Munich School of Robotics and Machine Intelligence, TUM
Talk Title: Form Building Robots to Bridging the Gap between Robotics and AI
Abstract: Major breakthroughs in lightweight articulated soft robots have enabled the real-world industrial use of adaptive impedance control, force control and high-speed collision handling under a unified framework. Together with smart reactive and user-centered programming systems that express a robot’s ability to physically interact with the world, a stage has been reached where even laymen are finally able to use cutting-edge robotics technology for tasks as complex as force-sensitive assembly or safe physical human-robot interaction. The first commercial robot system with these abilities is Franka Emika’s Panda, which “story in a nutshell” I will share. However, major challenges still remain in equiping these next generation robots with strong AI capabilities such as autonomous learning of physical human-robot interaction or human-level manipulation skills. Noticeably, neither pure control-based nor end-to-end learning algorithms have come anywhere close to human-level general purpose machine intelligence. In the meantime it has become clear by now that their algorithmic unification, i.e. bridging the gap between model-based nonlinear control algorithms and data-driven machine learning via a holistic approach, is a central open problem to be solved before we can leverage best of both disciplines. I will strenghen this statement with the help of two recent results: i.) the successful learning of exact articulated robot dynamics via the concept of first order principles networks and ii.) learning of human-performance generalizable manipulation skills by unifying adaptive impedance control and meta learning. To enable the robotics and AI community to build on our previous work and jointly move existing boundaries in manipulation, interaction and general AI-enhanced robotics, Panda was from the start on released with according research interfaces and modules. My sincere hope is that this step will give us the ability to join forces in order to tackle these grand challenges of robotics and AI research.
Bio: Sami Haddadin is Full Professor of Robotics Science and Systems Intelligence at TUM, Germany. He is also the founding director of TUM’s Munich School of Robotics and Machine Intelligence. His research topics include robot design, physical Human-Robot Interaction, nonlinear robot control, robot learning, real-time motion and reflex planning, optimal control, human motor control, and safe robotics. One of Sami’s goal is to bring human-safe and AI-enabled tactile robotics to the real world. His achievements encompass e.g. human-centered robots with an artificial sense of touch that can safely interact with the world and learn during these interactions how to self improve. His Munich-based robotics start-up FRANKA EMIKA pursues the vision of robots forming the next level of tools that make our lives easier and recently brought the high-performance, safe and affordable soft robotics platform Panda to market. Sami received degrees in EE, CS and Technology Management from TUM and LMU. He obtained his PhD from RWTH Aachen. He published more than 130 scientific articles and among others won several awards at the top robotics conferences and journals. He is a recipient of the 2015 IEEE/RAS Early Career Award, the 2015 RSS Early Career Spotlight, the 2015 Alfied-Krupp-Award and in 2017 of the German President’s Award for Innovation in Science and Technology.