by Matt Bunting
Pimp
My
Hexapod
Part 1
In the September ‘ 11 issue of
SERVO, the Geerhead column
was about my hexapod
design and implemented
projects. The platform itself
has been an evolutionary,
iterative design where I am
constantly upgrading parts.
Since it’s been been a year
and a half since the last major
upgrade, plans for a better
system are now designed to
alleviate the most common
issues.
I have been fascinated with robotics nearly since I can
remember. There was always something so intriguing to
me about building a small autonomous machine that
could interact with the world. The idea of building a robot
was akin to building a little creature. It has always
generated philosophical ideas, making me think about
56 SERVO 12.2011
how I accomplish a very mediocre task, and then how to
mimic that mathematically, electronically, and
mechanically. Very quickly it became apparent that even
some of the simplest tasks are nearly impossible to
accomplish with our current technology.
As a child, I had built many rover robots using LEGOS,
radio-controlled cars, and wireless cameras. The use of a
hot glue gun was key to much of my success. Though
none of my early robots were autonomous, I always
wanted to build an even more complex, elegant machine.
When I first saw a hexapod, the extraordinary number of
motors immediately piqued my interest. I thought to
myself that I would one day build a hexapod.
During high school, I dabbled here and there trying to
build a hexapod by scrounging up all the motors I had,
but I could not quite get to the point of making a strong
enough hexapod to walk. After working for a couple
years and saving as much money as I could, I was able to
splurge on some better motors. I was able to upgrade my
original design into a working, walking hexapod. I also
had a great excuse at the time, as I needed to build a
robot for a Cognitive Robotics class.
I resurrected what I had previously built and added
some better components, but it was nothing more than
an upgrade of my original build. The project turned out to
be pretty successful; I had the robot learn how to walk by
implementing Q-Learning, even though it took four
straight days for it to finally get it.
