IIT Home Page CNR Home Page

Seminario: "Optimizing performance in MIMO networks"

The problem we consider is that of maximizing throughput in a MIMO
network while accounting for variable rate streams on MIMO links.
The stream rates on a link depend on the channel conditions of the
link, and the manner in which the diversity-multiplexing tradeoff
is handled.  We identify and formally specify a series of problems,
namely feasibility, stream allocation, and one-shot scheduling, the
solutions of which can be combined to address the maximum throughput
problem.  While feasibility is a trivial problem under most network
models, we show that it is a special type of Boolean satisfiability
problem and is therefore computationally hard for MIMO networks.
However, we also identify several practical special cases, e.g. when
interference suppression is done only at the receiver side, for which
feasibility checking remains a polynomial-time operation.  For the
general feasibility problem, we present and evaluate a heuristic
algorithm, which is shown to be 95% accurate in single-collision-domain
networks with nodes having up to 12 antenna elements each.  We build on
this feasibility heuristic to develop an algorithm for stream allocation.
Simulation results show that our stream allocation heuristic is within
7% of optimal in single-collision-domain networks with up to 10 links
(in a typical case where the maximum concurrency allowed is 15 links).
Next, we evaluate several link selection heuristics, which can be
used with the stream allocation algorithm to provide a one-shot
scheduling algorithm for throughput maximization.  Finally, we
present preliminary results for multi-hop MIMO networks.  The
prior algorithms do not scale in performance nor in computation time
for multi-hop networks.  We have developed a hierarchical and
scalable algorithm for these networks.  Preliminary results show the
hierarchical algorithm performs well in MIMO multi-hop networks with
70 links.  We close with a discussion of several open problems in MIMO network optimization.

From 20/04/2010-11.00 to 20/04/2010-11.00 , Aula didattica dello IIT (Aula A32 - vicino biblioteca)

Speaker: Douglas Blough

Responsible: Paolo Santi

Note: NB: Il livello di approfondimento tecnico della presentazione e' volutamente lasciato ad alto livello, con l'obiettivo di rendere il seminario comprensibile a chiunque abbia un Computer Science/Computer Engineering background