Systems Seminar
Computation Codes - A New Tool for Multi-user Communication
Prof. Michael Gastpar
EECS Dept.
University of California at Berkeley
Abstract
A major potential in large wireless networks is cooperation: A
transmission from a single node is overheard not only by the intended
receiver, but by all other nearby nodes; by analogy, any receiver not
only captures the signal from the intended transmitter, but from all
other nearby transmitters. The pessimist's perspective on these facts
has shaped communication network designs of the past decades: Clever
algorithms and protocols have been devised to avoid interference.
Recent work, however, has revealed many scenarios under which
interference turns out to be beneficial, provided it is suitably
shaped. This is often referred to as physical-layer cooperation. Most
of the cooperation schemes that have been proposed to date harvest
the statistical dependence of the underlying signals. By contrast, in
this talk, we present novel codes that permit to exploit the
algebraic structure of the interference, enabling efficient and
reliable computation of functions of the involved messages.
Such codes will be referred to as computation codes. They are
of independent interest in applications that explicitly call for
computation, such as sensor networks.
More generally, the computation coding perspective is used to
develop a new framework for larger networks: Inside the network,
judiciously chosen functions of the messages (rather than the
messages themselves) are being passed around. As soon as a receiver
has sufficiently many functions, it can infer the underlying message
(i.e., the bits). This is reminiscent of so-called network coding,
with the important difference that in the new framework, the question
of which functions of the messages should be passed around is decided
according to the actual interference characteristics, which can lead
to significant gains.
This is joint work with Bobak Nazer (UC Berkeley).
Time and Place: Wed., Oct. 1, at 3:30 in 4610 Engr. Hall.
SYSTEMS SEMINAR WEB PAGE:
http://homepages.cae.wisc.edu/~gubner/seminar/schedule.html