Title:

High Performance Balancing on a Narrow Support

Abstract:

This talk will present recent advances in the physics and control of
high-performance balancing that we have made as part of the Skippy project.
Our approach has been to treat balancing as a physical process, rather than an
exercise in control theory, and to study it from a physical point of view.  As
a result, we have identified the physical properties that are important for
balancing; we can quantify how good a robot is at balancing, and therefore
design robots explicitly to be good at balancing; we can extract the
relatively simple dynamics of balancing from the potentially complex complete
dynamics of the whole robot; and we can design balance controllers that
exhibit higher levels of speed, accuracy and robustness than previously
reported.