AI for carbon reduction in ship navigation
AI for carbon reduction in ship navigation
The RRS Sir David Attenborough (SDA) provides an opportunity to revolutionise ship management using Artificial Intelligence. The SDA Digital Twin project will develop and integrate AI algorithms that exploit the diverse range of available environmental datasets and forecasts to develop a real time polar route-planning toolkit that can be used by the Captain for route-planning decision support on the bridge. These planned routes will be displayed via digital dashboards that enable proposed courses and parameter settings to be visualised in tandem with existing information services on the bridge.
BAS has committed to achieve Net Zero carbon emissions by 2040, aligning with NERC and UKRI sustainability commitments. Part of the BAS implementation plan comprises the reduction of shipping emissions by developing fuel-efficient routes that the Captain can use to minimise SDA fuel usage. To achieve these goals the route-planner will leverage Machine Learning-based sea ice/weather forecasts in order to provide both long horizon seasonal route-planning and real time route updates. The toolkit will provide the capability for the Captain to compare and evaluate different possible routes between destinations in terms of their fuel and carbon costs.
The route-planning functionality will comprise a fundamental component of a Digital Twin of the SDA which will both contribute to, and benefit from, other ongoing Digital Twin projects at BAS.
The project will provide initial demonstrators of the key components:
- Phase 1 – ending March 2023
- To provide an automated decision-making tool to optimise passage planning and overall carbon efficiency.
- Phase 2 – ending March 2025
- A risk-aware navigation system supporting route-planning in complex ice conditions.
- A cruise-planning tool that utilises the route-planner to optimise the selection of science goals and the scheduling of science and supply operations
- Phase 3 – beyond March 2025
- Deployment of the route-planning capability on the marine autonomous vehicles
- Extension of the cruise-planning system to include fleets of manned and unmanned vehicles including the SDA.
Phase 1 of this project will enable the first steps in the development of a Digital Twin of the SDA. We will focus mainly on external data sources for modelling the environment and informing decision-making. In phase 2 and beyond, we will use underway data (such as speeds and fuel requirements in different sea states, performance in ice, etc) to refine our models of how the vessel responds in different conditions. Initial demonstrators of the route-planning capability will be delivered which will form the foundations of further work to complete the Digital Twin of the vessel.
In the following we present a sequence of interactive plots that show the current capability of the route planner. The intention is for viewers to click through the interactive layers to explore the different features currently available.
Outlined below is a series of interactive examples of the work we are currently doing. Clicking the picture will take you to the individual interactive plots.
Example 1 – In this example we compare the information content of a mesh of Sea Ice Concentration (SIC) relative to the raw data. Try turning off and on the separate layers in the animation, found on the right of the plot, and zooming around the region. The greyed-out region represents land, and regions coloured in purple show the percentage of sea ice concentration, with darker purple a higher sea-ice concentration. Please click picture to go to interactive plot.
Example 2 – In this example we demonstrate the maximum ship speed and fuel usage across the region as two additional layers in the original interactive plot. In areas where the sea ice is too dense for transit are represented by an additional layer of ‘Extreme Sea Ice’, a region of no-go for the ship. Try turning on the separate layers and see how the sea ice concentration affects each of the vehicle specific layers. Please click picture to go to interactive plot.
Example 3 – In this example we demonstrate the routes from a location close to the BAS research station ‘Rothera’. In this example we show a series of travel-time optimized and fuel optimised routes from this location to a collection of points of interest, providing the travel-time and fuel usage of the routes in separate layers in the interactive plot. Try clicking on the lines and compare the travel-time and fuel usage differs for the fuel and travel time objective functions. Please click picture to go to interactive plot.
Example 4 – In this example we dive in on a single path between ‘Rothera’ & ‘SR4Bottom’ to inspect the differences between the fuel and travel time optimised routes. Notice that the fuel optimised route travels further away from the ice, increasing the transit time by 7hrs 42 minutes (0.324 Days) compared to the quickest travel-time route. However, the fuel route saves 6 tonnes of fuel (19 tonnes tCO2e, tonnes of carbon dioxide equivalent). This same carbon saving is the same as: heating ~9 houses for a year; ~11 single person flights from London to New York; or, driving a petrol car almost 3 times around the earth.
Jonathan D. Smith, Samuel Hall, George Coombs, James Byrne, Michael A. S. Thorne, J. Alexander Brearley, Derek Long, Michael Meredith, Maria Fox, (2022), Autonomous Passage Planning for a Polar Vessel, arXiv, https://arxiv.org/abs/2209.02389