This year, Scuderia Segfault secured a top placement at the 15th F1TENTH Autonomous Grand Prix, held from May 13th to 17th, as part of IEEE International Conference on Robotics and Automation (ICRA 2024) in Yokohama, Japan. About 20 teams from universities and research institutes worldwide competed in this autonomous racing challenge. Team Scuderia Segfault from TU Wien achieved 4th place in the final head-to-head race competition.
Preparation and development
To prepare for this autonomous racing challenge, the students of Scuderia Segfault invested countless hours in preparing the model racing car and developing the necessary software to race fully autonomously. First, an optimized hardware setup was needed, with carefully adjusted suspension, the right tires for optimal grip on the racetrack, and visual chassis indicators. Thanks to the Faculty of Informatics, the computing platform for this year’s competition was upgraded with a new powerful CPU and much more memory to run complex algorithms more efficiently. Most importantly, the students needed to implement software for fast localization on the racetrack, find the optimal race line, and precisely control the car. As the permanent room for testing was not yet available, students could only briefly test on weekends and evening hours in smaller seminar rooms and lecture halls, where all the equipment needed to be temporarily set up.
Arrival and setup in Japan
After a long journey to Japan, our team arrived early in the morning at the conference center to prepare all the hardware and software setups. The car needed to be re-assembled after transport, and the floor conditions in the competition hall were carefully evaluated for proper tire selection. When the track was set up by the organizers, a map of the racetrack was created so that the car could localize itself and determine its position. This map was also used to find the optimal race line. For this, our students developed a custom-made optimization tool to find the best speed and curve parameters. The first day was then used to test the cars on the track.
Time-trials: resolving last minute problems
On the next day, it was necessary to first demonstrate obstacle avoidance capabilities. Unfortunately, our team noticed severe problems in this software component. This could have led to a complete disqualification from the race. Urgent modifications were needed to identify the cause of this problem and fix it before the time-trial races. Our students mobilized all their problem-solving capabilities and worked in separate teams to test different scenarios and modify the software. This paid off, and the check was passed successfully.
In the time-trials, the objective was two-fold: the car should drive laps as fast as possible while also demonstrating its reliability. Therefore, the fastest lap and the most number of consecutive laps without crashing were counted. Our team was able to show good results in both categories. Compared to past competitions, it was noticeable that other teams had also greatly improved their cars and software. The fastest five teams were less than one second apart. Overall, Scuderia Segfault qualified in fifth place after the time-trials.
Final competition: head-to-head race
To prepare for the decisive head-to-head race, our team conducted further tests on the track. Unfortunately, our car was crashed into by another team from the back, and parts of the chassis mount broke. This required a full overnight repair to get the car running again. Several parts needed to be removed and fixed to solve the issue. For a complete checkup afterward, our team even tested the car in the underground parking lot of the hotel in the late evening. Furthermore, knowing the lap times of other teams, it was clear that the race software and parameters needed further tuning to stand a chance for a good placement. So further late hours were invested in working on this.
On the final race day, the first pairing was against Team Technion (Israel Institute of Technology). This was a clear win, as their car was not able to drive fast enough. However, it was important not to crash during overtaking maneuvers, so all the parameters and the setup needed careful adjustments for this race as well. The next race was against VAUL Renegade (Université Laval in Québec) – an extremely strong and fast team in the time-trials. It was clear to our team that this needed an all-in strategy. The lap time needed to be reduced by more than one second. They decided to use further critical adjustments, which were not tested yet, but absolutely necessary. This decision paid off fully, as the result was very close, with Team Scuderia Segfault winning and advancing to the semi-final. The next opponent was Dzik Team (Polish Academy of Sciences), which won very closely. The small final was against ForzaETH (ETH Zürich), and resulted in 4th place for Team Scuderia Segfault.
Thanks to our supporting partners
This race participation would not have been possible without the great support of Liquid AI. Our team is grateful for the generous support from TTTech Auto. They were further supported by HTU and Magna International.
