Data Science & the Seven Seas: Collision Avoidance
International regulations for preventing collisions at sea (COLREGS) codify long-established international norms for maritime navigation. However, many situations that are frequently encountered by mariners, such as three vessels converging, are not covered by COLREGS. Similar to automotive traffic laws, the safest maneuver, in a given situation, depends on many factors. This challenge will use data from ships underway on the high seas to develop algorithms to assist the Navy with preventing collisions for human-operated and autonomous vessels.
The Data and Environment:
For this challenge, the primary data is collected from the Automatic Identification System (AIS), which provides unique identification, position, course, and speed data to other ships and shore stations. AIS data is enormous and rich (the 2016 United States dataset alone has close to 10 billion records), but it was designed as a safety feature for local collision avoidance communication between ships at sea, not to be analyzed holistically. As such AIS data provides huge opportunity for understanding the contextual nature of interactions between ships at sea, but it also presents significant challenges.
Contestants are welcome to work with any tools they choose, but we will provide access to an AWS environment with tools such as Python, R, and ESRI’s ArcGIS mapping and analytics platform.
Challenge 1: Identifying COLREGS Interactions
Out of the billions of data points in the AIS dataset can you find the most efficient algorithm to identify potential COLREGS interactions?
Challenge 2: Contextualizing COLREGS Interactions
COLREGS are complicated and multifactorial. Can you apply contextual awareness to encounters between ships in order to help the Navy develop models for what constitutes “normal” behavior at sea?