pjk at design.eng.yale.edu
Sun Feb 25 18:00:07 EST 2001
Dear Colleagues -
This item from the current issue of The Economist
(text also below) set me thinking, not for the first time, about
modularity in engineering design. And more specifically, how it
relates to teaching engineering.
My own very practical work in electronics design over more than 30
years, has ingrained in me a modular approach, partitioning a
larger whole into functional modules, each described by an
input/output "cause & effect" behavior sufficiently resourceful,
but also sufficiently ideal, to allow the efficient assembly of
larger functions with such modular units. The "sufficiently ideal"
part allows the descriptive framework of the larger modular
assembly to be no more complex than that of an individual module.
Nature does not "design" this way, she evolves very slowly. The
result is a total fusion of function and form that we are right to
admire aspiringly, but can seldom take as a direct lesson. A blade
of grass is a totally integrated system of structure, fluid
transport and chemical reactor.
In designing the artifacts of our man-made world we seldom have the
leisure of slow evolution. When planning, designing and
fabricating are compressed in time, even encouraged to overlap, as
in "concurrent engineering", the role of modularity becomes
paramount. It keeps complex interactivity in check, a must in true
design thinking. Thus a growing design can still be managed by
either the individual or the small team. Being overly ready to
relegate growing complexity to large teams and/or computer
simulation leads to poor designs, to technological chimera. Today
such chimera are an encompassing enough presence to coerce the
attitudes of their users. They come to think of them as progress.
How to teach about modularity, about keeping design thinking
orderly and as simple as possible? There is a growing need for such
teaching, as I see a steady decline in the ability of arriving
students to do significant "systems thinking", to decompose larger
wholes into causally interactive component parts. In electrical
engineering even the very starting points of our enterprise are on
a quite abstract level. Partitioning electronic and software
systems into modular subunits seldom "speaks" very directly to
beginning engineering students.
That's why I found this story about modular robots so suggestive.
Mechanical systems project their constraints with more immediacy,
"speak" more directly, less veiled by mathematical frameworks and
computer simulations where the "magnitudes of things" are muted
beyond tangibility. That's why this article made me think again
that there ought to be an introductory course involving modular
mechanical systems, or mechanical/electronic ("mechatronic")
systems, a course common to all engineering programs, where
students get a telling exposure to systems thinking and the lessons
of simplicity and modularity.
Skill in such conceptualization is a defining characteristic of
engineering, yet increasingly neglected in many engineering
P.S.: A particularly interesting book related to these themes is
M.J. French "Invention and Evolution: Design in nature and
engineering" (Cambridge Univ. Press, 1988)
Your Flexible Friend
Feb 22nd 2001
>From The Economist print edition
IN THE film "Terminator 2" the villainous "liquid metal" P-1000
assassin-bot was able to change shape so that it could ooze through
narrow bars, or turn its extremities into blades. Mark Yim of
Xerox's Palo Alto Research Centre (PARC) has more limited and
benign ambitions. As he explained to the AAAS meeting, rather than
building murderous psychotic androids, he is trying to create a
robot for use in search and rescue operations, deep-sea mining and
space exploration. But like the P-1000, Dr Yim's machine can change
its outward form.
The PARC Polybot is made of a dozen or so identical modules. When
ordered to do so by its operator, it changes shape on the move by
rebuilding itself out of these modules. According to the terrain,
it can adopt one of three different arrangements. When crossing a
level surface it becomes a looped tractor tread. For travelling
down stairs or climbing over an obstacle it configures itself into
a caterpillar. On rough ground it changes into a four-legged
To achieve this trick, the segments talk to each other using
infra-red transceivers. They can then locate one another, align
themselves using small onboard motors, and lock and unlock from
each other at will. The computing power needed to control all this
is distributed among separate processorsone for each
segmentalthough there is a control centre in one module, which is
as close as the Polybot comes to having a brain.
Besides versatility, Dr Yim hopes the general idea of
reconfigurable modular robots will ultimately create cheap and
durable devices. Their robustness would come from creating a system
in which the failure of one or two modules would not matter.
Savings in cost would arise from the system's modularitythat is,
from having to manufacture only one sort of component in order to
build a wide variety of different devices.
Having proved the principle with a dozen modules, Dr Yim and his
colleagues are working on a more ambitious version. This machine,
which should be ready later this year, will have 200 modules.
Eventually, it will also be able to control its own behaviour. Let
us hope it does not develop any personality disorders.
Copyright © 1995-2001 The Economist Newspaper Group Ltd. All rights
More information about the EAS-INFO