The RoboRacer platform (formally known as F1Tenth) enables us to make significant contributions to autonomous navigation research through a scaled platform. Equipped with 2D and 3D LiDAR, stereo cameras, an IMU, powerful computing units, and robust motor control, it is a miniature version of the hardware used in our full-size vehicles. The RC platform can reach speeds of up to 100 km/h, allowing us to realistically simulate complex traffic situations such as intersections or traffic jams—at a fraction of the cost. At the same time, algorithms can be tested in challenging risk scenarios with minimal human supervision.
MIND-Stack: Modular, Interpretable, End-To-End Differentiability for Autonomous Navigation
F. Jahncke, J. Betz
IEEE Intelligent Vehicles Symposium (IV), 2025
doi: 10.1109/IV64158.2025.11097814 , PDF, Video, Code
GitHub Repository: RoboRacer-3DLiDAR
This repository provides the necessary drivers to operate the RoboRacer cars equipped with a 3D LiDAR sensor and, additionally, includes a robust and tested 3D LiDAR SLAM package.
Code
F1TENTH: Enhancing Autonomous Systems Education Through Hands-on Learning and Competition
J. Betz, H. Zheng, Z. Zang, F. Jahncke, F. Sauerbeck, Y. Zehng, J. Biswas, V. Krovi, R. Mangharam
IEEE Transactions on Intelligent Vehicles (T-IV), vol. 73, no. 8, pp. 10916-10931, 2024.
doi: 10.1109/TIV.2024.3495227, PDF, Video, Code
GitHub Repository: RoboRacer-Auxiliaries
A collection of essential helper tools and resources created to facilitate the experience of working with RoboRacer autonomous racing cars.
Code
Unifying F1TENTH Autonomous Racing: Survey, Methods and Benchmarks
B. Evans, R. Trumpp, M.Caccamo, F.Jahncke, J. Betz, H. Jordaan, H.Engelbrecht
arXiv Preprint, 2024
doi: https://arxiv.org/abs/2402.18558, PDF
Felix Jahncke Differentiable End-to-End Software Architectures | Email: felix.jahncke@tum.de |
Yuan Gao Foundation Models | Email: yuan_avs.gao@tum.de |
| Muhammad Talha Master's Thesis - Control | Email: ibad.talha@tum.de |
| Simone Dario Master's Thesis - Control | Email: simone.dario@studenti.univr.it |
| Nicolas Hanna Master's Thesis - Planning | Email: nicolas.hanna@tum.de |
| Michael Angerer Master's Thesis - Active Perception | Email: michael.josef.angerer@tum.de |
| Paul Klopfer Semester's Thesis - RoboRacer Dataset | Email: paul.klopfer@tum.de |
| Purujit Vasuki Semester's Thesis - RoboRacer Dataset | Email: purujit.vasuki@tum.de |
| Nils Benjamin Ohlert Bachelor's Thesis - Perception | Email: benjamin.ohlert@tum.de |
| Erich Efremov Bachelor's Thesis - Planning | Email: erich.efremov@tum.de |
1. Research:
PhD students use RoboRacer to test their algorithms. They can quickly begin initial real-world tests without spending too much time in simulation. The research covers the entire pipeline from perception to control, with a particular focus on learning-based and differentiable algorithms.
2. Thesis:
Students have access to a RoboRacer vehicle from the very beginning of their bachelor's, semester's, or master's thesis. Within six months, they work on their topic directly on the real hardware, gaining valuable practical experience. In addition to technical knowledge in embedded computing, Ubuntu, ROS2, Python, C++, and machine learning libraries such as PyTorch, they also acquire important skills that prepare them for the professional world.
3. Practical Lecture Course:
In the RoboRacer practical course, students learn the basics of autonomous driving step by step over one semester. Everything from understanding Ubuntu and ROS to independently developing algorithms for perception and control is taught in a hands-on manner. The internship concludes with an independent project in which students work out a problem, implement a solution, and finally present it and document it in a short paper.