USS John McCain Credit US navy web.jpg. Caption: Damage on USS Fitzgerald caused from collision with container ACX Crystal. Credit: US Navy.
Artificial intelligence-powered software could improve ship safety but there are questions about how the technology will affect crews’ working lives
Intelligent software powered by artificial intelligence (AI) is beginning to enter the shipping industry, with its proponents claiming it will optimise ship operations and make them safer and more efficient. The technology’s capability to analyse data from multiple sources without getting distracted or making errors of judgement promises to reduce instances of human error that are so often blamed for causing shipping accidents.
However, some industry voices are calling for a measured approach in adoption of the technology to ensure human error is not replaced by machine error.
The advanced technology shows great potential, but shipping must be careful it does not work against the skills and abilities of human navigators, David Patraiko, director of projects for the Nautical Institute and co-ordinator of its technical committee, told SAS. “The big challenge with AI across all industries is making sure it remains human-friendly. If this type of technology is going to be used to make systems more autonomous and therefore take people out of the loop when there is a problem, we cannot then expect people to jump back in and help when there is a problem.”
However, Frank Coles, Transas CEO, told SAS that the aim of its new A-Suite AI navigation tool was not to take crew out of the decision-making process, but to use machine intelligence to “augment the human in the loop”. The suite is available to end-users now and will become “fully operational” over the coming months. It comprises three core modules: advanced intelligent manoeuvring (AIM), advanced intelligent diagnostics (AID), and advanced intelligent routing (AIR).
AIM is a predictive system and “anti-collision support tool”, which helps predict and prevent incidents by learning the behaviour of crew members. It does this by comparing their actions with those previously collected from personnel sailing in the same location. This, combined with a hydrodynamic model of the vessel and anti-collision regulations that have been coded into the system, provides ‘advanced decision support’ for crew.
Critics of AI have warned against transferring decision-making away from skilled crew, but the change in approach could have positive legal impacts. Allan Graveson, senior national secretary at trade union Nautilus International, said, “The technology could help avoid the unwarranted criminalisation of seafarers. With decision-making comes responsibility and consequences when something goes wrong. Captains have lost their jobs, their reputations, and even their freedom when ships have got into trouble. Transferring some decision-making to a computer could also transfer liability.” Graveson claims the approach could bring shipping more in line with the aviation industry, where the captain of an aeroplane relies heavily on guidance from the ground and other remote monitoring systems, yet remains in command and in control of the aircraft.
Meanwhile, AID is designed to detect anomalies and support real-time decision-making, as well as a “more methodical” post-voyage analysis. By taking data from equipment and environmental sensors, it is able to spot excessive or unusual manoeuvring patterns, speed, and rates of turn, as well as unexpected changes in fuel consumption. It also records how and when operators interact with vessel controls. AIR is used for voyage planning and optimisation using real-time oceanographic and marine data including current, wave, sea level, and meteorological (metocean) data, hazards, and anticipated vessel traffic along the route and known bottlenecks.
Importantly, the system does not require crew to update special electronic charts (SENCs), weather data, and other navigational notices, thanks to an advanced data delivery (ADD), advanced remote maintenance (ARM), and AIR, which automates these procedures. This is all built upon Transas Harmonised Eco System of Integrated Solutions (THESIS), a scaleable platform launched in 2016 that enables ships, ship operations offices, and training facilities to communicate and share information to “reduce the administrative load” on ships.
In a bid to overcome ‘alarm fatigue’ experienced by crew bombarded by alerts from onboard technology, A-Suite is claimed to generate “far fewer real-time alerts than other systems”. The logic in A-Suite takes individual alerts when single parameters have exceeded their threshold values and then interprets them as part of a holistic assessment of the ship’s situation. “This means it consolidates multiple standard alerts and reduces cognitive load on the user,” Coles said, adding that it not only warns users about abnormal situations, but also provides recommendations on how to prevent the situation from escalating.
This aspect of the technology could help lighten the burden on navigators, whose jobs are becoming increasingly complex, said Patraiko. “Mariners have very complex lives right now; they have to make sense of a lot of information and check various systems against each other to detect errors. The problem is that humans can only do that with limited success. For example, if a human can check a satellite system against a visual system against a radar once every half hour, that is efficient, but if a machine can do that 30 times a second then alert the mariner if it detects something, that has to be a good thing. In addition, computers never get bored, unlike people, and they rarely make mistakes.”
“A core objective of THESIS, and by extension A-Suite, is to enable collaboration through the whole operational chain,” Coles said. “Alerts generated by A-Suite are automatically received on shore, which facilitates decision support from a vessel’s head office in critical situations.”
With the introduction of any new technology there comes a transition period, where crew must learn how to use the technology safely. Taking ECDIS as an example, this process is often not a smooth one, with the technology not as straightforward or as useable as it is touted to be. Coles stressed that training crew to use A-Suite “is much more straightforward than ECDIS” and is “highly configurable” so can be set up to provide users “with as little or as much information and alerts as they feel comfortable with and in alignment with a fleet’s individual operating procedures”.
A-Suite also includes an e-learning tool called Advanced Remote Training for Seafarers, which provides online access to manufacturer-approved, type-specific training courses for Transas ECDIS, compliant with SOLAS, ISM, and STCW requirements.
AI has become something of a buzzword over recent months as advances in machine learning and heavy-duty computing power have resulted in the first truly viable real-world applications across many sectors of industry.
It represents a fundamentally different approach to software whereby the machine learns from examples, rather than being explicitly programmed for a particular outcome. The technology has already made inroads into everyday life, providing an efficient means of searching for images on smart phones just by typing in the subject and it enables intelligent voice-activated assistants, such as Siri and Alexa, to hold conversations with humans.
AI has previously been touted as ‘the next big thing’, but the confluence of greater connectivity and availability of huge amounts of data, coupled with the inordinate computer processing power needed to recognise patterns in that data, has created a thriving environment for research and development.
Much debate in the shipping sector has focused on whether AI will ultimately take seafarers’ jobs away or serve to augment their skills. Exploring the technology with “eyes wide open” will ensure a positive outcome, said Patraiko, adding, “While I look forward to applications for AI that will assist mariners in performing complex tasks, the technology must be thoroughly tested. There have been cases with autonomous systems in cars that can cause an undue lapse in concentration or unforeseen circumstance. Without sufficient testing there is the danger we will introduce machine-based errors that need to be anticipated.”
Some critics have claimed that AI, coupled with the rise of autonomous shipping, could devastate seafarer employment. However, the transition could in fact improve their prospects, claimed Graveson. “It’s not a question of getting rid of jobs, it’s a question of the transfer of jobs. The life of the seafarer today is almost an 18th century existence in terms of working conditions. People are working over 90 hours a week and ruining their long-term health. Wouldn’t it be better if their roles were transferred to a centre on shore where they can monitor and control the ships working maybe 35 hours a week and potentially being paid a lot more?”