AI also runs (and wins)

Racing motorcycles don't have telemetry like Formula 1 cars. During the race, there's no constant communication between the rider and their team, who don't have real-time data from the machine. The only way to transmit any instructions is through a system that sends a text to a small screen on the motorcycle, although this must first be approved by race management. So anything that needs to be changed on a vehicle that can reach speeds of 350 kilometers per hour must come from training. And artificial intelligence has new solutions for two wheels.

“Wow... MotoGP technology has changed a lot in the last 10 years,” Marc Márquez exclaims, trying to explain how the sport is transforming. The eight-time world champion is riding one of the category's top machines this year, on the Ducati Lenovo team, with which he hopes to win the World Championship.

When Márquez climbs onto the bike on the first day of free practice for a Grand Prix, he has already attended a technical meeting in which the engineers have revealed a series of information that can help him on the different sections of the circuit. Strategy and analysis using high-performance computers are now closely linked.

“I like to race by instinct,” confesses Márquez, who is nonetheless delighted with the advantages computerized analysis gives him. “For example, in Qatar, the engineers used Lenovo technology to see which corners were most damaging to the front tire. They told me: ‘Watch this corner, this corner, and this corner, don't push, wait for such and such a corner to start pushing.’ They filter all this through for you, and then, at the end, on the bike, you have to remember it,” he notes.

At Ducati Lenovo, each Grand Prix day in the paddock begins with a technical meeting based on the analysis of data from the team's six bikes across the three categories (Moto GP, Moto 2, and Moto 3), as well as data from the other teams with Ducati bikes (BK8 and VR46). Plenty to analyze.

Márquez: “If you're tired, when you brake, you're one or two centimeters further forward; that changes the whole bike.”

Computers allow for increasingly precise simulations, but it's especially complicated on motorcycles. "The difficult thing about motorcycles, and the beautiful thing, is that there's no simulation possible like in Formula 1 in a simulator. It's impossible," says Marc Máquez, illustrating this with an example: "From first thing in the morning to last thing in the morning, you're already physically tired, and your body weight shifts one or two centimeters forward when braking. That changes the entire motorcycle, starting with the suspension and moving through the engine brake and the balance of the machine. This is very difficult to represent in a simulator."

Despite this complexity, data now rules. After each session on the track, the bikes are connected to a cable via a military-grade security connection that uploads everything captured by the 50 sensors on each machine to a computer. Around 30 GB of data.

Eighty percent of this information is processed on-site using Lenovo computers and software in about 10 minutes, but all the material is also sent to a supercomputer in Bologna, Italy, where more refined results can be extracted, allowing for better conclusions.

But to go further, data from the bike would have to be combined with data from the track. And that's more complicated. Ducati Lenovo has sought a solution by developing a robot that automatically navigates the track and precisely scans the entire layout using laser technology. Multiple simulations can be run on this model, although the technology still has limitations, such as simulating the complex effects of, for example, rain during a race.

The NTB-01 machine is a kind of Roomba-dog that creates a digital twin of the track, allowing it to simulate some of the conditions of a race without even taking the bike out of the pit. Scanning using LiDAR technology creates a digital twin of the track in the form of approximately 200 GB of collected data.

“We needed a suitable tool to save time, because resources are usually limited when we arrive at the track. So we decided with Lenovo to build this autonomous robot that should navigate the track autonomously, collect all this data, and provide us with the information we need about the track's geometry and layout,” says Nicolò Mancinelli, Head of Vehicle Performance at Ducati Corse.

The robot is still in the testing phase, but it will begin to be used at the Hungarian Grand Prix, from August 22nd to 24th, which will take place at the Balaton Park circuit, where races haven't been held for 30 years. The plan is to use it first on the lesser-known circuits and, eventually, on all of them.

The number of parameters that can be modified on the bikes is around 30. Computer analysis, with the help of artificial intelligence, allows the most appropriate configurations to be determined for each track or time of day. What you need for a race is not the same as what you need for qualifying. Once on the track, the rider has few options to change things. They only have three colored buttons—red, yellow, and green—to change the engine mapping.

Depending on which button they press, the engine control unit (ECU) programming—the same for all machines—can be modified to optimize the machine's performance. At speeds of over 300 kilometers per hour, there's no time to look at more complicated controls.

AI is also used for video analysis. All teams have someone filming each race. “With machine learning software,” Mancinelli says, “we can compare different drivers' driving styles and performance, and we show them this along with the data so they can see where they can improve.” “This has also yielded some results, and the drivers have started to trust what they see,” the head of engineering notes.

Read also

This relationship of trust between engineers and riders has been growing stronger. Gabriele Conti, Ducati Corse's electronic systems director, observes that the effort to reach an understanding is mutual. "Sometimes we tell them to do something, and the rider says, 'Okay, but if I do it, I'll crash.' That's the gap we're trying to bridge: discussing the data to find out what the bike's limits are compared to the riders' feelings."

“Over the past four years, data analysis has become essential. Now it's hard to think about going back to what it was before, because it's such a boon,” says Mancinelli. AI is running. And it's winning, too.