Williams F1 CIO: By 2020, AI Will Call Drivers In For Pitstops
INTERVIEW: Graeme Hackland has helped Williams F1 get the most from its data, now the team is looking towards automation to stay competitive
Like all other teams, Williams is restricted on how many teams it can send to the track. This is due not only to cost – there are nearly 20 races on 5 continents – but also FIA regulations. This makes communications all the more important.
Cloud services eliminate the cost and inconvenience of carting servers from track to track and a communications deal with BT means any new addition to the calendar, such as the European Grand Prix in Azerbaijan, can be covered.
Williams is an Avanade partner and uses Microsoft equipment, such as Surface tablets, and services like Office 365 and Skype for Business.
“If you set up a team from scratch you’d probably put everything in the cloud,” adds Hackland. “We still have some on-premise [equipment].
“We use some systems that are critical to us and we’ll keep on using them,” he continued, adding he dislikes the use of the word ‘legacy’ when describing such equipment.
Read More: How Caterham F1 created IT infrastructure from scratch
AI racing
Williams didn’t have the best of seasons in 2016 and Hackland is acutely aware of this. Williams finished fifth in the Constructor’s Championship, with the highest finish recorded by either of its drivers a 3rd place in Canada.
“It’s fair to say we’ve underperformed … but there are huge rule changes for next year and we’ve had to put a lot more effort into that,” he says.
But there is one area Williams leads the way in and that’s pit stops, which have been boosted by biometric analysis. No team can change the wheels of an F1 car faster than Williams, which recorded a stop of 1.92 seconds last season.
An ongoing drive towards automation will help improve pit stops further. Hackland says this will impact work at Grove and in the pits, hence his bold prediction about artificial intelligence (AI) prompted pit stops.
The idea is that through machine learning, a system will be able to determine the optimum time for pitting by assessing various metrics, such as tyre pressure and lap times, and observing various race variables, such as weather and opponent strategies. Likewise, an AI could choose what parts to build and how they are constructed by digesting the large amounts of data created.
“[There are] two areas we’re focusing on initially,” he explains. “New machines give us a huge new capability with smart factory and automating some of that decision making process. The machines will learn and we see that over the next few years.”
