Development of a Pipeline for the Comparison of Real-World and Synthetic Autonomous Driving Data

Lehrstuhl für Fahrzeugtechnik
Bachelorarbeit / Semesterarbeit /
experimentell / theoretisch /  

During the Indy Autonomous Challenge, team TUM Autonomous Motorsport captured large datasets of single and multi-vehicle racing at different racetracks. These datasets include Perception data from Lidar and Radar as well as Localization and odometry data such as the GPS position and velocities. The work packages of thesis can be split into two major parts: First, the already caputed real-world datasets have to be re-simulated on our existing Autonomous Driving Simulator. For this work package, a pipeline to convert the real-world data to the format used by the simulator and to synchronize different data sources has to be developed. Furthermore, the scenarios should be re-simulated and the synthetic perception data should be recorded. The second part of this thesis is the development of a pipeline for the comparison of the real-world data and the synthetic data from the first part of the thesis. This should be done using Object Detection algorithms trained with the real-world or synthetic datasets and comparing the performance with state-of-the-art metrics.

  • Programming experience (Python)
  • Ideally experience in working with neural networks and ROS(2)
  • Involved working attitude
Verwendete Technologien
Simulation, Object Detection, Neuronal Networks, Python, Git
FTM Studienarbeit, FTM IVS, FTM Huch, FTM Informatik
Möglicher Beginn
M. Sc. Sebastian Huch