Indy Autonomous Challenge
August 2020 - October 2021
The Indy Autonomous Challenge was my fast-paced, somewhat overwhelming introduction to robotics, and I wouldn't have it any other way. I'd recommend that anyone who wants to try something new, especially robotics, dive in head first and learn along the way. I was part of the University of Pittsburgh team joined by the Massachusetts Institute of Technology, the Rochester Institute of Technology, and the University of Waterloo to form MIT-PITT-RW. I learned a ton from this project, met some extremely talented people from across the world, and traveled to see the culmination of our team's work.
In this project, we developed a software stack for a full-scale autonomous Indy racecar to race head-to-head against other autonomous vehicles at the Indianapolis Motor Speedway. The high-speed nature of this competition (up to 200 mph) created constraints on the software stack's design, forcing us to solve the problem of processing sensor data quickly enough to act precisely, even when operating at the physical limit of the vehicle sensors. I worked on the perception sub-team where our goal was to find a method to detect and localize other racecars that balanced speed with accuracy to create a robust system. Specifically, I worked on a high-speed learning-based monocular 3D object detection method called KM3D, which we adapted and implemented. As the team expanded, I became the leader of the perception sub-team, where I managed the technical direction while learning more about robotic perception and full-stack integration from Pitt and CMU graduate students.
Fundraising was essential to this project, given the vehicle cost, travel for team members, equipment and services, etc. I also worked on the business team to help secure over $200K in funding from sponsors. Being a part of the business and management teams was a rewarding experience which taught me about the facets necessary for a successful robotics project outside of the technology itself.
In the summer of 2021, I traveled to Indianapolis to work in person. I helped out with testing on the Indianapolis Motor Speedway and the Lucas Oil Raceway. Up until then we had been testing our code in simulation, so it was incredible to not only see the car, but to work on it and test it. It was also the first time I had met many team members in person since most of the competition happened during the pandemic. Overall, it was a great experience where all the teams collaborated to figure out vehicle issues. We worked closely with the Korea Advanced Institute of Science and Technology to do much of our testing over the summer.
Race day was nervewracking even though I was there as a spectator and not working in the pit crew. The project scope decreased in the months leading up to the competition from a head-to-head race to a time trial with an obstacle avidance section. Although the competition had simplified, there was still a lot of pressure as we had worked on this for over a year. During our run, we almost completed our first lap, but crashed into a wall due to a localization issue. This crash was devastating in the moment, but we quickly came together as a team to support each other. Regardless of the results, we were one of the few teams that had made it all the way to the competition: raising enough funds, building a software stack and qualifying for the final race. On top of that, we were the only student lead team consisting entirely of graduate and undergraduate leadership.
I'm very thankful that I got to be part of this event and learn so much from everyone I worked with. I'm also very proud of the work I contributed to this project, and it's evident that I am a better engineer because of it. I left the team after the competition in October 2021 to focus on research and completing the final year of my undergraduate degree. The team has continued to compete and has shown growth every event. I continue to be a fan of MIT-PITT-RW and can't wait to see what they do next.
Images
