Newsgroups: comp.robotics
Path: cantaloupe.srv.cs.cmu.edu!europa.chnt.gtegsc.com!news.mathworks.com!uunet!world!pbrennan
From: pbrennan@world.std.com (Patrick M Brennan)
Subject: Re: Object Recognition Using basic sensors
Message-ID: <D8BD9C.DEn@world.std.com>
Organization: The World Public Access UNIX, Brookline, MA
X-Newsreader: TIN [version 1.2 PL2]
References: <hlong-0805951806040001@vernonst-79.dynamic.trincoll.edu>
Date: Tue, 9 May 1995 13:53:36 GMT
Lines: 31

Hasan Long (hlong@mail.trincoll.edu) wrote:
: I would like to build a mobil robot base with an array of sensors and
: possibly some manipulators, which can "learn" and "remember" how to
: negotiate different changing environments like an apartment with people
: and pets and open and closed doors.  It would catagorize things into
: stationary furniture, moveable furniture(likely to be moved by humans),
: moveable objects, walls, stairs, obstacles, doors, people, and pets.  Any
: ideas as to what sensors would be needed short of using video, to
: recongize these "features"?

Mister Long, I think you've put your finger on one of the Holy Grails
of the robotics movement, professional as well as hobbyist.  Object
recognition is an active and open area of research, and nobody has
a guaranteed algorithm for uniformly categorizing sensor input.  I do
think, however, that if you are going to _a_priori_ rule out using video, 
then you are putting yourself at a distinct disadvantage for object
categorization.  Indeed, perhaps you should rethink the reasons for which
you want to put a label on to objects.  You may come to the conclusion
that object "recognition" is unnecessary in order to generate "intelligent"
behavior.

Patrick

===================================================================
Check out my home page! http://world.std.com/~pbrennan
check out my robot!     http://world.std.com/~pbrennan/ryobot.html
===================================================================




