Intelligent Antenna Sharing and User Cooperation in Wireless Networks

Intelligent Antenna Sharing and User Cooperation in Wireless

Networks

Ph.D. Proposal, Program in Communication and Media Sciences

Aggelos Bletsas

Media LaboratoryMassachusetts Institute of Technology

20 Ames St, E15-495, Cambridge, MA 02139, [email protected]

December 9, 2004

Abstract

This thesis will study the issue of user cooperation and antenna sharing to improve wirelesscommunication, network (autonomous) time keeping and network (autonomous) topologyestimation. Practical schemes are proposed, analyzed and implemented using low costembedded radios. This thesis tests the following hypothesis: cooperative communicationprovides for more reliable communication under certain conditions, requires less transmis-sion power and decreases interference to neigboring communications, when compared topoint-to-point, non-cooperative communication. Therefore, cooperation can improve thescaling laws of wireless communication and networking.

Contents

1 Introduction 41.0.1 Four Distinctive Thesis Components . . . . . . . . . . . . . . . . . . 51.0.2 Thesis Proposal Outline . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.1 Antenna Sharing and User Cooperation . . . . . . . . . . . . . . . . . . . . 61.1.1 Three nodes... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91.1.2 More than three nodes... . . . . . . . . . . . . . . . . . . . . . . . . . 111.1.3 Supporting Technologies: Cooperative Timing and Positioning . . . 151.1.4 Progress and Deliverables . . . . . . . . . . . . . . . . . . . . . . . . 171.1.5 Prior Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1.2 Required Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221.3 Timetable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221.4 Short Bio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

1

List of Figures

1.1 Calculating the field of a single transmitter and a single, perfect reflector.It can be seen that depending on the phase of the direct transmission sig-nal and the reflected from the wall signal, the field might be stronger at apoint in space which is at a larger distance from the transmitter, comparedto a point closer to the transmitter. Therefore, the instantaneous receivedpower (proportional of the square of the electromagnetic field) is a functionof the environment with temporal and spatial fluctuations. User cooperativecommunication exploits that phenomenon by using antennas of “other” users. 7

1.2 Received power as a function of distance from the transmitter from actualmeasurements (figure 1.2(a)) or from artificial generation using Rayleigh fad-ing and propagation coefficient v estimated from measurements. The model(right figure) matches reality (left figure). . . . . . . . . . . . . . . . . . . . 8

1.3 Direct transmission, multi-hop transmission, cooperative transmission. . . . 91.4 Performance of cooperative communication compared to non-cooperative com-

munication in left figure (using 8-PSK and various propagation coefficients)and total transmission energy ratio for target Symbol Error Probability(SEP)=10−3 in right figure (using 8-PSK and v = 4), in Rayleigh wirelesschannels. Relay decodes and encodes (digital relay) and it is placed closerto the transmitter, 1/4 the distance between source and destination. We cansee that cooperative communication is more reliable compared to traditionalpoint-to-point communication, leading to higher reliability or transmissionenergy savings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.5 Regions of intermediate node location where it is advantageous to digitallyrelay to an intermediate node, instead of repetitively transmit. M=8 and thedepicted ratio is the ratio of Symbol Error Probability (SEP) of repetitivetransmission vs SEP of user cooperative digital communication. The coop-erative receiver optimally combines direct and relayed copy. Distances arenormalized to the point-to-point distance between transmitter and receiver. 11

1.6 Path diversity as a function of users and instantaneous channel conditions.Method of distributed timers allow for optimal path selection, adaptive tochannel changes, with minimal overhead. . . . . . . . . . . . . . . . . . . . . 11

1.7 Distributed selection of “best” relay path. The intermediate relay nodesoverhear the handshaking between Tx and Rx. Based on the method ofdistributed timers, the relay that has the best signal path from transmitterto relay and relay to receiver is picked with minimal overhead. The receivercombines direct and relayed transmission and displays the received text on astore display. The “best” relay signals with an orange light. The transmitteris fed with weather information from an 802.11-enabled pda. . . . . . . . . 12

2

List of Figures 3

1.8 During the first hop an information symbol is transmitted toward the re-ceiver and overhearing nodes. During the second hop, an amplified versionof the received information plus noise is forwarded to the receiver from theintermediate nodes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.9 Capacity histogram and expected value (dotted marker) of opportunistic re-laying (bottom plot) vs traditional approaches (top plot). Note that oppor-tunistic relaying, where the “best” relay path is used between transmitter andreceiver achieves higher capacity than proposed techniques in the literature(“All-relays case”), where all overhearing relays forward the information. . . 15

1.10 Visual proof of synchrony. A “heartbeat” pattern is synchronized over thenetwork and displayed at the edges. The distributed, server-free approach fornetwork synchronization resembles the decentralized coordination of coloniesof fireflies and inspired this work. . . . . . . . . . . . . . . . . . . . . . . . . 16

Chapter 1

Introduction

In the era of pervasive computing and communications, another thesis on wireless communi-cation and networking might seem obsolete or outdated. However, we have all experiencedbad reception while using our cell phone (also known as poor quality of service), we haveall forgotten to recharge the device during the night and subsequently be unable to useit during the day (energy/battery problems) and we have waited for too long for cellulartechnology to mature until we could start exchanging pictures or videos with our friendsusing our cell phones. Even in that case, data speed (throughput) is significantly less thanthe speed of Wi-Fi wireless technology we have been using in our homes. Finally, we haveall failed to talk to our friends using our cell phones in large venues such as the celebrationof 4th of July in front of Media Lab, when thousands of people alongside Charles rivergather to enjoy the spectacular fireworks but overload the statistically provisioned cellularnetwork.

Could we enhance the quality of service (QoS), increase the data speed (throughput)and/or reduce the required energy (and therefore increase battery life) for any wirelessuser, without overusing common resources such as spectrum or scarce resources like theavailable battery energy? Could we further reduce the transmission power levels of everybase station and therefore minimize public health risks due to electromagnetic radiation?Could we create wireless networking architectures that scale with increasing number of usersand if possible perform better the more users the system have?

Recent developments on multi-antenna transceivers (also known as Multi-Input MultiOutput systems) have shown that for the same bandwidth and power1 resources comparedto traditional single-antenna communication, MIMO systems could increase throughput(multiplexing gain) and/or increase reliability of communication (diversity gain). The ex-tra degree of freedom (apart from time and frequency) comes from space by exploiting thepossible statistical independence between transmitting-receiving antenna pairs. The statis-tics of the multi-antenna wireless channel could provide for independent, parallel spatialcommunication channels at the same carrier frequency and at the same time. In otherwords, MIMO systems exploit space and statistical properties of the wireless channel andtypically need intensive signal processing computation for channel estimation and informa-tion processing. Apart from extensive computing requirements, engineering and physicallimitations preclude the utilization of many antennas at the mobile terminal (typically nomore that two antennas at the cordless phone) and therefore multi-antenna transceivers are

1Energy and power will be used equivalently, since they are different by a multiplying factor, the infor-mation symbol duration.

4

5

typically utilized at the base station side.What happens when multiple antennas belong to different users? Could we exploit

multiple observations from users distributed in space of the same information signal, giventhe broadcast nature of the wireless medium? Could we earn the benefits of traditionalMIMO theory when the antennas belong to different users? In other words, this thesisexplores users in a network as an additional degree of freedom apart from time, frequencyand space, in combination with the intrinsic properties of the wireless channel.

The problem of user cooperation in wireless communication poses exciting challengesgiven that a) the computing (processing) capabilities of cooperating users are limited sincewe will assume that they are mobile with fixed computing capacity and energy consumptionb) cooperation basically means that one user will use her own battery to relay informationdestined for a different user (while the receiver will exploit the direct and the relayedtransmission) and therefore strong incentives should be inherent in the cooperative schemeand c) coordination at the network level among the cooperative nodes should be manifested,which is a radical change given the fact that all existing communication stacks have beenorganized according to point-to-point, non-cooperative communication links that mimicwires.

We are interested in practical schemes that address all the above issues and are ap-plicable using existing RF hardware architectures. To investigate performance, apart fromtheoretical analysis, we have also implemented proposed solutions, using low cost embeddedradios. Cooperation could lead to substantial total (network) transmission power savingsor increased spectral efficiency (in bits per second per hertz) under certain conditions. Thegoal of this thesis is to provide distributed and adaptive cooperation algorithms that canbe applied in practice.

Coordination algorithms required for user cooperative communication will be extensivelystudied in this thesis. However, the notion of cooperation can be extended to other impor-tant problems: if users in a network have strong incentives to cooperate for efficient wirelesscommunication, then they could use cooperative strategies for network time keeping andpositioning. We will show that cooperative communication networks could autonomouslymaintain a global clock (time keeping) and also be used for positioning estimation, us-ing local computation. Therefore the network becomes the timing and positioning systemwith specific accuracy and precision performance, again as a function of number of users.Efficient communication and autonomous timing and positioning are considered the mostimportant problems in future wireless sensor networks.

1.0.1 Four Distinctive Thesis Components

This thesis has four basic components:

1. Algorithms that react to the physics of the environment: Cooperative nodes in thenetwork adapt their behavior to instantaneous wireless channel realizations. Thealgorithms ought to a) scale with increasing number of cooperating users and b)require deterministic time to converge to a solution, well before the channel changes(well below the channel coherence time). Therefore, the network reacts to the physicsof the environment in real time, using measured characteristics of that space.

2. Distributed algorithms with unknown network topology: There is no central point ofcontrol that has global knowledge of the network (for example, there is no knowledge

1.1 Antenna Sharing and User Cooperation 6

on how many nodes cooperate in the network). There is no knowledge regarding thetopology of the network or distances to neighboring nodes.

3. Realistic wireless channel modeling: the channel model used in this work is based onexperimental measurements and excludes simplistic models of free space propagationor propagation within a constant radius sphere. The richness and complexity of wire-less propagation make wireless communication a challenging problem so any attemptto simplify the corresponding models could provide for unrealistic results.

4. Practical solutions for existing hardware: we tried to provide for signal processingtechniques, as well as modulation, transmission and coordination techniques thatcould be applied in existing hardware. Therefore, we don’t make the assumption thatwireless beamforming is feasible, where different wireless transmitters “phase” theirtransmissions so they can add constructively at the receiver. Implementing wirelessphased arrays is still an open area of research. Moreover, we will not assume thatany transmitter could fix its transmission radius to a given distance. Communicationrange is a function of transmission power as well as wireless channel characteristics,which are not user-defined.

The above components differentiate our work from existing approaches in the field, sinceprior research has focused on a subset of the above components. In this work, we devisesystem level solutions that would provide for distributed, infrastructure-free networks wherecommunications are improved with increasing number of cooperating nodes, using simplehardware and intelligent algorithms, applicable in practice.

1.0.2 Thesis Proposal Outline

We begin with the basic idea behind antenna sharing and we continue with analysis of usercooperation in wireless communication where there is one transmitter and receiver pair andone cooperating intermediate node (relay). We then continue with the case of several relays.Each section includes background information, current progress and proposed deliverables.The work will be completed within the proposed timetable and with the listed resources.

1.1 Antenna Sharing and User Cooperation

Envision inserting relays literally anywhere in the space near the receiver or transmitter.Our goal is to find one that is in a “hot spot” and can receive the signal well. If that relayis simultaneously in a hot spot with respect to the ultimate recipient of the information,then this relay can effectively support the communication. The more relays there are, themore likely be that we can find such intermediate.

Let’s start with a simple scenario: in figure 1.1, a transmitter is placed close to aconductive wall 1.1(a) and the received electromagnetic field amplitude is calculated at thefar field region, approximately one hundred wavelengths away. We assume transmissionof a single carrier and we observe that the received amplitude is not constant since thereis destructive or constructive addition of the direct and reflected signals. In this simplescenario, there might be locations in space where the field amplitude might be larger thanthat in locations closer to the transmitter (observe the circled point in figure 1.1(b)). Movingfrom constructive to destructive addition of the two rays, involves small physical movementsin space, on the order of a quarter of a wavelength. Antenna sharing techniques described in


(a) A transmitter is placed closeto a perfect reflector, that couldbe a conductive wall. Assum-ing no absorption from the wall(perfect reflection), we calculatethe electromagnetic field ampli-tude at specific region, at the farfield.

Received Field

(b) The calculated field amplitude as a function of space,for the case depicted in the previous picture. Depending onthe phase difference between the direct signal and the sig-nal reflected by the wall, there are locations far away fromthe transmitter, that have stronger field amplitude than lo-cations closer to the transmitter. Observe, for example thecircled points.

Figure 1.1: Calculating the field of a single transmitter and a single, perfect reflector. Itcan be seen that depending on the phase of the direct transmission signal and the reflectedfrom the wall signal, the field might be stronger at a point in space which is at a largerdistance from the transmitter, compared to a point closer to the transmitter. Therefore,the instantaneous received power (proportional of the square of the electromagnetic field)is a function of the environment with temporal and spatial fluctuations. User cooperativecommunication exploits that phenomenon by using antennas of “other” users.

this thesis, exploit in a distributed and decentralized way, cooperating users located at thosepoints where the wireless channel is as “good” as possible. Therefore, the more cooperatingusers, the higher the probability to find one of them in a “hot spot”.

In reality, wireless propagation is much more complex than the two-ray model describedabove. The wireless channel typically involves many reflectors, scatterers and obstructions.It changes at a rate interval (coherence time) that depends on wavelength and mobility. Alarge number of reflectors corresponds to a complex fading channel coefficient (2-dimensionalsince there are in-phase and quadrature-phase components) with a normal distribution2.The amplitude of a circularly symmetric complex Gaussian random variable correspondsto a Rayleigh-distributed random variable and the amplitude squared, corresponds to anexponential random variable.

If aij is the (complex) fading coefficient between transmitter i and receiver j, then fromfigure 1.2(a) we can make an estimate of E[|aij |2] as a function of distance. We assumethat received power Pr ∝ E[|a|2] ∝ 1/dvij where v is the propagation coefficient and showshow quickly power decreases as a function of distance. In free space, since electromagnetic

2according to the central limit theorem


v = 3.98

(a) Measurement of the received power profile asfunction of distance at 916MHz for an indoor en-vironment [16].

0 5 10 15 20 25 30 35 40 45-90

-80

-70

-60

-50

-40

-30

-20

-10

0

10

(b) Artificial generation of a similar profile, us-ing Rayleigh fading and propagation coefficient vtaken from measurements of the previous figure.

Figure 1.2: Received power as a function of distance from the transmitter from actual mea-surements (figure 1.2(a)) or from artificial generation using Rayleigh fading and propagationcoefficient v estimated from measurements. The model (right figure) matches reality (leftfigure).

field drops as 1/d, the received power drops as 1/d2 and v = 2. In practice, there is no freespace and we can see that v could be greater than two in highly reflective environments oreven less than two when RF propagation is waveguided.

Using the linear markers in figure 1.2(a), we can estimate v (v = 3.98). Then wecan artificially create received power profile according to Rayleigh fading, using E[|a|2] =1/d3.98

ij . Comparing the two plots, it appears that Rayleigh fading provides a realisticapproximation of wireless channels and further improvements could be made by adding aconstant term modeling the antenna gain between the transmitting and receiving antennasthat scales appropriately the results.

Cassioli, Win and Molisch [10] shown that the path loss can be modeled as a two slopefunction in a log-log scale with propagation coefficient v ' 2 for distances close to thetransmitter and v ' 7 for distances above a threshold. Several researchers have suggestedLognormal fading as a realistic model of wireless channel power loss while others havesuggested Nakagami fading from which Rayleigh fading can be seen as a special case. Forthe discussion in this proposal, we will be using Rayleigh fading with various propagationcoefficients v, since Rayleigh is the baseline model used in communication research and agood approximation of reality as can be seen from figures 1.2(a), 1.2(b).

It is interesting to note that in a free-space environment where transmit and receiveantennas are placed at different heights, it can be easily shown that received power dropsfaster than 1/d2 for large d, due to phase difference between direct signal and signal reflectedby the ground. Therefore, v = 4 is a very realistic assumption for both indoor or outdoorenvironments.

Having briefly described models for the wireless channel, we are ready to describe usercooperative communication.


1.1.1 Three nodes...

Approach

1

3

1

3

1

3

2 2

Figure 1.3: Direct transmission, multi-hop transmission, cooperative transmission.

In the case of a single transmitter, single relay and receiver (figure 1.3), cooperativecommunication exploits the direct transmission, as well as the relayed transmission from aneighboring relay. The receiver combines direct and relayed transmission to detect informa-tion. Where direct transmission is not possible (for example, the receiver is out of range),multi-hop communication can be viewed as a special case of cooperative communication.Two natural questions emerge: a) what kind of signal processing is needed at each relayto ensure reliable, end-to-end communication and b) under what conditions cooperativecommunication is more efficient than non-cooperative (point-to-point) communication. Inthe rest of this section, we explicitly address both questions.

Given the existence of a single relay, there are two basic options: it could either de-code and re-encode (digital regeneration) or simply, amplify and forward the received in-formation, plus its own noise (analog amplify and forward) and leave the decision at thedestination. Both strategies can be found in the literature and have different performance.Laneman and Wornell in [23] reported that analog amplify and forward is better than dig-ital regeneration for a relay located half way between the transmitter and receiver, when aMaximum Ratio Combining (MRC) receiver is used in Rayleigh fading, with various prop-agation coefficients (2 < v < 6). We managed to justify that result by showing that theareas of beneficial relay location are different for the above relaying techniques: analogrelaying areas are symmetrical half-way distance between transmitter and receiver, whiledigital regeneration is beneficial only closer to the transmitter, since the scheme is limitedby the probability of error in the communication from transmitter to relay.

In figure 1.4, we have calculated the Symbol Error Probability (SEP) for 8-PSK modu-lation in Rayleigh fading with various propagation coefficients v, when MRC combining isused at the receiver and total transmission energy is split in half between the transmitterand the relay. The relay regenerates (digital relay) and it is located 1/4 the distance betweentransmitter and receiver, closer to the transmitter. Performance is compared to direct (non-cooperative) transmission, where all the energy is used for direct, one-hop transmission. Wecan see that the cooperative scheme is more reliable for the same transmission energy used,or it needs less transmission energy for the same performance. For SEP=1/1000, the plot isinverted and transmission energy savings are depicted in the form of ratios between trans-mission energy needed in the non-cooperative case vs the transmission energy needed inthe cooperative case. The reduced total transmission energy used is crucial in terms ofscalability, since reduced transmission power results in reduced interference to neighboringcommunications.

Improved performance is also observed when a dense constellation is used, in combi-


0 5 10 15 20 25 30 35 4010

-7

10-6

10-5

10-4

10-3

10-2

10-1

100

E = E1 + E2

Symbol Error Probability for 8-PSK

noncooperativecooperative digital

v = 2

v = 3

v = 4

v = 5

Energy Gains

(a) SEP in 8-PSK for various environments andE = E1 + E2, E1 = E2.

0 1 2 3 4 5 6 7

5

10

15

20

25

30

35

40

v

Ratio of Energy without cooperation vs (Total energy with cooperation)

(b) corresponding ratio E/(E1 + E2) for

SEP=10−3

Figure 1.4: Performance of cooperative communication compared to non-cooperative com-munication in left figure (using 8-PSK and various propagation coefficients) and total trans-mission energy ratio for target Symbol Error Probability (SEP)=10−3 in right figure (using8-PSK and v = 4), in Rayleigh wireless channels. Relay decodes and encodes (digital relay)and it is placed closer to the transmitter, 1/4 the distance between source and destina-tion. We can see that cooperative communication is more reliable compared to traditionalpoint-to-point communication, leading to higher reliability or transmission energy savings.

nation with cooperation. For example, using a constellation of 3 bits per symbol (8-PSK)with cooperative transmission, performs more reliably than a constellation of 1 bit per sym-bol (2-PSK) of direct communication for rayleigh fading with v ≥ 3 and digital relaying(we have omitted the plots due to space restrictions). Therefore, cooperation can increasethroughput in uncoded systems by 50%, under certain conditions3.

From the above, it is concluded than under certain conditions, cooperative communica-tion could lead to substantial transmission energy reduction or increased spectral efficiency,under certain conditions. We also need to compare cooperative communication to the caseof point-to-point communication where the transmitter simply transmits twice each infor-mation symbol. In figure 1.5, the regions where digital relaying is beneficial compared torepetitive transmission are depicted, for the case of 8-PSK in Rayleigh fading and varioussignal-to-noise ratios (SNR) and two propagation coefficients v, normalized to point-to-pointdistance between transmitter and receiver. We can see that provided that there is a relayclose to transmitter, between transmitter and receiver, digital relaying (and consecutivelycooperation) is more beneficial at the low SNR regime in highly attenuating propagationenvironments (v ≥ 3), than repetitive, non-cooperative communication.

Notice that the above scheme is based on the assumption that there is a relay insidethe appropriate area and the transmitter knows that (i.e. the transmitter has decided thatrelaying is more beneficial than repetition). Such decision could be based on knowledgeof relay location. In the introduction, we have excluded any prior knowledge of network

33 bits per symbol, over two channel usages, one for direct and one for relayed transmission, result in 1.5bits per channel usage versus 1 bit per channel usage for binary constellation and direct transmission.


1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.81

0.5

0

0.5

1

1.5

Tx

Rxv=3

SNR=10 dB

SNR=20 dB

SNR=30 dB

(a) v=3

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.81

0.5

0

0.5

1

1.5

Tx

Rx

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.81

0.5

0

0.5

1

1.5

Tx

Rx

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.81

0.5

0

0.5

1

1.5

Tx

SNR=20 dB

v=5SNR=10 dB

SNR=30 dB

(b) v=5

Figure 1.5: Regions of intermediate node location where it is advantageous to digitallyrelay to an intermediate node, instead of repetitively transmit. M=8 and the depicted ratiois the ratio of Symbol Error Probability (SEP) of repetitive transmission vs SEP of usercooperative digital communication. The cooperative receiver optimally combines direct andrelayed copy. Distances are normalized to the point-to-point distance between transmitterand receiver.

topology, therefore by definition, the above schemes need additional steps of coordinationbetween transmitter and relay, to be implementable in practice.

In the following section we will describe a distributed and adaptive scheme where a deci-sion is made among a collection of nodes, about which node is more appropriate to forwardinformation, based on channel measurements. The proposed algorithm scales cooperationin more than one relays and resolves all the coordination issues among transmitter, relay(s)and receiver, with minimal overhead.

1.1.2 More than three nodes...

Approach

The physics of wireless propagation are in general unpredictable. In this work, we proposenetwork architectures that react to changes of the environment to enhance performance andscale with increasing number of users.

best path @ kT

best path @ (k+1)T

Direct Relayed

|as,j|2 |aj,d|2

Source Destination

Figure 1.6: Path diversity as a function of users and instantaneous channel conditions.Method of distributed timers allow for optimal path selection, adaptive to channel changes,with minimal overhead.


In figure 1.6, we can see that a specific path between transmitter-relay-receiver, mightbe “good” during a specific time interval, and a different path might be “better” during thenext time interval. The coherence time4 of any wireless channel usually describes how oftenthe channel varies and depends, as we have said previously, in environmental propertiesincluding wavelength and mobility.

If transmitter could find out in a distributed fashion5, which relay has the “best” channelconditions between transmitter-relay AND relay-receiver, then it could forward informationalongside that path. The more potential cooperating relays are available, the higher theprobability at least one relay has “appropriately good” channel conditions in relaying infor-mation from transmitter to receiver. Consecutively, performance improves with increasingnumber of relays (cooperating nodes).

Store Display

802.11- enabled PDA

Rx

Tx

Relays

Figure 1.7: Distributed selection of “best” relay path. The intermediate relay nodes over-hear the handshaking between Tx and Rx. Based on the method of distributed timers, therelay that has the best signal path from transmitter to relay and relay to receiver is pickedwith minimal overhead. The receiver combines direct and relayed transmission and displaysthe received text on a store display. The “best” relay signals with an orange light. Thetransmitter is fed with weather information from an 802.11-enabled pda.

In [8], we proposed the method of distributed timers that allows the network to se-lect the best path between transmitter and receiver: intermediate nodes that overhear theReady-to-Send (RTS) packet from transmitter to receiver and the Clear-to-Send (CTS) re-sponse from receiver to transmitter, are able to estimate their channel conditions towardtransmitter and receiver, according to the principle of reciprocity and assuming same fre-quency transmissions. A function of their channel conditions is used as a seed to a timerthat expires sooner for the relay that has the best end-to-end path between transmitter andreceiver. Then, that relay is selected and the other relays back off, either because they canlisten best relay’s transmission or because the receiver has informed the neighborhood viaa broadcast message.

In [8], we analytically calculate the probability of this scheme to fail for Rayleigh fadingand show that the average convergence time of the algorithm is user defined and boundedby a constant factor which is independent by the number of users. The scheme performsbetter with increasing convergence time which should be kept as small as possible and ofcourse smaller than the coherence time of the channel. The scheme needs no synchronization

4The coherence time is inversely proportional to Doppler shift, which is a function of speed and carrierfrequency.

5in the sense that coordination is simple and requires as small communication overhead as possible.


overhead and in [8], we show that with a cost of one CTS packet (with duration of a fewmicroseconds, as used in 802.11b) the scheme performs reasonably well.

The function h(asj , ajd) of channel parameters asj (between source-relay j), ajd (betweenrelay-destination), used as a seed to the timer Tj of each relay j, should balance between a“good” path toward the relay (so that the best relay could receive a high quality copy of thetransmitted information ) and from relay to destination (so that relaying is meaningful).

Tj = λ1

h(asj , ajd)(1.1)

hj = h(asj , ajd) =2

1|asj |2 + 1

|ajd|2(1.2)

The best relay is that with the highest hj function that will expire first (smallest Tj).Notice that the particular hj corresponds to the harmonic mean of |asj |2, |ajd|2 and we pickedthe above function since we noticed in [26] that outage probability (as an approximationof error probability in coded systems) increases with 1

E[|asj |2]+ 1

E[|ajd|2], for both digital or

analog relaying. Different functions hj could be used and similar performance obtained, ifthe best relay maximizes the following function:

hj = h(asj , ajd) = min(|asj |2, |ajd|2) (1.3)

The method of distributed timers [8] is synchronized by the point-to-point transmissionand requires simple amplitude measurements of channel conditions, easily acquired viaReceived Signal Strength Indications (RSSI), available in most existing radio architectures.We implemented a demonstration of the above technique in our lab, using custom, low cost,embedded radios (figure 1.7).


The above method can be easily extended in multi-hop communication, since multi-hopis a special case of cooperative communication, as we show in figure 1.3. Time synchroniza-tion at the packet level between transmitter and receiver is needed, in order to coordinatethe RTS/CTS exchange described above. Multi-hop synchronization at the packet level issimple and has been extensively studied within the context of this theses. Results could befound in [5], [9] for client-server schemes and in [7] for a decentralized solution.

The idea of mapping channel conditions to time and using that information to selectoptimal paths, is central in this thesis and unique so far in the field, to the extent of ourknowledge.

(a) First hop. (b) Second hop.

Figure 1.8: During the first hop an information symbol is transmitted toward the receiverand overhearing nodes. During the second hop, an amplified version of the received infor-mation plus noise is forwarded to the receiver from the intermediate nodes.

The above scheme raises the following question: why shouldn’t all relays that overheardthe communication during the first hop, relay during the second phase (figure 1.8)? Thatwas a suggestion followed in the case of analog amplify and forward relays in [42] or digitaldecode and re-encode relays in [25]. If we fix the total transmission power used by therelays, it turns that our scheme, the opportunistic relaying where all transmission poweris used by the single “best” relay, as selected above, outperforms the schemes found inthe literature where all relays transmit at the same time. The analysis has been omitteddue to space restrictions, however we present the capacity histogram for the two cases infigure 1.9. We can see that the average value (dotted line) is higher in the case of “best”relay compared to the “All relays” case proposed elsewhere. Our scheme, not only performsbetter, but also needs very simple error correction techniques since only one relay, the bestrelay, retransmits. In contrast, the digital and decode scheme requires more complex space-time coding techniques, to allow for an arbitrary number of relays to operate simultaneously.

The issue of receiver complexity should not be neglected, especially in battery operatedcommunication networks. In [28], [29] it was shown that error correction techniques usedin practice at the receiver side, are energy expensive to the extent that digital reception ofinformation needs energy comparable to transmission energy. Therefore, digital cooperativeschemes with high complexity could become an overkill for extended battery life, not becauseof retransmission (relaying) but because of reception. The scheme proposed above could bea viable solution for increased reliability and reduced receiver complexity due to its reducedoverhead and simplicity.


7 8 9 10 11 12 13 14 150

20

40

60

80

100

Histogram of bps/Hz capacity All Relays Case

7 8 9 10 11 12 13 14 150

100

200

300

400

Histogram of bps/Hz capacity Opportunistic Relay Case

* 1/10000

* 1/10000

Figure 1.9: Capacity histogram and expected value (dotted marker) of opportunistic relay-ing (bottom plot) vs traditional approaches (top plot). Note that opportunistic relaying,where the “best” relay path is used between transmitter and receiver achieves higher ca-pacity than proposed techniques in the literature (“All-relays case”), where all overhearingrelays forward the information.

Perhaps, the distributed sampling in space provided by the cooperating nodes, in com-bination with the adaptive reaction to environmental changes as described above, couldprovide for the necessary redundancy needed for error correction and consecutively simplifythe receiver6.

Notice that the average capacity for the both relay cases (dotted marker in figure 1.9),is smaller than the point-to-point non-cooperative capacity (rightmost blue marker). Thatis because we have not allowed the transmitter to transmit a different symbol, during thesecond phase of the algorithm (second hop in figure 1.8). In principle, we could allow thetransmitter transmit a new information symbol, while the relays forward the previous sym-bol. That was the case studied in [2], from an information theoretic point of view, in thehigh SNR regime. In practice, we could devise signal structures with “sufficiently” large du-ration allowing transmission and cooperative relaying within the same symbol period. Thatwas the case we devised in the context of OFDM-based WLANs, exploiting the propertiesof oversampling and IFFT [4].

1.1.3 Supporting Technologies: Cooperative Timing and Positioning

An important component of applications in cooperative networks entails incorporating po-sition and time into the information communicated. Lacking this, we are examining carsfor horsepower without considering the radio and heater or number of seats. This workdiscuss ways to determine both of the above in a distributed manner using the propagationinformation that is part of transport.

Specifically, we have exhausted the problem of network time keeping and results can6intelligence at the network, not at the individual nodes...


be found in [5], [9] for client-server schemes and in [7] for a decentralized solution. Thedecentralized approach was analyzed and implemented in our lab. We were inspired byways colonies of fireflies synchronize and we manage to establish a global clock and quantifyits error as a function of network diameter, when each node needs to exchange timestampswith neighbors only one hop away. In that way, global synchronization emerges out of localinteractions. In figure 1.10, a visual proof of synchrony at the edges of a network is providedwhen the nodes synchronize using our statistical, decentralized technique.

(a) (b) (c)

Figure 1.10: Visual proof of synchrony. A “heartbeat” pattern is synchronized over thenetwork and displayed at the edges. The distributed, server-free approach for networksynchronization resembles the decentralized coordination of colonies of fireflies and inspiredthis work.

The idea behind cooperative location estimation is that in high density networks, thesignal-to-noise ratio (SNR) of communication between a node and its neighbors might beseveral orders of magnitude better than that when the node communicates with a beacon ora satellite station at the edges of the network. If a node could estimate its location relativelyto its neighbors, and their neighbors relatively to their neighbors and so on, then the networkcould provide for topology estimation without any external infrastructure, much in the sameway it could provide time using cooperative statistical methods, discussed above.

It was surprising to see that the above cooperative idea has not been extensively stud-ied. That is because the problem of autonomous location estimation has been addressedseparately from the task of scalable wireless communication and therefore most researchershave underestimated or oversimplified models of RF propagation.

Luckily, the problem of three-dimensional structure determination using distance datahas a rich background in the context of protein structure determination, also known asmolecular distance geometry problem [11]. Distances between pairs of atoms in the proteinare found either using knowledge about bonds between atoms or through NMR experimen-tation. Then structure need to be computed. If all atom pair distances are known exactly,then the structure can be estimated using a holistic approach7 [11]. An interesting problemoccurs when inter-atomic distance data are sparse. In that case, either the missing dataare estimated and a holistic approach is used [11] or topology is estimated using the avail-able data in a neighbor-to-neighbor way [13], exploiting the graph connectivity and usingsimple lateration (i.e. triangulation) equations. The problem of structure determinationwith sparse and inaccurate distance measurements as that in wireless networks, remains

7Signular Value Decomposition in a matrix that includes all the inter-atomic distances.


unsolved and will be the focus of this work, from an algorithmic point of view.

1.1.4 Progress and Deliverables

This thesis tests the following hypothesis: cooperative communication provides for morereliable communication under certain conditions, requires less transmission power and de-creases interference to neigboring communications, when compared to point-to-point, non-cooperative communication. Therefore, cooperation can improve the scaling laws of wirelesscommunication and networking.

We have investigated so far the case where one stream of information is served, in atwo-hop scenario. We know the important regions where the intermediate relay should belocated for both strategies of analog-amplify and forward relay or digital regeneration relayand we have generalized the results for the case of multiple candidate relays, demonstrating adistributed algorithm to pick the “best” relay, in terms of channel conditions and RF-signalquality. A demonstration using low lost embedded radios has been constructed.

This thesis will show that the above “opportunistic” relaying scheme, essential formsa virtual antenna where individual antennas belong to different users. The simplicity ofthe scheme does not exclude optimality of performance. Specifically, we will show that thediversity-multiplexing trade off of opportunistic relaying is exactly the same with that inmore complex schemes proposed in literature such as those in [25]. The difference is thatopportunistic relaying is simpler and has been implemented in practice since it does notrequire complex space-time coding, unknown so far. This thesis will also test the scaling lawsof cooperative communication when compared to traditional, non-cooperative approaches.

Specifically, in this thesis:

• We present the diversity-multiplexing trade off in opportunistic relaying and show itsoptimality in certain regimes and its resemblance to other more complex approachesproposed so far.

• We provide a methodology to analyze the opportunistic relay selection algorithm, forany kind of wireless fading statistics, and show its reasonable performance in realisticconditions. The analysis quantifies the probability of successful relay selection, in thesame way Medium Access Control (MAC) protocols are analyzed.

• We show how opportunistic relaying can be extended to multi-hop schemes. Timesynchronization results, discussed above, will be helpful toward this direction.

• We extend the results obtained with one transmission channel to many in order to (a)determine the maximum throughput of the space and compare it between differentnon-cooperative communication schemes, (b) show in simulation that opportunisticrelaying could efficiently determine relays in distributed ways.

• We plot scaling of throughput with node count, assuming that any can be a transmitterand discuss energy consumption per node per received/transmitted bit.

• We propose and test in practice cooperative network timing algorithms. In terms ofcooperative network positioning, we will address the problem of structure determi-nation with sparse and inaccurate distance measurements, from an algorithmic pointof view and test proposed solutions via simulation. We have already demonstratedsub-centimeter accuracy in range estimation, using ultra low cost, short-range ( 1m)infrared transceivers.


Close-form analytical expressions are sought and in cases where that is not possible,extensive and reproducible simulations are performed.

1.1.5 Prior Art

Cooperative Communication

In their 2000 paper [23], Laneman and Wornell described a distributed diversity recep-tion scheme with three nodes, one transmitter, one relay and one receiver. In that work,they evaluated a simple modulation scheme (BPSK) in conjunction with maximum ratiocombining of direct and relayed transmission and showed significant (transmission) energygains of the cooperative scheme compared to direct communication, at the expense of re-duced rate, since symbol (bit) transmission would need two consecutive channel usages (onefor direct and one for relayed transmission) instead of one. In that work, they evaluatedamong others, two cases of relaying: i) digital decode and re-encode (regeneration) and ii)analog amplify-and-forward. These two cases of relaying were compared with the relay half-distance between transmitter and receiver and amplify-and-forward performed significantlybetter than the digital scheme. Channel known only at the receivers and Rayleigh fading,supplemented with geometry of the three nodes, was considered.

Several research questions emerged after Laneman/Wornell 2000 work: could coopera-tion increase throughput in uncoded or coded wireless communication? Why does analogrelaying perform better than digital regeneration and what are the conditions under whichthe cooperative scheme is more efficient than direct communication between any two points?In section 1.1.1, we showed that the region of successful cooperative digital communicationis not symmetric around the perpendicular, equidistant between transmitter and receiver(figure 1.5). That is due to the fact that digital regeneration is meaningful when digital re-ception at the relay is error free: in other words, the whole cooperative scheme performanceis based upon correct reception of information at the intermediate relay and it is natural toexpect that the region boundaries would be shifted toward the transmitter.

Consequently, the comparison of digital versus analog relaying, at half distance betweentransmitter and receiver, would give results in favor of analog relaying. We extended theresults of [23] in the case of M-PSK (instead of BPSK), found out the regions of cooperativecommunication and quantified the spectral efficiency increase of cooperative communicationfor the case of M-PSK uncoded communication in Rayleigh fading.

In thesis work [24], [26], Laneman followed an information theoretic approach and ana-lyzed the three-node scheme of transmitter, relay and receiver in terms of outage probabilityand spectral efficiency, at the high SNR regime. Such analysis facilitates outage probabilityas an approximation of probability of error, since when appropriate error correcting codesare used, fading is the limiting, deteriorating factor. In that case, fundamental limits of per-formance, “best achievable performance”, can be sought, without worrying about specificcoding schemes that could achieve such performance. In that framework, Laneman foundout that digital and analog relaying have similar performance in the high SNR regime.He also discussed adaptive protocols with limited feedback, analyzed them in the sameinformation theoretic context and found out improved performance. He did not discusspractical coordination schemes among transmitter, relay and receiver that could achievethose optimal bounds. This thesis comes to fill that gap.

In [25], the case of several relays cooperating in a 2-hop scenario with digital relayingis analyzed in the same outage probability-spectral efficiency context. Digital relays are


allowed to relay at the same channel, when their received SNR is above a threshold andit is shown that the diversity gain is on the order of number of relays that participate inthat scheme. Practical space-time codes that could achieve such performance are not de-tailed, even though there is a discussion that such codes could be found. Our opportunisticrelaying approach, discussed in section 1.1.2 is a practical manifestation of a scheme withseveral cooperating relays in a 2-hop scenario and we have shown that opportunistic re-laying provides higher capacity compared to the case where all relays retransmit, for fixedtotal transmission power. Opportunistic relaying as well as “all relays” case in [25] assumethat the transmitter does not transmit a new information symbol during the second phaseof cooperation when the relay(s) retransmit. If the receiver is allowed to transmit a dif-ferent symbol when relay(s) retransmit the previous symbol, then performance (obviously)improves and that was the case discussed in [2], again from an information theoretic pointof view, at the high SNR regime.

In [4] we provided a practical scheme in the context of OFDM wireless networks (like in802.11a) where cooperative communication could be employed without sacrificing one de-gree of freedom (one symbol period): direct and relayed transmission could happen withinthe same symbol period, due to the special structure of OFDM symbols and properties ofoversampling. Therefore, we could experience the benefits of cooperation, without addi-tional delay or reduced rate, at the cost of increased computation at each node.

In two different representations of the analog-amplify-and-forward cooperating channelin [42] and [31], it was shown that ergodic capacity can not be increased if total transmitpower is kept constant and channel side information is known only at the receivers, whencompared to direct communication. The same result was also reported in [24]. Constructiveaddition of superimposed signals at the receiver is needed to increase capacity and thatcan be done only when the transmitters have channel information and special hardware(beamforming). We are not assuming any kind of beamforming capability in our work.

For the case of multiple streams and several hops, it is difficult to come up with aconcrete formulation and an analytical solution. Significant work toward this direction hasbeen reported in [41] where the formulation of rate matrices for each transmit-receive pairin the network is introduced. The achievable regions for all feasible rates are numericallysearched, for various scenarios including multi-hop, power control, successive interferencecancellation, node mobility and time-varying fading. An interesting aspect of that work isthe introduction of negative rates for nodes that relay information initiated by other nodes.It is interesting to see what are the feasible rate regions (capacity regions) for the case ofcooperative communication.

In [39], spread spectrum communication is employed and it is shown that when it iscombined with local scheduling based on local time synchronization, then the network cansustain considerable data traffic. The argument there is that spread spectrum communica-tion, in contrast to TDMA/FDMA medium access schemes, could survive concurrent trans-missions up to a level where Signal-to-Noise-and-Interference-Ratio (SNIR) is not severelydegraded. By employing multi-hop, low-power transmissions instead of single-hop, high-power transmissions, higher volumes of traffic could travel larger distances.

Kumar and Gupta in [18] showed that throughput of each node, when n nodes arerandomly placed in a unit area disk, drops as 1/

√n log n instead of 1/n as one might

expect. That result is under the assumption of fixed radius transmission range. Moreover,if nodes are placed carefully on the disk, individual rates can drop even slower with 1/

√n

and the total distance-throughput of the network can scale with√n (in meters times bits

per second). This surprising result, is based on perfect scheduling of information routing


and non-realistic wireless channel assumptions. Therefore it serves as an upper bound of“best performance” in a wireless network8. It remains to be seen how closely cooperativecommunication can reach those bounds.

Finally, we should mention work on antenna selection in traditional MIMO systems andits performance as it is summarized in [30]. Antenna selection, where high-SNR signals areutilized while lower-SNR signals are discarded, could provide tools and intuition to studyantenna sharing among different users in wireless networking. References in [30] providestate-of-the-art work in MIMO systems in general.

Cooperative Timing and Positioning

Prior art in multi-hop time synchronization is presented in our paper [7]. Time sync is sensornetworks has attracted considerable attention over the last years, even though completelydecentralized techniques (as the one proposed in this thesis) have not been demonstrated.Due to space restrictions, we will not present prior art on distributed time sync here. Theinterested reader could refer to our paper [7].

In the field of wireless sensor networks, there has been tremendous interest on au-tonomous location estimation, although the problem remains still unsolved. The proposedsolutions so far, are largely based on custom infrastructure, in the form of specialized bea-cons with a-priori known locations that broadcast reference signals. The rest of the networknodes estimate their location by measuring time-of-flight, signal strength, carrier phase dif-ference, direction-of-arrival (DOA) or even just presence of signals coming from the referencebeacons. Therefore, even though there is a GPS-free way for nodes to estimate location,there is still a need for special infrastructure to be installed and calibrated.

A summarization of beacon-based systems could be found in [19]. As examples of suchsystems we could mention the Flying Karamazov system from MIT Media Lab with ultra-sonic time-of-flight (TOF) measurements, the ultrasonic time-of-flight Active Bats systemfrom AT&T, the ultrasonic/RF Cricket system from MIT LCS lab with RF used for synchro-nization and ID broadcasting and ultrasound for time-of-flight measurements, the Radarsystem from Microsoft Research using 802.11 base stations and received signal strengthindications (RSSI). Comparative performance for most of those systems has been reportedin [19].

Optical methods have also been used in the context of HCI research, such as the Lidarsystem reported in [33], [40]. In the same category of small number of beacons and limitedarea/space coverage, the Lighthouse location [36] system belongs: a single base station, thelighthouse, estimates optically the location of “smart dust” nodes scattered in a few squaremeters. Other interesting range-finding techniques have been reported in literature, basedon audio [15] or magnetic fields [35].

Even in the case of nodes estimating their location relatively to their neighbors, it isimportant to differentiate between neighbors with a priori known location and neighborswith unknown coordinates. If we made the assumption that a room or a floor had a highdensity of nodes with known locations, then such approach could not be categorized asautonomous, since it would require housekeeping overhead to precisely place and index allthe known location nodes.

Such approach was reported in [34], where it was demonstrated that four beacons atthe corners of 10x12m indoor space, could provide for location error on the order of 1

8similar scaling laws based on deterministic scheduling, as used in parallel computing, have been reportedin [22]


meter when spread spectrum power delay profiles are used to estimate time-of-flight or 2meters error when signal strength is used. The limiting factor in the case of signal strengthmeasurements was shadowing due to obstructions such as walls, that would increase thebias in the estimate. It is interesting to see how cooperative communication and positioningtechniques could be devised to pick the “best” anchor nodes, with clear LOS (line-of-sight)or at least with high SNR range measurements, in respect to which, relevant locationcould be estimated. Experimental data in [34] test assumptions regarding the probabilitydistributions on power loss and delay spread of the wireless channel and find the log-normaland normal distributions respectively, not perfect (as expected). Those findings should beincorporated in our modeling.

Ad-hoc positioning using linear programming was reported first in [12]. In that work, thetransmission radius of the radio interface provides an upper bound of distance between anytwo communicating nodes. If all the available communicating pairs are known to a centralprocessor, then a set of constraints about feasible locations could be found using linearprogramming techniques. This technique is centralized and therefore limited to small-sizednetworks. Moreover, it is based on unrealistic assumptions about wireless propagation.

In [38], the distance of each node to anchor points with known locations is calculatedthrough minimum hop-count paths to anchor points that have known locations. That rangeestimate consists of the sum of node-to-node distances along that path. Then a feasibleregion of location is computed using a low-complexity bounding-box algorithm. A similaridea is described in [32] where the number of hops is computed from anchor points to eachnode and range is estimated by multiplying the hop count with an appropriate distance perhop factor which, in turn, is calculated by known distances between anchor points. Thelocation of each node is then calculated using triangulation (lateration). Refinements ofthe estimates could be provided by associating a confidence metric with each node and usethem in a weighted-least squares solution [37].

In all the above techniques, an estimate of ranges to the anchor points (reference pointswith a priori known location) is estimated first and then from those estimates, an estimateof the location for the node is provided. That range estimate from each anchor pointbasically sums hop counts or node-to-node distance estimates alongside a minimum hop-length path. In that sense, errors in node-to-node distance estimation are accumulated andat the same time, important information regarding the quality of each node-to-node distancemeasurements (for example, the SNR of each measurement) is neglected. Instead of suchad-hoc approach, using minimum-hop path range estimates, each node should estimate itsrelative location to his “best” neighbors, where best neighbors are those with SNR distancemeasurements relatively high. The same procedure could be followed in a gradient way untilsome nodes estimate their location relative to anchor points with a priori known location.In that cooperative peer-to-peer approach, error propagation could be reduced, given thefact that each node uses high quality measurements for range estimation and lateration.

It was surprising to see that the above cooperative idea has not been yet pursued. Thatis because the problem of autonomous location estimation has been addressed separatelyfrom the task of scalable wireless communication and therefore most researchers have un-derestimated or oversimplified models of RF propagation.

Luckily, the problem of three-dimensional structure determination using distance datahas a rich background in the context of protein structure determination, also known asmolecular distance geometry problem [11]. Distances between pairs of atoms in the proteinare found either using knowledge about bonds between atoms or through NMR experimen-tation. Then structure need to be computed. If all atom pair distances are known exactly,

1.2 Required Resources 22

then the structure can be estimated using an holistic approach like the SVD approach pre-sented above [11]. An interesting problem occurs when inter-atomic distance data are sparse.In that case, either the missing data are estimated and a holistic approach is used [11] ortopology is estimated using the available data in a neighbor-to-neighbor way [13], exploitingthe graph connectivity with lateration similar to equations ??, given above. The problemof structure determination with sparse and inaccurate distance measurements as in wirelessnetworks remains unsolved and will be the focus of this work.

1.2 Required Resources

We will need a faster personal computer. Thesis will be based on analytical results. Nu-merical optimization techniques will be used as well, when required.

1.3 Timetable

At the fall semester of 2004, focus will be on cooperative communication in wireless networkswith deliverables as described in 1.1.4. At the first two months of 2005, focus will be oncooperative timing and positioning with deliverables as described in 1.1.4. Spring of 2005will be devoted in wrapping up and writing thesis and graduation is scheduled for MAY2005 (which is the end of 4th year in the phd program, end of 6th year in media lab).

1.4 Short Bio 23

1.4 Short Bio

Aggelos Bletsas explores the boundaries between wireless communication, wireless net-working and distributed, mobile computing. For his master’s thesis, he focused on timekeeping in distributed, chaotic environments and in his current research he is integrating hisfindings into the development of cooperative antenna arrays, applicable in practice. He hadbeen also involved in a nano-technology project to reverse engineer an atomic force micro-scope for nano-lithography purposes. He holds a MS from MIT and a Diploma in electricaland computer engineering with highest honors from Aristotle University of Thessaloniki.For his undergraduate thesis, he developed the first complete Hellenic text-to-speech system,which was immediately commercialized.

Bibliography

[1] M.S. Alouini, A.J. Goldsmith, A Unified Approach for Calculating Error Rates of Lin-early Modulated Signals over Generalized Fading Channels, IEEE Transactions on Com-munications, vol. 47, no 9, September 1999.

[2] K. Azarian, H. El Gamal, P. Schniter, On the achievable diversity-multiplexing tradeoffin half-duplex cooperative channels, submitted for publication to IEEE Transactions ofInformation Theory,available at http://www.ece.osu.edu/ schniter/research.html.

[3] M. Bhardwaj, Power Aware Systems, M.S. Thesis, Massachusetts Institute of Technol-ogy, May 2001.

[4] A. Bletsas, A. Lippman, “Efficient Collaborative (Viral) Communication in OFDM-based WLANs”, Proceedings of International Symposium on Advanced Radio Tech-nologie, Institute of Standards and Technology, March 2003.

[5] A. Bletsas, “Evaluation of Kalman Filtering for Network Time Keeping”, IEEE In-ternational Conference on Pervasive Computing and Communications (PerCom 2003),Dallas-Fort Worth Texas, March 23-26, 2003.

[6] A. Bletsas, A. Lippman, “Natural Spontaneous Order in Wireless Sensor Networks:Time Synchronization based on Entrainment”, Technical Report, MIT Media Lab, De-cember 2003.

[7] A. Bletsas, A. Lippman, “Spontaneous Synchronization in Multi-hop Embedded SensorNetworks: Demonstration of a Server-free Approach”, accepted for publication, IEEE2nd European Workshop on Wireless Sensor Networks 2005.

[8] A. Bletsas, A. Lippman, D. Reed, “A Simple Distributed Method for Relay Selection inCooperative Diversity Wireless Networks, based on Reciprocity and Channel Measure-ments”, accepted for publication, IEEE Vehicular Technology Conference 2005.

[9] A. Bletsas, Evaluation of Kalman Filtering for Network Time Keeping, submitted forpublication, IEEE Transactions of Ultrasonics, Ferromagnetics and Frequency Control.

[10] D. Cassioli, M. Win, A. Molisch, The Ultra-Wide Bandwidth Indoor Channel: FromStatistical Model to Simulations, IEEE Journal On Selected Areas of Communications,vol. 20, no. 6, August 2001.

[11] G.M. Crippen, T.F. Havel, Distance Geometry and Molecular Conformation, JohnWiley & Sons, New York 1988.

24

Bibliography 25

[12] L. Doherty, L.E. Ghaoui, S. J. Pister, “Convex position estimation in wireless sensornetworks”, IEEE Infocom, volume 3, pp. 1655-1663, April 2001.

[13] Q. Dong, Z. Wu, A Geometric Build-Up Algorithm for Solving the Molecular DistanceGeometry Problem with Sparse Distance Data, Journal of Global Optimization, no. 26,pp. 321-333, Kluwer Academic Publishers 2003.

[14] D. Ganesan, R. Govindan, S. Shenker and D. Estrin, Highly Resilient, Energy EfficientMultipath Routing in Wireless Sensor Networks, Mobile Computing and Communica-tions Review (MC2R), Vol 1., No. 2. 2002.

[15] L. Girod and D. Estrin, “Robust Range Estimation Using Acoustic And MultimodalSensing”, In Proceedings of the IEEE/RSJ International Conference on IntelligentRobots and Systems (IROS 2001), Maui, Hawaii, October 2001.

[16] P. Goud Jr, C. Schlegel, W.A. Krzymien, R. Hang, Multiple Antenna Com-munication Systems - An Emerging Technology, Canadian Journal of Elec-trical and Computer Engineering, accepted for publication. Also found inhttp://www.ece.ualberta.ca/ hcdc/publications.html

[17] M. Grossglauser, D. Tse, “Mobility increases the capacity of wireless networks”, Pro-ceedings of IEEE Infocom 2001, Anchorage 2001.

[18] P. Gupta, P.R. Kumar, The Capacity of Wireless Networks, Transactions on Informa-tion Theory, vol. 46, no. 2, March 2000.

[19] J. Hightower, G. Borriello, Location systems for ubiquitous computing, Computer Mag-azine, Volume: 34 Issue: 8, Aug 2001 Page(s): 57-66.

[20] M. Janani, A. Hedayat, T.E. Hunter, A. Nostratinia, Coded cooperation in wirelesscommunications: space-time transmission and iterative decoding, IEEE Transactions onSignal Processing, vol. 52, no. 2, February 2004.

[21] A. Khandani, J. Abounadi, E. Modiano, L. Zheng, “Cooperative Routing in WirelessNetworks”, Allerton Conference on Communications, Control and Computing, October2003.

[22] S. R. Kulkarni, P. Viswanath, A Deterministic Approach to Throughput Scaling inWireless Networks, Transactions on Information Theory, vol. 50, no. 6, pp. 1041-1049,June 2004.

[23] J.N. Laneman, G.W. Wornell, “Energy-Efficient Antenna Sharing and Relaying forWireless Networks”, Proc. IEEE Wireless Comm. and Networking Conf. (WCNC),Chicago, IL, Sept. 2000.

[24] J.N. Laneman, Cooperative Diversity in Wireless Networks: Algorithms and Architec-tures, Ph.D. dissertation, Massachusetts Institute of Technology, Cambridge, MA, Aug.2002.

[25] J.N. Laneman, G. Wornell, Distributed Space-Time Coded Protocols for Exploiting Co-operative Diversity in Wireless Networks, IEEE Transactions on Information Theory,vol. 49, no 10, October 2003.

Bibliography 26

[26] J.N. Laneman, D.N. Tse, G.W. Wornell, Cooperative Diversity in Wireless Networks:Efficient Protocols and Outage Behavior, accepted for publication, IEEE Transactionsof Information Theory, available at http://www.nd.edu/ jnl/pubs/

[27] K. Langendoen, N. Reijers, Distributed Localization in Wireless Sensor Networks: AQuantitative Comparison, Elsevier Computer Networks 43, 2003, pp. 499-518.

[28] R. Min, M. Bhardwaj, Seong-Hwan Cho, N. Ickes, Eugene Shih, Amit Sinha, Al-ice Wang, Anantha Chandrakasan, Energy-Centric Enabling Technologies for WirelessSensor Networks, IEEE Wireless Communications (formerly IEEE Personal Communi-cations), vol. 9, no. 4, August 2002, pp. 28-39.

[29] R. Min, Personal Communication based on his 2003 Ph.D. work.

[30] A. F. Molisch, M. Z. Win, MIMO Systems with Antenna Selection, IEEE MicrowaveMagazine, March 2004.

[31] R. U. Nabar, O. Oyman, H. Bolcskei, A. J. Paulraj, “Capacity Scaling Laws in MIMOWireless Networks”, Allerton Conference on Communication, Control, and Computing,Monticello, IL, pp. 378-389, Oct. 2003.

[32] D. Niculescu, B. Nath, “Ad hoc positioning system (APS)”, Proceedings of GLOBE-COM, San Antonio, November 2001.

[33] J. Paradiso, K. Hsiao, J. Strickon, J. Lifton, and A. Adler, Sensor Systems for In-teractive Surfaces, IBM Systems Journal, Volume 39, Nos. 3 & 4, October 2000, pp.892-914.

[34] N. Patwari, A.O. Hero III, M. Perkins, N. S. Correal, R. J. O’Dea, Relative LocationEstimation in Wireless Sensor Networks, IEEE Transactions on Signal Processing, vol.51, no. 8, August 2003.

[35] E. Prigge, J. How, “Signal architecture for a distributed magnetic local positioningsystem”, Sensors 2002, Proceedings of IEEE , Volume: 2 , Page(s): 1497-1504, 2002.

[36] K. Romer, “The Lighthouse Location System for Smart Dust”, Proceedings of MobiSys2003 (ACM/USENIX Conference on Mobile Systems, Applications, and Services), pp.15-30, San Francisco, CA, USA, May 2003.

[37] C. Savarese, J.M. Rabaey, K. Langendoen, “Robust Positioning Algorithms for Dis-tributed Ad-Hoc Wireless Sensor Networks”, Proceedings of 2002 USENIX AnnualTechnical Conference, pp. 317 - 327.

[38] A. Savvides , H. Park , M. B. Srivastava, “The bits and flops of the n-hop multilaterationprimitive for node localization problems”, Proceedings of the 1st ACM InternationalWorkshop on Wireless sensor Networks and Applications, Atlanta Georgia, September2002.

[39] T.J. Shepard, “A Channel Cccess Scheme for Large Dense Packet Radio Networks”,SIGCOMM 1996, Palo Alto California.

[40] J. Strickon, Design and HCI Applications of a Low-Cost Scanning Laser Rangefinder,M.S. Thesis, Massachusetts Institute of Technology, January 1999.

Bibliography 27

[41] S. Toumpis, A. Goldsmith, “Capacity regions for wireless ad hoc networks”, IEEEInternational Conference on Communications, 28 April-2 May 2002.

[42] A. Wittneben, B. Rankov, “Impact of Cooperative Relays on the Capacity of Rank-Deficient MIMO Channels”, Proceedings of the 12th IST Summit on Mobile and Wire-less Communications, Aveiro, Portugal, pp. 421-425, June 2003.

[43] Cosmic ray web site: http://www.mpi-hd.mpg.de/hfm/CosmicRay/Showers.html

Intelligent Antenna Sharing and User Cooperation in Wireless Networks

Documents