Abstract
This talk focuses on the
problem of designing optimal distributed
controllers for large-scale and
spatially distributed systems. The
search for the optimal controller
is restricted to the set of
controllers with a given
"local" communication architecture, e.g. those
with only access to information
from neighboring subsystems. The
general problem of finding optimal
distributed controllers under
arbitrary communication architectures
is yet unsolved. However, we
demonstrate that a certain choice of
the communication architecture
leads to a tractable, though
non-convex, optimization problem. We then
use a Steiglitz-McBride
(SM) relaxation to render the problem convex in
the controller parameters.