The Maritime Future Summit has highlighted what artificial intelligence could do for shipping and how top experts assess the opportunities for autonomous ships today – realistically speaking. The revolutionary tenor has gone, reports Felix Selzer
[ds_preview]The Maritime Future Summit has highlighted what artificial intelligence could do for shipping and how top experts assess the opportunities for autonomous ships today – realistically speaking. The revolutionary tenor is gone
The Maritime Future Summit (MFS) kicked off the conference program at the digital edition of SMM 2021. At the third edition of the MFS, »everything was a bit different,« as moderator Volker Bertram of the World Maritime University noted. Apart from the fact that this year‘s event was free to stream as an online conference, a different take on artificial intelligence (A.I.) and autonomous shipping was also new. Where bold visions of the future and enthusiasm usually prevailed on these topics in previous years, this time there was more sobriety and a look at the next – rather than the next after the next – steps in technological development.
Volker Bertram set the tone when he started the conference with an online poll: The avergage IQ is 100, Einstein’s was 170 – what is the average IQ of current artificial intelligence solutions? Correct answer: 40. So, what potentials does A.I. based technology have in store for us realistically speaking?
New solutions, new business
MFS-Speaker Pierre Sames, Group Technology and Research Director at DNV GL – Maritime, sees great opportunities in the use of AI for the development of new solutions and services. In 2019, the classification society established a dedicated department for A.I. solutions. For example, the classification society has used machine learning to improve the customer experience by shortening response times to their questions. Pattern identification and computer vision are being used to look at the future of vessel surveys. Currently, DNV GL is also training a new algorithm for battery health prediction using existing data, which now has become »a real asset« for the company. »But just having a core A.I. algorithm isn’t good enough. You need to embed it into an application or into your production system,« Sames said.
Assurance of digital assets is a new arena for the classification society, since A.I. based solutions come with new risks. »The question is: Have they be trained sufficiently, have they been tested accordingly and is the performance of those A.I. solutions according to our exceptions?« Sames said. »The principal problem is that we cannot ask the algorithm why it has given us a certain answer. So we need to find out how a decision has been made.« DNV GL has recently released a position paper on »trustworthy« A.I. solutions.
From gut instinct to data
Pierre Guillemin, VP Technology at Wärtsilä Voyage, said that the maritime ecoysystem poses certain challenges, e.g. the isolation of vessels from the network perspective. »Bandwidth, latency and cost are still barriers for A.I. deployment at sea. Other obstacles are questions around the access to data, data ownership, too many proprietary data formats and systems and a lack of standards and regulations in a »fragmented ecosystem,« Guillemin said.
Opportunities regarding operational and energy efficiency, decarbonisation and safety could be drivers for the adoption of A.I. technologies, he thinks.
But the maritime industry can benefit from other industries that have invested in the A.I. domain and are »already enjoying the fruits of these investments«, Guillemin said, especially with regard to vessel safety and autonomy. In his view, Artificial Intelligence offers the chance to take the step »from gut feeling to data and connected ships.«
»The business case is crucial«
Known as the »guru« of autonomous shipping, Oskar Levander, formerly at Rolls-Royce, now SVP Business Concept at Kongsberg Maritime, looked at things from a technical as well a economical perspective. He said that having already presented working autocrossing solutions, remotely controlled ship operation was not far off. »We already have control of the ship, now we need to know what’s going on around the ship,« he said. »Enhanced awareness« is the buzzword, and old-fashioned tools like radar will be augmented with cameras and other sensors to capture the ship‘s environment in all dimensions – »sensor fusion,« Levander calls it. Connecting such a system to the shore station will be the next step.
However, the business case is crucial for all discussions about the use of A.I. and ultimately remote-controlled or autonomous ships. What are the economic drivers?“ asks Levander. He sees a business case for one-third of the global fleet. Many ships would use artificial intelligence and decision support in the future, but 2/3 of the fleet is not suitable for unmanned shipping, he says, because of operational patterns or for technical reasons.
Reliable evaluation is crucial
Jilin Ma from China Classification Society (CCS) looked into the challenges and opportunities that come with the »unmanned bridge«, i.e. a remotely operated or autonomous vessel. Interestingly, he started by looking at the human side of things – no, not the »human element terms of human error. »Recruitment is becoming increasingly difficult as many young people are not attracted by the life at sea at all and there are no income advantages compared to jobs on shore,« Jilin said. Meanwhile, requirements of the IMO maritime labour convention increase manning costs. Crews also need accommodation, air conditioning and supplies. An unmanned bridge could also be repositioned, the design vessels like containerships could be changed allowing for more containers on board.
Why haven’t we achieved this goal in the last twenty years? How can a seafarer be replaced technically. Officers on watch processes data from a variety of sources (his own eyes, Radar, charts, AIS etc.), evaluate them and make decisions after communicating with other humans on board or at shore. Situational awareness through a combination of Radar, Lidar, cameras and other sensor systems has been enhanced in the past couple of years to an extent where it surpasses the ability of humans, said Jilin. »However, the interactive evaluation and communication can be a big challenge for machines,« he said. »If the A.I. evaluation can be proved reliable, execution may not be that difficult anymore. The recent past has shown that the technical problems can be solved. Does that mean it is time for us to fully adopt the unmanned bridge. In my opinion it is too early,« Jilin stated. CCS is very confident that the unmanned bridge will become reality at some point, but regarding the reliability problems with A.I. decision making as well as international regulatory issues, »it is still a long way to go.«
Time to analyse actual use cases
Jotun’s Regional Category Manager Hull Performance, Tom H. Envensen, demonstrated how A.I. translates to fuel savings. The coating manufacturer has already presented its hull cleaning solution »Hull Skater«, a robot that proactively removes the biofilm at the earliest stage of fouling. A fouling prediction algorithm was developed in-house using data from multiple sources: sailing data (area, speed, activity), idling data (period, depth, distance to shore), oceanographic data (temperature, nutrients, salinity) and protection data (coating technology, age, maintenance). Like Levander, Evensen calls it »data fusion«. The algorithm dictates when to employ the hull skater on a vessel. The company promises a 0,5% average speed loss over a five year period – or better.
Rodrigo Pérez Fernández of Spanish ship designer Sener looked at »paradigmatic cases for the application of A.I. in ship design«. One would be the routing of distributors – pipes, cables etc. – on board and the interaction of distributors with other elements of equipment, restrictions for routings, regulations, optimization of design and fabrication. Another would be the production block assembly at a shipyard, depending on maturity. An A.I. system keeping track of all alteration to the design, availability of materials etc. can help a production manager to assemble the blocks even while not completely outfitted. »The tools are here. Now it is time to analyse and establish our particular use cases.«
»Don’t be wowed, don’t be scared«
Annie Bekker from Stellenbosch University in South Africa, rounded of the conference programme by sharing experiences from her last expedition to the Weddell Sea in Antarctica. While the scientists used advanved robotics to survey the sea floor and search for the wreck of Ernest Shackleton’s ship »Endurance« that sunk in 1915. Bekker evaluated the technology used to get safely and efficiently through the ice to the wreck site, when experienced ice pilots were supported by digital solutions – A.I. versus human intelligence. In open water the machine learning algorithm and the human experts’ projections corresponded quite well. But with ice it is a different story as ice is complex and delivers »a lot of atypical data, which is very difficult to train an algorithm on,« Bekker said. Although technologies based on cameras and other sensors delivered at least some promising results measuring ice movement and thickness, the arctic conditions and the unpredictable ice and snow themselves made human intelligence indispensable. The expedition ended when the research vessel became trapped in ice, endangered to be crushed like the »Endurance« and an the underwater ROV was lost.
»A.I. is not smart in a human sense. It has no smart ideas, no curiosity. AI was formerly called numerical statistics – quite unsexy, huh? Today it is called A.I. but it is still numerical statistics – very sensible, very useful. Don’t be wowed, don’t be scared, A.I. is an idiot, but a useful one. We would be stupid not to use it,« moderator Volker Bertram summed it up quite well.
