Abstract – AI for ferries would be a game-changer
Can AI be the Ferry Godmother?
Artificial Intelligence (AI) for maritime applications is normally considered for larger shipping companies and major ports. These are where immediate and major advantages are easily envisaged. This tends to leave smaller operators of diverse fleets in multiple locations out of the picture.
Yet, it can be claimed they have even more need of AI assistance as much of the pollution, security problems, maritime damage and personnel and passenger accidents can be laid at this door. Indeed, improvement opportunities for maintenance and cost savings have greatest value here.
Ferries (Oxford Economics have estimates 15,400 vessels carrying 4.27 billion passengers plus 373 million cars, buses, trucks and trailers annually), fishing boats (estimates at 4.6 million by the UN) and coastal trading vessels (rough estimate of 15,000 vessels) comprise the majority of vessels on the oceans today, ignoring tugs which are addressed generally as port vessels.
These figures are assumed to be underestimations as many ferries and fishing vessels in the Far East and Africa are unregistered and unregulated. The advantages of applying AI to such a large swarm (a swarm is a single networked entity that directs itself using artificial intelligence) of vessels are incalculable.
Opportunities for AI amongst Ferry Operators
For the majority of folk in the western world, ferries are regarded as an exceptional form of transport but if we examine the number of archipelagos which are served by ferry transport, then an entirely different picture emerges. Countries such as below have island groups numbering many thousands and ferry services to match.
- Britain (6,289 islands and 200+ ferry companies though most are single vessel public entities)
- Canada (44,000 islands and 14 ferry companies)
- Finland (46,000 islands and 12 ferry companies)
- Indonesia (18,000 islands and probably around 50 companies)
- Philippines (7,500 islands and 30 ferry companies)
- Sweden (60,000 islands and 19 ferry companies)
Quite frequently the ferry services are controlled or owned and operated by single entities, either businesses or public bodies. There are a lot of single operators that might be encouraged to form co-operatives if the economic case can be made.
Examples of AI and Diversity Requirements
Stena Line in Sweden has already embarked on experimenting with autonomous ferries using AI to reduce fuel consumption and environmental impact. As an example of the diverse nature of ferry companies, this article will cite one serving the three island groups around the UK – the Hebrides, Orkney, and Shetland.
Of the over 200 ferry companies in the UK, 6 serve these islands and the largest of these is a company called Caledonian MacBrayne (Calmac) which has a fleet of 35 vessels serving 29 routes to over 50 destinations.
These range in size from a vessel that can carry one car and twelve passengers (MV Carvoria) to a vessel that can carry 143 cars and 700 passengers (MV Loch Seaforth).
This article is not about Calmac but about similar enterprises with diverse assets running diverse routes and how AI might assist them to save costs, reduce accidents and down-time and generally allow the companies to achieve greater returns with less risk.
How can AI help Ferry Operator Safety?
AI can take into account a large number of factors, such as currents, weather conditions, water depth and speed through water in more combinations than previously possible to do manually.
Despite recent developments, autonomous operation of freely moving vessels is never easy, especially in an environment which is complex, remembering that even small ferries are slow to change course or to stop. Hazards include leisure vessels, paddle boarders and swimmers who do not know or respect traffic rules and, of course, the obligatory man overboard scenario.
There is a common notion that AI will substantially decrease the number of sea accidents as many, if not most, can be attributed to human error (75-96%).
It will, but that fails to address the number of accidents that are avoided by human intervention. As a little aside, during WW2, fighter planes were returning with the wings and fuselage riddled with bullet holes but none on the engine and the powers that be thought that by reinforcing these areas with armour more planes would return.
A mathematician called Abraham Wald said the armour should go on the areas where the bullet holes were not. Wald’s notion was that the missing holes on the engines were on the missing planes. The reason planes were coming back with fewer hits to the engine is that planes that got hit in the engine were not coming back. To a mathematician, the structure underlying the bullet hole problem is a phenomenon called survivorship bias. In a ferry AI context we are looking for the accidents that did not happen because of expert human intervention.
AI for Ferry Operations
If we still need human crews, what is the use of AI? The answer is straightforward, operational efficiency.
Most ferry companies have good data collection about ferry usage but changes to reflect that usage normally take months to implement. AI would be able to reliably predict traffic requirements and select the optimum vessel configuration to address this possibly on a monthly basis.
The knowledge of any particular voyage and potential hazards is shared by AI to any vessel that is chosen to ply that route, by employing a variety of sensor types, including satellite navigation, cameras, standard radar, ‘laser radar’ LIDAR and microphones, along with ‘Automatic Identification System’ radio signals. These AIS signals transmit position, size and routing information. Information about local fixed infrastructure such as aquaculture farms or wind turbines and fixed hazards such as hidden wrecks, shallows and submerged rocks would also be included as chart updates can be slow.
It would appear at first glance that AI optimisation of maintenance and repair would be difficult if the fleet has many different engines, ancillary equipment and deck jewellery. In many areas of the world, ferries, and coastal trade, is carried out by ex-military landing craft and it is not uncommon to find ferries, even in the developed world, that are 20, 30 or even 40 years old. Sensor installation on these could be done fairly cheaply and AI analysis of the data collected would reap rewards far beyond the cost.
The Adoption of AI
The main obstacles for any ferry company are data integration, shortage of specific talent, allocation of sufficient time and energy and trust in AI by owners and operators. Data analysis as a tool to use past experience to improve decision-making is invaluable.
Some of the outstanding benefits of AI for ferry companies include automation, safety, route optimization to take account of tides and weather, and increased efficiency for loading and engine utilisation and performance forecasting. With alternative fuels becoming a necessity and therefore new engine types (which will already come with tens, if not hundreds, of embedded sensors), AI analysis becomes imperative.
Ferries invariably operate in congested areas where existing navigation tools are not suitable. Safety challenges include low situational awareness, shortage of experienced crew, no practical onboard training, and shortage of valuable data.