Date: Mon, 14 Jul 1997 22:47-EDT
From: Sanjiv.Singh@H.GP.cs.cmu.edu
To: vasc-seminar@CS.cmu.edu, frc-people@FRC2.FRC.ri.cmu.edu
Subject: FRC Seminar, Hans Moravec, July 15 4pm

Time:	 4pm
Place:	 FRC 100
Speaker: Hans Moravec, Robotics Institute
Food:	 Yes.



	Stereo Vision, 3D Grids, 1,000 MIPS, Vacuum Cleaning!
 
1,000 MIPS is enough to update a dense stereo-derived 3D evidence
grid of a mobile robot's surroundings about once per second, probably
adequate to navigate a small indoor robot with high competence and
reliability. One application is a compact very autonomous home vacuum
cleaner, ready to go to work out of the box.
 
An evolving program to demonstrate this now occasionally runs on our
Uranus robot, another version on a robot at Daimler Benz research in
Berlin.
 
In its first implementation last year, the program added several
spatial rays of evidence to a grid for each of about 2,500 local
image features chosen per stereo pair. It constructed a 256x256x64
grid, representing 6 by 6 by 2 meters, from a hand-collected test
set of twenty stereo image pairs of an office scene.  Fifty nine
stereo pairs of an 8 by 8 meter laboratory were also processed.
The positive (probably occupied) cells of the grids, viewed in
perspec tive, resemble dollhouse scenes. Details as small as the
curvature of chair armrests are discernible. The processing time,
on a sub 100 MIPS Sparc 20, was less than five seconds per stereo
pair, and total memory was under 16 megabytes. The results seem
abundantly adequate for very reliable navigation of freely roaming
mobile robots, and plausibly adequate for shape identification of
objects bigger than 10 centimeters.