Newsgroups: comp.robotics
Path: brunix!uunet!stanford.edu!ames!elroy.jpl.nasa.gov!forsight!gat
From: gat@forsight.jpl.nasa.gov (Erann Gat)
Subject: Re: MIT Insect Robots
Message-ID: <1992Jun4.172854.19251@elroy.jpl.nasa.gov>
Sender: news@elroy.jpl.nasa.gov (Usenet)
Nntp-Posting-Host: robotics.jpl.nasa.gov
Organization: Jet Propulsion Laboratory
References: <1992Jun1.142056.3477@advtech.uswest.com> <kkoehn.707442528@sfu.ca> <1992Jun3.213155.24343@seer.gentoo.com>
Date: Thu, 4 Jun 1992 17:28:54 GMT
Lines: 29

In article <1992Jun3.213155.24343@seer.gentoo.com> tomk@seer.gentoo.com (Tom Kunich) writes:
>Insects can evolve into a niche. Robotic insects can't. It takes
>human beings to design them for any particular niche. Then the
>humans have to be smart enough to understand the needs of these
>things.
>
>So I would guess that the answer is that nothing will happen at all.

Neither of these assertions is true.  Robotic insects could conceivably
be designed to "learn" and even mechanically "evolve".  There is even
some preliminary research being done in this direction.  Granted it is
still in the very early stages, but just because we don't know how to
do something today doesn't mean it's impossible.

As for nothing coming of human-engineered robotic insects, here at JPL
we have built a prototype autonomous Mars Microrover with a control
system based on Brooks' insects, and we are close to building another one.
There is a good chance (though by no means certain) that one of these
artificial insects will explore Mars before the turn of the century.

Erann Gat
gat@robotics.jpl.nasa.gov

P.S.  A paper about the Mars microrover will appear in this year's
AAAI.  Please don't ask me for copies; I am swamped right now.  For
references on learning and evolving robots, a recent paper by Mahadevan
and Connell makes a good starting point.  Fukuda has some papers about
modular robots that can assemble copies of themselves.

