The Institute of Vehicle Concepts (FK) of the German Aerospace Centre (DLR) is internationally recognised for the design of future road and rail vehicles that enable climate and environmentally friendly mobility while being affordable and user-friendly at the same time. We research and demonstrate the required key technologies and maintain close cooperation with other scientific institutions as well as industrial and political bodies.
What to expect
We invite you to join our team for a master’s thesis at the intersection of autonomous driving and foundation models. Your work will begin with a state-of-the-art review of foundation models in robotics, followed by the design of a vision-to-controller pipeline for docking and parking maneuvers.
Based on the literature, you will implement a simulation-based prototype in CARLA. Following, you will carry your work over into a proof-of-concept integration on the real vehicle.
Your tasks
- Survey (state of the art): Conduct a systematic review of foundation model frameworks, vison-to-action models, vision-language models (vision → language), language-to-design approaches (planning/trajectory generation), and design-to-controller methods (low-level execution).
- Taxonomy & comparison: Analyze agent roles (perception narrator, planner, safety critic, controller designer), communication patterns, knowledge integration (RAG), and safety/runtime checks.
- Prototyping (simulation): Develop a pipeline with: Vision → Language, scene description/key-fact extraction from images or video. Language → Design, converting plans/trajectories into declarative formats (e.g., waypoints, constraints). Design → Controller, passing designs to controllers. Retrieval/RAG, accessing policies, safety rules, and vehicle-specific information. Feedback, critic/verifier modules (collision checks, dynamics/comfort constraints)
- Stretch-Goal: Minimal integration on the test vehicle
- Documentation & presentation: Produce clean, reproducible documentation (repo/readme) and a final presentation.
Your profile
- Ongoing academic studies in Computer Science, Robotics, Mechanical Engineering, Automotive Engineering, Electrical Engineering, Mechatronics, Industrial Engineering, Data Science, Mathematics, or Physics or similar
- Strong interest in autonomous driving, LLMs/generative AI, and multi-agent systems
- Basic knowledge of Python and Git; familiarity with ROS2 or willingness to learn
- Experience with simulation (e.g., CARLA/Gazebo/SUMO) or path planning/control is an advantage
- Foundational understanding of ML/deep learning (datasets, training, inference)
- Ability to conduct structured literature research and write scientific texts
- Bonus skills: Docker/Linux, RAG/vector stores
Remuneration is up to the German TVöD 05 depending on qualifications and assigned tasks.
We look forward to getting to know you!
If you have any questions about this position (Vacancy-ID 3645) please contact:
Daniel Diegel
Tel.: 08153 28 4961
