Post-Doc: Ivana Maric
Channel model with two senders and two receivers permits various
forms of
partial transmitter cooperation. When cooperation is precluded, this
communication situation becomes the interference channel. We
examine
two-sender, two-receiver communication systems that allow for various
forms of
partial transmitter cooperation improving the network performance. For
example,
when there is a common part of the messages that is known to both
encoders, the
encoders can cooperate in sending the common part. In the very special
case when
only the common message exists, this scenario is a broadcast channel.
Another
interesting form of cooperation is conferencing
where the encoders communicate over links with finite capacities, a
model
proposed by Willems.
We are currently considering channel models in which one of the
transmitters has
the full or partial knowledge of the other user's message. We
refer to the
former channel as an interference channel with unidirectional
cooperation
(ICUC). This information allows one user to cooperate by forming
channel inputs
based on both users' messages, thus improving its own rate and the rate
of the
other user. For both channel models, we are developing outer bounds. We
are also
deriving an achievable rate region which generalizes the strategies
previously
proposed for the ICUC. For the special case of strong interference when
both
decoders can decode all the messages
with no penalty we have the capacity results.
The considered channel models capture some of the characteristics of
networks
with cognitive users. A cognitive radio can efficiently sense the
spectrum,
decode information from detected signals and use that knowledge to
improve the
overall system performance. The two considered channel models
assume that
one of the senders is cognitive. The assumption that the full message
of one
encoder is available to the cognitive user may be an over-idealized
model of the
cognitive network. Its capacity constitutes an outer bound on the
performance of
more realistic models. For that reason, we further relax such an
assumption and
consider a more general model in which
only a part of the message is known to the cognitive user. We seek to
understand
the capacity limits of networks with cognitive users and in general
networks
that incorporate cooperation among network nodes. Towards that end we
plan
to:
1) Further develop suitable channel models
2) Derive inner and outer bounds
3) Derive capacity results for specific scenarios and propose strategies that achieve them.
Student: Yifan Liang
The time-varying nature of the underlying channel is one of the most significant challenges in designing wireless communication systems. In particular, real-time media traffic typically has a stringent delay constraint, so the entire frame may fall into deep fading channel states. Furthermore, the receiver may have limited resources to feed back the estimated channel state information to the transmitter, which precludes adaptive transmission. There are generally two approaches to communication over time-varying channels: the pessimistic approach, where communication is based on the worst-case channel state; and the outage approach, which allows certain data loss in some channel states in exchange for higher rates in other states. The worst-case approach is the basis of Shannon capacity, and hence an information-theoretic study of other approaches requires generalizing the definition of channel capacity to include some data loss.
We propose more general definitions of channel capacity that includes no errors, outage, or expected capacity, where the transmitter uses a single encoder and the receiver can choose from a collection of decoders based on channel states. We will illustrate these general definitions with some examples. We then extend these ideas beyond data transmission over the channel to end-to-end transmission, i.e. data compression (source coding) as well as channel coding. In time-varying systems, in addition to channel variations, the source statistics may also vary over time. In particular, if the source or channel model is generalized to include non-ergodic statistics, it is natural to develop generalized end-to-end distortion metrics in addition to Shannon distortion, such as the distortion versus outage and the expected distortion. Shannon's renowned source-channel separation theorem enables separate design of source and channel codes with optimal performance. We show that the choice of end-to-end distortion metrics in a given system dictates whether source-channel separation optimality holds, and also which capacity metric coincides with the optimal transmission strategy. When separation is not optimal, the source and channel may communicate through an interface which allows multiple parameters to be agreed upon as opposed to a single number. Through the example of transmission of a binary symmetric source over a composite binary symmetric channel, we illustrate the performance enhancement under the more sophisticated interface.
M. Effros, A. Goldsmith and Y. Liang, Capacity Definitions for General Channels with Receiver Side Information, submitted to IEEE Transactions on Information Theory, April 2008.
Y. Liang, A. Goldsmith and M. Effros, Generalized Capacity and Source-Channel Coding for Packet Erasure Channels, To appear at IEEE Global Telecommunications Conference (GlobeCom), Dec. 2008, New Orleans, LA.
Y. Liang, A. Goldsmith and M. Effros, Distortion Metrics of Composite Channels with Receiver Side Information, IEEE Information Theory Workshop (ITW), September 2–6, 2007, Lake Tahoe, CA.
M. Effros, A. Goldsmith and Y. Liang, Capacity definitions of general channels with receiver side information, IEEE International Symposium on Information Theory (ISIT), June 24–29, 2007, Nice, France.
Student: Yifan Liang
Wyner’s pioneering work introduced cooperation among base stations (BS) as a new approach to increase the capacity of an infrastructure-based wireless networks such as cellular systems and wireless LANs. Full cooperation leads to the fundamental performance limit but is highly complicated. Study of local cooperation and resource allocation will reveal what sub optimal schemes may capture most of the performance gain from cooperation. Under the assumption of full base station cooperation, traditional design principles such as channel reuse also need to be reconsidered.
Effect of resource allocation (channel partitioning) in cooperative wireless networks. Cooperation among base stations has demonstrated substantial capacity gain in cellular systems. The optimal transmission scheme under full base station cooperation requires all users to transmit simultaneously and a central joint receiver for multi-user detection. We consider some sub-optimal but more practical schemes of orthogonal channel access either within a cell (intra-cell TDMA) or among cells (inter-cell time sharing), which correspond to a partitioning of overall channel resources. The effects of various schemes on the uplink capacity of a cellular system are then compared for a modified Wyner model.
Optimizing coverage spectral efficiency in cooperative cellular networks and wireless LANs: Coverage spectral efficiency (CSE) characterizes the tradeoff between efficient channel reuse and the achievable rates per cell, under the assumption of detection by a single base station and intra-cell FDMA. It is well known that intra-cell FDMA is not in general optimal. In this work we look at an alternative intra-cell wide-band scheme as well as the base station cooperation in detection, which has demonstrated potential capacity gain. The effects on CSE of different schemes are then compared and the optimal reuse distance is determined for each scheme.
Dynamic channel reuse in cooperative infrastructure-based wireless networks: In cellular systems a large reuse distance reduces co-channel interference while a small reuse distance favors system spectral efficiency. The optimal reuse distance is chosen to balance these two factors. Instead of applying a fixed reuse distance to the entire system, in this work we study the effect of adaptive channel reuse based on channel strength. Assuming the traditional single base station transmission, adaptive channel reuse under different propagation models, with or without fading, are analyzed. A new approach where base stations collaborate in transmission is also considered. Surprisingly, base station cooperation is not in general beneficial under optimized channel reuse.
Y. Liang and A. Goldsmith, Adaptive channel reuse in cellular systems, IEEE International Conference on Communications (ICC) , June 24–27, 2007, Glasgow, Scotland.
Y. Liang, T. Yoo and A. Goldsmith, Coverage Spectral Efficiency of Cellular Systems with Cooperative Base Stations, IEEE Global Telecommunications Conference (GlobeCom), San Francisco CA, Nov. 2006.
Y. Liang, T. Yoo and A. Goldsmith, Coverage Spectral Efficiency of Cellular Systems with Cooperative Base Stations, Asilomar Conference on Signals, Systems, and Computers, Pacific Grove CA, Oct. 2006. (Invited)
Y. Liang and A. Goldsmith, Symmetric Rate Capacity of Cellular Systems with Cooperative Base Stations, IEEE Global Telecommunications Conference (GlobeCom), San Francisco CA, Nov. 2006.
Student: Yifan Liang
Cellular systems are evolving to support high data-rate applications for the next generation network, but co-channel interference usually becomes the bottleneck for improvement. Base station cooperation has been demonstrated to increase capacity under the current network infrastructure. The main idea is to effectively form a virtual MIMO system with geographically dispersed antennas. We consider another evolving direction, namely denser base station deployment. Assuming a fixed number of users to be served in a certain geographical area, denser base deployment brings some idle bases to create “guard areas” among active users and effectively reduces their mutual interference. This approach focuses on infrastructure upgrade and can be regarded as a hardware approach while the virtual MIMO system formed through cooperation maintains the current network structure and focuses on advanced signal processing, which is a software approach. We study how the operating regime gradually shifts from interference-limited to noise-limited with larger base density. The comparison shows that denser base deployment outperforms sub optimal cooperation schemes (zero-forcing) when the base density exceeds a certain threshold, while a close-to-optimal cooperation scheme (zero-forcing with dirty-paper-coding) is always superior to denser deployment.
Y. Liang, A. Goldsmith, G. Foschini, R. Valenzuela and D. Chizhik, Evolution of Base Stations in Cellular Networks: Denser Deployment versus Coordination, IEEE International Conference on Communications (ICC), May 2008, Beijing China.
Y. Liang, R. Valenzuela, G. Foschini, D. Chizhik and A. Goldsmith, Interference Suppression in Wireless Cellular Networks through Picocells, Asilomar Conference on Signals, Systems, and Computers, Pacific Grove CA, Nov. 2007.
Student: Sachin Adlakha
Traditionally, the FCC has regulated spectrum quite closely: existing rules divide and govern the usage of spectrum from low to high frequencies. Several bands of spectrum have been deemed as “unlicensed” by the FCC, and made available for general purpose use; notable among these are the 900MHz, 2.4GHz, and 5.8GHz bands, which have supported a wide range of popular wireless devices including cordless phones, wireless LANs, and Bluetooth. This traditional model has come under increasing fire as spectrum becomes a more congested resource. The report of the FCC Spectrum Policy Task Force in 2002 found that while spectrum across the entire range regulated by the FCC may not be scarce, the regulatory procedures of the FCC have overly limited access and thus reduced utilization. The Task Force specifically recommended that the FCC move towards “flexible and market-oriented regulatory models.” However, such a move is not straightforward—both economic and technological challenges need to be overcome if deregulated spectrum use is to become a reality. If the FCC opens up large sections of spectrum for broad use, one potential outcome is a form of the “tragedy of the commons”: wireless devices that do not intelligently exploit holes in space, time, and frequency will interfere destructively with each other, eliminating any potential benefits of available bandwidth.
In this project, we will study a class of models to investigate the design of wireless protocols for use in a deregulated wideband wireless environment. Rather than assuming the status quo, our intention is to provide guidance for both competition and cooperation of wireless devices in such environments. Our methodology will combine components of stochastic and distributed control, game theory, and wireless system design to yield insights into how future wireless devices might exploit dynamic spectrum allocation.
Student: Sachin Adlakha
Control systems and communication networks are largely disparate research areas with little overlap or synergy to date. However, the ubiquitous deployment of wired and wireless communication networks enables compelling distributed control applications such as smart buildings, automated factories and highways, robot teaming, and automated security. Building a robust distributed control system closed over a communication network is a challenging task due to the different design principles inherent to these two disciplines; control theory often assumes feedback data that is accurate, timely and lossless, whereas random delay and packet loss are accepted and in fact unavoidable in most communication network designs. The optimal joint design of control and communication systems, along with separation principles whereby some aspects of the designs can be decoupled without loss of optimality, are wide open research problems. In addition, the application of control and optimization techniques to communication network design is emerging as a powerful tool to improve performance and robustness.
See American Control Conference 2007 Plenary Lecture by Prof. Goldsmith - PDF Slides
Caltech/Stanford/Berkeley Workshop on Control, Communications and Sensing, April 6-7, 2006 Stanford CA.
Cooperative Communications
Traditionally, a wireless network partitions its neighboring nodes into orthogonal non-interfering channels, which allows for tremendous reduction in complexity. However, partitioning a wireless network into non-interfering channels also results in sub optimal performance, and we have shown that significant capacity gain can be attained by allowing the neighboring transmitters and receivers to cooperate in their encoding, decoding, and routing of information [NgJGM06, NgG04b].
There are different forms of cooperation: for example, joint encoding, joint decoding, and relaying via decode-and-forward or compress-and-forward. We identified the most effective cooperation strategy based on the network geometry, number of cooperating nodes, the operating signal-to-noise ratio (SNR) [NgLG06], as well as the assumptions on the channel state information (CSI) and power allocation among different nodes in the network [NgG07, NgG06a, NgG05]. Another cooperative approach we plan to pursue is multi-round iterative conferencing which combines decode-and-forward coding schemes with hybrid digital-analog signal processing techniques. We have shown that iterative cooperation improves capacity over non-iterative schemes [NgMGSY06]. We expect no one single cooperation strategy excels at all operating conditions. Instead, we will derive the conditions under which a given cooperation scheme is effective, and devise combinations of the most promising cooperation techniques for a wide range of communication environments under realistic assumptions.
Student: Chris Ng
We have investigated the joint-optimization of user data compression in the application layer and channel coding in the physical layer. In the transmission of a delay-limited source without CSI at the transmitter, we solved the minimum expected distortion and optimal power distribution in Gaussian layered broadcast coding with successive refinement [NgGGE07c, NgGGE07b, NgGGE07a], for which previous results were only known in the high SNR regime or through numerical approximate solutions. Furthermore, the wireless medium is inherently time-varying due to changing channel conditions and user mobility. We have investigated time-varying broadcast channels, for which we derived the capacity regions subject to different rate and outage probability constraints [NgG04a], and we optimized the power allocation across fading states and users. We plan to incorporate additional network layers and study their benefits on the dynamic allocation of network resources. For example, in wireless sensor networks, it is common for the neighboring sensors to register correlated data; thus it is natural to additionally include the link layer and network layer to consider distributed compression and cooperative scheduling, routing, and transmission of the correlated source data based on channel conditions and end-to-end distortion requirements [GunNEG07].
Students: Taesang Yoo Sachin Adlakha
An ad-hoc wireless network is a collection of wireless nodes without a centralized control. In an ad-hoc network packets typically travel in a multi-hop fashion to reach their destinations. This type of network is suitable for the next generation wireless communication systems that require flexibility and a rapid deployment at a cheap cost. However, supporting real-time high data-rate traffic such as video over an ad-hoc wireless network is a challenging task due to limited resources (energy and bandwidth), mobility of the nodes, multi-user interference, etc. Traditional layered network architecture does not efficiently use the underlying flexibility of wireless ad-hoc network. In our work, we propose and develop cross-layer design approach, where the entire protocol stacks are co-designed for higher throughput and link reliability. In particular, we focus on the joint design of the physical, MAC, and network layers. Our simulation result shows that the cross-layer optimization results in 4 - 10 fold increase in the maximum supported data rate.
For a detailed description, please click here.
Building a distributed control system over a wireless network offers a lot of flexibility in terms of installation, mobility and maintenance. Yet this is a very challenging problem. Control systems and communication networks are typically designed using very different principles. Traditional control theory requires the feedback data to be accurate, timely and loss less. Conversely, random delay and packet loss are generally accepted in communication network design. Moreover, this delay and loss is much more pronounced in wireless networks than in wired networks due to limited spectrum and power, time-varying channel gains, and interference. Therefore, a joint design is necessary and we need a new approach for this joint design.
Joint design of control and communication is two-fold: the controller design needs to be robust and adaptive to the communication faults such as random delays and packet losses, while the network should be designed with the goal of optimizing the end-to-end control performance. The Kalman filtering in the presence of random packet losses [1] [2] is our first step in designing controllers that adapt to communication faults. In the effort to design the communication network to optimize the control performance, we first separately studied the link layer design tradeoffs [3] and the MAC layer design tradeoffs [4] [5]. Recently we propose a cross-layer framework to jointly design all the layers of the network to deliver the best end-to-end control performance [6].
For a detailed description, please click here.
In an energy-constrained wireless network where all the nodes operate on batteries, the total energy consumption should be minimized while satisfying certain throughput requirement in order to maximize the battery life. The traditional network design is based on layered structures, which is inherently non-optimal as far as the energy consumption is concerned. In this project, by jointly combing circuit level and system level design processes we try to come out with a cross-layer design methodology for an energy-constrained wireless network. I'm also working on energy-efficient joint estimation problems in sensor networks and optimal resource allocation problems in wireless networks. For a detailed description of my work, please click here.
Although the theoretical capacity gain of Multiple input multiple output (MIMO) channels is enormous, it requires perfect channel state information (CSI), which is rarely available in practice due to an increased number of channel parameters to estimate and to be fed back. In this research we have investigated the effect of imperfect CSI on the MIMO channel capacity. In particular, we provide answers for the following questions:
More and more control applications require a communication network to support the critical information exchange between distributed controllers. Examples of such applications include Automated Highway Systems (AHS) and the maneuvers of Unmanned Airborne Vehicles (UAVs). In many of these applications, a wireless communication network is essential since a wired network is unavailable or expensive to build. A wireless communication channel is an unpredictable and highly constrained medium due to the scarce radio spectrum and random power fluctuations. These power fluctuations cause intermittent connectivity and time-varying data rates. Thus, a wireless communication network inevitably introduces random delays and packet losses. Although modern controller designs are robust to modeling errors and external disturbances, they are normally not robust to the communication faults introduced by the network and the performance of the control system can be seriously compromised by these faults. However, although a wireless network can not guarantee an end-to-end quality of service, the wireless network can be designed in many aspects to adapt to the changing environment and resources can be dynamically allocated to meet the requirements of the control systems. We can also build in some robustness to the communication faults into the controller designs. We envision a joint optimization between control and communication will greatly improve the performance of a networked control system. In particular, we are interested in examining the tradeoffs between several critical network variables, such as data rate, delay and packet loss. We will evaluate the effects of these network variables on control system performance to design a wireless communication network for networked control applications.
Several projects in the Wireless Systems Lab are also studying turbo codes both as a topic in itself and also in conjunction with other system issues. Current research involves bounds for predicting the performance of turbo coded modulation systems. These bounds enable the study of adaptive turbo coded modulation. Space-time turbo codes are also being examined. Turbo coded modulation can be combined with many existing systems issues such as non-linear modulation and intersymbol interference. The study of turbo-coded CPM and the ability of TC-CPM to approach the capacity of a restricted input channel are also being researched.
Multirate DS-CDMA dominates the third generation wireless communications standards. WCDMA, CDMA2000 and IS95 HDR, all represent different evolutionary paths that rely on a multirate CDMA air interface. With new applications coming up, each with its own Quality of Service requirements, the need to support multiple data rates and a very diverse traffic distribution across users demands more sophisticated schemes to achieve high spectral efficiency. In particular we need to optimize the rate and power adaptation schemes. The importance of adaptation for improving the spectral efficiency is well recognized for single user wireless communications, and the single user optimum rate and power control strategy is well known. However, the general optimum rate and power adaptation strategy for multirate DS-CDMA systems using multiple codes, multiple processing gains, or multirate modulations, with traffic asymmetries remains relatively unexplored. Here at WSL, this forms the focus of our research in "Adaptive CDMA". So far, we have determined the required optimum adaptation strategies. The complexity and performance tradeoffs between optimum adaptation and multiuser detection schemes, impact of multiple antennas on adaptation schemes and the application of neural networks to achieve less complex adaptive structures form some of the future directions of research.
Recent advances in technology, have made feasible the use of Unmanned Air Vehicles (UAVs) in combat scenarios. Networks of such vehicles are expected to support battlefield operations, either by providing a communications network for ground forces or by gathering intelligence. UAV networks must therefore provide for the transport of data between the UAVs and between ground stations and the UAVs in a seamless and robust manner. Work in the lab is directed towards the design of such networks. We are currently focused on the characterization of the channel existing between two UAVs, and also in the investigation of theoretical limits in the performance of large networks of UAVs. Our work is supported by the Office of Naval Research (ONR).
Wireless networks consist of a number of nodes communicating over a wireless channel. Depending on their architecture, they can be roughly divided in two categories. In those following the cellular paradigm all nodes communicate directly with a base station that is responsible for controlling all transmissions and forwarding packets to the intended users. In those following the ad hoc paradigm all nodes have the same capabilities and responsibilities. Two nodes wishing to communicate can either do so directly, if possible, or route their packets through other nodes. Our work deals with this second type of networks. Lately, there has been work on determining the capacity of ad hoc networks. On a recent landmark paper, Gupta and Kumar derived upper bounds on the performance of a class of networks in the limit of a large number of nodes, in terms of a single figure of merit, the maximum uniformly achievable communication rate between all nodes and their selected destinations. In this work we define and investigate capacity regions for ad hoc networks with any number of nodes. These multidimensional regions are more descriptive, since they contain all achievable combinations of rates between the network nodes, under various transmission protocols.
Though the capacity of single-link Gaussian channels has been solved, the capacity of general multi-user networks is far from known. Gaussian multi-user networks are a very interesting and practical class of multi-user networks. We have discovered that an inherent duality exists between Gaussian broadcast and multiple-access channels and we have also been able to establish this duality for the multiple-antenna versions of these channels. Our results apply to non-fading and fading channels. We hope to establish a similar duality for arbitrary multi-user Gaussian networks. In related work, we are attempting to find the capacity regions of the Gaussian interference channel and the two sender, two receiver channel.
The Shannon capacity of fading channels can be defined in a number of
ways. Ergodic capacity appears to be the most straightforward
definition of Shannon capacity for fading channels, but there may be
infinite delay in the decoding process when achieving ergodic capacity.
Outage capacity and minimum rate capacity are two different notions of
Shannon capacity which we have been actively working on in WSL. Outage
capacity is concerned with maintaining a constant rate during
non-outage periods and not transmitting during outages. The decoding
delay during non-outages is independent of the rate of channel
variation. Minimum rate capacity is a combination of outage capacity
(with zero outage) and ergodic capacity: a minimum rate is achieved at
all times, while maximizing the average rates in excess of the minimum
rates. Since the minimum rate is maintained at all times, the decoding
delay of the minimum rate data is also independent of the rate of
channel variation. Outage and minimum rate capacity appear to be more
realistic performance measures due to the delay constraints of many
wireless applications such as voice transmission. We have characterized
the outage capacity and minimum rate capacity of Gaussian
single-receiver channels, as well as Gaussian broadcast and
multiple-access channels. We are currently investigating the minimum
rate capacity of the multiple-access channel in more detail and we are
also interested in finding even more realistic notions of Shannon
capacity for fading channels.