DPRG
DPRG List  



[DPRG] Robots in te news

Subject: [DPRG] Robots in te news
From: Chuck McManis cmcmanis at mcmanis.com
Date: Fri Nov 2 11:59:05 CST 2001

Blake wrote:
> > IMHO,
> > It seems much harder to build a robot that thinks on its own than to 
> build a
> > robot that you control.  (Anyone) could build a spinning piece of metal 
> and

And Ralph responded:
 > I'm with you! I don't consider the radio-controlled machines
 > to be robots, but I guess "robot" is a handy term to spare mass
 > media writers some effort when reporting these events.

Lee Felsenstein, the guy that invented the Sol-20 computer way back when, 
calls them 'Golems' rather than 'Robots' as a way of distinguishing their 
nature.

However, having been where Blake is (they're just armored R/C cars) and 
having gotten past that to actually build one, now four, I have found that 
I can no longer share Blake's opinion.

It is _not_ easier to design a mechanism to move energy from one platform 
into another.
It is _not_ easier to design control electronics for hundreds of amps of 
current.
It is _not_ easier to design pneumatic actuation systems that are impact 
resistant.

These are very hard problems to do well and are two of the three legs of 
the "True Robotics" tripod (Software, Hardware, Mechanics). They are just 
as hard as writing software that does path planning through a cluttered 
field, the difference is that they use another skill set.

To build a competitive combat "robot" is just as difficult and just as 
rewarding as building a "thinking" robot. The pejorative comment "(Anyone) 
could build a spinning piece of metal ..." might easily be turned around to 
say "Anyone could build a wall hugging maze solver." But the MicroMouse 
folks have taking maze solving to a place where its serious software to get 
right, and darn difficult to become competitive. The same is true for 
BattleBots type robots.

In one sense Blake is absolutely correct, anyone _can_ put a spinning piece 
of metal on a box with a couple of motors hooked up to an R/C receiver and 
a couple of COTS speed controllers. However, that mechanism will be quickly 
and easily defeated by someone who has developed some expertise in building 
these kinds of machines. Just like anyone could build a wall following maze 
solver but it would be beaten easily by a maze solving robot using full 
quadrant obstacle sensing.

So before you conclude, as I had, that its a "sport" of hooking up the 
pieces and fancy marketing, go through the exercise of designing a 
competitive bot. You may be surprised, I certainly was.

--Chuck



More information about the DPRG mailing list