Lotus F1 COO races from intuitive design to machine learning with Microsoft

Formula 1 is a technology and engineering sport. Constantly changing rules and regulations, combined with limited time frames, means racing teams that can innovate the fastest are often the ones that can gain a competitive edge. Seconds count in Formula 1. And those seconds are often gained through technology-enabled design.

Sitting down with Chief Operating Officer Thomas Mayer I quickly realized the huge effort required to improve the F1 car’s innovation development cycle. Mayer said:

Technology is a big enabler to perform better. We are always looking out for the latest technology and how we can apply it to make our business better. For us the key is to get more performance out of the car, to iterate faster - iteration for us brings performance. The more I can iterate, the better the car is, that’s where technology is the enabler to it.”

Twenty years ago we had very capable engineers, but it was a lot of intuition and 6th sense. Now it is all data-backed. We are producing terabytes and petabytes of data to give just a small glimpse of [innovation].

Lotus’ development begins with modelling designs using the team’s 38 teraflop super computer, which has approximately 2,900 cores. Aero shape designs are then shifted to the Lotus wind tunnel, where designs are reproduced to a 60% scale. Regulations insist that teams have a certain amount of time allocated to both these areas of development, meaning that Lotus has to figure out the optimum time to work in the CFD super computer versus time spent in the wind tunnel.

Mayer said:

I can mathematically model the car to the extent where I can model the screw I use. In terms of the data coming from the simulator it’s the same as the data coming from the car, it’s the same signals, same screens. Then I put the driver in and I change the model virtually and I haven’t produced any part yet. If the car gets faster then I go back to my engineers, telling them that they need to produce with this or that part. In the past it was trial and error, now it’s reverse engineering.

Cloud and machine learning

Not only does Lotus have to worry about the development cycle prior to races, but it also has to move huge amounts of data and infrastructure to 20+ tracks around the world so that development work can be carried out trackside during a race.

Although time is important on the track, it is equally important off the track. Lotus is working with Microsoft with the aim of improving its development timeframes. Microsoft’s Azure cloud could play a huge part in this, for example. Mayer’s point is that although Lotus can innovate and keep up with market trends at a decent pace, it doesn’t have the resources to do it at the pace that Microsoft can. He said:

The fixed infrastructure is huge, it’s a very high entry barrier.

Cloud - that’s what we are trying to do at the moment. As the regulations always change, you can see [the benefit] from a commercial point of view. I’m sure that somewhere in the world I could always have the latest infrastructure, because Microsoft can offer a lot of data centers. They put billions every year to data centers. Somewhere in the world I always have the latest chipset and the latest processor.

The benefits of casting off infrastructure to the team are clear. Mayer said:

For me, I would like to provision virtually all of this data, always close to the track. I still carry something like 36 servers with me around the world. [And although] latency is an issue, because I need [the data in something like 2 milliseconds], if I have a fibre connection and a data center in the same country, that’s not a problem.

Lotus has been a Microsoft Dynamics AX customer for some time and in 2014 was mulling Microsoft cloud. A the time, Mayer was concerned about security and latency in remote locations. Today, Lotus takes advantage of Office 365 and Power BI. It’s putting whatever it can into the cloud. Not only this, but over the past year or so Lotus has been making use of Microsoft’s cloud-hosted machine learning platform. Mayer said:

We are doing a lot now in terms of machine learning, because we have 20 years of data. We are able to correlate the weather data, the roughness of the tarmac and suddenly we can see patterns. We have a petabyte data lake that just looks at tyres.”

[The engine] looks at all these data points historically and recognises the pattern that you as a human being could not. There are too many data points. And then you feed it with the real data and it will tell you that the tyre will last another three laps or another 10 laps. It gets better and better, the more you feed it the more it learns.

Finally, Mayer said that he hopes that the team’s partnership with Microsoft will allow Lotus to take advantage of the company’s investments in augmented reality with HoloLens. Because so much of Lotus’ work is visual and design-based, Mayer sees a real opportunity for F1 to be a leader in this area. He said:

For me the next step is HoloLens. If you think about it, I do a lot of visualisation of aerodynamic flows. Imagine in a wind tunnel when you put the smoke in and you see how it flows - imagine if I could do this virtually. I can even combine the data, when changing a flap, for example, have the Holo Lens on and see how the air flow changes.

In our industry everything is visual. A race engineer looks at the scatter of different, what he looks at is pattern. It’s always flowing this data, it’s constantly flowing, that’s where I think we will use HoloLens.

Reflecting upon this conversation and thinking about what they said last year, I cannot help but admire the speed at which the Lotus F1 team has pursued its cloud strategy. Others could do well by learning from the Lotus F1 experience