SoC Distinguished Lecture Series
Visual Graphical SLAM
Henrik I Christensen
Robotics and Inteligent Machines
Georgia Institute of Technology
Host: John Hollerbach
Monday, January 14, 2008
3147 MEB
Refreshments 3:20 p.m.
Lecture 3:40 p.m.
Simultaneous Localization and Mapping (SLAM) is in theory a mature problem with a clear mathematical formulation. However, for real applications such as vision based mapping the methods are challenged by problems in data-association, linarizations and introduction of topological constraints. These problems are present for any SLAM application. When considered in the context of vision where feature discrimination typically is weak and parallax migth be small the problems are more pronounced. To provide a scalable solution and ensure robust processing in the presence of poor signal to noise ratios it is possible to adopt graphical solutions to estimation. A framework for robust SLAM is presented. The method allow flexible handling of data association ambiguities and easy introduction of topological constraints. Methods for representational optimization are also discussed. Example applications of the methods for indoor and outdoor scenarios are presented.
Bio
Henrik I Christensen is the KUKA Chair of Robotics and Director of the Center for Robotics and Intelligent Machines at Georgia Tech. Prior to Georgia Tech he has held position at Royal Institute of Technology (KTH), UPENN, and Aalborg University. He is was the founding chairman of the European Robotics Network (EURON). Dr. Christensen does research on estimation, human-robot interaction, and systems integration. He has published more than 250 contribution across vision, robotics and AI. He is the co-founder of 4 spin-off companies and serves on the editorial board of several leading journals in the field incl IJRR.
Return to 2006-2007 Events Calendar