ODD-aware decision-making for autonomous driving with real-world validation

Autonomous Driving Systems (ADS) are typically designed to operate within well-defined Operational Design Domains (ODDs), yet current behavior planning architectures rarely account for ODD constraints explicitly or adapt when those constraints are stressed. This project will investigate how behavior planning and decision-making can be made ODD-aware (capable of reasoning about operational boundaries and adjusting actions accordingly) in a principled and safety-conscious manner. The collaboration brings together complementary expertise: TUM-AVS contributes structured modular planning architectures, runtime safety verification, and the EDGAR research vehicle platform; CSIC-UPM AUTOPIA contributes situation awareness research, decision-making methodology, and real-traffic experimental infrastructure through its testbeds in Spain. Together, the two groups will explore how ODD constraints can be integrated into behavior planning frameworks, how planners should respond as operational conditions evolve or degrade, and how to evaluate such systems rigorously across simulation and real-world environments. The research visit will be structured around three phases. The first phase will begin prior to the in-person stay, through online meetings and remote collaboration, focusing on a joint review of the state of the art, alignment of research methodologies, and identification of the most promising integration points between the two groups existing tools and platforms. The second phase, coinciding with the physical visit, will involve the design and prototyping of ODD-aware behavior planning components, evaluated in simulation using shared benchmark scenarios. The third phase will consolidate results, define follow-up research directions, and produce initial dissemination outputs. Throughout the visit, the student will work embedded within the TUM-AVS group, with regular remote coordination with the AUTOPIA team. The visit will serve as the starting point of a longer collaboration, with the goal of producing initial results and laying the groundwork for future joint publications and funding applications.