Newsgroups: comp.robotics
Path: brunix!sgiblab!darwin.sura.net!news-feed-2.peachnet.edu!concert!news-feed-1.peachnet.edu!umn.edu!msc.edu!apctrc!hawkings!natos114!zjoc01
From: zjoc01@hou.amoco.com (Jack Coats)
Subject: Re: Global Positioning Question
Message-ID: <1993Oct8.130037.29996@amoco.com>
Sender: news@amoco.com
Reply-To: zjoc01@hou.amoco.com
Organization: Amoco
References: <CEKvLD.Duo.2@cs.cmu.edu>
Date: Fri, 8 Oct 1993 13:00:37 CDT
Lines: 33

In article 2@cs.cmu.edu, deadslug+@CS.CMU.EDU (Richard Voyles) writes:
>>|> can the starting point be reset.  Some experimental systems for cars
>>|> use inertial guidence systems to locate the car on a city map.  Maybe this
>>|> is the best technology for you.
>>|> 
>>|> --
>>
>>These systems depend on an "accurate" map known a priori to navigate.
>>They generally use a kalman filter to update the inertial position of the
>>car based on the known topology of where you are driving. (i.e. if the
>>inertial sensors say you're in the middle of a block but you turn right,
>>the kalman filter updates your position to the closest intersection --
>>simplified description, of course)
>>
>>An "accurate" map is a stringent requirement for an indoor environment
>>designed for humans.
>>

At a robo-fest in Austin TX earlier this year, some folks from Univ. of Texas
at Austin were doing research on not having maps, but having robots map a
'house' itself using a topographic description rather than a geographic
description of distances and directions.  The topo description was mainly of
'interesting' points (corners, doors, etc).  After wandering a while (and
only being able to 'see' 20" or so to each side, the robot was able to come
up with a fairly good description of the house.

Putting this in practice would make an intresting project for a sr robotics
work, or even a masters.

---
-- 
                         Ideas: mine              LawyerSpeak: theirs

