Clemson team SpeedTiger finished in Top 5 finalists at the F1TENTH IFAC World Congress Race. F1TENTH is an international community of researchers, engineers, and autonomous systems enthusiast. This year, due to the pandemic, it is the first virtual competition where teams would race head to head in a single-elimination bracket at the 1st Virtual International Federation of Automatic Control World Congress (IFAC-V 2020). Teams have to accomplish three benchmark tasks. The track was provided only two days before qualifying.
- Benchmark 1 is a single agent time trial without obstacle on the track. The objective is to achieve lower lap times.
- Benchmark 2 is a single agent task with unknown obstacles in the map beforehand. The objective is to finish laps without crashing.
- Benchmark 3 is a task where two agents compete simultaneously on the same track. The objective is to finish a certain number of laps before the other agent.
Of the 13 teams that submitted entries for qualifying, only 9 made it to the final round. Qualifying teams competed against each other and had to develop racing strategies overtaking opponents and avoiding crashes. Clemson Team SpeedTiger finished in Top 5 finalists at the Grand prix competition.
The team was comprised of Dr. Yiqiang Han, Faculty Advisor, Research Assistant Professor, Rongyao ”Tony” Wang, Mechanical Engineering (ME), Graduate Student, Alexander Krolicki, Mechanical Engineering (ME), Senior and Jonathan Daniel, Computer Science (SoC), Sophomore. This is the first time for the team to participate in such event.
For the autonomous racing competition, the team investigated the pure pursuit navigation algorithm and model predictive control. The team created heuristics based on apriori information to achieve faster lap times while still maintaining the ability to safely maneuver around dynamic obstacles for purposes such as collision avoidance and overtaking strategies. The team looks forward to participating again in the upcoming competition at the International Conference on Intelligent Robots and Systems (IROS) in October 2020. All of this was accomplished in just 15 days.
The research group has been focusing on both hands-on and coding aspects of the autonomous control development. Students learned how to configure and tune their algorithms to perform well with various embedded computing platforms and sensor configurations. The introduction of a virtual racing simulator also enabled the benchmarking of the proposed algorithms and opened access to the research frontier of embodied AI for students. The students plan to incorporate the latest development of Reinforcement Learning and deploy algorithms onto the actual car. The team has been growing over the past 2 years and is always looking to bring on new students with a multi-disciplinary background to learn more about building and testing autonomous systems.
