When Bits Absolutely, Positively Have to be There as Soon as Possible

SHARE:

Event Speaker: 

Greg Wornell, MIT

Event Location: 

32-155

Event Date/Time: 

Tuesday, April 2, 2013 - 4:00pm

Abstract:

In a growing number of applications, we need to reliably communicate at the
highest possible rate over highly dynamic and unpredictable channels. Rateless
coding is a natural solution to such problems, and indeed, information
theoretic analysis readily establishes that the desired capacity-achieving
rateless codes exist. As such, in recent years there has been growing interest
in the development of practical (i.e., low-complexity) codes for approaching
these fundamental limits. While there have been good low-complexity
capacity-approaching codes for implementation at the application layer, such
codes give up much performance in many (e.g., wireless) applications by not
directly controlling the available physical-layer resources. In this talk, I
describe a surprisingly simple framework for transforming standard good
off-the-shelf codes for the traditional additive white Gaussian noise (AWGN)
channel into good rateless codes. This framework is both capacity- and
complexity-preserving, and the result can be viewed as efficient joint design
of the physical and link layers in the network protocol stack. I will describe
two variants of this framework: one based on a simple layered architecture and
successive interference cancellation receivers; and the other based on novel
super-Nyquist signaling and decision-feedback equalization. In addition to
developing their performance characteristics and discussing their
implementation, I will mention some representative applications, including
recent experiments demonstrating their promise in next-generation underwater
acoustic modems. Finally, I will comment on the broader role this methodology
may ultimately play in the development of efficient interference-resilient and
jam-proof networks.
 
Based on joint work with Uri Erez and Mitchell D. Trott.

Bio:

Greg Wornell has been on the MIT faculty in EECS and RLE since 1991, where he
is the Sumitomo Professor of Engineering. He did his graduate work also at MIT
in EECS, and his undergraduate work at the University of British Columbia. His
research interests span a variety of aspects of signal processing, information
theory, digital communication, and statistical inference, and include
algorithms and architectures for wireless and sensor networks, multimedia data,
information security, and imaging systems, among other applications.