Out on their own: The world of autonomous shipping
Everyone asks – what about the guy in the kayak?
Victor Hugo wrote, ‘What leads and pulls the world along is not locomotives, it is ideas.” And as ideas go, autonomy is quite a big one.
But those involved are trying to minimise the conceptual leap. “Autonomy is a natural step – it’s nothing new,” says Peter Due, Kongsberg project manager for Yara Birkeland, the world’s first fully autonomous containership due for launch in 2018. He underscores this by saying: “In fact, we have been able to build a vessel like this for quite a long time.”
He has a point: not only does the company already manufacture the control kit for smaller autonomous surface craft, its well-known dynamic positioning systems have been stretched to include limited moves between wind farm towers and follow it with automatic gangway touchdown.
Still, the progression is being handled carefully: Yara Birkeland will initially operate as a manned vessel, moving to remote operation in 2019 and expected to be capable of performing fully autonomously from 2020.
It has to be said, unsurprisingly (and perhaps reassuringly) there’s a lot of shared ground with other ideas. Despite not being directly involved in the build stage, the 70m, open top box ship is DNV GL’s Re:Volt concept “pretty much as we envisioned it” explains Bjørn Johan Vartdal, Head of DNV GL Maritime Research. Re:Volt saw the potential for slow-going coastal operations “just ticking along” says Vartdal: travel wouldn’t be fast, “but regular as a conveyor-belt”. It’s worth noting that Yara isn’t particularly interested in vessels per se, just avoiding the safety and pollution issues that come from crossing a domestic area with 40,000 trucks a year. So, Birkeland is going to shuffle 100 to 150 containers at a time between Yara’s production plant at Herøya in Porsgrunn and Brevik port (7.5 nautical miles) and Larvik port (32 nautical miles) where the containers will be picked up by larger vessels.
The feeder will have two azimuthing pods and a pair of tunnel thrusters, powered by a 3.5 or 4MWh battery pack, giving it a 6-knot transit and 10 knot top speed. Fully electric vessels are not just in keeping with Yara’s green agenda, it also smoothes integration with the control systems - and as Mr Due points out, “electric propulsion requires less maintenance - a Tesla car engine has about 20 moving parts whereas a conventional engine has hundreds”.
The savings on autonomous vessels are quite clear: no bridge, no human support and the build will be “significantly” less costly than a standard ship says Mr Due, though he declines to say by how much at present. More importantly, they will also be much cheaper to operate: it won’t just benefit from lower maintenance, hotel loads and onboard crewing costs will be slashed.
But interestingly, as it will start off as a crewed vessel, Birkeland will be fitted with a detachable bridge based on a standard, 20ft container. “The reason for using a container is that you can use standard handling equipment quay-side to attach and detach the bridge unit,” explains Mr Due. As different areas are likely to take varying amounts of time to place appropriate legislation, this gives the design a fallback position. Further, he adds: “If a company owns several vessels and you need to sail to a destination outside your regular operating area, you can use the same bridge on all of them.”
While most people will admit that Kongsberg has mastered control technology and some will argue that dynamic positioning operations are halfway there by responding to wind, waves and current, other elements still leave them a little uneasy.
Both Due and Vartdal make the same point: Due underlines that “autonomous vessels must – must – be at least as safe as a manned ship” and Vartdal adds “in practice – probably a fair bit safer”. Therefore the input from a number of sensors are fused, including AIS, radar, LiDAR, and finally 360° and infrared cameras: Kongsberg’s sensors “can track a beer can at 150 metres,” says Mr Due.
Of course behind this, the ship has to make sense of even unlikely scenarios: “People always ask, ‘what about the guy in a kayak?’” says Due.
Firstly he points out that even macho guys in kayaks tend to stay out of shipping lanes. Secondly, the fused sensor input is tied together with recognition programmes and rigorous collision avoidance systems – and if the onboard software really can’t decide what to do, the ship will keep station on DP, throw a message back to the control centre onshore and get a human to guide it remotely, via the sensor input and cameras.
Still, Vartdal admits the ‘ethical’ issue of the ship choosing who to hit in a case where collision is unavoidable “is still a whole can of worms” requiring some hard programming, maybe against a sliding scale of vulnerability. He points out that while it’s more of a problem for the automotive industry, it’s not such an issue for slow moving marine traffic, especially since there will probably be another pair of eyes on the vessel in busy harbours. He added: “There’s no such thing as a risk-free operation. You have to evaluate, ask if this an acceptable risk or not. But if it’s proving to save a lot of human lives – even if only because it’s keeping people off the ships - then I believe it will be considered acceptable.”
However, these systems are being designed to educate themselves, and might, eventually, be able to name that beer can floating by. Or, at least, tell the difference between a can and a net float.
Machine learning will be a part of the ship’s planned capacity explains Mr Due. “If the vessel sees a floating log, and can’t identify it, it can send off images for analysis – and it will be able to start to build up recognition.” He adds that eventually, these ships will not just be able to learn from their own experience, “several vessels will also be able to learn from each other”.
Yara Birkeland is not the only autonomous ship involving Kongsberg to be launched next year. Hrönn, a 35m long by 10m beam offshore utility vessel being built under the auspices of DNV GL, is the brainchild of a Norwegian and UK consortium led by Automated Ships. Initially, the unmanned, light duty shuttle will service offshore installations as a remotely piloted ship: to this end, the control systems, including dynamic positioning, automation and ECDIS will be replicated at an onshore centre.
The same concept has many other purposes; in fact, Maritime Robotics – one of Kongsberg's partners - has just taken the lead on robotic inspection, monitoring and maintenance of fish farms.
However, it’s worthwhile noting Hrönn’s system isn’t yet completely evolved: Kongsberg say that the transition to full automation will take place “as the control algorithms are developed concurrently during remotely piloted operations”. It’s not a big issue says Tom Eysto, Vice President at Kongsberg Maritime, “as common technology is taking us a long way”, adding that it’s now possible to ‘teach’ the machines from video footage. He added the system still needs to be “sharpened... to the point where we can go to DNV GL for certification”.
Both these vessels’ remote phase will still put a lot of reliance on the control centre: is that actually feasible? In answer, Mr Due explains that back in the 90’s Kongsberg remotely operated other vessels. He also points to another Kongsberg strength: Norway relies a great deal on flights to connect its cities but the airport control towers have proved too expensive to man all the time, therefore the company has already developed around 13 remotely controlled airports. “The flight controllers say they actually get a better overview of the flights than if they were actually on the site’s control tower,” says Mr Due. “We are going to be using some of this technology in the remote marine centres.”
Despite that, both Vartdal and Päivi Haikkola of DIMECC, a Finnish innovation centre, points out that the apparent smooth transition from remote control to autonomy isn’t quite as straightforward as it seems: “In some places you don’t have the bandwidth from the satellites to cope with the huge amounts of data coming from the ship, while in other less trafficked areas, like the Arctic, you may not have signal coverage from the satellites at all.”
Although many believe that the areas of ‘thin’ connectivity will be addressed within the next five years, she points out “you always have to be prepared for the signal to go down”. As a result, any ship controlled from an onshore base, “has to be ready with a backup - and if you have discounted a human crew, then logically that backup system has to be autonomous”. For this reason, she says that the development might be faster than one might expect, leapfrogging some of the accepted stages.
Haikkola adds although a number of vessel types will eventually be aiming at full autonomy as that’s where most savings accrue, there are other reasons for wanting a slice without taking the whole pie.
“Cruise ships will never go to full autonomy... but they are very, very interested in safety,” she explains, pointing out that one of the Finnish players is cruise operator Meyer Turku. Ms Haikkola can see the case for “decision support”, first of all from remote stations, but later “completely from onboard systems” being compelling.
Further, Anderson Chaplow of Lloyd’s Register, which has been working on a new ‘unmanned’ code, says he sees many operations being far more fluid than originally conceived, possibly shifting between the different states “with full autonomy at sea, then possibly remote operation in the harbour”. Interestingly, the class society has also been part of the ‘remote tug’ trial involving Svitzer – another application that may eventually fold in some elements of the new technology.
According to Iiro Lindborg of Rolls-Royce, the acceptance of autonomy rests, at least partially, on a faith issue.
He points out Rolls-Royce, lead of the Advanced Autonomous Waterborne Applications Initiative (AAWA), has already been ascending the staircase toward autonomy since the start of data logging in back in 2005. So, he says it’s not such a big jump to turn the process around into a two-way street and give the ship the direct ability to respond to the data it’s producing. However, Lindborg points out that “most people still don’t trust technology for vital operations... we need to create that trust first”.
Therefore Rolls-Royce is feeding the market bite-sized chunks. Its Automatic Crossing System, due to be installed on no less than four, double-ended, battery-powered ferries developed by Multi Maritime for owner Fjord, combines existing technology to useful effect.
As Lindborg explains, the idea is the first ferry will be navigated out of the berth by the captain: he then presses a button to initiate the Automatic Crossing operation and the vessel accelerates on a predefined track. When it’s reached the desired speed, it switches over to transit propulsion mode.
Again, at a pre-determined point the ferry decelerates, activating the forward thruster and the captain is notified he has to take back control to dock. There are a couple of safety features: the speed can be adjusted during transit and the captain can disengage the system and return to manual handling by pressing a button. More, if the captain doesn’t take control when requested, the vessel will just stop.
One especially interesting point is that the kit itself is retrofittable to any vessel with an autopilot. Mr Lindborg explains: “Autopilots have fairly standard configurations... and so it made sense to use that interface rather than create another.”
It may not sound revolutionary – indeed, it’s not designed to. Mr Lindbord says that the benefit accrues from automatic adjustment of propeller pitch and rpm and especially the acceleration and deceleration phases: “These can be the wild cards on fuel efficiency,” he adds. It may not sound revolutionary – indeed, it’s not designed to. Mr Lindbord says that the benefit accrues from automatic adjustment of propeller pitch and rpm and especially the acceleration and deceleration phases: “These can be the wild cards on fuel efficiency,” he adds. Alongside this is Rolls-Royce’s development of the remote system for the Svitzer Hermod, which again bases itself on well-developed control technology – in this case the DP system.
Interestingly, the second phase of the ferry development will add an automatic docking system, defining the points of entry with a 3D map and using sensors to look out for obstacles on the way. “Then the next step is a situational awareness package that will assess other vessels - small and large - to help safety and crossing performance.” Like Kongsberg, the system will tie LiDAR together with optical cameras of different varieties along with other input. “It should recognise over a hundred different types of object... ships and other things like floating containers”, he says, adding that “fishing nets are actually fairly difficult”. Also, he admits that one challenge is incorporating still-novel technology: “LiDAR – Light Detection and Ranging - is useful as it can create a 3D map of its surroundings... but it is a totally new component, at least in the marine environment.”
Obviously, sensor fusion is key: therefore some elements are being trialled onboard Stena Line’s crossing between Sweden and Denmark: “You can see the sensors working together already,” says Päivi Haikkola, “I’d say these can catch as much as a human... maybe more.”
It has to be said, if it’s working as well as proposed, trust could be boosted by the simple expedient of a public viewing platform that allows us all to ‘see’ through the ferry’s sensors.
Although this development is ‘steady as she goes’, there’s been a stampede of closely-related concept ideas. Rolls-Royce has its own supply vessel drone concept (which superficially appears similar to Hrönn), that the company calculates could save a whopping great 21% per annum operating costs over a standard PSV. And QinetiQ has an 80m firefighting and rescue standby design which would hang around oil and gas platforms: usefully, this could get right into the heat zone far more effectively than a manned vessel.
There are also other, almost sci-fi applications that could be quite a lot closer to realisation than seemed possible just two years ago: Rolls-Royce has an idea for a ROV paired with a small, 20m autonomous surface vessel controlled from shore – after all, ROVs are remotely operated anyway. This would mean no crew at sea, higher utilization of skilled operators and again, no hotel functions so lower operating costs. There’s also the potential for lifting the whole caboodle onboard an OCV.
Further, it seems that Singapore’s Maritime Authority is beginning to be interested in a “master and slave” concept for tugs. It’s a possibility that has come a step closer with the Danish remote operation trial of Hermod.
Scandinavia, particularly Norway and Finland, is pitching itself as an incubator for autonomous technology. In some ways the two countries have similar drivers: both have a significant intellectual property investment – and both have suffered from having the primary market pulled from under their feet: Norway’s was oil and gas; Finland’s was Nokia.
And as the old truism goes, if you have a marching army, you have to find a fight: Päivi Haikkola explains: “When a lot of the industry got around the table to discuss where digitalisation could lead, there was just one answer - Autonomy.”
There are some differences in approach between the two countries, although everyone involved underscores the enormous amount of international collaboration. Finland’s ‘ecosystem’ has a number of big name companies supporting the Autonomous Shipping Alliance, including Wartsila, ABB, Meyer Turku and MacGregor and of course Rolls Royce (which has recently opened a Finnish Centre of Excellence covering the field) all of which have collaborated closely with agencies such as DIMMEC.
Norway has intellectual heavyweights including the Ocean Space Centre, the Norwegian University of Science and Technology (NTNU), plus the Marine Technology Research Institute (MARINTEK). However at present Norway’s cluster appears to be pulled along –fast – by just a couple of commercial names, heavily supported by the Norwegian Coastal Authority. Of these, Kongsberg has the largest profile, followed by AI expert Maritime Robotics.
Certainly Norway gained the world’s first test bed which opened in Trondheimsfjord last year: this has seen a 1:20 model of DNV GL’s Re:Volt concept in its quiet waters and the Kongsberg subsidiary Seatex start its initial testing. There’s also a practical application: NTNU is also working on an autonomous, on-demand, push button ferry service to get up to a dozen passengers and bicycles across Trondheim’s harbour.
Cheaper than a bridge, this ferry will have induction-charged batteries, GPS navigation and an anti-collision system. A pilot study is planned for this year, and it's expected to start real operations at the end of next year or the beginning of 2019.
However, there’s also now a second Norwegian test bed at Grenland which is much more heavily trafficked. According to Vartdal “considering the aggressive timelines of these projects, it’s necessary to trial these vessels in busy areas where you can test out the systems properly”.
On the other hand, Finland’s test bed area in the Gulf of Bothnia is an area noted for difficult ice and wind conditions which, Ms Haikkola points out, is just what you want if you are stress testing an autonomous craft, especially since many of the first vessels will operate around Scandinavia.
The Finnish test area also borders onto a ‘Smart Fairways’ project, which will have buoys relaying VTS information directly to the visiting ships. According to Ms Haikkola, this might eventually become two-way, system-to-system data transfer if both projects fulfil their potential.
The Scandinavian hotbeds won’t be alone for long.
As this goes to press, a testbed is being launched in the UK along the Solent, which, says Mike Woods of BAE Systems, will embrace both naval and commercial elements, giving businesses “as close to a plug and play” trial facility as possible.
A further hub promises to emerge near Singapore; the port authority has recently been given the go-ahead – and space – for a floating lab which will look at novel fuels and the potential tie-in for autonomous concepts.
Autonomy will, says Bjorn Johan Vartdal, “be a game changer” for the class societies. “Traditionally we have been involved in approving ship technology, but the operational part, we’ve left that to the crew. But in autonomous vessels, the operation is embedded in the technology, it’s the software that decides how it will manoeuvre, so we will have to approve the safety of the whole operation.”
“This is, of course, really challenging,” says Vartdal, and adds that they won’t come up with ideas set in stone: “It will have to be a case of continuous development.”
The thread is taken up by Lloyd’s Register’s Anderson Chaplow. The class society has come up with the Unmanned Marine Systems Code, which he explains has pre-empted its very first ‘goal-based’ code as the technology is wide open and, he explains, there’s a need to “support innovation alternatives” and give them room to evolve rapidly. However, it should be pointed out it still incorporates and builds on existing, prescriptive standards where appropriate.
Finally, some might be cynical about take up, given the realities of the lacklustre shipping market. But perversely, this might actually help push things along – after all, autoships are being pasted with a big sign saying “lower-cost build and operational savings here”.
Further, although developments are at present being kept within flag state boundaries, when the IMO rules are untangled autonomous ships promise to give the industry an alternative answer to the ‘economies of scale’ mantra that has both forced ship sizes up and up and induced unwieldy, global alliances to utilise onboard space.
By Stevie Knight