[Brown CS Talks] Brown CS Seminar: Henrik I. Christensen & Danica Kragic in Lubrano on
5/10/02 at 3 pm
talks-admin@list.cs.brown.edu
talks-admin@list.cs.brown.edu
Tue, 07 May 2002 12:22:18 -0400
CS Seminar
The Department of Computer Science
BROWN UNIVERSITY
presents
Henrik I. Christensen & Danica Kragic
Center for Autonomous Systems
Royal Institute of Technology
Stockholm, Sweden
Friday, May 10, 2002 at 3 pm
Lubrano Conference Room (CIT 4th floor)
Vision for handling of objects
Abstract
Demographics suggest that personal assistance might become a major new
business application. Personal robots are of tremendous value to allow
people to remain in their homes. Development of such machines requires
that the robots have facilities for interaction with humans and
methods for handling of objects. Handling of objects requires
recognition/localisation of a rich set of objects, servoing onto the
objects, and grasping them for manipulation. Visual servoing work has
traditionally focussed on single techniques with limited emphasis on
the overall task. The variability in geometry from distant objects to
close interaction implies that no single technique is adequate for
handling of objects. There is thus a need to integrate techniques for
localisation, 2D visual servoing, homography-based object alignment,
pose estimation, and 2.5D feature-based servoing to enable
manipulation of everyday objects.
In this presenation the overall system context is presented, the basic
methods for manipulation are outlined and it is described how these
techniques can be integrated into an operational framework for
operation in a realistic everyday environment.
Traditional computer vision research on visual servoing has emphasised
control performance and closing the loop in real-time. Recent progress
on biological systems suggests that natural systems might have a
stonger dependency on model-based predictive control rather than
closed-loop control. The integration of visual servoing, grasp
analysis and control are finally discussed in the context of a
biologically plausible strategy for manipulation and grasping.
Host: Professor Michael Black