In a recent significant international race in Las Vegas, the self-driving Indy car developed by a team of students from the University of Virginia (UVA) showcased its capabilities by advancing to the finals.

Surprisingly, the autonomous offering, named Cavalier Autonomous Racing vehicle, made headlines when it autonomously decided to break the rules at the Indy Autonomous Challenge racing event, ultimately securing second place. According to the UVA team, the car's unconventional move prevented its final opponent from passing, adding a unique twist to its performance earlier in this month's competition.

“It’s taken years of very meticulous preparation to get to this point. It was a very tall order, and our backs were against the wall to arrive at the (Consumer Electronics Show), qualify for race day, and prove ourselves, all within a span of four days," said Madhur Behl, professor at the School of Engineering and Applied Science professor at UVA, in a statement.  

Furthering technology

The team claims that the rigorous effort invested in pushing the technological boundaries of a 241km/h autonomous racer at a prominent track like the Las Vegas Motor Speedway serves a broader purpose beyond mere speed. 

The UVA team hopes their project will yield valuable insights that extend beyond the racetrack, aiming to enhance the safety of self-driving passenger cars. These advancements are crucial as the autonomous vehicle industry navigates ongoing challenges in recent years.

“Hopefully, all the research we are putting into this is going to pay off and kind of show that this new technology is not just a fluke. This is something that is going to be seen as reliable for the general public," said John Chrosniak, who graduated in December with a master’s in computer science, in a statement.  

In the ultimate showdown, the Cavalier Autonomous Racing vehicle, supported by a team of about 30 UVA students, faced off against TUM Autonomous Motorsport, a German team boasting a staff of 70 engineers.

Despite its unimpressive performances in prior years, the Cavalier team, starting race week without a seed, effortlessly navigated through a series of qualifying challenges. Their commendable efforts secured them the top seed position in the finals. 

After suffering two years of setbacks, the Cavalier team this year breezed through a series of qualifying challenges and earned a spot in the final race. Image: Autonomous Challenge.

Innovative proposition

Researchers at UVA are making strides in advancing autonomous vehicle (AV) technology, focusing on critical aspects such as decision-making, obstacle avoidance, end-to-end driving, path planning, and model predictive control.

In obstacle avoidance, the team implemented the Follow The Gap (FTG) method, which constructs a gap array around the vehicle and calculates the best heading angle for navigating into the centre of the maximum gap while considering the vehicle's goal. 

 

For end-to-end driving, the researchers combined a Long Short-Term Memory (LSTM) and a Convolutional Neural Network (CNN) to create a recurrent neural network (RNN). The model demonstrated accurate steering angle prediction, a crucial aspect of AV control.

The team also explored path planning, emphasising the structured nature of AV environments. They integrated static, or global, approaches, leveraging known information about the environment, like track layout and floor friction.

In model predictive control, the researchers introduced a local planner utilising a model predictive controller (MPC) and a learnt approximation of the policy it generates. The approach aims to replace planning components with learnt modules efficiently. 

Moreover, the researchers delved into Simultaneous Localization and Mapping (SLAM), Computer Vision, and an adaptive pure-pursuit controller for autonomous racing, showcasing the versatility of their research on the F1/10 testbed.

The F1/10 platform, designed by UVA, offers a comprehensive environment for experimenting with various aspects of AV technology, from algorithm development to real-world testing. According to the team, this multidimensional approach positions UVA at the forefront of autonomous vehicle research, promising advancements in safety and performance. 

Over the next few months, the Cavalier race car is set to undergo enhancements to its sensor array and drive-by-wire system in preparation for the team's upcoming race in June at the Formula One track in Monza, Italy. "We’re also planning another recruitment cycle this semester to welcome new students into the team,” says Behl.