Newsgroups: comp.robotics
Path: brunix!news.Brown.EDU!agate!howland.reston.ans.net!europa.eng.gtefsd.com!uunet!munnari.oz.au!metro!extro.ucc.su.OZ.AU!williamu
From: williamu@extro.ucc.su.OZ.AU (William Uther)
Subject: path planning - how to get an abstract representation of the world
Message-ID: <williamu.753666315@extro.ucc.su.OZ.AU>
Sender: news@ucc.su.OZ.AU
Nntp-Posting-Host: extro.ucc.su.oz.au
Organization: Information Services, Sydney University, Sydney, NSW, Australia
Date: Thu, 18 Nov 1993 23:45:15 GMT
Lines: 23

Hi,
  I have a small robot simulator that I'm trying to put a 'brain' into.  The
simulator passes knowledge about the world as a 2D array where each cell
contains the 'terrain' for that square.
  I would like to have some FAST way of planning a path between any two
points on the map.  I've tried A* search with each grid cell as a node and
remaining distance as a Heuristic, but this is about 100 X too slow.  I would
like to go a more abstract representation of the world so that I can try A*
(or a modification I have been working on) with the visible corners of objects
as the nodes.  There is also one added twist in that the map can change (there
is noise in the system), so there needs to be a way of updating the abstract
representation.
  Does anyone know of any algorithms for doing this form of conversion and,
if so, where there are any good descriptions of them.

     Thankyou

     Will       :-}
-- 
--------------"Nothing is so smiple that it can't be stuffed up!"---------------
                        O O                    williamu@extro.ucc.su.OZ.AU
  \x/ill Uther           |                     9106710@ugrad.cs.su.OZ.AU
                        \_/
