Many readers will have quite happily travelled in a driverless system. Maybe it wasn’t comprised of multiple carriages and it had a limited number of destinations, but once inside with the doors closed, you were under the protection of an autonomous system.
Recent advances in the automotive industry have produced cars with systems that autonomously parallel park and have collision avoidance systems. These brake the car when it considers the driver is not reacting fast enough and are two examples of where the human is removed from the control loop. The principles are similar and the technology is proven. Why then is it considered that trains on the main line, without a driver in the front, are not such a good idea?
So, what functions do drivers perform in a modern railway? As far back as 1988, Tony Hawker presented a paper at an Institution of Railway Signalling Engineers event titled Have we forgotten the driver?, the focus of the paper was that in making signalling systems more efficient (simplfying layouts, reversible operation etc.) meant that, while systems were safe and when the signalling system was operating correctly, it became more and more difficult to drive the train manually under degraded conditions. That was, of course, with a driver in the cab.
To take the scenario a stage further, should we forget the driver? If that is done, what happens under degraded modes conditions? However, initially, an examination is required of the skills needed of a driver on a day-to-day basis (in normal and degraded conditions). Once that is understood, the impact on risk of removing the driver can be better debated
What a driver does under normal conditions will depend on the level of automation of the train. In a manually driven train in general they manage the speed of the train to keep it at or below a safe speed and by observing and reacting to the signals. They give a safe separation between the trains as well as observing the track ahead for obstructions, especially at stations.

Is the driver more of a hindrance than a help?

On a number of modern systems (mostly metros) these are functions where technology already manages the safety of the train around the railway network. Systems have been around for many years that can easily manage and monitor the speed to a correct profile. Additionally, other systems manage the separation between trains and the stopping of trains at the correct place in platforms and in heavy traffic areas, the electronic system can manage the throughput of trains much more effectively than a driver can manage. So is the driver more of a hindrance than a help?
There are two issues to consider in support of the driver. Firstly, the driver is also the on-board front line mechanic for the train and has all sorts of overrides and bypasses which ensure that virtually all train failures can be overcome – even if the train has to proceed at very low speeds.
Secondly, the driver is human and able to adapt to each unusual situation. A transport system is there to move people. By their very nature, people are random in their reactions to the events around them. When there is a perturbation, a driver can cope with this (indeed is trained to cope); we may forecast a computer system able to adapt to this degree but not yet.
There are also downsides with too much automation. Valid concerns have been raised that if the driver lets the automated system control the train for too much of the time, the driver will lose the ‘feel’ of the train and thus the ability to effectively drive the train when the system is operating in degraded mode.
In fact, it is common among metro operators to ensure that the driver should operate the train manually for a period of time. For one operator, the figure of 30 per cent of the time in manual was tried and while it was appreciated how necessary it was for the driver to retain the competence to drive the train, the 30 per cent figure meant that in a lot of cases this manual driving had a negative effect on the train performance, which then impacted on headways and ultimately the system was unable to maintain the timetables. The 30 per cent figure, while desirable was then revised down quite significantly. On the Jubilee Line, the trains are only driven manually into the depot and that is really to react to events that may be occurring on the ground (shunters and depot staff walking in the area).
More relevant to mainline trains is the driver’s ability to sound an audible warning to announce the presence of the train – an essential safety function when staff are working on the line.
Another consideration is whether the autonomous vehicle concept is looking in the wrong area? Is there a case, based on the logic that some freight trains spend a lot of their time waiting for the availability of a path, that a better utilisation of driver resources (e.g. at the start and end of the journey) will be to make the freight trains autonomous and reduce the number of train crew needed.
However, after reviewing the published accident reports of the last few years, it is very evident that most derailments occur to freight trains, which would have to be a serious consideration if they were to be made driverless. Notwithstanding, driverless freight trains have been achieved on the Rio Tinto mineral trains in Western Australia, but they are on a dedicated line and are only freight and are all driverless. This is quite different to running driverless freight trains on the West Coast Main Lines among high speed and commuter services.

Are the railways ready for autonomous vehicles?

They do exist in some applications, mostly in low speed people movers on dedicated and protected infrastructure. There are some fully automatic metros, such as the DLR in the UK and in Denmark the Copenhagen lines Ml & M2. However, there are no mainline railways that are completely automated and without a driver, in the world. But, the Klang Valley Metro in Malaysia is migrating towards driverless (non-attended) operation.
The UK will have one of the few main lines with ATO when Thameslink and Crossrail are commissioned, but that is a few years away. The step to driverless and full automation is still considered a long way off for main line trains, unless they are specially designed for that purpose (in particular the infrastructure) and will require maximum protection to reduce access from errant people and vehicles.
So, in places like Britain with a massive legacy rail infrastructure it may be this very legacy which constrains us as we judge ourselves against those new build railways which continue to be constructed and for which driverless is more about why not rather than why.