AI is helping drive motorsports teams for General Motors

It will likely be a long time before motorsports organizations like NASCAR see self-driving cars compete, but that doesn’t mean artificial intelligence has no place in the sport.

At General Motors’ Charlotte Technical Center in North Carolina, a team of software and automotive experts harness AI to give GM racers an advantage in NASCAR, IndyCar, IMSA, and other races. The technology is used during races to do everything from delivering real-time transcriptions of driver radio conversations and interpreting race footage to giving recommendations on when it makes strategic sense for drivers to take a pit stop. 

It’s a relatively new addition to a fiercely competitive sporting tradition—and an area where GM takes pride in rolling out new advances at a pace much faster than automotive model years. 

“We’re specifically empowered to move at the pace of motorsports around here,” says Logan McLeod, director of motorsports software engineering at GM. McLeod’s team rolls out software updates on a one-week sprint cycle, all with the goal of helping racing teams make their own updates to strategies as quickly as possible come race day, when fractions of a second can make the difference between winning and losing. 

“What we’re doing is reducing the amount of time, dramatically, from the observance of an event to our ability to turn that into tangible outcomes on a racetrack,” he says. “We’re now used to doing that in real time—or near real time—within seconds of observation.”

During events, there’s steady communication from racing teams in the chaotic environment of the track to GM staff in the studied quiet of a pair of war room-like command centers within the Technical Center. There, floor-to-ceiling displays show driver and crew audio transcriptions, race images, and charts and visualizations of vehicle performance and positions. GM also operates three mobile trailers that can travel to races to provide technical support to Chevrolet and Cadillac racers and their teams.

Some of the challenges McLeod’s team has faced would be familiar to anyone who’s worked on AI and data projects in a large organization. That includes ensuring data from disparate parts of the operation is integrated and digitally accessible where it can be used to train, then query, AI models. Like many organizations, GM also makes decisions about when to build specialized software in-house and when to adapt off-the-shelf tools. 

Other aspects of the team’s AI work are very particular to the world of auto racing, like training computers to evaluate the complex set of factors that determine when drivers should hit the pit for a new tank of fuel and set of tires. Even something that seems like a familiar AI task, like translating speech into text, has special challenges when the audio is captured from the radios of adrenaline-fueled drivers and the spotters who guide them around the track—and must be transcribed as accurately and as close to instantaneously as possible. 

Algorithms also automatically highlight photos of vehicle damage that needs to be evaluated by a human eye. Ultimately, the key decisions are still made by racing teams, not computers, and part of the engineers’ job is following up with teams to determine why they sometimes don’t heed automated recommendations. Sometimes, it’ll be an “old school gut check” based on experience overriding the machines, McLeod says, and sometimes, decisions will be influenced by observable technical factors that can be interpreted into future versions of the model.

As the AI continues to build up its quite literal “track record” of good advice, it can free up human race-team members to focus more of their attention on those things only humans can do, McLeod says. The motorsports operation also serves as an incubator for technology including simulation and machine learning software that goes into developing GM’s performance cars—and as a training ground for human engineers who can take the skills and practices they’ve learned working in the competitive racing environment and apply them elsewhere across the company.

“What GM Motorsports is as an incubator is not just about the technology,” McLeod says. “It’s also about the people and the process side.”

https://www.fastcompany.com/91122342/ai-motorsports-general-motors-nascar?partner=rss&utm_source=rss&utm_medium=feed&utm_campaign=rss+fastcompany&utm_content=rss