Visions of Control

Tariq Samad

T echnological disciplines are defined

by their visions no less than their prac-
tices. The long-term, even futuristic, pro-
Population

jections of the former serve as beacons.

illuminating (it is hoped) the trail for re-
searchers and practitioners who, at any
time, are likely to be mired in their quo-
tidian tasks. This is particularly true in the
re-engineered world, where the next dead-
r
1
Fertility
line, milestone, or deliverable may be-
come an all-consuming pursuit-a Desired Food
succession of impermanent efforts that are Per Caoita
likely to produce a permanent myopia.
Immediate concerns may be necessary
stepping stones, but grand visions can be-
come the casualties of their reductive, ana-
lytic pursuits. Paradoxically, the greater
the successes on the staging grounds, the
more remote the ultimate goal can appear.
And even research, the activity explicitly
chartered with maintaining the long-term
perspective, can be compromised.
I Industri; Output Pollution

Vision vs. Practice in Controls

These general remarks fit the controls
field rather well in my view: the concor-
dance between research practice and vi-
sion in our discipline leaves much to be
desired. Key conferences and journals are
ready sources for identifying current re-
search directions. While a broad range of
topics can be observed, there are a few that
dominate the research agenda (and the
forums for presenting its results). Linear
systems theory, a veritable discipline in
itself and virtually synonymous with clas-
sical control, is one of these. Advances in
linear systems continue to be generated,
furnishing important, if by now largely
incremental, improvements to the analysis
and synthesis of feedback control under
small-signal assumptions. Another popu-
The author is with the Honeywell Technology
Center; avid is a Technical Associate Editor
of the Magazine.
I I Industrial
Capital
Investment Average Lifetime
Rate of Capital
Fig. 1. Feedback loops ofpopulation, capital, agriculture, and pollution. From 12,p . 581 as
adapted from [3, p . 971. Published with permission of Potornac Associates, Washington,
D. C., U.S.A.
lar trend is riding the “nouveau control” tems’. This very feature renders these
wave: neural networks, fuzzy control, ge- technologies less amenable to rigorous
netic algorithms, chaos theory, etc. These theoretical treatment, and the develop-
technologies are often heralded by their ments in this area are often empirically
proponents as enablers for solving the demonstrated on standard problems such
next generation of control problems. Typi-
cally, one distinguishing feature is singled ‘I know of one, fundamental, exception to this osten-
sible fact: the wave function in quantum mechanics is
out-a capability to directly handle the linear, apparently by necessity and not simplification [ 1,
nonlinearities inherent in all real sys- pp. 85-89].

88 IEEE Control Systems


r4
Ore
Grinding
Water
r loop, which can be measured in weeks.
ri
Heat Exchange
I I
9 Calcination
Alumina
Fig. 2. The liquor loop in aluminumprocessing. Figure courtesy of Neil Freeman, Honeywell
Australia.
as a stirred tank reactor or the inverted
pendulum.
For a view of the vision of control, as
distinct from its practice, let us turn not to
journal articles but to an undergraduate
controls textbook. Introductory texts are
(among other things) repositories of a
field’s vision, articulated by the authors
and seconded by the faculty that rely on
the texts. These books, and the visions
therein, are directed primarily at the col-
lege sophomore, whom they hope to at-
tract to the field by emphasizing its
breadth of application and its relevance for
problems that can be appreciated by a
non-technical readership. But they are
also useful to us seasoned professionals,
as reminders of the excitement we felt on
first exposure. Indeed, the idea for this
column arose as I was browsing through
Februa y 1997
+,
the textbook from my undergraduate days
and came across the illustration shown in
Fig. 1. Many similar examples can be
found in all introductions to control sys-
tems. The message being imparted is that
there is no limit to the systems we can
analyze, design, and regulate. Inverted
pendulums and linear models are just
stepping stones, albeit necessary ones.
Renewing the Vision
Preoccupation with small-scale appli-
cations is at odds not only with the vision
for control, but also with industrial trends.
Advances in single-loop or low-level con-
trol will always be of interest, but in an
increasing number of domains the real
opportunities for impact are at much
higher levels of the system. Optimizing
the eco-sphere will remain a dream (or
nightmare) for some time yet, but organi-
zations are starting to focus on problems
of a scale seldom addressed by our com-
munity. In process control, for example,
improved methods for PID tuning have
become a secondary concern; plant-wide
control and optimization are the buz-
zwords. In aluminum processing, the ap-
plication can encompass over a half-dozen
processes, each with its own array of PID
or MIMO controllers (Fig. 2). Numerous
significant complications arise, such as
the extensive transport delays in the liquor
This is far from an isolated instance. Pe-
troleum refiners are seeking refinery-wide
optimization-plant models can have a
million nonlinear equations in this case,
over which an iterative optimization algo-
rithm must be executed to solve for near-
optimal control settings. Other examples
include building-wide optimization for
minimizing energy consumption, free-
flight air traffic control, and intelligent
highway systems.
Considerations of such applications
can be found in controls conferences, but
they are relatively peripheral. One reason,
no doubt, is that controls-by any defini-
tion-is only part of the solution, not the
whole parcel. Yet I would maintain that
controls is the right umbrella discipline for
the multidisciplinary solutions required.
The issue, after all, is the regulation of the
dynamical behavior of a system, and the
control system is the de facto platform for
implementing the necessary algorithms.
Ultimately, the argument for renewing
the vision in controls is a pragmatic one:
the future growth of our discipline is prob-
lematical otherwise. Let us not overlook,
though, the revitalizing effect of new tech-
nical challenges. The intellectual rewards
can be equally substantial.
References
[I] Dreams of a Final Theory, Steven Weinberg.
New York Pantheon Books, 1992.
[2] Modern Control Systems, Richard C . Dorf.
Reading, MA: Addison-Wesley Publishing Com-
pany, 1974.
[ 3 ] The Limits to Growth: A Report f o r the Club
ofRome’s Project on the Predicament ofMunkind,
2nd edition, DonellaH. Meadows et al. New York
Universe Books, 1972.
89