On State-Dependent Multiaccess Channels: Capacity Results and Optimal State Compressions

Abdellatif Zaidi

Universite Paris-Est Marne-La-Vallee, France

Abstract:

The study of channels that are controlled by random states has spurred much interest, due to its importance from both information-theoretic and communications aspects. For example, state-dependent channels may model communication in random fading environments or in the presence of interference imposed by adjacent users. The channel states may be known in a strictly-causal, causal or noncausal manner, to all or only a subset of the encoders.

There is a connection between the role of states known strictly causally at an encoder and that of output feedback given to that encoder. In single-user channels, it is now well known that strictly causal feedback does not increase the capacity. In multiuser channels or networks, however, the situation changes drastically, and output feedback can be beneficial --- but its role is still highly missunderstood. One has a similar picture with strictly causal states at the encoder. In single-user channels, independent and identically distributed states available only in a strictly causal manner at the encoder have no effect on the capacity. In multiuser channels or networks, however, like feedback, strictly causal states in general increase the capacity.

In this talk, we study a two-user state-dependent cooperative multiaccess channel (with degraded message sets), with the states revealed, depending on the scenario, strictly-causally, causally or non-causally, to both or only one encoder. We discuss optimal ways of exploiting the knowledge of the states, and find explicit characterizations of the capacity region for some of the studied models, in both discrete memoryless and memoryless Gaussian cases. In particular, we show that, like feedback, outdated states at the encoders are in general useful and increase the capacity.

When the state is observed strictly causally, at both or only one encoder, a key ingredient of the coding schemes that we develop is the idea of quantizing the state (jointly, or individually) and conveying it to the receiver. However, critical in this problem is the way the state compression should be performed. We show that both compression \`a-la noisy network coding, i.e., with no binning, and compression using Wyner-Ziv binning can be optimal. The scheme that uses Wyner-Ziv binning shares elements with Cover and El Gamal original compress-and-forward, but differs from it mainly in that backward decoding is employed instead of forward decoding and the compression indices are not decoded uniquely.

Biography:

Abdellatif Zaidi received the B.S. degree in Electrical Engineering from École Nationale Supérieure de Techniques Avancés, ENSTA ParisTech, France in 2002 and the M. Sc. and Ph.D. degrees in Electrical Engineering from École Nationale Supérieure des Télécommunications, TELECOM ParisTech, Paris, France in 2002 and 2006, respectively.

From 2003 to 2006, he was with the Communications an Electronics Dept., TELECOM ParisTech, Paris, France and the Signals and Systems Lab., CNRS/Supélec, France pursuing his PhD degree. From 2006 to September 2010, he was at École Polytechnique de Louvain, Université Catholique de Louvain, Belgium, working as a reaserch assistant. Dr. Zaidi was "Research Visitor" at the University of Notre Dame, Indiana, USA, during fall 2007 and Spring 2008. He is now, an associate professor at Universite Paris-Est Marne-La-Vallee, France.

His research interests cover a broad ange of topics from signal processing for communication and multi-user information theory. Of particular interest are the problems of relaying and cooperation, network coding, interference mitigation, secure communication, coding and interference mitigation in multi-user channels, source coding and side-informed problems, with application to sensor networking and ad-hoc wireless networks. A. Zaidi is an Associate Editor of the Eurasip Journal on Wireless Communications and Networking (EURASIP JWCN)