Machine Learning Will Automate Maritime Domain Awareness

Saildrone uncrewed surface vehicles (USVs) have conducted extensive data collection missions around the world, continuously collecting meteorological and oceanographic variables above and below the sea surface and building an unprecedented at-sea image library of some four million images. This image library is the foundation of Saildrone’s machine learning software, unlocking new maritime domain awareness (MDA) capabilities at sea, a challenging environment where every pixel is moving across every frame.

MDA is the effective understanding of anything associated with the safety and security of the global maritime domain—illegal fishing, drug enforcement, and limiting intrusion into protected marine sanctuaries. The United States Coast Guard was tasked with exploring the feasibility, costs, and benefits of improving maritime domain awareness using a low-cost autonomous system like the saildrone. Throughout the month of October, Saildrone is demonstrating how its platform could be used to conduct MDA missions offshore and in remote areas of the ocean.

Saildrone USVs are equipped with a specially built 360° camera system integrated with a GPU. Computer vision is based on deep neural networks; the patterns within the collected data set are represented by numbers, which in turn are mathematically mapped to define a model that a computer can be trained to recognize. Saildrone achieved this over several years, leveraging Amazon Web Services (AWS)’s industrial-strength large-scale cloud-based compute infrastructure.

The Saildrone MDA solution combines the cameras, automated identification system (AIS) receivers, and optional radar or infrared cameras for night-time capabilities. The cameras capture images on a very high frequency and the crucial ML software fuses the data from all sensors, recognizes and identifies targets of interest, and automatically alerts the end-user in real time.

Each vehicle can operate for up to 12 months at sea. They use wind power for forward propulsion and are virtually silent while operating, sailing autonomously from waypoint to waypoint. Alternately, saildrones can be tasked to sail in a “racetrack” formation or hold station. The key to machine learning is continuous training of the model; the more saildrones that are deployed and the longer they are deployed, the better the Saildrone model becomes.

During the October demonstration, a fleet of Saildrone USVs will illustrate a variety of use cases and real-world examples to show the system’s effectiveness for improving maritime domain awareness in remote areas of the Pacific Ocean.

Read more: Saildrone Begins Demo of Autonomous MDA Capabilities for USCG

The United States Coast Guard and the Role of Autonomy at Sea

Combating IUU Fishing with Autonomous Vehicles