Dr. Bitmead, Robert R.
Department of Mechanical Aerospace and Engineering
UCSD
Seminar Information
A romp through the philosophy of modeling for feedback control design
awaits the audience. The target material will focus on the stringent requirements of
experiment design and data content management to achieve models just barely
adequate for high-performance control design. The crux of the matter is that, for this
case, the data need to be informative for a new purpose and its associated novel
closed-loop operating regime. This necessitates bringing in information about the
specific application, notably the performance expectations and the prevailing
disturbances. That is, there is a part of modeling that requires posing hypotheses and
deducing their consequences; then contriving experiments to falsify them altogether. In
many areas of data-driven modeling, even for control, these issues remain unaddressed
or deliberately ignored, perhaps because they are case dependent and very difficult.
More widely than control design, the dearth of analysis of data quality or fitness for
purpose plagues big-banner subjects such as machine learning and artificial
intelligence. Often, the sole and assumptive property of the data is that it is BIG. In data-
driven control methods, this usually relates to the rank of a mosaic Hankel matrix
composed from the data, and hence generically full-rank. Bob will draw on his
background in system identification, based on statistical time-series analysis, to whale
into criticism – positive and negative – of these nouveau subjects willfully proscribing
hypotheses and deduction in search of The Algorithm for fitting the data, context
notwithstanding. Right-brain thinking being absent, these practitioners lack imagination
and a sense of fun. They should be pitied.
Bob Bitmead has been a fixture at UCSD’s MAE Department since last
century. He hails from Australia, where he got his start on the hard slog that is industrial
modeling for control design. His steep learning curve, driven by fear of embarrassment
and failure, continues to this day. Bob regards himself as a system theorist with interests
across control, signal processing and telecommunications. He was President of the
IEEE Control Systems Society in 2019 and spent six years on the Council of the
international Federation of Automatic Control. At UCSD, he was inaugural Associate
Vice-Chancellor for Academic Personnel from 2006-2009. He has a long history in
management, which area he regards as his imaginary axis.