Newsgroups: comp.robotics
Path: brunix!sgiblab!sdd.hp.com!ihnp4.ucsd.edu!agate!doc.ic.ac.uk!uknet!EU.net!Germany.EU.net!news.dfn.de!zeus.rbi.informatik.uni-frankfurt.de!terra.wiwi.uni-frankfurt.de!news.th-darmstadt.de!surz03.hrz.Uni-Marburg.DE!news.belwue.de!news.uni-ulm.de!rz.uni-karlsruhe.de!subnet.sub.net!flatlin!faber!jr
From: jr@faber.ka.sub.org (Juerg Reinhart)
Subject: Q: coping with uncertainty due to incomplete knowledge
Date: Sun, 27 Feb 1994 15:43:19 GMT
Message-ID: <CLw3o8.L4w@faber.ka.sub.org>
Organization: -
Lines: 26

Coping with unstructured environments is one of the challenges of
robotics. One issue is that the robot cannot know everything in
advance. I am interested in approaches to model the uncertainty due to
incomplete information (as opposed to, e.g., incorrect sensory
measurements). In the mobile robot domain the presence, shape, or
trajectory of an obstacle may be unknown. As for the geometric aspect
I am aware of the certainty grid technique. But there may be a lot
more of interesting approaches which go beyond the traditional idea to
assume a closed world and plan only with those obstacles noted in the
terrain map which is updated only in the case an unexpected obstacle
is meet. Perhaps there are also approaches which directly incorporate
sensing into the navigation function.

Uncertainty due to incomplete knowledge comes in several modalities. A
box someone forgot to remove which the robot bumps into is something
else than a chair which may stand right before a table or may be moved
off. There are numerous cases of this kind or another. Perhaps,
someone has set up a classification scheme. My question is: What is
the best way to cope with any of the possible appearances of
uncertainty due to incomplete knowledge. I would be happy to hear of
any idea you have.

Thank you for your attention and help

- JR

