DON26BZ01-NV027 TITLE: Automated Ice Detection and Polar Navigation Tool (PolarNav)

OUSW (R&E) CRITICAL TECHNOLOGY AREA(S): Applied Artificial Intelligence (AAI)

COMPONENT TECHNOLOGY PRIORITY AREA(S): Human-Machine Interfaces;Integrated Sensing and Cyber;Trusted AI and Autonomy

PROJECTED CMMC LEVEL REQUIREMENT: Level 2 (Self)

OBJECTIVE: Develop a prototype for a system that integrates information on sea ice conditions from a diverse set of sources, including shipboard instruments, airborne and spaceborne sensors, and sea ice model output, to yield optimized route options as a planning aid for navigation through ice-infested waters in polar regions.

DESCRIPTION: Recent trends of warming in the Arctic have led to a steady decrease in the extent of multi-year sea ice, a corresponding increase in seasonal sea ice, and an overall lengthening of the navigable season [Refs 1, 2], thereby making the Arctic increasingly open to maritime traffic. Vessels operating in and near sea ice must make navigation decisions that balance the capabilities of the ship with the objectives of their voyage. Such route planning is complicated by the dynamic nature of sea ice, as it is subject to movements caused by a number of factors such as the Beaufort Gyre, transpolar drift, and weather events, which are even more pronounced on the thinner, seasonal ice. A system capable of aiding navigation teams in route planning based on ice observations and forecasts over time scales on the order of hours to days is essential for safe navigation through polar regions.

Currently, ice navigation relies heavily on manual processes. A majority of route planning information comes from satellite imagery, either optical or synthetic aperture radar (SAR), or from forecast information from entities like the U.S. National Ice Center. Due to dynamic weather conditions and rapid movement, the operational value of overhead imagery is sometimes temporally limited. These longer-range data sources are augmented by shipboard systems, such as onboard radar systems for icebergs, in-situ ice floe, and pack ice detection, which typically have detection ranges on the order of a few tens of kilometers. These close-range systems help inform tactical navigation decisions and near-term route planning.

Key aspects of ice analysis, whether conducted onboard or remotely, are ice edge definition, identification of ice types (e.g., seasonal ice, multi-year ice) and concentration, and detection of ice features such as ridges and icebergs. This analysis is then presented to the navigation team and command who assess the current and planned route and make course adjustments as necessary. Current ice forecasts do not always adequately account for projected ice movement over the next 12-96 hours, which is crucial for effective route planning. Moreover, the analysis and route planning are often separate functions, each conducted by distinct teams based on their own personal experience and knowledge. This separation can lead to suboptimal decisions and increased risk.

The goal of this SBIR topic is to develop a prototype tool that helps ships make safe navigation decisions in the Arctic. The tool should leverage established ice prediction models and incorporate other available sources to assimilate models and improve forecasts. These additional sources may include:

• Onboard sensors: Radar, thermal cameras (forward-looking infrared), and microwave sensors on the ship.

• Aircraft sensors: Sensors on airplanes and unmanned aerial systems (if available).

• Satellites: Optical and SAR data, dynamically updated with every new overpass.

• Iceberg records: Historical data on where icebergs have been seen/located (e.g., from the U.S. Coast Guard's International Ice Patrol).

The envisioned product is a geographic-information-system-based tool that uses artificial intelligence, first-principles algorithms, and automated data processing schemes to combine information from the above sources, update model-based predictions, provide 12–96-hour sea ice forecasts, and suggest potential navigation routes. Route options should consider vessel specifications, such as ice resistance characteristics and fuel consumption rate, and provide options for fastest route to destination, shortest route to destination, route with minimal wear/tear on vessel and crew, and maximum safe speed based on ship hull type/construction. Ultimate route decisions should be left to the vessel’s navigation team.

PHASE I: Draft a conceptual framework for dynamic route planning based on sea ice characterization and forecasts from data fused and integrated from disparate sources. Define and develop in detail the concept and methodologies for extracting and combining data from diverse sources to provide nowcasts and forecasts of sea ice conditions and route planning options at tactically relevant spatial and temporal scales. Prepare a report containing preliminary results of retrieving sea ice characteristics using fused multi-modal satellite imagery, a framework for improving predictions through assimilation of data from diverse sources, and a framework for dynamic route planning which will be the basis for the proposed tool. If the Phase I Option is exercised, carry out a simple demonstration using multi-temporal and sequential datasets from multiple satellites and/or in situ measurements and modeled sea ice predictions for a specific region to test approaches for improving predictions of sea ice distribution and movement and use the combined information for route planning in hindcast and forecast modes.

PHASE II: Develop a prototype data analysis and route planning software tool that can be tested operationally on a vessel and is in the form of a standalone system with a display interface showing the latest satellite imagery of the ocean in the vicinity of the vessel, akin to the display on a Global Positioning System (GPS) based automobile navigation system. This prototype should be able to connect to data streams from instruments onboard the vessel, near-real-time satellite data that could be downloaded through the vessel’s onboard communication system (e.g., satellite communications), and sea ice model output; produces nowcasts and 12-96-hour forecasts of sea ice conditions; and provides multiple route options for navigation, optimized for the fastest route, shortest route, the most fuel-efficient route, or the route with the least ice encounter.

PHASE III DUAL USE APPLICATIONS: Further develop the prototype into a commercial tool for integration onto a U.S. Coast Guard icebreaker or an ice-hardened Navy vessel. The tool, operating onboard the vessel in a place of the navigation team and command’s choosing, shall provide real-time updates of sea ice conditions in the vicinity of the vessel and route options that the vessel navigation and command team can use to make informed decisions for sailing through ice-infested waters. The tool will also find its use in commercial industries such as shipping, fishing, and tourism in the polar regions.

REFERENCES:

1. Babb, D. G. et al. "The Stepwise Reduction of Multiyear Sea Ice Area in the Arctic Ocean Since 1980." Journal of Geophysical Research: Oceans, 128, e2023JC020157, 2023. https://doi.org/10.1029/2023JC020157
2. Stroeve, J. C. and Notz, D. "Changing State of Arctic Sea Ice Across All Seasons." Environmental Research Letters, 13(10), 103001, 2018. https://doi.org/10.1088/1748-9326/aade56
3. Stroeve, J. C.; Markus, T.; Boisvert, L; Miller, J. and Barrett, A. "Changes in Arctic Melt Season and Implications for Sea Ice Loss." Geophysical Research Letters, 41, 2014, pp. 1216-1225, https://doi.org/10.1002/2013GL058951

KEYWORDS: Polar navigation; artificial intelligence/machine learning; AI/ML; sea ice forecast; route planning; satellite imagery; data fusion; Arctic; Antarctic; ice identification; ice classification; ice prediction; ice analyst; Meteorology and Oceanography; METOC; remote sensing; modeling; shipboard sensors; human-machine interface; big data

TPOC 1
Wesley Moses
wesley.j.moses.civ@us.navy.mil

TPOC 2
John Woods
john.e.woods.civ@us.navy.mil