Professor Ulrike Thomas studied Computer Science at the Technical University of Braunschweig, Germany, and at the University of Edinburgh, Scotland. She received her Diploma in 2000 from the Technical University of Braunschweig. From 2000 to 2007 she was assistant researcher at the Institute of Robotics and Process Control at the Technical University of Braunschweig. She worked in robot programming, sensor integration for assembly, assembly sequence planning and real-time robot control systems. In 2008 she received the PhD with summa cum laude. From 2007 until 2015 she worked at the German Aerospace Center (DLR) at the Institute of Robotics and Mechatronics in Oberpfaffenhofen in several research projects ranging from remote sensing and image processing to robot control. Since April 2015 she is full professor at Chemnitz University of Technology, where she leads the robotics and human machine interaction group. Her main research fields are tele-robotics, human-robot interaction, locomotion and perception with a focus in production and every-day services. With her research group she develops new intelligent systems for better humanoid robots, which can be used in manufacturing, service robotics, telemanipulation and surgery or in rescue situations.
In this presentation, different methods for robot-assisted assembly will be introduced. Initial methods were based on planning, whereas today's systems are data-driven and learn many processes. Methods for assembly planning are first presented, ranging from purely plan-based methods to game theory and learning-based methods. For assembly with uncertainties, Bayesian estimation methods can be applied, or the movement patterns can be learnt directly. Auto-decoder-encoder networks are shown for this purpose. Examples in human-robot interaction show that learning-based methods are replacing plan-based methods and a combination of them forces success.