The idea of an autonomous vehicle is almost automatically associated with cars, but railways could offer one of the most realistic grounds for large-scale automation. The reason is simple: a train does not have to interpret a chaotic intersection, predict a cyclist's behavior, or suddenly choose an alternative route. It moves along a constrained trajectory within a regulated network, with safety margins already governed by signaling systems. This does not make autonomous rail operation easy, but it fundamentally changes the nature of the challenge.
In Europe, this transformation revolves around the progressive digitalization of railway control, ERTMS, and Automatic Train Operation—a family of technologies that can automate acceleration, braking, and schedule adherence while leaving safety limits to protection systems. The industrial goal is not simply to remove a driver from the cab: it is to run more trains on the same infrastructure, with more consistent speeds and lower energy consumption.
From the metro to the national network
Automated metros have existed for years, proving that driverless rail service is technically feasible when the environment is highly controlled. However, applying the same principle to open railways is far more complex: there are level crossings, mixed freight and passenger traffic, diverse stations, construction sites, weather conditions, and networks built across different eras.
This is why rail automation progresses in levels. In the most common systems, the train driver remains in charge while software monitors speed and protection; at higher levels, ATO can manage most of the driving, leaving human intervention to focus on exceptions. The complete absence of on-board staff represents the end point, and not necessarily the most useful one for every line.
The real prize is capacity
A congested railway network cannot always be expanded by building new tracks, especially in major cities. Safely and precisely reducing the headway between trains can therefore unlock additional capacity without equivalent infrastructure works. Furthermore, if every train accelerates and brakes according to a coordinated profile, it reduces the ripple delays that propagate along the line when a train falls behind schedule.
It is a principle similar to that of computer networks: value stems from orchestration. An autonomous train in isolation is far less compelling than a network capable of tracking the position, speed, and destination of trains and optimizing them as a unified system.
ERTMS is the common language
European strategy has long relied on the European Rail Traffic Management System to replace fragmented national signaling systems with interoperable standards. ETCS ensures that the train complies with movement authorities and speed limits; ATO can leverage that safety baseline to automate driving. The combination is essential because autonomy cannot be built as an isolated vehicle function.
Deployment, however, remains costly. Both tracks and rolling stock must be upgraded, software with exceptionally high safety requirements must be certified, and compatibility must be maintained throughout a transition that will take years. In the rail industry, a software error is not merely an interface bug: it can involve tons of moving mass.
Automation can make trains more efficient
Automated operation can select speed profiles that adhere to the schedule while avoiding unnecessary acceleration and late braking. Across thousands of runs, even small percentage energy savings become significant. The system can also adapt travel to real-world network conditions, making up for delays when necessary and prioritizing efficiency when there is margin.
This is probably one of the least spectacular yet most concrete applications of distributed intelligence in transport: not simply replacing a human decision, but coordinating decisions that are currently made locally.
The train driver will not disappear tomorrow
Framing automation as a plan to eliminate drivers would be misleading. For many years, mixed networks will continue to need staff capable of handling anomalies, emergencies, obstacles, and equipment failures. The job, however, is likely to evolve, with an increasing share of routine driving handed over to systems and the human role focused more heavily on supervision and unforeseen situations.
The history of aviation offers a useful parallel: commercial aircraft are highly automated, but this did not immediately make the pilot obsolete. It changed which skills are needed and which phases of the work are delegated to the machine.
A technology less flashy than robotaxis, but perhaps closer
Railways have a structural advantage over roads: the environment is already designed to control movement. For this reason, automation could deliver widespread benefits before a fully autonomous car is able to navigate anywhere and in any condition.
The most important outcome might not be an empty train cab, but a network where trains run more punctually, consume less energy, and operate with greater frequency. It would be an uncinematic revolution, but exactly the kind of innovation that truly transforms an infrastructure.
