Driverless rail vehicles
Autonomous travel across the country and around the city

There are still a number of technical and regulatory obstacles to overcome before autonomous cars are ready for mass production. But when it comes to rail transport, things are much further ahead. Self-driving trains are already a common mode of transport. But how does autonomous driving on rails work?

mask Nahaufnahme einer Schiene

We are still some years away from self-driving vehicles being a routine sight on our roads. Yet on the rails, this is already a reality — and has been for almost 40 years. The first driverless metro train was brought into service in the French city of Lille back in 1983. Today, there are more than 60 metro systems around the world that are controlled fully automatically via a central control room. All of the tasks that would otherwise be carried out by people — setting off, coming to a halt at the stop, opening the doors and monitoring the route — are controlled by a computer system.

How the driverless metro system in Nuremberg works

The U2 and U3 driverless metro lines in Nuremberg are the only ones of their kind in Germany. In operation for over ten years now, they transport more than 200,000 people to their destinations on an average workday. This is made possible through the networking of the vehicles, track and signal towers. Computers are installed in the vehicles and along the track to control the service. Camera and radar systems ensure safety on the platform. If unauthorized persons access the tunnel, sensors immediately trigger the alarm. An obstacle detection system on the track bed ensures that the train brakes in good time if anyone falls onto the line. Intelligent doors detect if people or objects have got caught in them and prevent the train from setting off. However, this system doesn’t run entirely without human assistance. Employees in the central control room monitor the service and can intervene in an emergency. Operator VAG Nürnberg is delighted with how the system works. Passengers are transported reliably, punctually and safely — and there haven’t been any problems with acceptance either. All things considered, the computer even drives better, according to a spokesperson.

The advantages of driverless systems

The costs of additional technology, infrastructure and maintenance are nevertheless high. So what are the advantages of driverless systems? Self-driving trains increase passenger safety. The monitoring systems react faster to unforeseen events than people. Waiting times are also shorter, as an automated service enables trains to run more frequently. For example, the Nuremberg metro runs every 100 seconds — that’s twice as often as the conventional service. If there are large numbers of passengers, trains can be coupled automatically, doubling their length — and all remotely from the control room.

The speed of the trains can also be increased: Optimum braking and acceleration times are calculated with the aid of artificial intelligence (AI), so the trains start and stop at exactly the right time. Energy consumption is also reduced thanks to the more efficient driving style that is adopted.

Fraunhofer IIS was involved in planning departure and arrival times for the Nuremberg metro that would enable more energy to be saved. These methods are being developed further in the ADA Lovelace Center for Analytics, Data and Applications. Fraunhofer IKS is also part of this competence center and is conducting research into new data analytics processes and algorithms for concrete AI applications that offer added value for industry, data-driven services and research alike.

Center of competence for data analytics and artificial intelligence in industry

As a project partner in the ADA Lovelace Center for Analytics, Data and Applications, a cooperation platform on data analytics for science and industry in Bavaria, Fraunhofer IKS is researching new data analytics methods and algorithms for concrete AI applications.

Find out more Pfeil nach rechts

Autonomous systems even in long-distance transport

Driverless systems are mainly deployed in metro systems, as their line networks are less complex and the same types of vehicle are generally used. Thanks to the tunnel system, they have a relatively closed infrastructure with manageable external influences. However, autonomous rail transport above ground is possible in principle too. Although train drivers still sit in the long-distance trains operating on the German rail network, contact with the control room is theoretically already sufficient to enable it to control the train. However, legal hurdles mean that passenger trains in Germany are not yet allowed to be driven without drivers. This may change soon though, and the required infrastructure is already being prepared: The European Train Control System (ETCS) is set to become the uniform standard for automatic train control systems across Europe. On routes equipped with ETCS, sensors monitor the speed and position of the train. What’s more, checks are carried out to see whether the track is clear — and if it is, an automatic clearance to travel is issued. This system is a basic prerequisite for an autonomous service. Experts anticipate that driverless trains that also transport people will be on the tracks in Germany in the next five years.

Dependable perception for safe applications

The correct detection of the environment is essential for the safe operation of cognitive systems. Autonomous vehicles, for example, continuously record their environment using various sensor technologies such as camera, radar or lidar systems.

Find out more Pfeil nach rechts

Perception — research required

The foundation for this advance is flawless machine-based perception of the environment. Perception is generated from collected and merged data from cameras and sensors involving radar and lidar technology, for example. Fraunhofer IKS is working on making AI more reliable when cameras are used to perceive the surroundings. A wide range of obstacles need to be able to be reliably detected and differentiated. AI also has a major role to play in checking the tracks for defects. To enhance the performance of AI for the application, the best possible machine learning processes and the associated hyperparameters have to be selected. To enable the quality of the predictions to also be improved, the uncertainties of AI also need to be determined precisely during operation. Under the banner Safe AI, scientists from Fraunhofer IKS are conducting research into AI assurance and developing systems that are proven to operate reliably.

Read next

Future mobility — market analysis, part 1
Autonomous locomotion in business use

Anna Sophie Kreipp
Anna-Sophie Kreipp
Smart farming / Fraunhofer IKS
Smart farming