attention because they only spent about $20,000 in cash
for their entry and not one person on the team was paid.
They only prepared for the DRC for a very short six months,
as opposed to Team IHMC who spent about three years.
Humanoid Robots —
the Slow Advance of an
Electro-mechanical Science
Modern humanoid robotics engineering has had several
false starts — first in the 1940s with Electro, and later with
the myriad of toy and hobby rolling robots that were being
designed, built, and sold in the 1980s — followed by
walking robots about five years ago.
The mechanics were the easy part. The programming,
not so easy. The biggest barrier robotics had even 10 years
ago was sheer computer processing power. With so many
sensors bringing in so much data and movements
happening in real time — in three dimensions — older
computers could not keep up. I remember when I was
working as a technician for Androbot in the mid ‘80s that
one of the engineers told me that he watched a rolling
robot take three hours to navigate its way around a box.
Navigation, balance, finding objects, and voice recognition
all depend upon fast computers.
Our modern computers might just be getting close to
what is needed. Or, perhaps we just need better
algorithms.
Your Robot Overlords are
Here Now and Taking Over
the World! If Only They Could
Just Stop Falling Down ...
At the DRC, there were quadrupeds and biped robots.
Walking on four legs is relatively easy — just ask any dog or
baby. Walking on two legs is another story. Between 1937
and 1939, one of the first “walking” (it was more like a
shuffle) two-legged robots built was Westinghouse Electric’s
“Electro.” At seven feet tall — with both voice input and
output — Electro captured the imagination of science fiction
writers and inventors alike. Still, fiction outpaced science
and walking bipedal robots didn’t advance much until
Honda’s humanoid robot development which began in the
mid-1980s.
Honda, with a lot of hubris, a lot of money, and a lot
of falling down (just go to honda.com and watch the early
robots fall, fall, and fall again), got its robots walking. Now,
its Asimo robot runs, punts a soccer ball, and climbs stairs,
proving that with a “modest” budget of say a $100 million
or so, anything is possible. Now, Honda was never known
for sharing its technology much, so aside from Asimo,
most of the bipedal advances happened independently of
the huge car maker.
In spite of all the advances in walking robots, during
the challenge, there were more than 10 robots that fell
over — some more than once. Many sustained damage, but
were able to continue. One man described the walking
robots as sure footed as toddlers. After watching a
You Tube “fall compilation” video that went viral, I prefer to
think of them as surefooted as someone who has had too
much to drink.
Machine Intelligence or
Human Intelligence?
Based on what I could see, it looked like some teams
used a brute force and dumb luck method of competing,
while others — particularly the college teams — used a pure
math and engineering method. Based on the teams that
won, I am guessing that the engineers and mathematician
teams might have had a large advantage over the others.
After doing my own research, I discovered that the
amount of human intervention varied greatly from team to
team. Still, theverge.com had this to say about the
American Team IHMC that came in second: “ ... (The IHMC
team) focused on balancing and walking algorithms and
left the rest to ‘humans.’” “All the real cognition was inside
64 SERVO 08.2015
