Behind Saildrone: The Company Sending Autonomous Sailboats Into the Eye of Hurricanes
Driven by wind and solar power, storm-chasing Saildrone vessels are autonomously collecting valuable scientific data for up to a year at a time.
In 2021, an uncrewed surface vehicle (USV) called Saildrone Explorer SD 1045 charged through Hurricane Sam, battling 100-mph winds and towering waves. The 23-foot-long vessel was headed directly into the eye of the storm to collect unprecedented data about the interactions between the atmosphere and ocean, prospectively yielding new insights about hurricane intensification.
This summer, the SD 1045's maker Saildrone has partnered with the National Oceanic and Atmospheric Administration (NOAA) to launch 12 more USVs for hurricane research. Through hurricanes and tropical storms, these autonomous sail drones will continuously send high-frequency data, including barometric pressure, wind speed and direction, air temperature, humidity, water temperature and salinity, wave height, and sea surface temperature, to researchers at the Pacific Marine Environmental Laboratory (PMEL) and NOAA’s Atlantic Oceanographic & Meteorological Laboratory (AOML).
The Saildrone Explorer SD1045 autonomously sails into the eye of a hurricane, all the while sending data back for scientists to better understand ocean-atmosphere interaction. Image used courtesy of Saildrone
This article takes a look at the edge AI, sensors, and software that enables Saildrone to spend months at sea with little human intervention and the techniques that allow climate scientists to read and interpret ocean data.
Learning the Motion of the Ocean
Saildrone's goal is to autonomously and safely guide its USVs to gather data in harsh ocean environments—an environment notoriously inhospitable to electronics. To this end, Saildrone equips each vessel with four onboard cameras, an automatic identification system (AIS) transceiver, a radar reflector, navigation lights, and bright orange wings.
Powered exclusively by solar energy for onboard computers and sensors and the wind for propulsion, the Saildrone Explorer can deliver maritime data continuously for up to a year to meet various science, mapping, and security objectives.
Numerous sensors are used to measure ocean metrics autonomously, while the onboard AI computing hardware combines the data streams and performs the necessary computation at the edge. Image used courtesy of Saildrone
Using the Saildrone Mission Portal software, Saildrone USVs can continuously present meteorological and oceanographic (metocean) data in a real-time, turnkey format. Instead of requiring a costly startup process, this cookie-cutter solution can also be adapted to gather target data.
Saildrone Mission Portal software. Image used courtesy of Saildrone
Both the vessel hardware and the processing software are based on a standalone architecture, leveraging edge AI technology to process all video and sensor data at the edge and providing a clear user interface, whether a single vessel or hundreds are used.
The USVs perform the majority of processing onboard the vessel using NVIDIA Jetson modules. By using the Jetson for edge processing instead of a central processing server, considerably less data must be transmitted from the vessels. This provides easier, real-time data visualization with Saildrone’s Mission Portal and additional power savings, which is important for long-term missions powered only by wind and solar sources.
The Saildrone vessels make use of NVIDIA Jetson modules to perform AI computing at the edge in a power-efficient manner. Image used courtesy of NVIDIA
A USV for Any Need
Saildrone offers three different USVs for different target applications: the Explorer, the Voyager, and the Surveyor. The smallest of the three vessels, the Explorer, is targeted primarily for ocean and climate data collection. Metocean metrics such as salinity or wind speed can be autonomously collected and processed onboard before being sent to a data storage and visualization server.
The Voyager and Surveyor, on the other hand, are better suited for shorter missions, up to 180 days. These include ocean mapping missions that use radar onboard the USVs. Saildrone's USVs have sailed many waters, helping private companies and governments gather scientific data—from tracking the seasonal movements of Alaska red king crab to melting ice and CO2 levels.
While commercial utilitarian robots have certainly come a long way since the introduction of the Roomba, the level of autonomy achieved by the Saildrone system is a remarkable feat. And while the onboard electronics are certainly more complex than a home vacuum cleaner, the amount of data collected by the Saildrone vessels is indicative of a broader trend toward distributed AI computing.
