated navigation,network video games, and other peer-to-peer interactions These network needs

can be efficiently supported by an “opportunistic” multihop wireless network

among cars which spans the urban road grid and which extends to intercityhighways This ad hoc network can alleviate the overload of the fixed wirelessinfrastructures (3G and hotspot networks) It can also offer an emergencybackup in case of massive fixed infrastructure failure (e.g., terrorist attack, act

of war, natural or industrial disaster, etc) The coupling of car multihop network,on-board PAN and cellular wireless infrastructure represents a good example

of hybrid wireless network aimed at cost savings, performance improvements

and enhanced resilience to failures An example of such network is illustrated

in Figure 1.2

In the above application the vehicle is a communications hub where the tensive resources of the fixed radio infrastructure and the highly mobile ad hocradio capabilities meet to provide the necessary services New networking andradio technologies are needed when operations occur in the “extreme” condi-tions, namely, extreme mobility (radio and networking), strict delay attributesfor safety applications (networking and radio), flexible resource managementand reliability (adaptive networks), and extreme throughput (radios) Extremelyflexible radio implementations are needed to realize this goal Moreover, crosslayer adaptation is necessary to explore the tradeoffs between transmission rate,reliability, and error control in these environments and to allow the network togradually adapt as the channel and the application behaviors are better appraisedthrough measurements

ex-Another interesting scenario is the Campus, where the term “Campus” heretakes the more general meaning of a place where people congregate for various

Figure 1.2 An example opportunistic ad hoc network.

cultural and social (possibly group) activities, thus including Amusement Park,Industrial Campus, Shopping Mall, etc On a typical Campus today wirelessLAN access points in shops, hallways, street crossings, etc., enable nomadicaccess to the Internet from various portable devices (e.g., laptops, notebooks,PDAs, etc.) However, not all areas of a Campus or Mall are covered by depart-ment/shop wireless LANs Thus, other wireless media (e.g., GPRS, 1xRTT,3G) may become useful to fill the gaps There is a clear opportunity for multi-ple interfaces or agile radios that can automatically connect to the best availableservice The Campus will also be ideal environment where group networkingwill emerge For example, on a University Campus students will form smallworkgroups to exchange files and to share presentations, results, etc In anAmusement Park groups of young visitors will interconnect to play networkgames, etc Their parents will network to exchange photo shots and videoclips To satisfy this type of close range networking applications, PersonalArea Networks such as Bluetooth and IEEE 802.15 may be brought into thepicture Finally, “opportunistic” ad hoc networking will become a cost-effectivealternative to extend the coverage of access points Again, as already observed

in the vehicular network example, the above “extensions” of the basic tructure network model require exactly the technologies recommended in thisreport, namely: multimode radios, cross layer interaction (to select the bestradio interface) and some form of hybrid networking

infras-These are just simple examples of networked, mobile applications drawnfrom our everyday lives There is a wealth of more sophisticated and demand-ing applications (for example, in the areas of pervasive computing, sensor net-

works, battlefield, civilian preparedness, disaster recovery, etc) that will soon

be enabled and spun off by the new radio and network technologies

1.3 Design Challenges

As mentioned earlier, ad hoc networks pose a host of new research problemswith respect to conventional wireless infrastructure networks This book in factaddresses these challenges and each chapter is focused on a particular designissue at one of the layers of the protocol stack We will provide a review of thechapters shortly First, we wish to report on some design challenges that cutacross the layers and should be kept in mind while reading about specific layersolutions in the other chapters These are: cross layer interaction; mobility,and; scalability

1.3.1 Cross Layer Interaction

Cross Layer Interaction/Optimization is a loaded word today, with many ferent meanings In ad hoc networks it is however a very appropriate way torefer the fact that it is virtually impossible to design a “universal” protocol (rout-ing, MAC, multicast, transport, etc) and expect that it will function correctlyand efficiently in all situations In fact, pre-defined protocol layers a’ la Internetwork reasonably well in wired nets (e.g., routing, addressing, DNS etc workfor large and small.) For example, the physical and MAC layers of the wiredE-net are the uncontested reference for of all Internet designs In contrast, inthe wireless LAN (the closest relative of the E-net), there is convergence not

dif-to one, but dif-to a family of standards, from 802.16 dif-to 15 dif-to 11, each standardaddressing different environments etc Even within the 802.11 family a broadrange of versions have been defined, to address different needs

In ad hoc network design the importance of tuning the network protocols tothe radios and the applications to the network protocols is even more critical,given the extreme range of variability of the systems parameters Clearly, therouting scheme that works best for network of a dozen students roaming theCampus may not be suitable for the urban grid with thousand of cars or thebattlefield with an extreme range of node speeds and capabilities Even moreimportant is the concept that in these cases the MAC, routing and applicationsmust be jointly designed Moreover, as some parameters (eg, radio propaga-tion, hostile interference, traffic demands, etc) may dynamically change, theprotocols must be adaptively tuned Proper tuning requires exchange of infor-mation across layers For example in a MIMO (Multi Input, Multi Output) radiosystem the antenna and MAC parameters and possibly routes are dynamicallyreconfigured based on the state of the channel, which is learned from periodicchannel measurements Thus, interaction between radio channel and protocols

is mandatory to achieve an efficient operating point Video adaptation is another

example of cross layer interaction: the video rate stipulated at session ization cannot be maintained if channel conditions deteriorate The proper rateadjustment requires careful interplay of end to end probing (eg, RTCP) as wellmeasurements from channel and routing.

initial-1.3.2 Mobility and Scaling

Mobility and reconfiguration is what uniquely distinguished ad hoc networksfrom other networks Thus, being able to cope with nodes in motion is anessential requirement Large scale is also common in ad hoc networks, asbattlefield and emergency recovery operations often involve thousands of nodes.The two aspects - mobility and scale - are actually intertwined: anybody canfind a workable ad hoc routing solution, say, for 10 nodes, no matter howfast they move; and anybody can find a workable (albeit inefficient) solution(for routing, addressing, service discovery etc) for a completely static ad hocnetwork with 10,000 of nodes, say (just consider the Internet)! The problemsarise when the 10,000 nodes move at various speeds, in various directions over

a heterogeneous terrain In this case, a fixed routing hierarchy such as in theInternet does not work That is when you have to take out the “big guns” tohandle the problem

Mobility is often viewed as the #1 enemy of the wireless ad hoc networkdesigner However, mobility, if properly characterized, modeled, predictedand taken into account, can be of tremendous help in the design of scaleableprotocols In the sequel we offer a few examples where mobility actually helps

1.3.2.1 An example: Team Communications among Airborne Agents using LANMAR LANMAR is a scalable routing protocol for large, mobile,

“flat” ad hoc wireless networks It has been implemented in the Minuteman work under ONR support [1] LANMAR assumes that the network is groupedinto logical subnets in which the members have a commonality of interests andare likely to move as a “group” (e.g., a team of co-workers at a convention; ortanks in a battalion, or UAVs in an unmanned scouting mission) The logicalgroups are efficiently reflected in the addressing scheme We assume that a twolevel, IP like MANET (Mobile Ad hoc NET) address is used consisting of agroup ID (or subnet ID) and a host ID, i.e <Group ID, Host ID> The group

net-ID tells us which nodes are part of the same group Group assocoation maychange from time to time as a node is reassigned to a different group (e.g taskforce in a military scenario) The Host ID is fixed and typically corresponds

to the hardwired device address Such MANET address uniquely identifies therole (and position) of each node in the network Similar to an IP network, thepacket is routed to the group first, and then to the Host within the group Thechallenge is to “find” the group in a large, mobile network

LANMAR uses the notion of landmarks to keep track of such logical groups.Each logical group has one node serving as “landmark” The landmark adver-tises the route to itself by propagating a Distance Vector, e.g DSDV (Destina-tion Sequences Distance Vector) [3] Further, the LANMAR routing scheme

is always combined with a local routing algorithm, e.g Fisheye State Routing(FSR) [2] FSR is a link state routing algorithm with limited “scope” featurefor local, low overhead operation Namely, FSR knows the routes to all nodeswithin a predefined Fisheye scope (e.g., 3 hops) from the source For nodesoutside of the Fisheye scope, the landmark distance vector must be inspectedfor directions As a result, each node has detailed topology information aboutnodes within its Fisheye scope and knows distance and routing vector (i.e., di-rection) to all landmarks An example of LANMAR routing implementation isshown in Figure 1.3

Figure 1.3 An example of LANMAR implementation.

When a node needs to relay a packet to a destination that is within its Fisheyescope, it obtains accurate routing information from the Fisheye Routing Tables.The packet will be forwarded directly Otherwise, the packet will be routedtowards the landmark corresponding to the destination logical subnet, which

is read from the logical address field in the MANET address Thus, when thepacket arrives within the scope of the destination, it may be routed to it directlywithout ever going through the landmark In summary, the hierarchical LAN-MAR setup does the scalability trick - it reduces routing table size and routeupdate overhead making the scheme practical for a network with practicallyunlimited number of nodes (as long as nodes move in groups of increasingsize).The latter assumption is actually well validated in ad hoc networks asso-ciated with large scale, cooperative operations (eg, battlefield) If nodes aremoving randomly and in a non coordinated fashion (like perhaps the customers

in a shopping mall) other techniques can be used to achieve scalability in arandom motion scenario Along these lines, recently proposed routing and

resource discovery schemes such as “last encounter routing”, and “epidemicdissemination” exploit the fact that, with random motion, the destination that

I want to reach “has been seen” some time ago by some nodes that now havemoved close to me This is a perfect example of symbiosis of mechanical in-formation transport and electronic information relay It allows me to find thedestination through a “motion assisted” search which eliminates the need for acostly (and definitely non scalable) full search

1.4 Evaluating Ad Hoc Network Protocols - the Case for a

WHYNET is a wireless networking testbed that can be used to evaluate theimpact of emerging technologies that are going to shape the nature of wireless,mobile communications in the next decade The eventual impact of this researchtestbed will be to redefine how specific innovations in wireless communicationtechnologies are evaluated in terms of their potential to improve application-level performance as well as how alternative approaches are compared witheach other

WHYNET differs from existing testbeds both in its scope and approach.

Its primary objective is to provide researchers at every layer of the protocolstack, from physical devices to transport protocols, a testbed to evaluate theimpact of their technology on application level performance, using scalableand realistic operational scenarios To achieve this objective, WHYNET willuse a geographically-distributed, hybrid networking testbed that combines therealism of physical testing with the scalability of multi-mode simulations.The primary deliverable from WHYNET will be a set of tools and method-ologies encapsulated in a well-defined evaluation framework, a set of studiesthat demonstrate its suitability for evaluation of emerging network technolo-gies, and a repository of networking scenarios, measurements, and models.The design and development of the testbed will require coordinated efforts

of a multi-disciplinary, multi-institution team of researchers from academia,government, and industry This effort will substantially leverage existing net-