Newsgroups: comp.ai.alife
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!spool.mu.edu!news.clark.edu!netnews.nwnet.net!ns1.nodak.edu!plains!altenbur
From: altenbur@plains.NoDak.edu (Karl Altenburg)
Subject: Re: Reactive moving systems
Sender: usenet@ns1.nodak.edu (Usenet login)
Message-ID: <D34v7J.F7v@ns1.nodak.edu>
Date: Sat, 28 Jan 1995 20:50:07 GMT
References: <3g8db6$hhl@vishnu.jussieu.fr>
Nntp-Posting-Host: plains.nodak.edu
Organization: North Dakota Higher Education Computing Network
X-Newsreader: TIN [version 1.2 PL2]
Lines: 63

Emmanuel ORAIN (orain@ecoledoc.ibp.fr) wrote:

: The question I ask is : "What if the environment is such that no quasi-
: linear trajectory is possible ??"

: For instance, what a such reactive travelling agent could do with the 
: following environment ? :


:    
:                    OOOOOOOOOOOO
:                               O
:      Start                    O                    Target
:        X                      O                       X
:                               O
:                               O
:                    OOOOOOOOOOOO


: Does anybody has examples of reactive systems in such situations ?
: How could they solve the problem without building a map ?
: Could they even do that ?

I've seen this type of behavior in my personal work with reactive robots
and in papers and simulations done by others.

Is the target detectable behind the walled area?

If not, than simple random wandering or wall following should get
the robot out of this cul de sac.

If the target is detectable, then a reactive robot may spend
a little time bumping into the walls.  However with a little
memory of the frequency of encounters with the wall, the robot could
change its behavior to overcome the local obstacle.

Two types of behavioral changes come to mind.

One involves increasing the magnitude of the obstacle avoidance 
behavior, so that at first, the robot bumps gently along the 
back of the wall, however through time it eventually does a 180 degree
turn and races out of the cul de sac.

This appraoch is used in Beer's artificial cockroach simulation (I think),
and I beleive his work on Avoiding Behavioral Loops addresses this
problem.

An example of the other behavioral change would be to become less and 
less attracted to the target so that lower level reactive behaviors
such as wandering or wall following take over.  In the extreme, the
detection of the target could eventually serve as a repulsive force
long enough for the robot to clear the cul de sac.

I'm not sure about the last question, but if you are asking if a map
could be made, the answer is yes, and there are several approaches to the
problem.  Mataric, for example found a solution that was purely
reactive, while others have developed more hybrid type solutions.

--
Karl R Altenburg				altenbur@plains.NoDak.edu
North Dakota State University, Fargo, ND  58105

All things are artificial, for nature is the art of God.  SIR THOMAS BROWNE
