MIDS Capstone Project Spring 2021

Safe navigation by sea is critically important.  Roughly 90% of global commerce travels by sea.  The world’s oceans provide humanity with 15% of its animal protein intake, and fully 10% of people rely on fisheries for their livelihoods.  Collisions at sea hamper commerce and kill or injure mariners.  The European Maritime Safety Agency (EMSA) recorded between 250-350 collisions at sea each year between 2011 and 2018.  In 2017, two U.S. Navy destroyers collided with merchant vessels in separate instances resulting in the loss of 17 sailors and nearly a billion dollars in damage.  Current collision avoidance tools work well in the open ocean but fail to provide accurate warnings when vessels are maneuvering.  This is simply inadequate to produce adequate predictions in and around ports.  We built a tool to fix this problem by providing watch officers on the bridge of a ship with detailed, probabilistic predictions of how the vessels around them will maneuver.

Our tool predicts the evolution of vessel traffic over 30 minutes using only six minutes of positional data as input.  We trained our deep learning model using automatic identification system (AIS) transponder data for vessels in San Francisco Bay captured by the U.S. Coast Guard between 2015 and 2017.  Our intuitive user interface allows a mariner to build a quick and comprehensive understanding of the vessels which will pass nearby.  Armed with this information, the mariner can act, whether by adjusting course and speed, by arranging safe passage by radio, or simply focusing attention on the most important vessels in a crowded port.

Our solution will make marine transportation safer, reduce insurance costs, and save lives.  Interested?  Check us out at!

More Information Logo

Video Demo Demo

If you require video captions for accessibility and this video does not have captions, click here to request video captioning.

Last updated:

April 14, 2021