Matthew Sheahan

COULD TECHNOLOGY BEING DEVELOPED FOR THE WORLD’S FASTEST OCEANGOING YACHTS HELP ALL SAILORS AVOID COLLISIONS AT SEA?

I’m addicted to offshore machines and regularly suffer an uncontrollable urge to watch hours of Ultime and IMOCA 60 footage. On the big multihulls in particular, I still don’t get how anybody manages the mental stress of riding a guided missile at 35-45 knots in the dark, let alone how they physically handle these machines alone or doublehanded. How do you live aboard a jaw-shattering carbon platform that is trying to hurl you across the cabin or off the deck like a rag doll with every wave it slams into?

And what really doesn’t bear thinking about is the thought of running into something at these speeds.

Few of the sailors want to talk about that but, as we saw most recently with Alex Thomson’s Hugo Boss, it happens. Under their super-cool exteriors, thse skippers are kept awake at night by the thought of a collision at speed. So many of them have recently bought into a new technology and are working collaboratively with their rivals to help better the odds of staying safe at speed.

The technology is called Oscar and has already been adopted across more than half of the IMOCA fleet and is on four of the Ultime trimarans. Oscar is a sophisticated collision avoidance system that uses a combination of thermal imaging and daylight cameras mounted in a single unit at the masthead to scan the sea ahead for objects in the boat’s path. When it sees a risky object it sounds an alarm, but the system is soon to be able to talk to an autopilot to take appropriate avoiding action.

The concept sounds simple, but when I spoke to Raphael Biancale, Oscar’s inventor, it became clear that this has been a huge and complex project that has only been possible through the collaboration of some of the top teams and offshore rock stars such as Jean Le Cam, Vincent Riou, François Gabart, Armel Le Cléac’h and many others.

“As far as the hardware goes, the masthead unit has two thermal cameras, each with a 50° field of view. They are mounted side by side to create one image of sufficiently high resolution to identify an object of 1m or bigger,” explained

Biancale. “The unit also includes one daylight camera that has a field of view of 120°. Combining these images with data from a gyro sensor, a system for detecting the horizon, and a means of stabilising the image, we can map every pixel to a position in space.”

It’s an impressive task when you consider the amount of movement at the top of the mast at speed, let alone the amount of data and the speed with which it has to be processed. But that’s just the start.

It is based around the principle of machine learning, and Oscar’s processor, which is mounted below decks, compares what it sees with images that have been catalogued from previous runs. This image database is created from files that are regularly sent back for analysis after each sailing session from boats that have Oscar fitted. “We already have a library of around 60 million pictures and are annotating 40,000 images a week to train the neural network to identify targets,” Biancale explains.

The next stage is teaching the system to recognise certain objects such as what a boat is. “For this we have to annotate images of a variety of boats in different conditions: with and without sails, in different sea states, with different sail plans set and so on. We need to present more than 10,000 images to get the system to learn what a boat is.”

Of course, it’s not only boats that are a threat; big fish, containers and general debris are all potential hazards that need to be created and catalogued in the database.

Even then, identifying a potential threat is one thing. Knowing what to do about it and being able to take avoiding action is another, and takes the processing onto another, challenging level.

“Our top of the range system can identify an object at 1,000m and by 600m it has calculated what the collision risk is,” Biancale continues. “The two experts in autopilot control systems, Madintec and Pixhawk, are creating the means to instruct autopilots to avoid collisions based on the information that our system sends.”

Work is already being done on a version for the broader market. If it proves to be a success for those hurtling through the dark at breakneck speeds, the future looks good for the rest of us when it comes to the worry of hitting something unseen in the water.

‘It can identify an object of 1m or bigger’