Abstract: Cognitive robots need to perceive their environment to act in a goal-directed way. This includes simultaneous localization and mapping, semantic categorization of surfaces, object detection and pose estimation, as well as the perception of persons. In the talk, I will report on efficient methods that we developed for environment perception, based on local multiresolution representations, deep learning, and object-centered representations. The resulting models form the basis for locomotion and manipulation planning, for which we developed efficient methods based on local multiresolution and abstraction. My team integrated perception and planning for autonomous control of multiple cognitive robot systems and demonstrated their abilities in complex environments, including humanoid soccer, domestic service, space exploration, disaster-response, bin picking, and industrial inspection. To improve behavior, learning is crucial. I will discuss ideas on how structured models for efficient perception and planning can be learned from limited experience.
Bio: Prof. Dr. Sven Behnke holds since 2008 the chair for Autonomous Intelligent Systems at the University of Bonn, Germany, and heads there the Computer Science Institute VI – Intelligent Systems and Robotics. He received a PhD in computer science from Freie Universität Berlin in 2002. In his dissertation “Hierarchical Neural Networks for Image Interpretation” he extended forward deep learning models to recurrent models for visual perception. In 2003 he did postdoctoral research on robust speech recognition at the International Computer Science Institute in Berkeley, CA. 2004-2008 Professor Behnke led the Research Group “Humanoid Robots” at Albert-Ludwigs-Universität Freiburg. His research interests include cognitive robotics, computer vision, and machine learning. His team NimbRo has won numerous robot competitions (RoboCup Humanoid Soccer, RoboCup@Home, MBZIRC, ANA Avatar XPRIZE).
Event Timeslots (1)
Room K3 – Koncerthuset
15 March - Academic Track