Realizing The Union of Information Theory and Communication Networks
Rockey Luo, Colorado State University
Monday, November 16, 2009
4:00 p.m., 223 Weber
Modern communication networking simultaneously demands high resource efficiency in terms of achieving high throughput with limited energy and bandwidth as well as modularized architecture
to enable large system operations.
Providing its fundamental design guidance therefore requires the integration of information theory and network theory. In this presentation, we investigate two topics that represent some major incompatibilities between classical information theory and classical network theory.
In both topics, we extend information theoretic channel coding to networking scenaria by revising its basic definitions and assumptions.
In the first topic, we consider random multiple access networks where users access the channel randomly without knowing whether parallel transmission will cause destructive interference or not.
Assume users independently choose their communication rates (the number of data bits encoded in a packet) in each time slot.
We consider channel coding within one time slot and characterize performance limitation of the multiaccess channel using an achievable rate region in the following sense:
the receiver reliably decodes the packets if the joint rate vector is within the region; the receiver reliably outputs a collision if the joint rate vector is outside the region.
We show the achievable rate region equals Shannon’s information rate region without a convex hull operation.
In the second topic, we consider reliable data transmission in an overlay network.
The source message is encoded into packets, which are then transmitted through a virtual network link with arbitrary distortions and erasures.
Without knowing the statistics of distortions and erasures, we show that good rate and error performance can be achieved using a fountain communication model that determines communication duration at the receiver.
We introduce a concatenated fountain channel coding scheme and show that it can achieve near optimal rate and error performance with a linear coding complexity.
We also show the coding scheme has several key properties appealing to content sharing applications.