For more than 120 years, racing has been used to improve the performance and safety of automobiles. Along the way, numerous innovations developed for use on the race track have trickled down to road cars. That tradition continues today, as engineers push the boundaries of autonomous systems technology.

Thanks to a big-time collegiate competition called F1TENTH, some of those breakthroughs are occurring on a small scale. The organization was founded in 2016 at the University of Pennsylvania to foster interest and innovation in the field of autonomous systems.

Students from around the world develop algorithms and program battery-powered 1:10 scale cars equipped with sensors that enable autonomous operation. Each team uses a standardized chassis that includes a battery, controller, motor, drivetrain and wheels.

But, software programming is what sets each team apart. It enables the vehicles to compete in head-to-head competition over challenging courses full of twists and turns that emulate real-life grand prix racing circuits.

“We focus on racing, because it is both competitive and fun,” says Rahul Mangharam, Ph.D., an associate professor in the department of electrical and systems engineering at the University of Pennsylvania, and one of the developers of the F1TENTH program. “It pushes racers to the limits of their performance.

“[F1TENTH] introduces universities to the idea of autonomy and studying across multiple disciplines,” explains Mangharam. “Education is traditionally siloed. You study mechanical engineering, but you may also want to learn about control engineering.

“When you design a racecar for F1TENTH, you need to know all different elements of engineering,” notes Mangharam. “This program allows students to develop a well-rounded set of skills that sets [them] up for future robotics-focused positions in various companies.”

In fact, Penn engineers who participated in autonomous racing in the past are now working at companies such as Amazon, Honda, Nvidia, Rivian and Tesla.

“We push the limits of agility and handling, while balancing safety and performance,” claims Mangharam. “Racing with full-size cars is expensive and dangerous. Instead, we use a low-cost, open-source platform that enables students to learn. We also use simulation systems to predict how the cars will perform in a race.

“F1TENTH is a platform that allows [students] to develop prototypes on a smaller scale,” adds Mangharam, who also heads up Penn’s General Robotics, Automation, Sensing and Perception (GRASP) Lab and the Autoware Center of Excellence located at the Pennovation Center. “It is a stepping stone for students to develop larger autonomous commercial and industrial vehicles.”

The main focus of the F1TENTH organization is on designing and maintaining a powerful and versatile open-source platform for autonomous systems research and education. This includes numerous autonomous racecar competitions that bring together an international community of students and educators.

“In motorsports, there is a saying that ‘if everything seems under control, then you are not going fast enough,’” says Mangharam. “Expert racing drivers have split-second reaction times and routinely drive at the limits of control, traction and agility under high speed and close proximity.

“Autonomous racing presents unique opportunities and challenges in designing algorithms and hardware that can operate firmly on the limits of perception, planning and control,” explains Mangharam. “Autonomous racing has the potential to serve as the litmus test for self-driving software.

“While a large portion of autonomous vehicle research and development is focused on handling routine driving situations, achieving safety benefits also requires a focus on driving at the limits of the control of the vehicle,” notes Mangharam.

According to Mangharam, balancing performance and safety are crucial to deploying autonomous vehicles in multi-agent environments. “In particular, autonomous racing is a domain that penalizes safe but conservative policies, highlighting the need for robust, adaptive strategies,” he points out. “If you are too conservative, you will lose the race. But, if you are too aggressive, you can crash.”

Mangharam teaches a popular 15-week course at Penn that focuses on F1TENTH racing. Engineering students engage in hands-on projects that are focused on applied machine learning, artificial intelligence, control theory, motion planning, sensor technology and other topics.

They learn about the fundamentals of autonomous driving, such as localization, mapping, object detection, perception and trajectory path planning. By focusing on a racing environment, students discover how to develop algorithms that operate on the edge of vehicle dynamics with high accelerations, high velocities and high computation frequencies.

In addition to F1TENTH, Mangharam is also involved in a collegiate competition called AV Gokart that involves autonomous go kart racing.

Each F1TENTH chassis is 16 inches long and weighs approximately 6 pounds. It is available as either a kit or built-up version from a company called RACECAR/J. The chassis contains mounting points for cameras and other sensors. Teams also use state-of-the-art lidar technology donated by SICK.

“Lidar is the dominant sensor on board the racecar,” says Mangharam. “It uses a point cloud to determine how fast it can be driven around the racetrack and still navigate around others.”

The most recent F1TENTH event in the United States was held in Philadelphia in May 2022 in conjunction with the IEEE International Conference on Robotics and Automation (ICRA). It featured 23 teams from a variety of schools, including Carnegie Mellon, Clemson, Lehigh, Penn, Rice, the University of Texas, Virginia Tech and the University of California at San Diego. The race also attracted participants from Austria, Canada, Italy, Malaysia, Poland, South Korea and Switzerland.

Similar races were held recently in Germany and South Korea. The next international event will take place in London as part of ICRA on May 29-June 2.