MSRI Workshop: Mathematics of Relaying and Cooperation in Communication Networks April 10-12, 2006

MSRI Workshop: Mathematics of Relaying and Cooperation in Communication Networks

April 10-12, 2006

 

Talk Abstracts
Monday, April 10

Introduction of the Relay Channel

Edward van der Meulen, Leuven

Abstract:

I will first describe the historical context of the work I did on the relay channel in the period 1965-1968. Then I will present the main results of my 1968 Berkeley Ph.D. thesis and my 1971 article in the Advances of Applied Probability. I will put these early results in perspective with the more advanced results obtained later by other authors. Next I will present some results, obtained around 1990, on the deterministic relay channel concerning capacity calculation and code construction. I will conclude with a discussion of the almost Gaussian degraded relay channel.

Slides [pdf]

Diversity-Multiplexing Tradeoff for Cooperative Systems: Characterization and Impact on Joint Source-Channel Coding

Elza Erkip, Brooklyn Poly

Abstract:

In this talk, we first investigate the diversity-multiplexing tradeoff for cooperative wireless networks. We study various network configurations such as multiple relays, multiple source/destination pairs, multiple antenna users, different user locations, as well as the performance achievable by different relaying strategies. We are particularly interested in whether the virtual multi-input multi-output (MIMO) system created by cooperation mimics a physical MIMO. We next show how the diversity-multiplexing tradeoff can be utilized for efficient transmission of an analog source over a slowly fading cooperative wireless channel. The performance metric is the distortion exponent, which is the signal-to-noise exponent of the average end-to-end distortion caused by compression and channel errors. We discuss source and channel coding strategies that compress the source signal in multiple layers and then transmit these layers either successively in time or superimpose them for simultaneous transmission. We also illustrate the benefits of unequal error protection provided by cooperation.

Joint work with Deniz Gunduz and Melda Yuksel.

Slides [ppt]

Cognition, Collaboration and Competition in Space and Time: From Theory To Implementation

Vahid Tarokh, Harvard

Abstract:

Elements/clusters of a sensor network operating on the same band canoperate using three different paradigms: i) Competition: This is information theoretically casted in the framework of interference channels. ii) Collaboration: Silent transmitters/receivers can help active transmitters/receivers in the transmission/reception of their messages, but have to extract this message from the underlying transmission or by other methods, and iii) Cognitive Radio Transmission: Some devices extract the message of other transmitters from their signals or by other methods, and use it tominimize interference from/to their own transmitted signals. Competition has been well-studied in the literature. Collaboration has been less studied and Cognitive Radio Transmission has not been much studied. For the case of collaboration, we demonstrate that most of the MIMO space-time gain can also be achieved through collaborative communications with single-antenna/multiple-antenna nodes when there is one receiving agent. In particular, for the single antenna case, we consider communication to take place between clusters of nearby nodes. We show the existence of collaborative codes for communications for which the intra-cluster negotiation penalty is in principle small and almost all the diversity gain of traditional space-time codes may be realized. For example, for a single transmitter node with two collaborators and one receiver node, if the collaborators have as little as 10 dB pathloss advantage over the receiver, the penalty for collaboration over traditional space-time systems is negligible. For the cognitive radio transmission, we consider a channel defined as an n-transmitter, m-receiver interference channel in which sender i obtains (causally or uncasually) the messages senders 1 through i-1 plan to transmit. The two sender, two receiver case is considered. In this scenario, one user, a cognitive radio, obtains (genie assisted,or causally) knowledge of the data to be transmitted by the other user. The cognitive radio may then simultaneously transmit over the same channel, as opposed to waiting for an idle channel as in a traditional cognitive radio channel protocol. Dirty-paper coding and ideas from achievable region constructions for the interference channel are used, and an achievable region for the cognitive radio channel is computed. It is shown that in the Gaussian case, the described achievable region approaches the upper bounds provided by the 2x2 Gaussian MIMO broadcast channel, and an interference-free channel. Finally, we demonstrate/discuss the implementation of these ideas on an SDR platform.

This is a joint work with Natasha Devroye, Patrick Mitran, Hideki Ochiai (Harvard) and John Chapin's Team (Vanu Systems).

Slides [ppt]

Throughput Scaling in Wideband Sensory Relay Networks: Cooperative Relaying, Power Allocation and Scaling Laws

Junshan Zhang, Arizona State

Abstract:

A sensory relay network consists of one source node, one destination node and many relay nodes. We investigate the achievable rates and the scaling laws of power- constrained sensory relay networks in the wideband regime, assuming that relay nodes have no a priori knowledge of channel state information.

We first study narrowband relay networks in the low SNR regime. We examine the achievable rates in the joint asymptotic regime of the number of relay nodes n, the channel coherence interval and the SNR per link. We propose an information relaying scheme, namely amplify-and-forward (AF) with network training, in which each relay node carries out channel estimation and then uses AF relaying to relay data. We provide an equivalent source-to-destination channel model, and characterize the corresponding achievable rate. Our findings show that when the transmission energy in each fading block is bounded below, the achievable rate has the same scaling order as in coherent relaying, thus enabling us to characterize the scaling law. We then generalize the study to power-constrained wideband relay networks, where frequency-selective fading is accounted for. Again, the focus is on the achievable rates by using AF with network training for information relaying. In particular, we examine the scaling behavior of the achievable rates corresponding to two power allocation policies applied at relay nodes, namely, a simple equal power allocation policy and the optimal power allocation policy. We identify the conditions under which the scaling law of the wideband relay networks can be achieved by both power allocation policies. Somewhat surprising, our findings indicate that these two power allocation policies result in achievable rates of the same scaling order, and the scaling law can be characterized under the condition that the energy per fading block per subband is bounded below and that the bandwidth W is sub-linear in n.

We will also talk about our ongoing work on the tradeoff between sensing and communications, and discuss cooperative routing in multi-hop relay networks.

Slides [ppt]

Network Layer View of Cooperation

Tony Ephremides, Maryland

Abstract:

In this presentation we are demonstrating two forms of user cooperation that is happening at the MAC and the Routing Layer respectively. One utilizes idle periods in a TDMA schedule-based protocol that utilizes a relay and the other one is based on a collaborative routing strategy in a sensor network deployed for event detection. They both are somewhat different than usual forms of cooperation that have been considered mostly at the physical layer, but they have similar performance improvement results and indicate that cooperation can be exploited in diverse , but always beneficial, ways.

Slides [ppt]

Throughput Optimal Control of Cooperative Relay Networks

Edmund Yeh, Yale

Abstract:

We give a model for cooperative communication in a parallel relay network that includes the stochastic arrival of packets and queueing. For this model we provide a throughput optimal network control policy which stabilizes the network for any arrival rate in its stability region. This policy is a generalization of maximum differential backpressure policies which takes into account the potential cooperative gains in the network. Generalizations of our model are discussed.

This is joint work with Randall Berry.

Slides [pdf]

Optimal Cooperative Communication in the Low SNR Regime

David Tse, UC-Berkeley

Abstract:

In wireless fading channels, a primary benefit of cooperative communication is to provide spatial diversity without the need of multiple antennas at each of the nodes. In such a setting, the performance limit is characterized by the outage capacity of the underlying relay fading channel. We compute the outage capacity in the low SNR regime and present a simple relaying scheme that achieves it. The basic features of this scheme are: 1) analog amplify and forward, 2) low duty cycle flashy transmission, and 3) non-coherent energy combining at the destination.

This is joint work with Amir Salman Avestimehr.

Slides [ppt]

Resource Allocation for Fading Orthogonal Relay Channels

Venugopal V. Veeravalli

Abstract:

An orthogonal relay channel model is studied, where the source transmits to the relay and destination in one channel, and the relay transmits to the destination in an orthogonal channel. Separate power constraints are assumed at the source and relay. The three links are subject to fading, and the fading state information is assumed to be known at both the transmitter and receiver. The source and relay can hence allocate their power adaptively according to the instantaneous channel state information. An achievable rate for this model is derived, and is optimized over the power allocation at the source and relay. This optimization is a max-min problem. A general technique for solving this max-min problem is introduced and applied to derive the optimal power allocation. The analysis is extended to the joint optimization of power and channel resource (time and bandwidth) allocations. In these scenarios, the maximum capacity of the channel (optimized over all possible power and channel resource allocations) is obtained for some special cases.

Joint work with Yingbin Liang

Slides [pdf]

Efficient Rumor Spreading: Cooperative Multicasting in Wireless Networks

Gregory Wornell

Abstract:

We consider the problem of multicast transmission (one source, multiple destinations) in wireless networks, and develop a meaningful notion of Shannon capacity in slow fading multipath-rich environments when the number of nodes (network size) is large. We further show that at any desired signal-to-noise ratio (SNR), capacity is achieved via a simple two-phase cooperative network protocol. Finally, we introduce the notion of a network ``scaling exponent'' to quantify the rate of decay of error probability with network size as a function of the targeted fraction of the capacity.

Joint work with Ashish Khisti and Uri Erez.

Slides [pdf]

Tuesday, April 11

Capacity Theorems for the Relay Channel

Abbas El Gamal, Stanford

Abstract:

Basic results on the capacity of discrete-memoryless relay channels with and without delay are reviewed. These will include: the cut-set bound, partial decode and forward coding, side-information coding, and instantaneous coding. I will also review results on the capacity of various AWGN relay channel models, including using linear relaying. I will end my presentation with a discussion of possible directions for future research.

Slides [pdf]

Cooperation, Multihopping and Relaying in Underwater Networks

Urbashi Mitra, USC

Abstract:

There has been significant activity in answering information and communication theoretic questions for wireless networks. A largely ignored arena is that of underwater networks. Furthermore, while there has been focus on establishing methods for point-to-point links in underwater systems -- the underwater communications network has been not been rigorously treated. Underwater communication channels are characterized by very significant delay spreads (often on the order of hundreds of symbols), very sparse multipath, transmission distance dependent effective bandwidth, and high Doppler. In this talk, we review several analyses and algorithm developments for the underwater acoustic communication network. We show how the underwater channel features impact designs for cooperative communication systems and sensor networks where distributed space-time coding, equalization, multihopping/relaying are considered.

Slides [pdf]

Discrete Memoryless Multiple Access Channels with Correlated Sources and Feedback

Sriram Vishwanath, UT-Austin

Abstract:

An achievable region for the discrete-memoryless multiple access channel (DM-MAC) with correlated sources and feedback is characterized, based ona modification of the Cover-El Gamal-Salehi approach for DM-MAC without correlated sources. Two different sufficient conditions for this region to be the optimal joint source-channel coding region are provided alongw ith an example scenario where each condition holds. The achievable region is subsequently generalized to the case of discrete-memoryless interference networks with feedback. The special case of Gaussian multiple with correlated sources and feedback is also investigated, with a modified Schalkwijk-Kailath coding scheme used to devise an achievableregion for the system.

Slides [ppt]

A Multi-Source Multi-Relay Scheme

Liang-Liang Xie, Waterloo

Abstract:

A multi-relay scheme is developed for networks with multiple sources. By this scheme, each source is sent to its destination via a multi-relay route, and the multiple multi-relay routes from the multiple sources can operate at the same time even if they intersect with each other, in the same spirit of the code-division multiple-access (CDMA). It is found that in the generalization to multiple sources, the backward decoding achieves higher rates than the sliding-window decoding. Potential applications of this scheme in sensor networks are discussed, where capacity results in certain scenarios are established.

Joint work with P.R. Kumar

Slides [ppt]

"Crystallization" in Large Fading Networks

Helmut Boelcskei, ETH

Abstract:

We study large wireless networks and show that the fading nature of the network leads to fundamentally new effects not present in the Gupta-Kumar model. In particular, it is demonstrated that the network can be made to "crystallize" in the large number of nodes limit. We analyze the impact of protocols, cooperation between nodes and processing at the individual nodes on the crystallization effect and describe a framework for characterizing the rate at which the network crystallizes.

Slides [pdf]

Cooperation and Competition

Randall Berry, Northwestern

Abstract:

In this talk we study wireless networks in which each node may have its own traffic to send. While cooperation may improve global performance, all users may not benefit from these gains and so not have an incentive to cooperate. One approach for modeling such issues is through cooperative game theory. We review some of the the basic ideas of this theory and present some simple game-theoretic models for cooperation and relaying in wireless networks.

Slides [pdf]

Results for Two Random Network Models

Babak Hassibi, Caltech

Abstract:

We will discuss some capacity results for two classes of wireless networks. The first are wireless erasure networks where links are possibly correlated erasure channels. Such models may be appropriate for systems where communication is packet-based and where some form of error correction is used to detect packet erasures. For problems with single source and multiple destinations all requiring the same information, we obtain an exact capacity result independent of the network topology. We will discuss interpretations and implications of the result, as well as connections to areas such as network coding. The second are so-called random wireless networks where each link is a fading channel drawn iid from a given distribution. Such models may be appropriate for networks of small physical size where the strength of a connection depends more on a random event (such as the existence of an obstacle), rather than on the distance between the transmitter and receiver. These networks are closely related to the classical Erdos random graph, and we show achievable throughputs that scale almost linearly in the number of nodes, a significant improvement over results obtained from the geometric Gupta-Kumar models. We discuss scenarios where such models may be reasonable, as well a multi-scale network model where the fading coefficients are iid at a local scale and obey a path-loss model at a global scale.

Slides [pdf]

Wednesday, April 12, 12-2:30

The Multiple-Access Channel with Cribbing Encoders Revisited

Frans M.J. Willems, Eindhoven

Abstract:

We first discuss the capacity regions for various communication situations in which one or both encoders for a multiple-access channel crib from the other encoder and learn the channel input(s) (to be) emitted by this encoder. Most of the achievability proofs for cribbing encoders hinge upon the concept of backward decoding. Also the notion of Shannon strategies seems to be of crucial importance. After having reviewed these concepts we discuss a recent development related to cribbing encoders that corresponds to the relay-channel without delay.

Slides [pdf]

Multiple Access Channels with Bi-Directional Links

Ashutosh Sabharwal, Rice

Abstract:

Most networks allow bi-directional communication, i.e., allowing receivers to send messages back to the transmitters but with the feedback link sharing the same bandwidth as the forward link. In this talk, we will study multiple access channels with bi-directional links and explore the role of receiver participation.

Slides [ppt]

Cooperation in the Low Power Regime: Capacity and Coding

Anders Host-Madsen, Hawaii

Abstract:

In this presentation we consider cooperative communications in the low power regime. The low power regime can be defined as the limit where the power spectral density of the transmitted signal goes to zero, which can be achieved by either letting the bandwidth converge towards infinity or letting the rate converge towards zero. There are a number of advantages to communications in the low power regime: in a non-cooperative channel the energy consumption (energy per bit) is minimized in the low power regime, the communication is interference-free, the modulation is simple, and the communication is covert. In the cooperative channel, energy per bit is not necessarily minimized in the low power regime, but it is close to the minimum, as will be shown, and the other advantages remain. This makes low-power communications ideal for applications like sensor networks. In the presentation, we develop capacity for low power communications for a number of channels. We also consider coding methods that are close to capacity. In particular, we consider two block Markov coding schemes, namely, multiplexed coding and superposition coding schemes. The multiplexed coding using fully multiplexed (FMP) codes, which outperforms the superposition coding theoretically, however, is difficult to implement with practical error-correction codes. We then introduce a partially multiplexed (PMP) coding scheme for the coderate R < 1/2, and propose a PMP code design method using irregular repeat accumulate (IRA) codes. The design method can be extended to build the low-rate PMP codes using irregular repeat zigzag Hadamard (IRZH) codes. The outage analysis shows that the cooperation schemes provide significant gains over the non-cooperative MAC in low-powerregime. Also, the PMP-IRZH coding performs only 0.5 dB away from low-power outage capacity for multiplexed coding.

Joint work with Zigui Yang (Hawaii), Guosen Yue (NEC Labs), and Xiaodong Wang (Columbia).

Slides [ppt]

Capacity and Cooperation in Wireless Networks

Andrea Goldsmith, Stanford

Abstract:

We consider fundamental capacity limits in wireless networks where nodes can cooperate in transmission and reception. We propose novel cooperation techniques that approach the capacity bounds, including virtual MIMO transmission, relaying, and conferencing. The optimal cooperation strategy is cross-layer in nature and depends on the network topology, the channel SNR, and the channel side information available at the nodes. Extending cooperation ideas to multiuser channels, multihop networks, and cooperative compression is also discussed.

Slides [ppt]

Backward Decoding Strategies for the Relay Channel

Mehul Motani, Singapore

Abstract:

In this talk, we will discuss coding strategies for the general three- node relay channel. We start with the well known achievability results of Cover and El Gamal, which superimpose decode-and-forward (DF) and compress-and-forward (CF). We then describe several new strategies which also superimpose DF and CF, and make use of sequential and simultaneous backward decoding. The rates with the simultaneous backward decoding strategy are shown to contain those of Cover and El Gamal.

Slides [pdf]

Codes for Half Duplex Relay Channels

Behnaam Aazhang, Rice

Abstract:

We propose a scheme to implement half-duplex decode-and-forward and estimate-and-forward relaying. We will first present an achievable rate for the estimate-and-forward protocol in the context of half-duplex relaying. Guided by the coding scheme for the achievable rates, we propose a practically implementable scheme for decode-and-forward and estimate-and-forward relaying. We introduce several simplifications to facilitate implementation. We will also present some simulation results on the performance of the codes in a relay channel.

Joint work with Arnab Chakrabarti, Alexandre de Baynast, and Ashutosh Sabharwal.

Slides [ppt]

Wireless Relay Networks

Massimo Franceschetti, UC-San Diego

Abstract:

It is well known that the throughput capacity of wireless networks scales poorly as the network size increases and all nodes wish to communicate simultaneously. In this talk we focus on the scenario when a constant density of nodes is randomly scattered in an increasing area and only two nodes wish to communicate at any given time, while all others can be used as possible relays. Using a percolation argument and a simple TDMA construction it is shown that in this case any pair, among an arbitrarily large fraction of the nodes, can maintain a constant bit-rate. On the other hand, an ergodic argument shows that this fraction can never be one, irrespective of the nodes cooperation strategies. Namely, there always exists a positive fraction of poorly connected nodes, among any two of which the rate must tend to zero as network size tends to infinity.

Stated informally, the conclusion is that in a wireless network it is not possible to have full-connectivity (i.e. a non-vanishing rate among any pair of nodes), but it is indeed possible to have almost-full connectivity (i.e. a non-vanishing rate among almost any pair of nodes). The constant rate in this latter case clearly depends on the fraction of connected nodes.

This is joint work with Olivier Dousse and Patrick Thiran.

Slides [ppt]

Capacity Results for Finite-Field Relay Networks

Piyush Gupta, Bell Labs

Abstract:

Recent efforts in analyzing the capacity of wireless networks have focused on modeling the broadcast nature of the medium but not interference. This is due to the challenging nature of studying the latter -- for even simple network configurations, such as the single-relay channel or the interference channel, the capacity regions are not yet known. Thus we study a simplified model, where the network operates over a finite field, but which incorporates both the broadcast and interference aspects of the medium, as well as allows for fading. For this model, we obtain an upper bound on the single-source multicast capacity, and present a network coding strategy that achieves rates arbitrarily close to the upper bound under uniform i.i.d. fading. We further show that channel fading in conjunction with network coding can lead to large gains in the multicast capacity over no fading.

Joint work with Sandeep Bhadra and Sanjay Shakkottai (UT-Austin).

Slides [pdf]

On the Capacity of Wireless Channels with Selfish Users

Hesham El Gamal, Ohio State

Abstract:

We adopt a game theoretic approach for the analysis of wireless channels where the users are modeled as rational and selfish players interested in maximizing the utilities they obtain from the network. In the fading multiple access channel, we first show that the sum-rate optimal point on the boundary of the capacity region is the unique Nash Equilibrium of the corresponding water-filling game. By introducing the base-station as a player interested in maximizing a weighted sum of the individual rates, we then propose a Stackelberg formulation in which the base-station is the designated game leader. We show that this formulation allows for achieving all the corner points of the capacity region, in addition to the sum-rate optimal point. Then we consider the cooperative multiple access channel, where we show that non-cooperation is the only Nash Equilibrium. By introducing a relay node designated as the leader of a Stackelberg game, we show that cooperation emerges at the equilibrium point under some typical operating scenarios. Finally, we establish similar results for the interference channel.

Slides [ppt]