Dr.
Oliver Urbann

Oliver Urbann is Team Lead in AI and Robotics at the Fraunhofer Institute for Material Flow and Logistics in Dortmund. He received his Diploma (equiv. M.Sc.) in Computer Science from TU Dortmund University in 2010 and his PhD (Dr.-Ing., summa cum laude) in Computer Science from TU Dortmund in 2017, working on model-based real-time control of humanoid robot locomotion. With nearly two decades of experience in robotics, including many years at the Institute of Robotics Research at TU Dortmund, his work spans humanoid and mobile robotics, autonomous vehicles and transport systems, and the automation of complex, physically interactive systems.

His research focuses on (humanoid) robotics, learning-based control, in particular reinforcement learning for dynamic robot locomotion and navigation, resource-aware machine learning on embedded platforms, and perception and simulation methods that narrow the sim-to-real gap. Applications range from intralogistics and production to healthcare and human-robot collaboration. He has coordinated and contributed to numerous national and national and European research projects and serves as Associate Editor for the IEEE Robotics & Automation Magazine and the IEEE-RAS International Conference on Humanoid Robots.

Current research topics:

• Learning-based locomotion for dynamic robots (humanoids, wheeled and hybrid) in real-world environments
• Reinforcement learning and sim-to-real methods for robust autonomy of mobile robots and vehicle fleets in logistics and other domains
• Resource-aware machine learning and perception for embedded and edge systems, including tracking, localization and vision
• Human-centered Embodied AI for intuitive human-robot collaboration and cognitive capabilities in autonomous systems

Special interests:

Long-standing involvement in RoboCup humanoid robot soccer, including a world championship title in 2016 (SPL Outdoor), has strongly shaped my view of robotics research and education. I am particularly interested in how competition-driven, physically grounded experiments force us to build truly robust systems that can handle uncertainty, real-time constraints and complex team behaviour. I believe that teaching and supervising robotics projects benefits enormously from this mindset: students should experience the full pipeline from simulation and algorithms to reliable performance on real robots, under realistic constraints and with clear success criteria.