AROS: An Intelligent Arctic Route Optimization System Integrating Machine Learning with Environmental Risk Evaluation

Abstract: Abstract—Shrinking Arctic ice has opened corridors that were impassable until recently, and operators are paying atten-tion. The Northern Sea Route cuts voyage distances between northern Europe and northeast Asia by 30 to 40 percent compared to the Suez Canal path — a saving hard to ignore. But sailing these waters is a different challenge altogether. Ice concentrations can shift from clear to dangerous in a matter of hours, icebergs calve off Greenlandic glaciers and drift on paths that are notoriously hard to predict, polar lows can deepen faster than almost any mid-latitude storm, and congestion data from the high Arctic often arrives too late to be useful. We built AROS to tackle all of this at once rather than piece by piece. It is a browser-only voyage planning tool, written in React and TypeScript, that chains four ML inference modules into a single workflow: a TensorFlow.js neural network scoring route risk from a ten-dimensional environmental feature vector; a Random Forest regressor estimating weather from location and departure date; a K-Means model mapping active vessel positions into congestion zones; and a binary network outputting iceberg collision probabilities for individual ship-iceberg pairs. Ten ports across Arctic Russia, Norway, Greenland, Canada, and Antarctica are supported. Waypoints are interpolated along geodetic arcs and validated against sea-region bounding boxes to keep routes over water throughout. Testing across all supported port pairings confirmed correct risk classification over the full range of ice concentration and iceberg count values, with every computation finishing inside a single browser render cycle. The outcome shows that a client-only architecture is sufficient for operationally useful Arctic maritime decision support.

Keywords: Arctic maritime routing, sea ice concentration, iceberg col-lision risk, Random Forest, K-Means clustering, TensorFlow.js, OpenLayers, route optimization, vessel traffic analysis.