A popular introductory slide in presentations about railway signalling at industry conferences is a photo of an empty track held up alongside a congested motorway. The caption says it is two transport systems at full capacity.

The shot of the track emphasises the limitations of the railway. Fixed infrastructure with often long fixed block sections and the extended braking distances of trains demand high levels of safety, restricting the number of trains able to operate.

The principle of controlling entry of a single train into a block section has been in place since the first token-based signalling systems and continued with lineside optical signalling. Generations of signalling engineers have attempted to crack this conundrum. However, it is only recently that the technology to do so has emerged.

Communications-based train control (CBTC) for metros was among the first innovations to deploy a moving block principle, whereby the protected block moves for a specific distance behind the train, increasing the flexibility of operation and providing the opportunity to cut distances between trains.

The technology also facilitates automatic and driverless operation. However, it was not the first to offer such functionality. London Underground’s Victoria Line, which opened in 1968, pioneered Automatic Train Control where the train operates automatically, but a driver is present in a supervisory capacity. The Port Kobe Line in Japan, which opened in 1981, and the VAL system deployed initially in Lille in 1983, were the first entirely driverless urban transport systems.

CBTC made its debut on the driverless Vancouver SkyTrain in 1986, and subsequently on the Detroit peoplemover and London’s Docklands Light Railway (DLR) in 1987. These early applications were based on inducted loop technology developed by Alcatel-SEL, now Thales, and introduced as an alternative to track circuit-based communication. Subsequent radio-based systems provided by other suppliers in the early 2000s increased reliability and led to a surge in the use of CBTC; all new metro lines and networks now tend to be fitted with CBTC and deploy a degree of automation, with the majority driverless at Grade of Automation 4 (GoA 4).

CBTC is an optimal application for metros because these tend to have little or no interfaces with other lines, reducing the complexity of operation. Mainline networks are far more complicated with many interfaces and varying speeds of operation. However, work is progressing to bring the benefits of operation pioneered with CBTC to the mainline.

The transition to in-cab rather than lineside signalling began with various national Automatic Train Protection (ATP) systems based on electronic interlockings and developed in response to several accidents during the 1960s and 70s.

The dawn of high-speed in Japan in the 1960s and Europe in the 1980s led to a steady rollout of cab-based signalling. The development of the European Train Control System (ETCS), the signalling component of the European Rail Traffic Management System (ERTMS), in the mid-1990s was intended to supersede national in-cab systems by providing an interoperable signalling system that permits seamless cross-border operation.

However, with projects proving expensive and time consuming, and loyalty to national systems strong, progress has been slow, much to the frustration of many. Indeed, more kilometres of railway are fitted with ETCS outside of than inside Europe.

London’s Thameslink project was the first commercial deployment of ATO over ETCS at GoA 2 in 2018.

Nevertheless, this gap is steadily closing and the signs at the beginning of the 2020s are more positive.

Improvements to software and the steady switch from relay-based to direct drive interlockings during the 2000s was followed by the expansion of ETCS Level 2 during the 2010s to match almost universal GSM-R telecoms coverage in Europe.

Adoption on key corridors has been accelerated by European Union (EU) funding, and a number of countries now have national rollout plans. This began with smaller networks in Denmark, Belgium, Luxembourg and Norway, which is embracing a ‘one country, one interlocking’ principle where the entire network is controlled from a single control centre rather than 300 separate interlockings. Italy and Spain have also been passionate adopters of ETCS for their high-speed, and increasingly mainline networks, while traditional resistors Germany and France are also beginning the process of wider adoption.

Indeed, ETCS is the foundation of French and German projects to introduce semi and full automation of mainline operation.

London’s Thameslink project was the first commercial deployment of ATO over ETCS at GoA 2 in 2018. Using systems supplied by Siemens, trains operate autonomously on a central section, which has increased throughput to 24 trains per hour per direction.

Siemens has followed up the London project with a contract to deploy ETCS at GoA 2 on Sydney’s metropolitan rail network while Thales has GoA 2 pilot projects in France and Germany. In addition, Alstom worked with Dutch infrastructure manager ProRail and Rotterdam Rail Feeding to install GoA 2 on freight trains using the Betuweroute in the Netherlands in 2018.

Alstom is leading a project to test a GoA 3+4 operation on the Braunschweig - Wolfsburg line. Photo: Keith Fender

Alstom is also working with German regional operator Metronom, the Regional Association of Greater Braunschweig, the German Aerospace Centre (DLR) and Technical University of Berlin to test GoA 3 and 4 on the Braunschweig - Wolfsburg line. The project envisages operation at GoA 3 with the driver remaining in the cab, and GoA 4 for driverless operation in depots. The first trials with passengers are expected to take place in 2023.

Elsewhere in Germany, Siemens is working with the city of Hamburg and German Rail (DB) to introduce ATO over ETCS on the S-Bahn network and main lines in and around the city. Thales is also partnering with Albtal Transport (AVG) to develop GoA 3 and GoA 4 operation at the depot of Karlsruhe Transport (VBK) with a view to extending autonomous operation to the city’s tram-train network. Furthermore, trials of ATO over ETCS by Siemens with Swiss Federal Railways (SBB) are the first to comply with Unisig standards. SBB’s tests with regional trains confirmed an energy saving potential of up to 37% and a 30% increase in capacity.


French National Railways’ (SNCF) ambitious mainline automation programme commenced in 2018. Spilt into passenger and freight projects, the objective is to introduce commercial operation on the mainline network, including high-speed lines, from 2025.

Mr Luc Laroche, director of the autonomous train project at SNCF, says the project passed two important milestones in 2020: successful completion of the first obstacle detection tests at 100km/h, and successful operation of a freight train at GoA 2 on the Longwy - Longuyon line in eastern France at the end of October. “The purpose of these tests was to test the first brick of the future GoA 4 train,” Laroche says. “The ATO GoA 2, which complies with European standards, demonstrated correct functionality. The tests also helped to define the entire organisation for ambitious trials in the future.”

Among these is the next major challenge facing the project: identifying and reading lineside signals on which Laroche reports steady progress. He also says monitoring the railway environment is a major challenge and an area where significant headway is expected.

Further milestones are expected in spring 2021 with the final demonstration of the telecontrol project to successfully pilot a train from a remote location following initial demonstrations in June 2019. Testing of the autonomous passenger train project is also scheduled to begin at the end of April. A TER Regio2N EMU has been equipped to conduct the tests, beginning like the freight project at GoA 2, with the goal of reaching GoA 4 trials by 2023.

SNCF is consulting closely with DB on autonomous trains, working together to guide the European specifications for GoA 2 and GoA 4, and sharing work and thoughts on obstacle detection. Laroche says regular exchanges are also held with Italian and Russian counterparts. Europe’s Shift2Rail (S2R) project, of which SNCF and DB are members, is also playing a critical role in shaping future consensual regulations and specifications between operators and the industry.

Main line automation projects are not limited to Europe. GoA 3 operation is already in place on the Moscow Central Diameter railway and further work is underway on GoA 4 applications, with the first results expected by the end of the year. China is similarly embracing GoA 2 for main line operation, opening the 174km Beijing North - Zhangjiakou line on December 30 2019, the world’s first automated high-speed line. Japanese railways are also exploring the benefits of ATO; JR East is set to introduce ATO over ATC at GoA2 on the Joban Line this month. Work to introduce automation on the Shinkansen network is also underway.

Positive Train Control

In the United States, a major milestone for railway signalling was achieved on December 29 2020 when the Federal Railroad Administration (FRA) confirmed that all 41 freight and passenger railways had met the extended deadline for the rollout of Positive Train Control (PTC) across 92,575 mandated route-km.

The FRA heralded completion as a landmark achievement. PTC implementation was required by the federal government following a head-on collision between a Metrolink passenger train and a UP freight train on a single-track line in Chattsworth, California, in 2008. The railways have subsequently spent billions on developing a system that meets the required safety standards and on subsequent rollouts across the country.

For proponents, PTC reinforces the existence of an inter-operable American rail network, improving safety and efficiency of operation. However, for critics, it is a missed opportunity. They argue that specifications were developed before the suppliers were ready with the technology. Development also overlooked reducing and maintaining tight headways.
Yet PTC is the foundation of new efforts to improve operational efficiency among freight and passenger railways.

With a 27,500 route-km network, UP was responsible for the US’ largest PTC deployment. The railway is set to reach 480 million km of operation using PTC this year. And, crucially, it has accumulated multi petabyte-sized datasets of every piece of information retrieved from the system since it began operating.

“Our big data environment was intentional. It’s real-time. And it’s pretty key to what benefit we will get out of PTC.”

Michael Newcomb, UP’s assistant vice-president for transportation service

Mr Michael Newcomb, UP’s assistant vice-president for transportation service, says from this Big Data environment, the railway has a better understanding of the physics of its network than it ever did. This is helping to inform decision making and reduce the impact of exceptional events on overall network operation - for example by limiting the number of false positive braking events, which has a concurrent impact on other traffic.

“Our big data environment was intentional,” Newcomb says. “It’s real-time. And it’s pretty key to what benefit we will get out of PTC. You can logically assume that we would do everything, from optimising train movements to repairs, to other kinds of things based on that kind of data, and it’s making a difference. Now, would I say does it overcome the cost of PTC? No. But does it help us? Yes.”

As an overlay system, PTC is also providing the foundation for several other technologies that will improve performance in the future.

Newcomb serves as chairman of the Train Control, Communications and Operations Committee (TCCO) at the Association of American Railroads (AAR). The committee is made up of AAR members including the Class 1s as well as the American Passenger and Transit Association (Apta), American Shortline Railroad Association (ASLRA) and terminal railways. Together they are working to identify the priority projects which make sure no element of the system is left behind.

In total, the committee and its railroads have around 25 active projects and 50 that are ready to go. All of which are agreed by the committee and follow its strategy of improving safety, efficiency of operations, undoing the negative impacts of PTC - such as increased maintenance - and supporting the industry lifecycle. “These projects are all geared to meeting these objectives,” Newcomb says.

Among the highlights of the projects underway at UP are the rollout of Energy Management Systems (EMS), which manage the key functions of the train from stop-start to air and dynamic brakes and will be installed on 3100 UP locomotives by the end of the year; Positive Train Location, which will supersede GPS and dramatically enhance train location, particularly in tunnels and canyons - it could also facilitate automation through use in an updated end-of-train device; Quasi Moving Block (QMB) a form of moving block operation which retains track circuits; and a standardised interface for the Locomotive Command and Control Module (LCCM), which will support EMS.

Yet the key to future automation according to Newcomb isn’t the technology. “It’s variability,” he says.

The variability challenge is apparent when comparing UP with the current holy grail for automated freight operation, Rio Tinto’s Autohaul project, the world’s first driverless freight railway. The signalling technology deployed by Hitachi Rail STS is helping Rio Tinto to increase train speeds and reduce run time variation to the extent that it is now carrying 300 million tonnes of iron-ore out of the Pilbara every year. However, it’s a much simpler operation than UP. A single train consist type is used to operate the 1500km point-to-point network, which has 44 level crossings. UP’s 51,000km network has 31,000.

“I already have 2600 locomotives that for 90% of the trip operate without human intervention, so that’s not my challenge,” Newcomb says. “My challenge is all the things that stop trains. If I had automation today, and I had to take a car to go and rescue a train, and I took 30 minutes to do it, I couldn’t use it because the variability is too high.”

Future evolutions

Further evolutions of current generations of ETCS technology in Europe concern work on satellite-based positioning, improvements to cybersecurity, and the development of digital interlockings, including cloud-based solutions. Siemens has successfully deployed its first cloud-based interlocking at Achau, Austria, using its Distributed Smart Safe System (DS3).

Mr Amaury Jourdan, vice-president technical and innovation at Thales Ground Transportation Systems, which is also working on similar concepts, says this approach increases the scope of control over signalling functions, meaning a centralised control centre is able to oversee greater proportions of the overall network.

Copenhagen metro is equipped with CBTC supplied by Hitachi Rail STS and operates at GoA 4. Photo: Shutterstock/Michael Donnelly

Similar evolutions are taking place with CBTC. While this is now a mature technology and deployment of GoA 4 commonplace, work on the eighth generation of CBTC at Thales is focusing on improving performance capabilities, specifically running many more trains using less hardware from a centralised location.

ETCS Level 3 replicates CBTC’s moving block principle and has been around for more than a decade - Bombardier presented the inaugural system on the 134km Västerdal freight line in central Sweden, which went live in 2012. However, limitations with the technology, specifically train length detection, halted further rollouts.

This could change with hybrid Level 3, which Jourdan says will start to go beyond the limits of fixed blocks. Trials led by Network Rail and ProRail in 2018 hinted at the potential of the application. Work in S2R should also aid development and make it a viable solution for the future.

The gradual process of digitalisation is also set to benefit railway signalling. The rollout of the Internet of Things (IoT) and the collection of data from more sensors and touchpoints can contribute to improvements in efficiency through the development of Digital Twins. Artificial Intelligence (AI) is also set to play a role in aggregating this data.

Indeed, AI based on deep learning and machine learning algorithms that form neural networks is already present in the railway environment in predictive maintenance, video analytics, and operational support and decision making. Integration with signalling, however, is a little more difficult according to Jourdan due to the need to comply with Cenelec standards.

“To go from point A, which is the data you inject, to point B, which is the result, is often unexplainable, you simply cannot explain the result that is obtained, and that’s a big no go for Cenelec for good reason,” he says. “Because of that the first generation of autonomous products that we are working on are not using deep learning. It is not necessarily simpler but with much more deterministic algorithms that we can prove against safety validation.

“In parallel we are working on a topic relating to AI and safety, which is called ‘Trustable AI.’ Trustable AI means you have a way to control the result, you will get a learning principle. At that point we can revisit Cenelec.”

“I already have 2600 locomotives that for 90% of the trip operate without human intervention, so that’s not my challenge. My challenge is all the things that stop trains.”

Michael Newcomb

Thales’s work here emphasises another point made by Jourdan: that rail innovation is not really geared for quick surges and big failures. “Right or wrong, I am not taking sides here, it is a fact,” he says. “This means it is more of a consumption market than a technology revolution market.”

While rail automation will certainly benefit from innovations pioneered in other sectors, it has to look beyond adapting innovations from the road space. In the United States, Newcomb’s says the most complicated area of ATO is not controlling the trains, it is the sensor package required to work in all weather conditions at the distance required, which is very different from the needs of road vehicles because they do not need to see as far ahead.

“We find that the better partner is the military industrial complex,” he says. “The US Navy has created a GPS-denied warfighting capability and the vendor community that built that fits with our needs. That’s an area where there’s a lot of research going on in what is the higher technical risk area for ATO.”

Ultimately the decision to proceed with different technologies will come down to economics. Rio Tinto was the first freight railway to proceed with full automation and paid a hefty price for doing so, which was way and beyond the estimate at the start of the project. However, the increased efficiency that it has unlocked is already offering an economic pay back. It is similar with mainline signalling automation. Technologies are now more readily available and proven to enhance rail’s viability as a mode of mass transport by providing greater flexibility, enhancing the business case for adoption.

For example, combined with measurements of passenger density, it is possible to easily adapt automated services to spikes in demand at minimal extra cost - it will be possible to deploy fleets of autonomous trains as soon as the final whistle blows at a football match, for example. Automation will also reduce overheads and increase the financial viability of lesser used lines by offering a more frequent service that is more conducive to passenger needs, helping to convince more people to use the train.

This is the thinking behind a project by Italian Rail Network (RFI) to develop its pure Control Command System (CCSp) Regional ETCS Level 3 concept. ‘Pure’ means the system does not overlap with the national train control and traditional interlocking systems and is based on fixed virtual blocks and virtual track circuits with the option to deploy moving block to increase capacity.

The emphasis on intelligent trains rather than intelligent infrastructure instructing trains is indicative of a future shift in the basic concept of signalling away from the central wayside controller.

Jourdan says Thales is working on these concepts for both CBTC and mainline applications. This includes developments in train positioning using radar-based sensors onboard metro trains and satellite-based positioning for mainline trains. Enhanced obstacle detection is another area of work, which has already been pioneered in light rail vehicles. Overall, he expects significant progress here within the next decade.

Many of these concepts and technologies will require enhanced telecommunications to function effectively. GSM-R has been a great success, but it is 2G and will begin to be withdrawn in 2030. Work led by the UIC on the 5G-based Future Railway Mobile Communications System (FRMCS) is encouraging and reached a key stage in November with the start of the 5GRail project to develop the first FRMCS prototypes. The objective is to begin line trials in France and Germany in 2023. Japan is also trialling the capabilities of 5G on its Shinkansen network while China is actively developing the technology.

In the United States, Newcomb says work is underway to update UP’s communications capacity and capability. New radios will soon be available that will support 20 channels rather than six on existing 220MHz bands, which will meet future demand for more capacity in busy areas like Chicago. In addition, the railway is set to move data and voice communications for train control into the 160MHz band in the next decade.

This will also involve the use of additional cellular or low-earth orbiting satellites to boost coverage in areas which are out of reach of existing cell coverage, effectively creating a dedicated private network.

That these two technological evolutions are taking place in parallel is particularly encouraging. Progress will take time. But with pressure building to get more trains on track, the railway sector needs to make more of what is has. Enhanced signalling is the best way to do this and many of the projects now underway offer a window of what future automated networks will look like.