Two Lunchtime Seminars: Information Structures, Convexity, and Linear Optimality; and Using Game Theory for Distributed Control Engineering

Mike Rotkowitz, Fellow, Department of Electrical and Electronic Engineering, The University of Melbourne

Anders Rantzer, Professor, Department of Automatic Control, Lund University

11:30-12:30: "Information Structures, Convexity, and Linear Optimality" by Mike Rotkowitz

ABSTRACT: The synthesis of optimal decentralized controllers is a long-standing open problem. Conventional controls analysis breaks down when multiple controllers have access to different information. The Witsenhausen   Counterexample 40 years ago showed that linear controllers may be highly suboptimal for an LQG problem, and that the optimal control problem may be nonconvex. With the advent of complex interconnected systems, what has long been recognized as a difficult mathematical problem is now a                  
pertinent one as well.

We review results showing that when a simple condition holds, the optimal controllers may be found via convex programming.  This condition unified the few previously identified tractable problems, and elucidated many new ones.  The implications for several examples will be shown, including control over networks, and including new results showing the implications for spatio-temporal systems.  We also touch on work which not only extends these ideas to nonlinear control and time-varying   control, but potentially beyond the field of decentralized control, to all types of constrained control problems.                                                                      

We lastly discuss recent work exploring the connections (and lack thereof) between when optimal controllers are linear, and when finding the optimal controller is a convex problem, and see that one of the  questions posed in Witsenhausen's original paper can largely be answered.


MICHAEL ROTKOWITZ TALK
tac06_repub.pdf
cdc05_proc.pdf
acc06_distn.pdf
cdc06_distn.pdf


BRIEF BIO:
Michael Rotkowitz is with the University of Melbourne as the Future Generation Fellow in the Department of Electrical and Electronic  Engineering, and an Honorary Fellow in the Department of Mathematics and  Statistics.  He obtained his PhD from Stanford University in 2005, and  has also held positions at the Royal Institute of Technology in Stockholm and the Australian National University in Canberra.  His awards include the 2005 CDC Best Student-Paper Award, the 2008 IFAC World Congress Young Author Prize, and the 2007 Axelby Award.  

---

12:30-1:30: "Using Game Theory for Distributed Control Engineering" by Anders Rantzer

No Abstract available

ANDERS RANTZER TALK
RantzerGamees2008final.pdf

---

Related information can also be found at the following links:
http://www.control.lth.se/~rantzer/                                                    
http://www.ee.unimelb.edu.au/people/mcrotk/