From Course to Competition: RoboRacer Students’ Journey

_{^{Written by Jillian Mallon}}

_{^{Caption for featured image: MEAM MSE ’26 graduate Cedric Hollande (left) and ROBO MSE ’26 graduate Dhyey Shah (right) with their vehicle at the ICRA 27th RoboRacer Autonomous Racing Competition.}}

Ever since the RoboRacer course was added to the Electrical and Systems Engineering curriculum by Professor Rahul Mangharam a decade ago in 2016, it has garnered attention from both potential graduate students interested in the class as well as a global network of engineers in the field of autonomous systems. The course, originally called F1TENTH, focused on developing the best control, perception and planning algorithms for cars at one tenth of their actual scale. It has since expanded outside of the classroom into autonomous racing competitions held at various conferences throughout the years.

Recently graduated Robotics master’s student Dhyey Shah and MEAM Master’s student Cedric Hollande enrolled in the course in Spring 2026. After just a month of instruction, they knew they wanted to try to take this experience to the next level and compete in the 27th RoboRacer Autonomous Racing Competition at the 2026 IEEE Conference on Robotics and Automation (ICRA 2026) in June.

“RoboRacer/F1TENTH was the first class I heard of at Penn, which really got me excited about the program, and both Cedric and I follow the actual F1 as well,” recalled Shah. “I was interested in the competition aspect of it, so we were trying from the first month of the class to get the logistics in place for competing.”

Hollande echoed this statement, sharing, “I always wanted to take this class, but never realized that it would become the topic I am most interested in regarding robotics. I loved seeing hardware and our algorithms work during the class, so getting the chance to go to a conference, meet a lot of people, and see what the competition is like in a huge setting would take it to another level.”

One of the main reasons that the pair decided to compete together was to implement some of the more advanced model-based control algorithms that they learned and developed throughout the RoboRacer class. Hollande incorporated a Model Predictive Path Integral (MPPI) algorithm into his controller for trajectory optimization during the course, and Shah had tested a Reinforcement Learning (RL) pipeline as his racing stack.

“When the organizers released the plan for the ICRA competition, we encountered new obstacles that we had not seen in the races at Penn, including a long bridge that cars had to go under and above, making our localization more challenging with a 2D LiDAR. The competition level at Penn was amazing, but at ICRA we were faced with teams that had years of experience in competitive races like this one,” Shah explained. “Generally, a lot of the RoboRacer competition teams use pure pursuit, which is a geometrical controller, which doesn’t incorporate the dynamics of the vehicle. We chose MPPI because we were some of the few people who were actually aware of such algorithms and that was because of the class.”

After five months of planning, Shah and Hollande made it to Vienna, Austria to compete in the ICRA RoboRacer competition as the UPenn Autonomous Racing team. After two days of training and practice sessions on June 1st and 2nd, the time trial competition took place on the 3rd, with UPenn Autonomous Racing winning second place with a best lap time of 16.179 seconds.

The team was poised to perform just as well at the Master Cup the following day, but due to a mishap at the bridge portion of the track, they placed in fifth.

“I honestly think we had the best algorithm, but we did not have the best hardware. We were going really fast, and you could tell that it was interesting and different. The community is starting to see smarter controllers by choosing model-based approaches,” Hollande reflected.

While their novel controller approach did not result in a competition win, it did result in the two of them being invited to present their methods to the attendees.

“The competition felt like a full conference experience because we definitely had a big task on our hands as compared to presenting a single paper or attending a particular workshop in the conference. We didn’t end up in the top 3 for ICRA, but we ended up presenting our work,” Shah said. “Exchanging new ideas with other teams is really important as a learning lesson, and also it attracts a huge crowd, so there are networking opportunities. We had people coming up to us during the conference asking us about our implementations. I think the real value is actually testing your theoretical knowledge on the actual car, and comparing it to that of different teams from a lot of different places. It was awesome to see that RoboRacer is in over 90+ universities worldwide and competitions are a great way to be a part of the global community for autonomous machines.”

Instead of stopping at ICRA, Shah and Hollande were determined to prove the efficiency of their methods and signed up to compete at the 28th RoboRacer Autonomous Racing Competition during the 37th IEEE Intelligent Vehicles Symposium (IV 2026) in Detroit, Michigan in late June. With more preparation and planning time at their disposal, they competed as separate one-person teams, with Hollande using the MPPI controller while Shah used a RL-based controller. Hollande received first place in the competition. Two Robotics master’s students who also took the RoboRacer course the same semester, Milan Manoj and Manasi Srekhar, also competed together as a team at the IV 2026 competition and received the second place prize.

Hollande has also taken the helm of planning the upcoming 31st RoboRacer Autonomous Racing Competition, which will be taking place at the 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2026) from September 27th to 30th.

Overall, both Shah and Hollande would highly recommend that future RoboRacer students take on the challenge of competing at RoboRacer conference competitions.

Hollande shared, “What I found interesting is that the teams at the Penn RoboRacer course are really advanced in terms of algorithms. That is one thing that stood out to us, how the class is really high quality. I think that all of the teams we competed against at Penn would have performed really well at ICRA after only four months of racing in the RoboRacer class.”