SMART BOATING with
SMART BOATING with machine learning – which is basically advanced statistics – can find those patterns in a much more systematic and logical way.” “It’s about trying to fill in information where there is none,” says Muñoz, “we use this real-time data to monitor and identify causes for failures and then piecing this together to get a bigger overall picture.” To maximise input and output and gain that bigger picture, the analysis stretches far beyond what happens on the boat during a single training session, Muñoz explains. “We’re looking beyond the traditional day-to-day performance analysis approach,” he says. “So we crunch not only today’s data but compare it to what we generated a few months ago, hoping that seeing the trends over time can lead to a more informed decision- making process in the development of the boat.” But it’s not as simple as it sounds. “One of the fundamental things that you have to understand about sailing is that unlike Formula 1, the problem is very ill-defined,” says Muñoz. “The boat’s performance can potentially differ quite dramatically from day to day, even if all the variables you can control on the boat remain the same.” Indeed, these variables force the very best out of the Portsmouth team and the algorithms they use as they work to cover a growing range of scenarios and influences in their analyses; what’s defined as a 10-knot wind whipping across the sea surface can actually vary between 7 and 13 knots; underwater currents constantly shift; and things tend to move fast on an America’s Cup Class boat. Combine that with the variables the racing team can control – like switching out different hydrofoils – and it becomes complicated. “If you modify or change a component on the boat, you’re trying to understand the impact of that specific change and remove the background noise,” says Johnston. “The sheer wealth of information being generated by the boat’s sensors makes a huge difference. The machine learning techniques we have designed help filter through this morass of information to find and monitor elements THE 35TH AMERICA’S CUP Dates: 26 May to 27 June Location: Bermuda Number of teams: Six (one Defender and five Challengers) Defending Champion: ORACLE TEAM USA (Skipper: Jimmy Spithill) Boat Class: All America’s Cup boats are subject to specific design rules, the so-called ‘Class Rules’. All teams must adhere to these specific design rules and present their own boats, built specially for the race. The Land Rover BAR data team is based in Portsmouth, UK, and has been created to boost performance on the waters: pictured here are Jim Johnston (left), Richard Hopkirk (below left) and Mauricio Muñoz (below right) “THE FIGHT TO WIN that are constantly changing.” By using these ongoing miniscule THE AMERICA’S tweaks to improve the boat’s CUP IS ON A NEW, design, the team based in Portsmouth and their counterparts MORE ADVANCED, training in Bermuda have as such been working in harmony and FRONTIER: BIG DATA” aligning their findings with the common goal of optimising the R1 Land Rover BAR America’s Cup Class boat as far as possible ahead of the competition. The effort comes from a relatively young team, competing against opposition with decades of experience. Still, the Portsmouth crew has been able to benefit from the technologies developed by Jaguar Land Rover deriving from car design and research. This helps accelerate the learning curve, and ultimately, progress. Johnston adds; “It was Jaguar Land Rover’s experience in managing and analysing data for our self-learning car technology that helped us create an expert team able to understand and develop algorithms to disseminate the data. This includes aerodynamics, control systems and of course the machine learning power, combined with workstreams we developed.” “We’ve got a definitive end goal,” Hopkirk says. “In Bermuda, we have to get everything spot-on. Though there’s no way of knowing what other teams are focusing on, the large-scale analysis of performance data is potentially a world-first in the racing world and it is our goal to show just how effective this can be. We are constantly looking for new ways of harnessing that all-important data to feed the algorithms and generate new insights.” “It’s all about winning and bringing the cup home,” Johnston says. “That’s the key driver.” Though that won’t be the end of it,” Hopkirk adds. “We’ll be using the insights we are gathering all through this cup, into the next competition, and far beyond.” 44
SMART BOATING 300 CHANNELS TO HELP GATHER REAL-TIME INSIGHTS FROM THE BOAT 175 GIGABYTES OF RAW DATA GATHERED PER SIMULATION PHOTOGRAPHY: HARRY KENNEY-HERBERT (1) 45
