Innotrans is off. The next edition will take place in 2020, in Berlin. At this date, will we talk more about the autonomous train that is announced everywhere? We are going to peruse the latest innovations on this theme.
Automatic metros have been around for a long time. There are some examples in London with the Jubilee Line and the Dockland Railway, as well as Paris on line 14 or Lille with VAL. By the end of 2013, there were 48 fully automated public metro systems in use in 32 countries, according UITP. The major innovation is the autonomous car. Why ?
Because the examples of automatic metros show that they work in fact in closed loops.‘All these systems have no obstacle detection and assume a free line’ explains Burkhard Stadlmann, a professor of at the University of Applied Sciences Upper Austria. When the trains are the same length, run all day long according to repetitive criteria and stop at all stations that have the same length of perron, then the automation becomes relatively “easy”. But none of this exists with the concept of autonomous cars. Indeed, the ability to drive autonomously in heterogeneous environments without GPS, pattern identification (e.g. road following), or artificial landmarks is key to field robotics. To address this challenge, it is necessary to use technological building blocks in the form of GPS, radar/LiDAR, infrared and ultrasonic sensors, cameras, inertial systems and more. Automation software must be developed for autonomous vehicle process flows.
We are running here to more complexity field than an ‘simple’ automatic metro. Autonomy means indeed instantly recognize its environment, which changes every meter, that the computer must translate to take a decision immediately. So, thousands of data must be captured, assembled and decoded in a few seconds. The main challenge is to recognize if there is an obstacle in front of you, and what decision you must take. When there is a vehicle in front of you, either it rolls at the same speed as you, and just follow it, or it brakes, and you also immediately have to break to avoid a collision. Currently, the only signal available that shows a vehicle braking in front of you are the rear red headlights, and it’s only your eyes that “grab the message”. With the autonomous vehicle, the rear red headlights are of no use. So something else was needed. This “thing”, that’s a permanent calculation of distance and approach between the vehicles. It is therefore necessary to answer in thousandths of a second so that, as soon as the vehicle detected in front of you brakes, immediately your autonomous vehicle brakes too. Unfortunately, this is not possible with trains! Why ?
Because trains have a much higher distance between them than in road vehicle flows. No LIDAR or sensors can measure the train that is far ahead of you. Currently, trains run through blocks (2 km or more), in which they are alone. Once a block is free, the next train can enter. Each block is protected by a lineside signal. This system is still in effect, even with visual signaling is available in the driver’s cab, as on high-speed trains or in ETCS level 2: one train per block. As long as the block in front of you is busy, you do not enter, the signal is red and the speedometer shows you a speed of “zero”. You are stopped until the ‘freedom’ of the block. The information that tells you that the block is free is delivered by a track circuit in the rails. When it no longer detects metal masses in the rails, it means that there is no train in the block, with a certainty of almost 100%. So you can enter the next block, its signal is green or yellow, and the ETCS level 2 speedometer tells you with which speed you can enter.
At most the blocks have a small length, at most you can send trains on a line, generally between 8 and 12 trains per hour and per direction. On small local lines, some blocks have a very large length, which means that the flow is much lower, for example 2 to 3 trains per hour.
What conclusions can be drawn from this ?
The first element is that train detection means a lot of on-line equipment and cables to lay along the track. This requires maintenance and major purchases. Everyone knows that electrical and electronic equipment are very expensive, even when you buy a high quantity. The electronics and electricity sector is very lucrative. As a result, a railway line is de facto very expensive, and even more so if it is electrified. Of course, these investments are designed for the long term.
The second element is that the strict obedience to the signaling is fully supported by the human factor, even in case of ETCS level 2. Of course, the current equipment can detect some faults. If you do not brake within 3-4 seconds when your ETCS speedometer requires it, the computer will engage the emergency brake until the train stops. But it’s not enough. And often it’s too late. « Train drivers have little room for decision-making, » says Jürgen Siegmann, professor of rail transport and railway operations at the Technical University of Berlin.
The third element is obviously the factor ‘cost of railway workers’, which is combined with operational factors. The railroad is known for its large labours needs, while for other transport, one man is sometimes enough. We are thinking of maneuvers in industrial installations or in marshalling yards. In some rail public services, wages eat more than half of the turnover. Financial aspects should not be underestimated. According to some experts, the cost coverage of an autonomous regional train could increase by 60%, which obviously interests the transport authorities and the State. This is unfortunately not verifiable at the moment.
The conclusion is that the railways are looking for a reduction in operational costs, through new signaling with less equipment, and reduction of the human factor, where is possible. In Germany, an expert report at the end of September 2018 showed that to absorb the growth and the modal-shift expected of in the future by a greater number of trains on the network, it would be necessary to multiply the tracks on the congested railway lines. This solution would cost almost twice as much as digitization, taken in all its components, not only by the autonomous train. One point on which they will not procrastinate: the security. Let’s take a look at what’s going on with the autonomous train.
Rio Tinto operates about 200 locomotives on over 1,700km of track in the Pilbara, in Australia, allowing it to transport ore from 16 mines to four port terminals. On July 10, a train, consisting of three locomotives and described by Rio Tinto as ‘the world’s largest robot’, travelled over 280km from the company’s mining operations in Tom Price to the port of Cape Lambert. The train was remotely monitored by Rio Tinto’s Operations Centre in Perth more than 1,500km away. The locomotives are equipped with AutoHaul software and are fitted with on-board cameras for monitoring from the centre. Of course, this train only ran alone on a single-track line in a desert region, where the probability of an obstacle was certainly low, despite the presence of some crossing-levels.
‘We are working closely with drivers during this transition period as we prepare our employees for new ways of working as a result of automation’ explains Ivan Vella, Rio Tinto Iron Ore managing director for Rail, Port and Core Services. According another director of Rio Tinto ‘AutoHaul has shown in trials that the autonomous trains delivered the product to the port nearly 20% faster than a manned train.’ The proof that all of this is taken seriously is that the Office of the National Rail Safety Regulator (ONRSR), in Australia, has fully approved the technology which underpins the entire system, AutoHaul.
ProRail and Rotterdam Rail Feeding (RRF)
The Dutch railway infrastructure manager ProRail announced on TEN-T days in Rotterdam that it wanted to create the conditions to test the automated operation of freight trains on the Betuwe Line connection, a railway line reserved for freight traffic between Rotterdam and Germany, which operates only with ETCS level 2 (cab signaling).
As part of the ERTMS Corridor A, this line is fully equipped with ETCS 2 (SRS 2.3.0), so without lineside signals, like high speed lines. This is a minimal requirement for an automated operation, even though originally, ETCS was specified for manual mode operation. Alstom signed an agreement with ProRail and Rotterdam Rail Feeding (RRF) to carry out the tests. It is planned that an RRF locomotive will run approximately 100 km from the port of Rotterdam to the CUP Valburg freight terminal using ATO on the sections of the route which Alstom has previously equipped with ETCS Levels 1 and 2. Here, the train is controlled by computer, but the driver is always on board, operates the doors, starts the train and can take control if necessary.
Rio Tinto and Pro Rail tests are only for freight trains only. What about passenger trains? As in the case of cars, there is also an international classification of levels of automation in public transport, or “automation levels” (GoA). Four levels of automation are available, and the tests are initially oriented on levels 2 and 3. For level 2, the system supports the journey from start to finish, but the driver is still responsible for the operation of the doors and the starting of the vehicle. For level 3, the train operates without a driver, but an onboard attendant always checks the doors and can move the train via an emergency system if necessary.
The European rail champion is currently transforming a section of several kilometers near Chemnitz into an trial site for autonomous train. DB Regio, a subsidiary of the DB, has upgraded a self-propelled train, with cameras and sensors in its Chemnitz workshop. The system must detect obstacles and stop the train if problems. By autumn 2018, this train should be operated in part automatically. Only the approval of the German federal railway authority (EBA) is still lacking.
‘Autonomous driving is complex. The rail system, where fast and slow passenger trains and freight trains run and are mixed, is more difficult than a metro – but it is possible. The first pilot projects are underway, and we have set up a test area on the Erzgebirgsbahn. Fully automatic rail driving is the next big step in development and a matter of time.’ explained former CEO Rüdiger Grube in 2016. Since then, the DB has signed an agreement with SNCF concerning the autonomous train.
In 2015, however, after a bad year with strikes, the DB stressed that ‘In our safety philosophy, train drivers remain a strong pillar.’ Customer surveys have shown that passengers do not want to abandon train drivers. Three years later, is it still the case?
SBB (public railways in Switzerland)
‘What might surprise you is that we are also a big software company,’ said Erik Nygren, a business analysis and AI researcher at the company. Switzerland also studies on the autonomous train. On the night of 5 December 2017, SBB tested for the first time an autonomous train on the Bern-Olten line. This train has braked and accelerated independently from any action of the driver. The driver only controls the processes and function of the systems, just like pilots in an airplane cockpit. The constructor Stadler Rail also sees it as a huge advantage: “This trip was a first and it is proof that you must continue to compete in the highly competitive rail market,” explains Peter Spuhler, CEO of Stadler. In other words, it is a question of guarding against competition (Chinese?) and, for Switzerland, the autonomous train is part of a broader strategy for exporting technologies of the country. The Confederation is in the top 5 countries relying on artificial intelligence, and this also explains this policy.
Contrary to Germany’s caution, SBB’s plans for the future “SmartRail 4.0” strategy show that the partial automation of trains would be planned for “the coming years” and that operation of fully automated trains would start in the period after 2025. Optimistic? We’ll see.
Austria is also at the forefront of progress. With its Swiss and German neighbors, we can see that it’s the whole German-Alpine region is embarking to the digital rail technology. The line is situated between Oberwart in the Austrian state of Burgenland and Friedberg in Stiermarken. It passes eight stations, twelve railway crossings and a 524-meter long tunnel. This variety of environmental factors make it possible to test different situations that can occur during a train journey, in the context of a project of autonomous train made by ÖBB, the Federal railways.
A old Emu from the Traunseebahn operated by Stern & Hafferl Verkehrsgesellschaft was rebuilt and equipped with various sensors (laser scanner, mono and stereo video, radar, infrared and ultrasound as well as location sensor technology in conjunction with algorithms) for obstacle detection and an automatic control system. With the help of the developed software system, the railcar can drive completely autonomously, can control breakpoints and brake before many obstacles. The system is based on a digital train protection and control solution that Siemens Austria has developed together with the FH Upper Austria (University of Wels, Research Group Rail Automation).
The current project called “autoBAHN2020” aims at a demonstrator system and associated simulation environment that can serve as a basis for future concrete product and approved developments for the public transport in order to facilitate the autonomous trains on secondary railway lines. Siemens Austria is involved in the project in questions of system approval as well as driving and braking control..
In France ?
Shortly before Innotrans 2018, France’s national railway operator SNCF has announced plans to introduce prototypes of driverless mainline trains for passengers and freight by 2023. SNCF will be partnering up with rolling stock specialists Alstom and Bombardier who will be heading up consortia for freight and passenger traffic, respectively. According the CEO Guillaume Pepy : ‘With autonomous trains, all the trains will run in a harmonised way and at the same speed. The train system will become more fluid.’ It is difficult to know where this statement comes from, which remains to this day unverified and unverifiable. One can understand indirectly that the second European railway carrier does not want to be left behind by its neighbors. The French rail operator said it was talking to German operator Deutsche Bahn about promoting a European standard for driverless trains.
Let’s end with a real test in real conditions. Govia Thameslink Railway, which owns the Thameslink franchise, a north-south line running through London, started a first test in March 2018. After almost 18 months of testing, the first commuter train in automatic operation was Monday’s 9.46am Thameslink service from Peterborough to Horsham. Shortly after 11.08am, the driver, Howard Weir, pressed the yellow button in the cab that allowed the train’s computer to do the driving between St Pancras and Blackfriars.
Gerry McFadden, directeur technique de la société mère de Thameslink, rassure tout le monde : ‘Nous aurons toujours besoin d’un chauffeur dans la cabine, mais cette technologie nous permet de faire circuler plus de trains, plus souvent que nous ne pourrions le faire manuellement. Pour les voyageurs, les trajets n’auront jamais été aussi fluide.’ Avec 24 trains par heure en heure de pointe, Thameslink n’a aucun intérêt à se tromper. Nous sommes à Londres, sur l’un des réseaux ferroviaires les plus encombrés du monde. Et dernièrement, de nombreux couacs sont venus perturber le quotidien des navetteurs de la capitale britannique, pour d’autres raisons. L’heure n’est donc plus aux tergiversations : il faut que cela fonctionne !
Gerry McFadden, Technical Director of Thameslink parent company, reassures everyone: « We’ll always need a driver in the cab, but this technology allows us to run more trains, more frequently than we could by driving the trains manually. For passengers, the trip will be as smooth as ever. » With 24 trains per hour in peak, Thameslink has no interest in making a mistake. We are in London, on one of the most congested rail networks in the world. And lately, many problems have come to disrupt the daily lives of commuters in the British capital, for other reasons. The time for procrastination is now over : it must work!
We are only at the beginning. Autonomous trains concern two separate branches of the railways: freight on the one hand, and passengers on the other. We are pretty sure that under certain conditions, like in Australia, the freight train will benefit first from the autonomous locomotive. With these recent developments, one might wonder if it is easier to bring autonomous passenger trains to the mainstream before self-driving cars could make it to the traffic. But that’s not going to happen anytime soon. Achieving full automation would require advanced image processing technology relaying information at high speeds to the control units at all times. These systems must also be constantly maintained by highly trained personnel, adding more costs to the implementation.
Those who say that would perhaps do well to reread Schumpeter ….
- Autonomous Cars: The most disruptive innovation ever
- Why Don’t We Have Driverless Trains Yet?
- Zukunft Mobilität (site web en allemand)
- AI Think I Can: Why GPUs Could Be the Engine Leading Us to Autonomous Trains
- Rio Tinto setzt zum Eisenerztransport in Australien auf fahrerlose Güterzüge
- Rio Tinto achieves first autonomous train iron ore delivery
- Annaberg-Buchholz and TU Chemnitz Intend to Establish Research Campus for Automated Train Operation
- Autonome Züge: Deutsche Bahn startet Pilotprojekt
- Wenn der Computer den Zug steuert
- Roboterzüge: Verlieren SBB-Lokführer bald den Job?
- Zug ohne Lokführer ist keine Vision
- Alstom to perform automatic train operation test drive with Prorail and RRF
- Autonome Züge kommen! Aus der Traum vom Lokführer?
- Test track for autonomous trains in Austria
- Fully autonomous trains are better suited for moving ores than people
- Höhere Effizienz ohne Lokführer?
- Regionalbahnen auf dem Weg in die Zukunft
- Digitalisierung im öffentlichen Verkehr