Taxi firm Addison Lee announced in mid October that it would be introducing driverless autonomous cars in London by 2021 (its über-competitor Uber has previously announced similar plans). That’s all very interesting, but, apart from delivering passengers to stations and then collecting them again, what has it got to do with rail?
Autonomous driving for road vehicles, when it happens – experts differ as to whether it is as imminent as Addison Lee think or a decade or more away – will have a big impact on the relative economics of existing urban rail services, especially light rail (trams).
Historically, light rail has been seen as ‘greener’ and more effective that its main on-street competitor – the bus. Light rail is, after all, (normally) electric powered and so doesn’t have emissions problems from internal combustion engines. In addition, by running on track, whether dedicated or shared with road vehicles, light rail is seen as both safer and more productive, offering higher capacity for its footprint of road space than any bus.
However, things are changing. Sales of electric buses increase month-on-month in cities around the world and autonomous self-driving buses exist – they are mostly small shuttle vehicles at present but they are in use on test in cities in many countries right now.
Bus manufacturers are working hard to develop autonomous buses that can handle city streets and, potentially, be more productive in use of road space. The challenge for light rail will be to remain cost effective, especially in terms of running costs. If, in the future, both buses and trams are electrically powered, they will both share the ‘green’ credentials.
The move to autonomous vehicles is not just something that is being driven by high-tech firms in the USA (Tesla and Google being well known proponents, along with Uber). In the UK, the Department for Transport funds the Centre for Connected and Autonomous Vehicles and, since 2014, the UK government has invested £120 million in autonomous (road) vehicle projects (with a further £68 million coming from industry contributions).
LIDAR tram front corner sensor and radar (square above), and tram front LIDAR (big rectangle) and radar sensor (square below).
Siemens presented what it called the “world’s first autonomous tram” in Potsdam, just west of Berlin, on the morning the InnoTrans show started on 18 September. Siemens has worked with Potsdam transport operator Verkehrsbetrieb Potsdam (ViP) to develop an autonomous self-driving tram using a Siemens-built Combino prototype/demonstrator vehicle dating from 1996.
The autonomous light-rail technology is a combination of software and algorithms, created by Siemens and housed onboard the trial vehicle in a large computer cabinet, plus a range of hardware, much supplied by automotive parts specialists, attached to the tram and linked to the computer.
Some parts of the system (the high resolution cameras and radar) have already been used by Siemens in its ADAS (Advanced Driver Assistance System), now in service as “Siemens Tram Assistant” in Ulm in southern Germany in new Avenio trams and on order for use in several other European cities including Den Haag in the Netherlands. In addition to the cameras and radar, the other key hardware items are LIDAR (Light Detection and Ranging ) laser-based measurement systems, a very precise GPS system and actuators to control functions such as braking and power control based upon the computer’s analysis of the data from the sensors.
The system uses a digital map of the network and the software relies on the vehicle (or more precisely its onboard computer) being ‘trained’ to know a specific route. When in use, it continuously uses various sensors and GPS to establish where it is, where it is going, at what speed and where it should stop – either for passenger stops or, in emergencies, stopping immediately if the track is blocked by people, vehicles or other obstructions.
Using its cameras, the system even reads lineside signals, the images being processed by the computer into actionable data – to stop or to proceed – as the tram signalling system only has these two options.
The tram’s stopping accuracy is worth mentioning. The system is designed so the vehicle, which is 26.4 metres long, will stop within a 50cm tolerance at tram stops. In practice, Siemens reported that tests have shown the tram can stop to the 50cm accuracy at stops and then drive from one stop to the next without human intervention.
The autonomous tram prototype has been tested by Siemens and ViP since the summer, operating at up to 50km/h on a section of normal tramway in the southern suburbs of Potsdam, based at the Babelsberg depot. The six-kilometre section chosen includes multiple level road and footpath crossings and tests have operated with a human driver in the cab for supervision of the computers for legal reasons. ViP says the tram driver has not yet had to intervene as the sensors have detected obstructions such as cars, people or cyclists and, via the computer system, used the trams’ braking and power control system to stop it safely.
The autonomous Combino demonstrated it can travel at 50km/h without a driver and proved able to stop precisely in platforms whilst also detecting and, where necessary, braking for pedestrians and vehicles in its path, including some deliberately ‘foolish’ pedestrians with prams arranged by ViP especially to prove this!
The system aims to detect all possible obstructions at 100 metres and can stop the vehicle in less than 80 metres, even at full speed.
The system has been ‘taught’ what requires immediate reaction and what doesn’t. This was demonstrated on the test run when, completely by chance, a fairly large bird (a Hooded Crow) landed on the rails about 50 metres in front of the moving tram – the system did not react as its been taught that items this small do not merit attention (the crow sensed the tram and flew off just in time).
Siemens provided onboard screens so the various data from the sensors being analysed by the onboard computer were presented visually for the benefit of the humans onboard.
Unlike traditional communications-based train control (CBTC) type systems (as used on the DLR for example), which send operating signals to the train, the Siemens autonomous tram is ‘intelligent’ enough to know where it is going (having been ‘taught’ the routes) and can ‘read’ signals provided for human operators using its cameras, so does not need signalling information transmitted to it (as CBTC does).
Many cities have automated metros using CBTC but these have dedicated infrastructure and, in many cases, segregated platform spaces (with platform edge doors), so the Potsdam test is a very different operating environment.
Siemens and ViP plan to expand the trial area to more of the Potsdam network, including the city centre, and may, as a first stage, trial completely unmanned operation in the tram depot.
The current legal framework under which trams operate in Germany (BOStrab rules) makes passenger operation without a driver unlikely (although not impossible – automated U-Bahn metro trains already operate under the same legal framework in Nuremburg) and this test vehicle is not designed for public use.
Legal changes to permit autonomous operation of light rail systems would probably follow similar legal changes currently being considered in many countries, including the UK, to permit autonomous self-driving cars and other road vehicles – in most countries light rail regulations are a hybrid mix of road traffic and railway operating rules.
Siemens may have separately developed the Potsdam autonomous tram and sold its ADAS driver-assistance package to several operators, but there is competition emerging for rail drivers’ assistance systems. Bombardier has been supplying Frankfurt’s tram operator VFG with a ‘Driver Assistance System’ using forward looking cameras since 2015.
In addition, German equipment supplier ZF launched a passive collision-avoidance driver-assistance system for trams using artificial intelligence at Innotrans. ZF claims that its ProAI computer system is capable of ‘deep learning’ and using data from onboard radar, cameras and LIDAR to identify and warn the driver about potentially dangerous situations at tram stops, where many people are often moving around the exterior of the tram.
Driver assistance (or replacement?)
Fully autonomous trains are already in operation – in the largely unpopulated Pilbara region of Western Australia where mining company Rio Tinto has been operating 28,000 tonne freight trains without drivers since the summer of 2018. This solution uses an ATO over ETCS L2 solution from Ansaldo STS, combined with onboard sensors for location and speed plus lineside cameras fed to a central control centre for the few locations where human activity may intersect with the railway, such as at level crossings.
The same overlay of ETCS and ATO (as used for the Thameslink ‘core’ in London) is seen by many as the approach for main line rail automation to make it possible in densely populated cities rather than the Australian outback, although many issues remain unresolved.
The Potsdam trial certainly shows the technology has the potential to act as a supervisor for human drivers, preventing, for example, over-speed operation on curves or avoiding collisions with pedestrians or vehicles which, for whatever reason, stray into the path of the vehicle at the last minute.
Whether the technology demonstrated in Potsdam could actually safely replace human drivers/supervisors for unfenced light rail systems anytime in the near future is not clear; however drivers make up the largest part of a light rail operators wage bill (in Potsdam ViP has 116 tram drivers for 53 trams). If buses and taxis migrate to autonomous operation in cities, light rail will probably need to as well so as to avoid being substantially more expensive to operate (and therefore unlikely to attract capital expenditure for fleet or infrastructure renewals).
Arguably, if future legislation and technology permits autonomous buses and cars to share roads with pedestrians and cyclists, then a rail guided vehicle – such as a tram – should be easier to drive autonomously in safety. Unlike a self driving bus, its route is fixed by the rails and pedestrians and other road users are able to see them and thus be aware of the potential for a tram to appear.
Written by Keith Fender.
Read more: The benefits of adopting new technologies