With old Pacers still dragging their heels along many rail routes across the country, talks of automation can feel like a distant pipe dream. But are they, really? Dr Amine Arezki, Thales’ head of strategy for autonomous trains, talks to Quadrant Transport about the future of travel
When it comes to autonomous technology, you would be forgiven for thinking that self-driving cars are leading the way. Look up ‘autonomous’ online and the top results are consistently related to cars: a quick experiment shows me that Ford is building a new autonomous car factory in Michigan, and that Lyft’s CFO has told investors that the California-based Uber competitor will soon be phasing in self-driving cars on simple routes. Even tech tycoon Elon Musk has promised a truly self-driving Tesla by 2020. Put simply, we’re told autonomous cars are all the rage.
But that could not be further from the truth. In fact, the rail industry could even welcome in autonomous vehicles before these ever safely and reliably hit the highways.
The age of the digital train
The Fourth Industrial Revolution is coming thick and fast across all sectors, and rail is no exception: the government has made clear its intention to prioritise digital innovation as a way to boost capacity without necessarily having to buy more trains or build brand-new infrastructure. Investment in integrated communication systems, modern signalling, and predictive maintenance are just some of the exciting developments taking place, stretching all the way to more advanced resources such as artificial intelligence (AI), Internet of Things (IoT), and deep learning.
As part of these new developments, there has been a global stride towards enhancing automation in rail networks. For all intents and purposes, it’s the future.
The rail industry could even welcome in autonomous vehicles before CAVs ever safely and reliably hit the highways
But not every automation is made the same. Services can be categorised into four levels, with the first being the lowest – consisting of Automation Train Protection (ATP) and a driver at the helm – and the fourth being the highest, reaching a peak of Unattended Train Operation (UTO). Within that range, services can run under anything from Automatic Train Operation (ATO) – essentially meaning partial or complete automatic train piloting and driverless functionalities, but with an attendant present to close doors – or completely driverless, but still supervised by someone. For those interested in the more specific differences between each grade, the International Association of Public Transport has a handy online guide with all the facts and figures.
Automatic train technology in itself is nothing new. Take London, for example: on the Underground, several lines run with ATO, and others are in the process of catching up; all lines of the DLR run as Grade of Automation 3 (GoA3) ever since the network opened; and the Thameslink core network recently made UK history with the first-ever commercial deployment of ATO over ETCS Level 2 on a mainline rail.
The safety factor
So does this all mean the rail industry is already blazing a trail on autonomous travel technology? Unfortunately, not yet. These services may be automatic, but they are not yet fully autonomous. Metros are a good example of this: they may use communications-based train control (CBTC) technology, which does mean computers are responsible for governing the trains, but they are not capable of making decisions by themselves. External systems are needed to send instructions according to the data received from the train.
According to Thales, an autonomous train must be capable of three things: first, you must know where it is in the network; second, the vehicle must be able to identify obstacles at a distance; and third, it must be able to make decisions about whether to travel or not. In layman terms, it must be a fully-functioning robot with a mind of its own, connected to the minds of many others like it across the network, all tied together by a similarly fully-functioning traffic management system.
It is perhaps this straightforward description of the vehicle as a robot with no driver at the wheel that fuels public fears over automation. Once again using cars as an example – after all, the automotive industry is much better at advertising itself in the media – it is not uncommon to come across scaremongering headlines calling into question the validity of a machine which, when faced with a life-or-death decision, is essentially programmed to choose who to kill. And with incidents like Uber’s autonomous Volvo XC90 killing a pedestrian in Arizona last year and Tesla’s separate but equally fatal crashes in Florida and California, it doesn’t take much to see that public faith in autonomous technology isn’t exactly thriving.
We’re not driving only one or two people like cars do – we’re driving thousands
The rail industry is keen to dispel that danger myth. It’s as Dr Amine Arezki, Thales’ head of strategy for autonomous trains, told me: when we talk about trains, we talk about safety-critical technology – namely, Safety Integrity Levels (SILs). That forms the basis of the entire network.
“Without this, we don’t have railways, we don’t have trains, we don’t have people travelling on trains,” he explained. “And why do we take this so seriously? Because we’re not driving only one or two people like cars do – we’re driving thousands.”
The selling point is, of course, improved punctuality, reliability, cost-effectiveness, and availability. But these goals will only be achieved if safety is part and parcel of the overall vision. “What we are working on is to avoid this question between safety and availability, and to have both: a system which is available all the time – so whenever we want to travel, we can do so – and which is always safe,” explained Amine.
If you were around this time last year, when the arguably disastrous May timetables were rolled out, you’ll understand the benefits of any technology that can reduce disruption across the network and boost the reliability of services. If anything, investing in autonomous travel in rail is even more important than in the automotive industry, at least from a utilitarian point of view: it would bring the greatest amount of good to the greatest amount of people.
“We’re not investing in this just because the car industry is doing autonomous cars,” Dr Arezki agreed. “We’re doing it because we see the benefits and the huge potential behind it – and the technology today is ready to move to that direction.
“When we talk about trains, we talk about increasing capacity, improving punctuality – this is what the operators are looking for. Most importantly, this is what the passengers are looking for at the end of the day; they want to start on time and arrive on time, as expected.”
Amine himself argues that the rail network will benefit from this innovation before cars do. From his perspective, rail is less complex in terms of scenarios – the multitude of unexpected events that could happen during any given journey – and in terms of responsibility. With regular vehicles, there is an ongoing debate around who is responsible if anything goes wrong: is it the supplier who provided the technology, or the driver who was meant to supervise it and take control if needed?
“It’s very foggy in terms of responsibility. But in the train industry, we never talk about that,” he explained. “It’s always the responsibility of the technology. If the driver is responsible for driving the train, and there’s no damages in the system, then he’s responsible for sure. But if the train is fully automatic or autonomous, the supplier is responsible.”
The fearmongering factor also comes into play in the car vs. train debate. Regular news readers are always shown news stories about self-driving cars being programmed to decide between, say, crushing a child who is crossing the road or turning the other way and smashing into a truck carrying hazardous materials.
“When people hear these kinds of scenarios, they think it’s never going to be 100% safe. There always will be decisions made by a machine, which are going to be either accepted by people or not. Everyone will have their own opinion,” said Amine. “In the rail industry, we’re not thinking at all like that. We’re talking about avoiding any incidents. Of course, the number of scenarios is quite limited on railways: we talk about one dimension, and not two or three dimensions. This is much simpler. At the same time, we have traffic management orchestrating all the trains, knowing their position, and solving any conflicts, with an overview of the whole network.
“In the car industry, they are talking only about the car itself. But even if you have much more advanced sensors around the vehicle, it’s not going to know what’s happening in other roads, if there is an incident, or if it needs to anticipate anything. There are two different philosophies there, and I think the car industry can learn a lot from that orchestration element.”
Driving in the open space
It’s not all plain sailing in the rail industry, though, especially when it comes to mainline networks. Automated metros have a much easier time because trains operate in a closed, known environment – there is no risk of adverse weather or people or animals crossing over the track, for example. If there ever is an obstacle in front of the train, it’ll hit it; there is no intelligence or interconnected sensors to detect these events, because they are very unlikely to happen.
The mainline, on the other hand, is an open environment where unexpected scenarios can always spring up: animals, cars, trees falling, heavy snowfall, fog and poor visibility, collisions, you name it. What’s more, due to higher travelling speeds, sensors would need to be able to detect issues from at least one kilometre away, if not more – requiring a much more advanced type of technology compared to the automotive industry, for example, where the detection distance tends to be closer to 100 metres.
“When you have obstacles, the train will be capable of detecting them, analysing the danger, and then taking the right decision – either to stop, to continue, or to send a report to other trains to avoid delays,” Dr Arezki said. “In terms of anticipation, this is very important – it keeps everything on track.” (No pun intended, I’m sure.)
These things are solved in metros, and the building blocks are being used as a contribution for the mainline
That’s not to say the mainline can’t learn from the successes of the metro, though. Automated metros, whose deployment has so far been successful across the globe, still have to consider shared scenarios: how to bring the train from a depot to a platform, how to start the engine, how to pass rigorous safety tests, and so on.
“These things are solved in metros, and the building blocks are being used as a contribution for the mainline,” explained the automation lead at Thales, which itself is leading on self-driving metro technology worldwide. “Especially in terms of the way of driving: the interface with the train, acceleration, acceleration curves, deceleration curves, etc. This is very important to driving smoothly and ensuring passenger comfort.”
Turnkey solutions are just around the corner
Much like mainline travel, light rail is a different beast compared to metros. Cutting through the heart of towns and cities, tramways face extremely intricate scenarios more closely comparable to cars, and which can prove even more difficult to automate than heavy rail. At present, there are no fully autonomous trams – just advisory systems.
But this could soon change. At InnoTrans last year, Thales Deutschland and the managing director of AVG (Albtal-Verkehrs-Gesellschaft) signed an agreement to, working together, develop an electric city railway that drives autonomously. The initial phase of the project included a demonstrator in December at a German depot, with an old tram-train used as a proof of concept for driving in GoA3 and GoA4. The train will then be taken outside for an open-environment drive later this year in what promises to be a major milestone in the two-year scheme.
Albeit based in Germany, the good news is that the final goal is to develop a turnkey solution that can be replicated elsewhere. “It’s going to be a standard solution rather than specific to that station, that depot, or that type of train,” said Dr Arezki. “We are going to reproduce this in many other countries and many other projects; that’s our target.”
Also in Germany, Deutsche Bahn has made clear its intentions to introduce driverless trains by as early as 2021. Pilots at a test field in Saxony were already underway as far back as 2016. In Japan, East Japan Railway has begun testing ATO on Tokyo’s 34km-long Yamanote Line – chosen because it does not share tracks with other lines – and hopes to roll out the driverless operation at an undisclosed date in order to protect the industry against future staff shortages and to cut losses in rural areas.
In France, Thales is working with SNCF and other partners – including sector leaders Alstom and Bosch – on a five-year project to test mainline automation. Dr Arezki’s French counterparts are responsible for the intelligent onboard system designed to make key decisions, as well as satellite positioning, cyber security, communications, and digital maps. This is all part of an ambition to roll out heavy-rail automation by 2023 with two prototypes – one for passengers and another for freight – all coming in at €57m. According to SNCF’s deputy head of rail technologies, Pierre Izard, existing trains will then get equipped with autonomous technology on a larger scale from 2025 – but, much like in Germany, the goal is to develop a generic solution which can also be implemented everywhere else.
That is perhaps the beauty of autonomous rail travel: vehicles don’t have to be made from scratch, and technology can just be layered on top of what organisations already have in place. In France, for example, they are layering tech over ETCS, which would then be UK compatible. “We can add this layer easily on top, with already existing interfaces, so you can just upgrade your system to make it autonomous when it’s needed,” Amine told me. “The deployment will be much, much easier.”
When it comes to ETCS development, the UK is leading the way. As part of a project between Network Rail and ProRail, Thales UK – in partnership with its Holland and Germany divisions – successfully demonstrated the benefits of ETCS Hybrid Level 3 technology at Hitchin last year in what Amine described as a “really, really good base for implementing autonomous trains” in the future.
Thales’ work with Network Rail to develop traffic management systems also bodes well for the future of autonomous travel in the UK. The company was awarded a £28m contract back in 2014 to deploy its ARAMIS TMS solution for two regional operating centres in Romford and Cardiff. At the time of the contract award, ARAMIS was already in place in 11 countries, controlling more than 60,000 trains; this has grown since then, with Norway recently choosing the company to provide its next-generation system as part of a large-scale modernisation programme.
Traffic management is going to have a huge role, especially when we talk about autonomous trains
According to Amine, we don’t talk about traffic management systems – and their role in automation – quite enough. It is an extremely advanced digital platform that uses AI to make decisions designed to optimise traffic flow, but it’s often put in the backburner because conversations around automation can often be train-centric. That is, centred around the physical trackside elements that will be replaced with more sophisticated digital solutions inside the train itself, and forgetting about the fact that this information all comes from a central system.
“Traffic management is going to have a huge role, especially when we talk about autonomous trains,” he told me. “One of the main targets is to increase capacity, which means denser traffic. And denser traffic means that if there’s any obstacle disturbing the traffic, there might be a lot of impact in terms of schedule. If traffic management isn’t taking this into account and having an overview of the system, orchestrating trains, and finding new routes, it could be problematic. If you don’t see globally what’s happening around you, it’s going to be difficult to solve conflicts.”
The role of track and train integration in all of this
So what does this all mean for the UK? Realistically, it’s unlikely that autonomous technology will be in place here while prototypes are under development elsewhere. But this doesn’t mean operators and Network Rail can’t invest in their own prototypes, running projects in parallel so that they can learn from each other in symbiosis. “There are always overlaps between projects, but that is OK – at least you get to test the systems with the different customers and then get the right feedback to develop something suitable for all markets,” said Amine.
One thing that can help speed up the process in the UK is greater cooperation between operators and infrastructure managers, a relatively recent phenomenon best epitomised by partnerships such as Transport for Wales and ScotRail Alliance.
“We always talk about cost and price, but we’re not focusing on the value that is going to be brought by this kind of technology. The value is really huge; you can create completely new kinds of services around autonomous trains. But today, this solution is going to be split between either operators or infrastructure managers,” Amine lamented. “Therefore, I think the two need to talk between each other in order to make the right split, or to cooperate and understand the benefits for both parties.”
We always talk about cost and price, but we’re not focusing on the value
Cooperation could also help motivate partners to take the leap and experiment with automation first-hand. “It’s not that they’re not interested; it’s just that they want to see something working first, and then we can help them to define projects and show them the benefits,” said Dr Arezki. “This is why we try to show with some proofs of concept what we can do, what the possibilities of today’s technology are, and what kind of certifications we can do.”
Closer working between TOCs and Network Rail is a key part of what Amine describes as “a clear, open-book discussion” about value – both in savings and benefits. This also means having frank discussions with tech suppliers: it’s not necessarily about procuring generic solutions and applying them to the network, but rather about looking at operational use cases and existing problems that could be solved by bespoke technology.
“This is why we’re interested in working with customers and end users,” he added. “Because we need to make sure we’re understanding the problem correctly, starting from the user or from the operator, and then going back to the technology.”
And, of course, the elephant in the room: jobs
As well as debates around safety fears, the mere invocation of the word ‘autonomous’ can fire up emotional conversations about the potential job losses that theoretically come hand-in-hand with technological innovation. In the UK, even preliminary discussions around driver-only operation – where the train driver is responsible for dispatch and opening and closing doors – has resulted in heated clashes between rail unions and TOC bosses in separate but interconnected rows often running for months at a time. The Office of Rail and Road has already ruled that driver-controlled operation is safe, but this has not proven sufficient in quelling redundancy fears. We can therefore expect that, even if companies can guarantee a 100% safe autonomous rail network, a dispute around jobs will linger on.
Research on whether automation will lead to job cuts varies considerably. Some argue that robots are coming for all of our jobs – and to be fair, a 2017 study by McKinsey Global Institute suggested that between 75 million to 375 million workers could be displaced by 2030, although this only represents three to 14% of the global workforce. Others claim that only certain industries will take a hit, while others will learn to adapt by changing the skills required of its employees – much like what happened during the Industrial Revolution, sans the terrible working conditions. Indeed, examples of this can already be observed in pockets of the rail industry, with advances such as digitisation already pushing companies to focus on digital skills when training the workforce of tomorrow.
The idea is to help the driver to have a better quality of life, and with better work conditions
The latter school of thought is much closer to Amine’s take on the future of rail in a automated world. In his opinion, autonomous technology will help drivers lead better lives; rather than making them redundant, their job profiles would simply change to fit new specs.
“The idea is to help the driver to have a better quality of life, and with better work conditions. That will be much more efficient. This needs to be, of course, discussed with the unions to explain the different roles that the drivers of today will have – what they will still do, what they won’t do anymore, and what the benefits will be to them,” he explained.
“If the capacity in the rail network is going to increase, autonomy is going to help the operators too, because the trains can drive by themselves and the experts – the drivers of today – are mainly going to supervise the trains either remotely or in the cabin. Essentially, instead of sitting on a train with all the conditions you may have in the cabin, the driver will sit comfortably in an office. There are many positive stories on the improvement of life and work of drivers in the future, once autonomy is available. I see it as a very positive thing – but it needs to be communicated and discussed openly.”
If you have all safety controls in check and union agreement locked down, it seems there is nothing that an autonomous rail industry of the future can’t do. Maybe it’s time we, too, invest in media exposure showing off autonomous high-speed trains travelling up and down the permanent way, so that the wider public can see for themselves what the industry is capable of.