The answer is blowing in the wind
_{How we use LiDAR at the America’s Cup to make the invisible, visible}

The wind is important for sailing, and never more so that in the America’s Cup, a competition that attracts top sailors, deep pocketed hosts and sponsors, and an estimated 1 billion viewers. Understanding the capricious nature of the wind is key to understanding the race, both for sailors and their audience.

For the first time in its history, broadcasters of the 2024 America’s Cup in Barcelona could show viewers a graphical overlay of the wind superimposed over real time video images of the racecourse. Viewers could ‘see’ the wind and how it affected the race. It allowed the broadcasters to compare the paths taken against optimal routes, and predict what boats on the water should do next.

This was all thanks to WindSight IQ™, a wind sensing and visualization system developed by Capgemini with AC Media. So, what did it take to achieve all this?

Taking the measure of the wind

It all starts with LiDAR, or Light Detection and Ranging, the lesser-known cousin of radar. Whereas radar bounces radio waves of objects to detect them, LiDAR bounces light, normally in the form of focused pulsed laser beams, off aerosols in the air. Applications of LiDAR and radar are everywhere around us – from obvious applications in the automotive and aerospace world, to less obvious applications in mapping and 3D scanning.

A Doppler LiDAR adds an extra dimension to the signal that bounces back, using the Doppler effect (how squashed or elongated the retuning waves are) to measure how fast that object is moving towards or away from us.
It is this effect that powers WindSight IQ™. Each Doppler LiDAR shoots laser beams out into the air, around 10,000 pulses per second, some of which bounce off aerosols and impurities in the air and return an indication of where they are and how fast that air is moving towards or away from us.
By combining multiple LiDAR’s perceptions of this towards-vs-away movement at a particular point, we can calculate both the wind speed and direction at that point.
A huge amount of mathematical modeling is then required to turn those raw measurements into a usable wind-field, but it all starts with these Doppler measurements of the speed of the air.

Choosing locations to measure the wind

To correctly calculate the wind speed and direction, we must “see” the same bit of wind from multiple different angles. The America’s Cup racecourse also moves daily (and even within races), so we need to ensure we can cover all possible scenarios.

In an ideal world, we would just place LiDAR units uniformly around the racecourse, but this is where the real world gets in the way of our plan – we can’t just put the LiDARs wherever we would like, because there are obstructions, reflections, legal issues and – in some cases – a distinct lack of land where we would like to place a unit.

Early on in the project, we recognized that the site selection of our LiDAR units would be key to the success of the whole project. This is a classic constrained optimization problem – we have many different parameters, some which we can control and some which we can’t, and we need to find the optimal configuration that maximizes the performance and robustness of the system, and minimizes cost and complexity.

We even created a dedicated analysis tool specifically to assist our team with this site selection problem. See the video below.

Eventually, we settled on a site to the north of the race area, near Parc del Fòrum, one in the South at the Baleria Terminal, and later one in the middle, near the San Sebastian Beach. Installing the LiDARs, each of which is 100kg of carefully calibrated scientific equipment, required a diverse mix of engineering and logistical support from our local teams, as well as the cooperation of multiple building owners – who allowed access to their roofs as sites for our LiDARs.

Finally came the most exciting part of any geeky innovation project – deciding on names for our tools. This is often a bit of fun in any tech project, but also serves an important practical purpose – providing instantly recognizable names for the hardware that won’t be confused in a busy operational setting. We settled on names derived from Greco-Roman gods – the northern LiDAR was named Borea, after the god of the North Wind, the southern LiDAR was dubbed Notus, after the god of the South Wind, and in the middle is Zephyr, after the god of the West Wind. Each of the LiDARs have their own personalities, our operators were very fond of Zephyr due to its very good range and consistent performance, and I’m sad to say that Notus is often the “black sheep” of our LiDAR family.

A gale-force stream of wind data

When you’re scanning a 6x5km area at high speed and high resolution, you produce a lot of data. Each LiDAR generates approximately 10 megabytes per second, and must feed that data back to our operations center in as close to real time as possible. However, the LiDARs are several kilometers from the operations center – across beaches, water, and streets, so running a cable between them is not an option.

Luckily, as a major world center of telecoms, Barcelona is blessed with an excellent and reliable 5G cellular network, so each LiDAR is paired with a ruggedized mobile router to push the data over a custom UDP protocol and VPN back to base. However, mobile networks are notorious for congestion around major events, and with the America’s Cup estimated to bring an extra 2.5 million visitors to Barcelona over the race period, we needed to ensure that we had a viable backup to fail-over to, if necessary. This was achieved using Starlink units, giving us the option of a satellite uplink if required.

How to keep a LiDAR happy

Many things affect the quality of the data we get from our LiDARs. As stated earlier, the LiDAR’s laser beams bounce off impurities in the air – which could be aerosols, dust, or even pollution. We need there to be “stuff” in the air in order to see the air.

Anyone who’s ever used a laser pointer will know that the beam itself is not visible – it’s only when that beam hits something that we can see it. This is why laser light shows always use smoke machines to make the lasers visible, and why movie thieves always blow smoke into a laser security grid to see where the beams are.

Essentially, we need the air to be a bit dirty to see it, but not too dirty. Rain is our worst enemy – during rainfall, the LiDAR beams are stopped dead and we can’t see the wind at all. However, even when the rain stops we can still have a problem. Immediately after rainfall, we sometimes perceive that the air feels fresh and clean, and that’s exactly true – all the particulate matter has been washed from the air and our laser beams just keep going into the atmosphere and don’t bounce back enough to detect. We must trade off other parameters in this situation, potentially sacrificing range and/or accuracy to get a strong enough signal to detect the wind.

In a fast changing and unpredictable environment like a sporting competition, where real-time insight matters, the tools need to respond to changes in input data in real time. For many intensive months, our team worked with these LiDAR signals, building algorithms on top of them. That has given them a near sixth-sense where they interpret this raw data effortlessly and rapidly tweak and reconfigure the settings to maximize range and accuracy in response to changing weather patterns.

When disaster strikes

A few weeks into the America’s Cup events, disaster struck. Borea got sick. A never-before-seen hardware fault developed and it would not start up. We were down to two LiDARs with races just hours away from starting. Luckily the WindSight IQ™ algorithm took this situation in its stride. The algorithm was always designed to use as much or as little information as was available in its calculations. When new data sources became available, the algorithm swallowed it up and used that new data to increase the accuracy and confidence in the windfield. Likewise, when there was not as much data as expected, the algorithm creates the best estimate possible with the data it does have.
For example, when we got access to live wind telemetry from the boats and marker buoys, WindSight IQ™ ingested that too and used it to increase its confidence in those areas of the windfield. It always builds a model of the wind that uses any information at its disposal, without relying on any one source completely. Nobody watching the TV coverage knew that WindSight IQ™ was running with 33% less data than normal during those two days before Borea could be replaced. The inherent scalability of this algorithm is what now allows us to consider creating windfields over much larger areas for future applications.

Setting the sails and a new precedent

The introduction of WindSight IQ™ is not just a first in the sailing world, it’s a first in the world, full stop. Our ability to combine multiple Doppler LiDARs into a single, coherent, accurate picture of the wind at high resolution and in real-time changes how we think about measuring the wind.

Previously, we were limited to sampling a handful of point measurements to give us a vague understanding of the wind – like fumbling around in a darkened room trying to figure out the layout. Now, we have flicked on the light switch in that room, and need no longer guess – we can just see it directly. That has already shown its value on the seas, but detailed wind sensing and visualization could soon have huge benefits in many other areas, including airport safety and operations optimization, optimizing windfarms, precision farming, high rise construction, disaster response, and the design of vehicles, boats and aircraft. Having proven its worth in the challenging environment of a live global broadcast of a big sporting competition, we are excited to work with partners in all these industries to explore what else it can do next.