Safe, automated and energy-optimal inland lake navigation based on scene and intention prediction under uncertainty

• Applicants: Professor Oliver Dürr, Ph.D.; Professor Dr.-Ing. Johannes Reuter; Professor Dr. Georg Umlauf
• Subject Area: Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 564133053

The planned research project aims to develop new methodological foundations for safe, automated and energy-efficient navigation in inland lakes. The focus is on probabilistic modeling approaches, advanced sensor fusion methods, mathematical optimization techniques and control methods for safety-critical systems. In the long term, the resulting concepts should serve as a well-founded toolkit for making autonomous navigation systems robust and versatile. One focus is on the integration of heterogeneous sensor data (e.g. cameras, LiDAR, radar) into a uniform, probabilistic model. The combination of classical statistical signal processing, deep learning-based object recognition and advanced scene analysis algorithms creates a coherent, error-robust image of the environment. This flexible model is able to realistically take uncertainties into account and map different environmental conditions. Building on this, novel planning and decision-making strategies are designed that use probabilistic optimization approaches, Bayesian inference and Markov decision processes. In this way, navigation routes can be generated that keep collision risks to a minimum and react to changing situations. These dynamic planning processes take into account both uncertainties in perception and the evolution of the scene. The result is adaptive methods that make robust decisions even with incomplete and faulty information. At the same time, the focus is on minimizing energy requirements. By combining nonlinear optimization methods and reinforcement learning methods, a resource-efficient control concept is developed that reduces energy consumption without compromising on safety. This creates a theoretically sound repertoire of methods that extends far beyond the actual navigation task and can be transferred to other areas of autonomous robotics and autonomous driving. Overall, the project creates the basis for a new generation of highly autonomous and energy-optimized navigation systems that can be flexibly adapted to different requirements. These approaches, developed at a methodological level, have the potential to have an impact far beyond the specific field of application.

• DFG Programme: Research Grants (Transfer Project)
• Application Partner: Shiptec AG