Anemomind has great ambitions to leverage advanced technology to help sailors sail faster. Where is this technology coming from, and how did the company Anemomind come to life?
Computer Vision for the America’s Cup
Everything starts in 2003, in Lausanne, Switzerland. At the time, Alinghi, the Swiss team for the America’s Cup, had won. When Julien Pilet started to work on his PhD at the Computer Vision Lab of EPFL, this victory offered a perfect opportunity for him to join his two passions: computer science and sailing. Julien’s research challenge was to measure the deformation of a spinnaker based on pictures. A few years after, the CVLab provided Alinghi with a working solution for their 2007 campaign.
Early data recording
In parallel, Julien was racing Irene, a 6 meter Rule boat. In 2007, he plugged a GPS and an old Palm PDA to Irene’s onboard instruments in order to record navigation data. The system, although quite rough, would allow to visualize the trajectory and the measured wind on a map.
The project came to a pause between 2009 and 2012: Julien moved to Japan for a post-doc and then to Zurich, to work for Google.
Automatic extraction of VMG target
In 2013, after his time at Google, Julien wrote new algorithms to extract useful information from data previously recorded during Irene’s navigations. Since Irene is only sailing wind oriented races, knowing at what speed it is supposed to climb up or slide down the wind is key. After a searching for a way to clean the data, Julien invented an algorithm that produced a clean and expressive curve that would allow the sailors to know the target VMG given the true wind speed. Immediately, Julien wanted a device to read that curve automatically during navigation. At the end of 2013, thank to EPFL support, this hobby project turned into a serious start-up project.
True wind and calibration
When testing the system, it became clear that one limitation was the accuracy of the true wind estimated by the instruments. The calibration procedure is a pain to follow, and the result is usually not very convincing. With the help of Jonas Östlund who joined the project at the end of 2013, Julien imagined a “big data” approach to calibration: if we feed enough navigation data to the system, the system should be able to understand how to properly estimate where the wind is blowing from and at what speed. In 2014, the idea became real and the first results showed good improvements in the true wind accuracy.