Adaptive autonomy is a central research focus of the chair. The goal is to develop autonomy for robotic systems that can flexibly adapt their perception, decision-making, and action behaviors to changing environmental conditions, tasks, and system states. This adaptation occurs both in a data-driven manner based on current sensor data (bottom-up) and in a situational and knowledge-based manner through context, mission objectives, and experience (top-down).
The department focuses on the development of reliable autonomous robotic systems that operate reliably, safely, and in a traceable manner despite uncertainties, disturbances, and incomplete information. The underlying engineering approach encompasses both software and hardware development and considers the entire life cycle of autonomous systems.
Behavioral networks and methods of behavior-based robotics are a central concept in the chair’s research on the modeling and implementation of adaptive and reliable behavior in autonomous robotic systems. The central idea involves a strict decomposition of autonomy into behavioral modules, as well as the targeted design of their respective interactions. This enables decentralized adaptation of control and perception, allowing autonomous systems to respond flexibly to changes in the environment, the task, and the system’s state.
Autonomics describes a young, interdisciplinary field of study that comprehensively addresses the design, understanding, and deployment of autonomous systems. It focuses not only on traditional technical aspects such as perception, decision-making, and action in autonomous systems, but also on overarching issues in science and society. These include issues of reliability, safety, human-machine interaction, ethical and legal frameworks, and the integration of autonomous systems into complex, real-world environments. The goal of Autonomics is to holistically research and validate autonomous systems and to sustainably translate them into innovative applications.
The off-road environment places particularly high demands on autonomous robotic systems, as it is characterized by unstructured, dynamic, and only partially predictable environments. Ground conditions, vegetation, weather conditions, and visibility can change continuously, leading to highly variable demands on perception, decision-making, and control. The chair’s research addresses these challenges through adaptive autonomous concepts, robust perception and control strategies, and reliable system architectures that enable the safe and efficient deployment of autonomous systems even under changing operating conditions.
Finroc is a powerful, modular software framework for the development and operation of complex robotics and control systems in research and practical applications, which has been under development since 2008. Its objective is to efficiently build autonomous systems from reusable, quality-tested components. In addition, a wide range of tools is available—such as those for visualization, data acquisition, and simulation support—which significantly facilitate the development, analysis, and deployment of robotics software. Finroc is used in both research and industry and is constantly being further developed.
The REACTiON architecture is a modular, behavior-based system design that was specifically developed to make the perception and decision-making capabilities of autonomous robots significantly more robust and adaptive in unstructured off-road environments. The core idea is not to apply classical data-driven algorithms such as neural networks in isolation, but rather to process and dynamically evaluate their results in the context of behavior, environmental knowledge, and situational meaning. The architecture allows for automatic reconfiguration and parameterization of the perception and processing processes according to the characteristics of different robot platforms and sensor systems, enabling high reusability and flexibility across various systems and application areas.
This research data repository provides multimodal datasets that were collected during extensive off-road field trials under demanding real-world operating conditions. Data was collected using a Unimog equipped with a synchronized sensor system consisting of a laser scanner, LiDAR, cameras, an IMU, and GNSS. The datasets document complex scenarios across a variety of terrain and environmental conditions. As such, the repository serves as a valuable foundation for research in the fields of autonomous off-road navigation, sensor fusion, localization, and robust terrain perception.
Our department drives innovation through research projects that combine theoretical foundations with practical applications. We are currently focusing on three funded core projects:
• TEACH-R (2025–2026): Use of real robots in teaching to develop innovative teaching formats.
• IMARO (2025–2028): Development of a middleware solution for cooperative robot swarms.
• AI-SafeHumanRobot (2026–2028): Semantic environmental perception and situational safety assurance for humanoid robots in challenging environments.
"With these initiatives, we are continuously pushing the boundaries of what adaptive autonomous systems can achieve."
Discover our latest findings. We transparently share our research findings on adaptive systems, autonomy, and off-road technologies with the global scientific and industrial communities.
Source: Robotics Institute Germany / German Robotics Conference. Original video on YouTube: https://www.youtube.com/watch?v=6vi1EP3phRE
