MIT Department of Electrical Engineering & Computer Science

SPECIAL EECS SEMINAR

Monday, April 10, 1995
RLE Conference Room, 36-428

Refreshments at 2:30 PM
Talk at 2:45 PM

New Directions in Robust Nonlinear Control

Randy A. Freeman
University of California, Santa Barbara

Robust nonlinear control theory attempts to face two of the most significant challenges of model-based design: nonlinearity and uncertainty. Thus far, none of the three pillars of nonlinear control theory -- geometric control theory, Lyapunov stability theory, and optimal control/game theory -- have led to designs which adequately address each of the issues of robustness, performance and computation. Geometric control theory reveals the structure of a system, but designs based on this theory (such as feedback linearization) can lead to poor robustness and performance. Lyapunov designs can guarantee robustness but they require knowledge of a Lyapunov function and do not address performance. Optimal controllers have good robustness and performance properties by design, but they are extremely difficult, if not impossible, to compute.

Our results on inverse optimality and recursive Lyapunov design unite these theories, thereby incorporating their benefits and avoiding their drawbacks. New constructive design procedures narrow the gap between theory and application by concurrently addressing the issues of robustness, performance, and computation. These advances promise to bring robust nonlinear control closer to the forefront of practical design.