Interpreting Range Data for Mobile Robot Feature Detection
Speaker: Dr Martin Adams, Nanyang Technological University
When: 2003-02-24 11:30:00
Venue: 78-420
Host: Dr Gordon Wyeth
Abstract:Outdoor mobile robotic systems often utilise either RADAR or LADAR range
data to aid autonomous navigation. A key research component of the SLAM
(Simultaneous Localisation and Map Building) problem is that of
reliable feature detection and subsequent association.
This talk will demonstrate theoretical analyses and experiments which have
been carried out on various commercially available time-of-flight (TOF)
LADARS. This will show the importance of correctly modelling the
sources of systematic and random range errors committed by these
devices, and the impact that such errors can have on various feature
detection algorithms. It will be shown that if the amplitude of the
received signal is available, then a unique range variance value can
be estimated and assigned to each range data sample.
On-line feature detection algorithms will also be demonstrated,
which are capable of detecting and distinguishing between the edges
of planar surfaces, cylinders/trees and more generally changes in
surface radius of curvature in outdoor environments.
Finally, initial work aimed at solving the data association problem,
necessary as a prerequesite for reliable SLAM, will be
presented. Each individual range scan can contain many tens of
features, and these must be correctly associated and pruned when such
scans have been recorded from differing robot positions. Results with
real laser range data will be presented.
Biography:(biography unavailable)
Type: ITEE Seminar
Contact:Dr Gordon Wyeth, seminar host (wyeth@itee.uq.edu.au)
or Guido Governatori (ITEE seminar co-ordinator)
(guido@itee.uq.edu.au)