Navigating the stormy waters of autonomous systems in maritime
A new wave of opportunities
The introduction of autonomous systems presents a plethora of possibilities for many sectors, and maritime is no exception. These systems could improve the maritime industry by reducing fuel consumption through better optimisation of routes, supporting decarbonisation and reducing environmental impact, including deciding when to switch fuels. This could result in a reduction of operating costs and a drop in price of goods being transported by sea which benefits consumers. However, the assurance challenges and other large costs associated with the introduction of new technology means any economic upside will take time to be realised.
Autonomous systems could also bring more indirect benefits to a long-standing challenge in the sector – staffing. Around the world, recruitment in the maritime sector is under strain, with shipping consultants Drewry stating that the “2023 officer availability gap has widened to a deficit equating to about 9% of the global pool…the highest level since it first started analysing the seafarer market 17 years ago.”[1] Safety-wise, the need for smaller crews on vessels means fewer people are put at risk – but there is an attendant challenge of assuring the safety of the autonomous capabilities.
As highlighted in the Global Maritime Trends report produced by Lloyd’s Register and Lloyd’s Register Foundation, even with automation, there will still need to be people on board ships to deal with the safety requirements. The report explains how “technology initially slowed the growth in the number of seafarers needed, but global collaboration ensured that overall trade volumes increased sufficiently to prevent the loss of jobs.”[2] With more seafarers available, it is hoped that more time can be spent by crews maintaining the ship and applying themselves to their work with greater knowledge in the safest way possible. Autonomy was ultimately developed to make the sector safer for employees, something that should be remembered as technology advances.
Safety is paramount
In recent months, debates over the use of AI have been extensive, with many expressing concerns over potential data breaches or bias. One area, however, should not be overlooked: physical safety and the role of safety assurance in the development and deployment of AI-enabled autonomous systems. If vessels have autonomous functions but still carry crew (and/or passengers) then little is different in terms of the objectives for safety and environmental protection. However, if the use of AI, and more particularly, machine learning (ML) provides autonomous functions on vessels, the method of providing assurance should reflect this. ML refers to a branch of AI and computer science that focuses on using data and algorithms to enable AI to imitate the way that humans learn.[3] This is where regulations and standards are lacking – but there is a growing understanding in how to address these “gaps”.
The Centre for Assuring Autonomy (CfAA), a partnership between Lloyd’s Register Foundation and the University of York, and its predecessor, the Assuring Autonomy International Programme (AAIP), has pioneered work on assurance of AI, ML and autonomy. It now has systematic approaches to assurance known as SACE (for systems) and AMLAS (for the ML components), which are being used in several domains, including maritime. Both SACE and AMLAS are tools developed to help safety engineers assess and showcase the safety of both ML components and systems. This information can then be linked into a system safety case, providing a coherent approach to demonstrating the safety of the autonomous capability and its AI/ML components.
Variation in regulation
The state of autonomous system regulations varies across sectors, with (in the UK at least) road vehicles the furthest forward due to the passage of the Automated Vehicles act.[4] There are, however, a vast number of standards in development through organisations such as the International Organization for Standardization[5], for the verification and validation of AI in autonomous vehicles[6] for example.
The differences in regulatory approaches across sectors are likely due to cultural issues rather than government hesitancy. For example, when it comes to autonomous driving, the US approach to regulation is much more reactive[7] than precautionary, contrasting starkly with the UK’s approach. In maritime, the International Maritime Organization (IMO)[8] started work on regulations for maritime autonomous surface ships some time ago and are evolving a ‘code’ for such vessels. However, the IMO has around 175 nations and several other organisations as members, which can lead to slow progress. Meanwhile, individual nation states are making their own rules – which they can do within their own territorial waters – to accelerate trials and the introduction of maritime autonomy in their jurisdiction.
Regulations are often the responsibility of governments and international bodies such as the IMO. However, businesses, including class societies, can get involved in the development of guidelines on how to meet regulations. For example, bodies such as Lloyd’s Register Group in the UK and Det Norske Veritas in Norway[9] have produced guidance on assessment and assurance of software and autonomous functions.
Balancing responsibility and ethical deployment
When it comes to responsible and ethical deployment of AI and autonomous systems in maritime, the issues are really about the possible extent of harm – be it loss of life or environmental damage. For example, if a vessel was to switch from highly sulphurous to clean fuels on entry into national waters, but does so too late, it will cause pollution and likely lead to fines for the shipowner.
Responsible and ethical development involves not just looking at vessel operations, but at the whole life cycle of the maritime autonomy infrastructure, including robotics and cognitive systems. People involved in the development and training of AI, for example labelling training images, often work unreasonable hours, in poor lighting conditions, which is harmful to health. There are further questions which also need answering when it comes to design and development. How can incidents involving vessels be managed, including the recovery of a vessel, without putting the rescue crew at risk? How can robotics and cognitive systems and remote operations be defined to avoid placing unjustified responsibility (blame) on remote operators? How can maintenance be done safely when vessel functions (or even the whole vessel) operate autonomously?
Such questions need to be addressed in design and development of autonomous systems to minimise the risks during operations. But as the world changes, new ships are developed, and new technology is deployed, the questions of responsible and ethical innovation need to be kept under constant review – and these are issues that the CfAA is working on in conjunction with industry and regulators, with the aim of providing impartial advice to all stakeholders.
[1] https://splash247.com/seafarer-labour-market-tightness-at-highest-levels-recorded/
[2] Global Maritime Trends 2050 | LR
[3] What Is Machine Learning (ML)? | IBM
[4] https://www.legislation.gov.uk/ukpga/2024/10/contents/enacted
[5] https://www.iso.org/home.html
[6] https://www.iso.org/standard/83303.html
[9] https://www.dnv.com/rules-standards/
Related Articles
ALTERNATIVE FUELS & LUBRICANTS
ALTERNATIVE FUELS & LUBRICANTS
OPINION