While Formula One, under the management of Liberty Media, has made significant inroads into the American market with three grand prix now situated in the US, it is Nascar that continues to see a rise in its television viewership, contrasted with a small dip in audiences for both F1 and IndyCar series.
European racing enthusiasts often dismiss stock car racing, yet the allure of a simple, powerful V8 engine—displacing 358 cubic inches (5.8 liters) delivering 670 horsepower—circling an oval track has a unique appeal that other racing formats seem to have overlooked or abandoned.
Nascar teams, however, are constantly seeking technological edges. Richard Childress Racing, one of the prominent groups in the fray, partners with Lenovo to refine their pit stop strategies. Use of artificial intelligence aids in achieving real-time insights during frequent refueling stops—especially crucial given the five to twelve pit stops a race may necessitate depending on the specific track and unfolding race conditions.
In Nascar, managing fuel effectively is both a critical tactical element and a dramatic aspect of the racing spectacle. Unlike F1, where refueling during races has been prohibited since 2010 due to cost and safety, Nascar cars do not even have fuel gauges in the cockpit. It thus falls to the team strategists to meticulously monitor fuel levels and consumption throughout the race.
Fuel consumption in motor racing varies significantly based on factors such as track layout, race duration, and the speeds pursued by drivers. Additionally, strategic use of “cautions” or slowdown intervals during the race enables vehicles to roughly halve their usual rate of fuel usage.
In Nascar, a common strategy called “drafting” allows drivers to conserve fuel by reducing throttle usage while staying close to other cars. This technique leads to less fuel use and fewer pit stops, enhancing overall efficiency. On average, a car in the Nascar Cup series consumes about 100 gallons (380 liters) of fuel per event, despite not being particularly fuel efficient.
Lenovo’s AI team is tasked with refining these measurements to make the fuel management during races nearly as precise as possible. By capturing the duration that fuel cans are connected to the cars during pit stops, more accurate fuel usage data can be assessed and strategized.
Addressing this challenge, Lenovo developed a method involving in-car transponders alongside a camera positioned above the pit area of RCR (Richard Childress Racing). This setup identifies when cars enter the pit and initiates a real-time video feed to monitor the process accurately.
“An AI engine analyzes each frame to determine whether the fuel can is connected or disconnected,” explains Sachin Wani, a data scientist at Lenovo. “We process 30 frames per second, so the accuracy is about 0.03 seconds. Previously, the fuel attendant estimated about seven seconds for refueling—without any technological assistance due to safety concerns.”
“Essentially, it was about mental estimations, leading to seven seconds potentially extending to eight or nine, or dropping to five or six. This could disrupt the strategy, resulting in insufficient refueling and the need for an additional pit stop,” Wani adds.
Racing involves precise calculations, now aided significantly by Lenovo’s AI technology at pit stops. With the fuel can attached to the vehicle, about 11 gallons of high-performance fuel are delivered during a service. As tires are changed from left to right, the fuel attendant can switch to a backup can. A Nascar fuel cell has a capacity of 20 gallons (76 liters) and a liter of race fuel weighs around 0.75 kilograms. In racing, small differences matter; even an additional second in refueling can critically impact the vehicle’s weight and competitiveness. Lighter equates to quicker speeds.
“This AI gives teams greater trust in their fuel metrics,” states Eric Kominek, technical director at RCR. “Telemetry, combined with precise measurements of added fuel, helps teams precisely calculate fuel mileage down to 100 feet. Whether aiming to avoid fuel delays in future stops or optimizing their position at the end of a race stage, the AI’s fuel data is extremely useful.”
Sachin Wani from Lenovo revealed that the company’s NASCAR AI assistant was developed using a proprietary dataset with enhancements for precision. “We considered various race conditions, including night events, to train the AI rigorously for the 2024 NASCAR season throughout 2023,” explained Wani.
The team didn’t just stop after initial training. They continuously improve the AI, focusing on critical aspects such as the exact angle for fuel can attachment to optimize refueling. Future plans include using GoPros on the pit crew to ensure that tire lug nuts are adequately tightened during stops, added Wani.
David Ellison, Lenovo’s chief data scientist, shared, “None of us were big NASCAR fans before the project, but we have gained a profound appreciation for the sport and the team dynamics involved.”
