Race to the Apex: Benchmarking Multi-Vehicle Trajectory Planning on Real Autonomous Race Cars
- Institut
- Professur für autonome Fahrzeugsysteme (TUM-ED)
- Typ
- Semesterarbeit
- Inhalt
- experimentell theoretisch konstruktiv
- Beschreibung
SEMESTER THESIS PROPOSAL
Race to the Apex: Benchmarking Multi-Vehicle Trajectory Planning on Real Autonomous Race Cars
Where: Professorship of Autonomous Vehicle Systems (AVS)
Supervisor: Dr. Mattia Piccinini (mattia.piccinini@tum.de)
Type: Semester thesis, hands-on and hardware-oriented
Start: as soon as possible------
WHY THIS THESIS?
Driving fast on an empty track is hard. Driving fast with an opponent right next to you is much harder. In head-to-head autonomous racing, a planner has to find the racing line, predict what the other car will do, decide when to attack and when to defend, and still respect the physical limits of the car. All of this happens in a fraction of a second.
Multi-vehicle planning is one of the open challenges in autonomous driving, and racing is a good place to study it: the scenarios are aggressive, the margins are small, and every lap gives a clear answer about which algorithm works better.
You will study this on the RoboRacer (https://roboracer.ai), a 1:10-scale autonomous racing platform. RoboRacer is an open-source community with more than 80 partner universities worldwide. The cars are smaller, but the problems are the same as on full-scale vehicles: perception, state estimation, planning, and control at the handling limits.
This is a hands-on project. Your algorithms will not stay in simulation: they will drive real cars, on a real track, against real opponents. You will gain a lot of real-world experience with autonomous driving algorithms.
------
WHAT YOU WILL WORK ON
- Collaborate with the AVS team to design and refine trajectory planning algorithms for multi-vehicle scenarios: overtaking, defending, and racing wheel-to-wheel. We have very good baselines ready to go, and we can improve them together.
- Deploy them on real cars. Implement the algorithms on our RoboRacers and run them in our indoor, configurable test track.
- Race them head-to-head. Evaluate the algorithms in competitive head-to-head real-world scenarios and build a benchmark that shows which approaches win, and why.The experimental setup is fully ready. Deploying algorithms on the RoboRacer is easy and user-friendly, so you can spend your time on what matters: testing algorithms on the track!
------
WHAT YOU WILL GAIN
- Real-world experience: you will see your code drive a real autonomous car, and learn what it takes to move from simulation to hardware.
- Solid skills in algorithm development: trajectory planning, multi-agent interaction, and real-time implementation.
- Transferable skills: what you learn applies to many other mobile robotics problems, from delivery robots to drones to full-scale autonomous vehicles.
- Close supervision in an international team working on autonomous vehicles and autonomous racing.
- Access to the RoboRacer community, with more than 80 partner universities.------
HOW TO APPLY
Send a short email to mattia.piccinini@tum.de with:
- your CV;
- your transcript of records (including all exams with marks).Ready to race? We look forward to hearing from you!
- Voraussetzungen
Required:
- Good programming skills in Python and/or C++
- Some experience with ROS / ROS2 and real hardware.
- Motivation to work with real hardware, and the curiosity to find out why an algorithm wins or loses a race!Nice to have:
- Background in trajectory planning and/or vehicle dynamics.- Tags
- AVS Piccinini
- Möglicher Beginn
- sofort
- Kontakt
-
Mattia Piccinini
mattia.piccininitum.de