Lantao Liu and his student team push the boundaries of driverless-car technology with their Indiana University cream-and-crimson racer at the Indianapolis Motor Speedway.
The Fast Track to Autonomous Vehicle Success
The start of the Indianapolis 500 bombards the senses. Eardrum-rattling sounds, blurring sights, and organ-rumbling vibrations combine in a sensory deluge that must be experienced to be fully understood.
Thirty-three cars aligned in 11 rows of three come off Turn 4 and accelerate toward the start/finish line and the green flag. In the stands, the roar of more than 300,000 fans competes with the cacophony of turbo-charged engines revving up to produce as much as 750 horsepower of forward force. By the time the cars complete one two-and-a-half-mile circuit, speeds can approach 230 mph, all orchestrated by drivers riding on a razor’s edge between control and disaster.
Now take away the drivers.
That is the experience Indiana University associate professor Lantao Liu and his student team recently experienced during the Indy Autonomous Challenge (IAC). Liu is part of IU’s Luddy School of Informatics, Computing, and Engineering. He also directs the university’s Vehicle Autonomy and Intelligence Lab (VAIL), which was at the heart of the IAC effort.
That effort isn’t directed at taking the jobs of the highly skilled race car drivers of today or erasing the accomplishments of the historical icons of the sport — names like Mario Andretti, A.J. Foyt, Rick Mears, and Emerson Fittipaldi. Rather, Liu and his team of student researchers are working toward improving the safety and adoption of driverless street vehicles.
As Liu explains it, autonomous race cars operate at extreme speeds and require software that functions with even more extreme precision. That software must be highly optimized because racing depends not only on speed but on accuracy, with minimal margins separating victory from a catastrophic crash.
“Applying these principles to autonomous vehicles can significantly enhance their performance, efficiency, and safety,” Liu says. “The advancements in high-speed decision-making and optimization in racing can translate into improved responsiveness and reliability in everyday autonomous vehicles.”
While that might be the ultimate goal, that doesn’t mean there’s not fun to be had along the way.
‘This is something else’
The 2024 Indy Autonomous Challenge was held September. 6. A total of 10 teams with students from 19 schools competed to determine the world’s fastest autonomous car.
First on the fabled Indianapolis Motor Speedway was the IU Luddy team’s entrant. As thousands of fans watched, the cream-and-crimson vehicle that looked in all ways like a typical Indycar — save for the absence of a driver — made its way down pit lane and onto the track.
“If you haven’t seen this, folks, this is something else,” said well-known racing announcer and pit reporter Greg Creamer as the vehicle wound up to full speed. It was indeed just that — something else — for Liu and his team.
“Being on the track at the Indianapolis Motor Speedway is an extraordinary experience,” Liu says. “We watched with immense pride as our car completed high-speed laps, witnessing our hard work pay off.”
Pushing the Limits of Autonomous Racing
Description of the video:
Ladies and gentlemen, start your software. The Indy Autonomous Challenge is an initiative that is focusing on improving the technology for autonomous driving cars. So, the key challenge here is about the high speed of the driving. Our goal for the IAC is to improve the AI driver capability. We were pressured with the timeline because thecar was delivered to us in late-July. From that moment, we needed to implement all of our past
code and setting up the sensors and installing our software stack on the car. The time that we
spent really on the real practice is only three weeks. It could spend more than a couple seasons
or a couple years to get the team really to the state like achieve high speed racing. Many people
are amazed for our progress this year. During our practice we had a crash two days before the
race day. We had to spend time to figure out what's wrong and what caused our car and we have
to make the car run again two days later on the final race. Applying what you have learned, you
know, in classes and approaches that you already implemented with those kind of, you know, robots
and transfer this knowledge to very high speed autonomous vehicle. This is one of the biggest
challenges we had. For the arrows I would like to design it with the motif of it is like speed and precision,
which is very essential qualities of AI racing. This the first time I work on such a big
project. It was really a good experience for me and I got some hands-on experience on the camera
and the computer vision algorithms as well. Since I led design for our dashboard UX/UI design, I really
wanted to make sure the dashboard worked perfectly and synched well with the racing car. I have like three or four cups of coffee per day. Coffee and excitement. That makes me work. I would say the most exciting part was definitely on race day when we finally got to hear like the what those engines could do. Gonna see what this third lap is going to be looking like here. A ten mile per hour jump lap 1 to lap 2. Will they match that or better it even here? They've come across the yard of bricks
- another 3 mile an hour jump. We like from beginning super nervous, don't know what to do.
And then during the race time and then just cannot breathe very well. And after the race, ah yeah, we
are done this year. Pretty good job. All the team members were proud of what they did and how that led to us completing some successful laps. So, that looks like that's going to be their fastest
lap time there - 124, which is outstanding. It's a, whoa, that's impressive. And again, when
we talk about one of the newer teams in this whole operation here to be up approaching
125 miles-an-hour as an average speed, that's saying something! I could see everyone's faces. All of us were very happy. After the race, we realized we really needed to document
our procedures and the approaches we took to insure that the next time we run the car, it runs
smoother. In terms of the overall software, we have some issues we need to work on. Most advanced technology usually started from the racing cars. And that's why I would say that all the technology - tried, tested, and proved on the racetrack will be immediately transferrable to the normal self-driving technology. The impact will be huge. AI is coming up and doing a lot of cool things. I think this is also our time to show what AI can do for robotics and also
for society. Our challenge next step is in early January in 2025 in Las Vegas. And so the timeline for the next race will be just a couple of months. It's just our starting point and we will be
better in the future. It's always challenging so you should not take a step back. Always take a step forward and face the challenges.
The IU Luddy team includes eight PhD students who lead various subteams to develop the systems for the car, as well as around a dozen master’s and undergraduate students. These students specialize in everything from intelligent systems engineering to computer science and informatics.
They also showed their ability to specialize within a short time frame. While some schools spent years working with their car, the IU Luddy team had a matter of weeks. But they didn't let that disadvantage slow them down.
The team spent a month testing and refining the car's software at the Kentucky Speedway and Indianapolis Motor Speedway. At the IAC, the students completed their three highest-speed laps ever with the car before safely returning it to its pit-lane stall. The fastest lap came in at an average of 124.767 mph.
It was a momentous accomplishment for the IU Luddy team.
“Some students said they wanted to cry after our car successfully returned,” Liu says.
‘Making a measurable impact’
Liu and his student team are not resting on their Indy success.
Liu is proud to see students who worked with VAIL move on to highly sought-after positions in the field. This is a testament to the rigor of their training and the practical relevance of their skills, Liu says.
Those students’ departure left positions open inside VAIL for new students to come in and keep the progress racing forward.
Liu’s short-term goal is for VAIL to take on new and exciting projects that address real-world challenges and prepare students for top-tier positions in academia, AI, robotics, and machine learning. Long term, Liu wants VAIL to be recognized not only in Indiana and nationally but also internationally as a center for applied AI-research that produces groundbreaking work to push the boundaries of scientific knowledge and address critical societal challenges.
Liu’s ultimate goal? To have VAIL alumni be influential leaders in the AI community, with their achievements continuing to inspire and shape the future of AI research and its positive contributions to society.
“I aim to cultivate a legacy of mentorship and collaboration within VAIL, creating an environment where students feel empowered to pursue ambitious research initiatives,” Liu says. “I envision VAIL as a hub where innovative solutions are developed for real-world applications, making a measurable impact on industry and society.”
Pushing the Limits of Autonomous Racing at the Luddy School
Description of the video:
My lab's name is Vehicle Autonomyand Intelligence Lab or VAIL, in short. One project that is very relevant to
this one is autonomous racing. Autonomous racing is something similar to what
we are doing as autonomous Navigation. But racing has different rules
and also different challenges, because you are scaling up
to a higher speed. Our team
now has around 30 students in total, and that 30 students include eight
Ph.D technical leads, and each lead has their own expertise
for a very important component. We are developing a full stack
software system and that's very challenging. This is actually interesting, fun, and give us professional experience. There are multiple perspectives
for our goal. We want to make some breakthrough to really push the boundary
for the racing capability. And through this program, we have a lot of
components working relevant to the AI. And this way then we can include many,
many students from different interests and their different
background or expertise. So the first stage is about
the simulation. Each round
we have different like added challenges to make it more
and more realistic to the final racing scenario. We have a track, a track
that is a replicate of the real environment. And then we put our car there.
The objective was to have like the minimum lap time
to complete one lap in a minimum time. For this around the simulation,
we got a rank of fourth across around 20 teams globally and all other three
teams before us they are long-time players. So I would say that I'm very happy
with these results for our first trial
and we got such kind of amazing rank. We are the number one in the US.
And then the second round of simulation, which happened around two weeks
after the first round. In the environment a new car, although it's also
virtual, is added in the track. So we are competing with AI car. Our car,
will need to avoid not only about infrastructure on the track,
but also the AI car. This is for first time for us. so we overtake four opponents,
but we crashed in the fifth one. But this was a good progress. And then the third round
is about other challenge where our sensor information
might not be reliable. In this way. then for example,
our GPS signal might dropped out periodically. They wanted to mimic the real scenario
where like the car like moving under bridge or there is like trees,
then you have a bad measurement. For the third sim race
we improved our local planning. So for this one we didn't have any crash
and that was good. We are trying to take one step forward now and we are trying to use one of our
small robots as a testing platform. So we convert our code to the, the robot
and then we're trying to make all the modules running even if it is low speed,
like it's only two meters per second. But for autonomous racing,
the car can reach up to 186 mile per hour. If we get all the stuff working in the real world then now
we have like a smaller challenge which is converting from low
speed to high speed in real world instead of having like both together
in April. Luddy already ordered the race car for us and I think we will get the race car
on the 1st of April. And we will have the time
to access the Putnam Park track so we can do some experiment to run
the real car, to see the real output from the real sensors. Based on the timeline
from the racing organization, which is the IAC, the Indianapolis
Autonomous Challenge, the real car racing event will be held in Milano, Italy, in June
this year. I wait for that day to come.
How that will look like, I don't know. But for sure it will be very excited. This was really exciting
and also a challenging task for me. Put our code in a real car and,
see how we compete like with other teams.