VASC Seminar Announcement
				=========================

Date: Monday, 2/15/99
Time: 3:30-4:30
Place: Smith Hall 2nd Floor Common Area

Speaker: Frank Dellaert
         Carnegie Mellon University

Title:  Vision-Based Monte Carlo Localization using Large Scale Ceiling Mosaics

Abstract: 

Monte Carlo Localization (MCL) is a revolutionary new method to solve the
mobile robot localization problem, based on a particle filter (known in the
CV community as the Condensation algorithm). It is easy to implement, yet is
an order of magnitude faster than comparable robot localization methods,
while being more accurate. In my talk I will discuss how to use MCL in 
conjunction with large scale ceiling mosaics and a simple vision based 
sensor. The first half of the talk will focus on building these mosaics 
from a large set of noisy and distorted images, gathered by the robot. 
In the second half I discuss how this mosaic is used in the MCL framework, 
and I present experimental results from the Minerva robotic tour-guide project.


-----------------------------------------------------