UC Berkeley autonomous racing team sets new road course record

Featured from left to right Allen Yang UCB, Siddharth Saha UCSD, Gary Passon UH, Raymond Song UCSD, Tianlun Zhang UCB, Haoru Xue CMU, Wade Koglin UH, C.K. Wolfe UCB (Photo credit Jordan Esley, Indy Autonomous Challenge) | Credit: ROAR

The AI Racing Tech Team led by UC Berkeley’s Robot Open Autonomous Racing (ROAR) program achieved the fastest autonomous lap time ever recorded on the U.S. Road Course track, at 1 minute 27 seconds at the Putnam Park Road Course in Greencastle, Indiana on August 31. The team’s custom Dallara AV-21 race car, which is considered the most advanced autonomous car ever built and valued at over a million dollars, broke the previous autonomous course record of 2 minutes 9 seconds.

The performance adds to a long history of success for the most decorated U.S.-based AI Racing team in the International series of the Indy Autonomous Challenge (IAC), a DARPA-inspired competition designed to encourage the next generation of STEM talent and innovation.

The AI Racing Tech Team’s record-setting performance was a triumph of developing novel sim-to-real learning tools and machine learning technologies to optimize the safe performance of an autonomous driving system, adapting in real-time to individual road courses with variations in terrain, configurations, and track characteristics. To put this accomplishment into perspective, the theoretical computational lowest limit for the fully optimized Dallara AV-21 car at Putnam Park Road Course is 1 minute 12 seconds.

“The sim-to-real transfer really pushed us,” says S. Shankar Sastry, UC Berkeley professor, ROAR faculty director, and racing team advisor. “You design in-silico, train extensively in simulation, and then see how the system behaves under the pressure of competition. Going 150 miles an hour really focuses the mind – and tests the resilience of the algorithms and neuro-symbolic design tools.”

The AI Racing Tech Team is made up of artificial intelligence and autonomy researchers and undergraduate and graduate students at four universities: Carnegie Mellon University, the University of Hawai’i, and the University of California, San Diego, in addition to UC Berkeley. Last month’s record-setting performance builds on the team’s past racing successes, coming in second at the 2022 IAC at Texas Motor Speedway and third at the Autonomous Challenge at CES 2023 in Las Vegas, in competition with eight other international IAC teams.

This record-setting exercise at Putnam Park serves as a training ground for two upcoming Challenge events: The Oval race at CES 2024 in Las Vegas, NV in January, and in June 2024 at the MIMO Road Course race in Monza, Italy at the famous Autodromo Nazionale Monza so familiar to Formula One fans.

This pioneering competition for extreme autonomous robotics signals a transformative era for industries grappling with the challenges of pushing automated systems to the limit and corner cases of design. The race environment mirrors the hardware challenges encountered in application domains such as aerospace, defense, and healthcare. These challenges also include expensive equipment, limited access, unknown conditions, and the need for safety and performance assurances.

“Extreme robotics needs provably correct performance guarantees, and the AI Racing Tech Team is among the first to deploy and experiment on these theories using full-scale race cars,” says Allen Yang, UC Berkeley FHL Vive Center Executive Director, ROAR Principal Investigator, and UC Berkeley racing team lead.

Follow the AI Racing Tech Team’s journey to prepare for the 2024 race season at airacingtech.com and social media.

The UC Berkeley Robot Open Autonomous Racing (ROAR) program

The UC Berkeley Robot Open Autonomous Racing program was launched in 2019 to advance solutions of Autonomous Systems, Intelligent Machines, and Human-in-the-Loop Control for extreme robotics applications. Its faculty and students have extensive experience in developing high-performance algorithms for 3D Perception, Model Predictive Control, Reinforcement Learning, Generative AI, and Simulation and Virtual Reality, and have received major funding from NSF, ONR, ARL, DARPA, and various industry sponsors.

The AI Racing Tech Team

The AI Racing Tech Team, the U.S.-based team with the most podium finishes in last year’s Indy Autonomous Challenge, pushes untested boundaries and drives research to ensure the highest caliber of safety for the future of commercial autonomy. Co-led by Team Principal Gary Passon of the University of Hawaii and UC Berkeley ROAR Director Allen Yang, the team is made up of faculty and students from UC Berkeley, the University of Hawaii, Carnegie Mellon University, and UC San Diego. The technical co-leads are Siddharth Saha, Haoru Xue, and C.K. Wolfe.

Editors note: This story originally appeared on the UC Berkeley website: https://roar.berkeley.edu/uc-berkeley-autonomous-indy-racing-team-sets-new-u-s-road-course-record/

mike oitzman headshot.
Written by

Mike Oitzman

Mike Oitzman is Senior Editor of WTWH's Robotics Group, cohost of The Robot Report Podcast, and founder of the Mobile Robot Guide. Oitzman is a robotics industry veteran with 25-plus years of experience at various high-tech companies in the roles of marketing, sales and product management. He can be reached at moitzman@wtwhmedia.com.