-----------------------------------------------------------------
24 June 1992
REPORT ON
FIRST INTERNATIONAL ROBOT OLYMPICS
UNIVERSITY OF STRATHCLYDE
HOSTED BY THE TURING INSTITUTE
27th TO 28th SEPTEMBER 1990
REPORT COMPILED BY
THE SHADOW GROUP
29 JAN 1991
SHADOW GROUP: 357 LIVERPOOL ROAD LONDON N1 1NL 071 609 8522
1st International Robot Olympics
--------------------------------
The 1st International Robot Olympics was held on 27th to 28th
September 1990 at the University of Strathclyde in Glasgow. The
Event was conceived by Dr Mowforth of the Turing Institute,
Glasgow as a British 'Robotics Weekend' on the lines of a weekend
event held the previous year at the Artificial Intelligence
Laboratory of The Massachusetts Institute of Technology. However
perhaps because Glasgow was City of Culture 1990 the scale of the
event changed and it grew to a major international meeting.
Impression of Olympics
----------------------
At first sight the Olympics seemed to be a large well attended
international event and judging by the intense media coverage one
at which startling advances in robotics were going to be
revealed.
However a French t.v. journalist confided that he had asked
around among the other reporters and they all agreed that they
were somewhat disappointed at the level of technology. (He
thought the Shadow walker was the most interesting looking
entry.) He was also disappointed that there were no French
entries. Nevertheless this view can not have been the only one
since Philipe Dufay reporting on the Olympics in the Figaro
colour magazine remarks; "It is high time to finish drilling the
Channel Tunnel: our neighbours have now too many secrets..."
As was to be expected the level of technology varied with the
level of funding, but irrespective of cost, most designs were
first used over ten years ago. The University entries had
obviously had considerable funds expended to produce
nicely-finished engineering presentations. With few exceptions,
however, the level of performance did not match their appearance.
The entries from amateurs and schools were, on the whole,
developed with minimal expense, some deliberately, others from
necessity. Their engineering tended to be more primitive, but
showed several neat methods of solving design problems.
Most entries suffered from being moved to Glasgow, and perhaps
more effort should have been applied to improving reliability.
(Perhaps an additional day should have been allowed for setting
up, free from Press involvement.)
The event was held (as might be expected) in a sports hall but it
was well disguised with the traditional exhibition carpet and
equipped with numerous tables and chairs. The carpet later proved
to be exceedingly troublesome to most of the robots; the small
ones got stuck in the pile, larger ones navigating by
ded-reckoning can only have been confused by the pile induced
drift (much as a rug on a carpet seems to have a mind of its own
migrating purposely in some inconvenient direction) and the
largest wheeled machines appeared to have been designed to run on
nothing less than reinforced concrete. Virtually every robot at
the Olympics required electricity, for itself and/or its
controlling computer, money spent on the carpet could well have
been put to better use in providing an adequate electricity
supply. Sports halls are not noted for this need and many
competitors seemed to have difficulty with the number of sockets
provided and a few complained of computer malfunction attributed
to an unreliable supply.
Little prior thought seemed to have been given to the actual
competition events themselves. In traditional athletics the first
past the post is invariably the winner but with today's robots of
limited versatility the outcome is not quite so clear cut
especially when competitors did not know what was to be tested
until after they arrived! This lack of adaptability ought to have
been foreseen and had some bearing on the events and on the
selection of judges for the events.
It was interesting to compare the differing attitudes of the
teams. Half, in charge of robots which didn't look so much
designed as assembled in the University workshop from parts
chosen at random from a machinery catalogue were apparently
bored to death by the whole affair. Whilst the others in charge
of machines, working or not, which they had either built
themselves or been closely involved in the building of (for
example Paul Channon of Cardiff) were enthusiastically
demonstrating, discussing or repairing them. This latter half
were definitely involved in the sharper end of the technology.
Suitability of Events
---------------------
The competitions were really too difficult for real robots, and
conditions did not adequately take account of their limitations.
Although providing some comic relief, they did not serve the real
purpose of pushing development forward. For example, in the
walking race there was no handicapping for size, which left a
four foot high machine having to 'run' as far as a machine one
foot high.
Salient Technology
------------------
Walking:-
The one major robot from Britain that was demonstrated reliably
was the biped walker from the University of Wales College of
Cardiff which tottered backwards and forward on command but
didn't seem to be able to turn corners. Despite the fact that
getting the machine to walk at all was a major achievement the
robot should be viewed against the global background. The machine
performs about as well as the Biper range of walking robots from
the University of Tokyo Japan. The Biper range was started about
10 years ago and each machine probably cost about one hundredth
that of the Cardiff machine.
The Shadow Group biped walker while being low cost is using a
form of air powered muscles which Professor Kato of Washida
University Japan has been unable to make work successfully.
However the walker although complete has yet to be demonstrated
under control.
MIT were represented by two small hexapod walkers one working the
other not yet finished. These really were state of the art
machines. The subsumption software on the former allowed
incredibly fluid walking even over some telephone directories!
The latter had eight microcontrollers on board, one to each leg
with one co-ordinating those six, while the eighth was used for
picture analysis from an onboard miniature (one half inch cube)
t.v. camera and range finder.
Arthur Collie From Portsmouth Polytechnic along with a team from
Russia were demonstrating wall climbing robots. The Russian
machines were very primitive and seemed not to draw on knowledge
of previous wall climbers over the last twenty years. Arthur
Collie's machines on the other hand were novel robust and worked
well, the largest using suction pads on the end of its legs was
able to step over small irregularities on its way up the wall.
Its robustness was inadvertent demonstrated when after falling
about ten feet to the floor due to someone turning off its supply
it was immediately without any repairs being necessary put back
to climbing the wall.
Other walking machines can be ignored as either not working or
based on ancient designs, e.g.. Penelope an eight legged machine
from the University of Edinburgh was an implementation of a
design from the late sixties.
Wheeled locomotion:-
Once again these robots demonstrated the old adage that one has
got to learn how to walk before one can run. Most robots seem to
last only one or two years before being discarded. With that
approach everybody has to learn everything for themselves
starting from scratch and nowadays that really does mean from
scratch without the benefit of childhood experience obtained
through experimental construction using for example Meccano or
Balsawood. The Japanese entry Yamabico won the overall prize for
the best robot, it is now about ten or twelve years old and was
refined over the first half of its life until now it can be taken
out of its cupboard, dusted of and demonstrated with reasonable
surety of success.
Two of the nicest mobiles were commercial items from Real World
Interfaces in the USA, they have been available for about the
last five years, unfortunately the people from the University of
Salford seemed to be using them as radio controlled toys. Tagg's
early mobiles built round Atari ST computers have in the past
bumbled around but at the Olympics for some reason were strangely
silent.
Dr Goodhead of Warwick University was demonstrating a fast highly
manoeuvrable omni-directional robot base with a novel drive
arrangement using inclined wheels to overcome power steer, but it
was one of those that kept trying to tear-up the carpet. However
in all fairness it is being developed for factories and
warehouses.
The much vaunted Trolleyman from the NEL was not working (despite
the fake t.v. shots on Tomorrows World) and was hidden away in a
corner!
In front of the Shadow Group stand a machine called Jon's Jalopy
which looked like it had been designed by the proverbial Roland
Emmett actually hid a video camera and range finder linked by an
umbilical to an Acorn Archimedes computer which was using the
video picture information to guide the vehicle to a
pre-designated object.
Several smaller robots were about (many on the Shadow Group
stand). Most of them were very rudimentary autonomous vehicles
with perhaps only one or two sensors. Dave Bisset, from the
University of Kent, put in two small buggies which won a silver
and a gold medal. One vehicle Icarus built to a 1963 design by D
Buckley contained as its brain only a photocell a transistor and
a relay yet won a Silver medal! Most of them would have been
state of the art 10, 20, 30, or even 40 years ago. Yet they
should not be dismissed; some of them embody Neural Net type
control architectures which is again, after being out of favour
for too long, showing promise of giving robustness to control
strategies for mobile robots.
However Yamabico aside non of these machines are anywhere near
being a self contained autonomous robot which knows where it has
been and to where it is going.
Special purpose machines:-
A great deal of effort had been expended in machining large
pieces of aluminium to construct massive inverted single and
double pendulums with the object of learning how to balance them
under computer control. These pieces were not seen working!
Pendulums are not at the edge of technological knowledge and
these projects should only be seen as interesting if expensive
student projects.
Dr Todd of Edinburgh University had a large wheeled robot which
was designed to open a special test door which once the problems
caused by transport were sorted out it did. The handle had though
to be in exactly the right place.
Conclusions
-----------
Three things were very evident from the Games. Britain is not at
the sharp end of robot technology, enthusiasm from a mere handful
of researchers is not enough: there just is not sufficient
backing both for resources and the time necessary to build state
of the art machines: and certainly in Britain's universities
there is not the necessary intellectual infrastructure and
knowledge base.
It is important to realise that despite the tones to the contrary
from one or two prominent researchers in the field Britain has a
lot of catching up to do. On the whole British universities are
about ten years behind the wavefront and this is not going to be
changed overnight.
Fortunately there does seem to be a growing awareness amongst our
researchers that it is not possible to build advanced research
robots 'on paper', and that actual working hardware is vital for
understanding of the problems involved. This awareness must be
nurtured.
SHADOW GROUP Robot Olympics Report