Start of funding 01.01.2007

Dynamic 3D object maps for autonomous mobile robots

Prof. Dr. Michael Beetz
Technische Universität München
Intelligent Autonomous Systems Group

Prof. Dr. Kurt Konolige
SRI International
Artifical Intelligence Center



The field of autonomous robotics has seen tremendous advances over the last decade ranging from mobile robots navigating as tour guides through museums, autonomous vehicles successfully conducting races through deserts, and wheeled robots skillfully playing robot football. In all cases the robots use informative maps of the environments they operate in and in many cases acquire these models automatically.

However, while these maps that can be acquired automatically are accurate and enable very challenging navigation tasks they are clearly insufficient to perform manipulation tasks. In order to navigate it suffices to consider the environment structure as a set of landmarks for navigation and as an obstacle that must be avoided. In contrast, to perform manipulation the environment has to be - at least partly - structured into objects of interests and the objects have to be considered as affordances for the robot to carry out its service tasks.

This is where the proposers see the field of automatic robotic model acquisition to be developing to. Thus, the proposers plan to coordinate their research activities and conduct common projects that investigate the automatic acquisition of object maps of the robots' environments where the objects are seen as affordances to perform the robots' tasks. Thus, objects in the environment maps are labeled with the roles they can take in the robots' intended activities.