Mechanical
Aspects of
Robot Design
As I have discussed so
many times in this column,
the mechanical design of a
robot seems to be the
designer’s greatest
stumbling block. It seems to
be fairly easy for a beginner
to take two modified servos
and mount them to a small
plastic or metal platform; add two
wheels to the servos and a caster for
a third wheel; add a Propeller, BASIC
Stamp, or Arduino microcontroller
board and some AA cells ... and you
have a robot. This is a great way to
start learning about robots.
Not long after, a person wants to
go bigger with a robot that can do
more than just move around and
respond to its sensors. The builder
wants a robot that can “do
something” when he/she is asked
about the bot. Now all of a sudden,
servos have to be replaced by
gearmotors, and larger batteries are
required. Mounting sensors and vision
systems require special bracketry.
Larger motors also need very secure
mounting techniques for the drive
systems. The addition of an arm or
two, and the builder is faced with
new metal structural needs. Cutting
and bending sheet metal or
machining metal becomes a bit scary
for many folks.
A builder might go the way of
using metal channels and associated
bracketry made by ServoCity/
Actobotics or similar suppliers as
structure for their larger robot. Many
manufacturers have used these
popular structural components for
several iterations of prototypes, but
eventually desire to go with bending
sheet metal or using machined metal
parts for a final product. Only a few
experimental robot builders ever get
to this point. Most experimenters cut
sheet metal and attach pieces
together with brackets or by welding
to form final structural components.
This is also the area where a lot
of potential manufacturers face design
and manufacturing problems and
begin to cut back on desired features.
Funding becomes an issue and
multiple design reviews sometimes
turn ‘great’ robots into ‘good’ robots.
DRC Robots are
Great Examples
My September column
highlighted the DRC this past June
and the 25 amazing entrants. Though
six of the entrants used previously-developed million-dollar plus Atlas
platforms, there was tremendous
ingenuity evident in the various
robots. Team Tartan’s entry, CHIMP
(shown in Figure 10) used rollers on
its four appendages as testing proved
that basic walking might be difficult in
a disaster area. Yes, it was very
problematic for all the entrants. The
winner — KAIST’s modified Hubo —
did the same thing with wheels on its
lower extremities. Team GRIT, Team
Robosimian, and Team Nimbro Rescue
also deviated from a basic bipedal
humanoid. Why? Because walking is
really difficult for robots in uncertain
environments.
Marc Raibert of Boston Dynamics
(who I mentioned earlier) has long
been the world’s expert on legged
robots. I first saw his one-legged
“Raibert Hopper” back in ’ 81 or ’ 82
that’s shown in Figure 11. It was
sitting on a rack in his lab, but I saw a
keep from falling over, but
it was an excellent example
of balancing techniques —
long before we had the
MEMS accelerometers and
gyros of today. Raibert and
his team went on to design
Big Dog, a robot cheetah,
much of the Atlas robot’s
systems, and other legged
robots.