Maybe the number of incidents at level crossings has something to do with it, though they have been slowly dropping over the last few years.

Money.


Or to put it another way:


Money money money.


Technically it's not an issue, we've had ATO (Automatic Train Operation) for a long time now. Julie's point is interesting: It's important to realise that a train driver can VERY rarely avoid hitting an obstruction by applying the brakes, but they are there to deal with the aftermath. It's like aircraft, simplistically modern airliners are almost completely flown under autopilot, the pilot's there to cope when things go wrong.


We will see higher levels of automation (possibly full automation) as ERTMS level 3 (in-cab) signalling gets rolled out, but it is expensive and it will take time.


It is a very interesting question, from my point of view the other way around. As James says, on the whole autonomous trains are much easier - they have defined routes, the signalling system knows where they all are and where they are going, and the installation and maintenance is very heavily regulated to maintain the safety. (Although there is an issue in the length of time they take to stop - I've just been reading an incident report where a train travelled 3.22 MILES after the brakes were applied at 64 mph. Which is why it is a really bad idea to cross in front of a train at a level crossing!)


So my question would be: why do we think autonomous road vehicles will be so much easier? I have a feeling that after a few high profile accidents involving autonomous vehicles we might see a slowing down of the programme. Or maybe, as a society, we will just accept a level of risk on the roads (compared to the railways) that is even higher than it is now.


This is because overall there are much bigger reasons to make road vehicles autonomous than rail vehicles. On large high speed trains, as I mentioned above, you will probably still want a person in the cab anyway to deal with the unexpected, so you need a good reason for them NOT to be driving. And when you take the actual number of fatalities due to train driver error in the UK (for example none afaik in 2015-16) compared to road fatalities, together with the cost of installing automated systems, it becomes a difficult argument.


Whereas with cars there is a huge advantage in autonomous vehicles: it means you can tap away on your laptop, have lunch, take a nap. The enormous (and, I believe, hugely underestimated) challenge now is to make such road vehicles no more unsafe than they already are with a human driver! 


P.S. Although I work as a functional safety assessor, the views expressed above are entirely my personal opinions.

Andy, as you say the convention has been to improve the infrastructure, (signalling), and the tools to assist the driver. New techniques being tried out for autonomous road vehicles could easily be applied to rail vehicles. In fact I suspect it wouldn't cost very much, less than a driver's annual salary. Unlike other signalling/control improvements where the line and the train have to be fitted this approach only requires changes to the train and on an individual basis if money is tight.


While it is true that the existing system is (more than?) acceptably safe the more I think about it 'ALARP' screams out at me! Technically it seems so 'simple' to do, the problems have been solved elsewhere. If a 'speed' cameras on a motorway can capture a moving number plate surely a camera on a train can capture a (literally) fixed signal? The lateral location of the train will be known with centimetric precision and the longitudinal position can be determined by wheel revolutions, sleeper counting and GPS. Trackside structures (anything) can be used to re-calibrate the train equipment. Abnormalities in the environment (blocked level crossings?) can be identified. Given favourable conditions a camera system can see beyond any driver and can use thermal imaging too if it could be beneficial. Adding these features as driver aids would just swamp a human driver.


The problem of unexpected low adhesion is an interesting one (braking at 64 mph/3.22 miles to stop). A continuous 'mu' measure could be made by measuring the torque needed to 'slip' a non-driven wheel. That could be an automatic process. I understand that the current advice is that drivers periodically attempt a full brake application if they suspect slippy conditions. A driverless train would make the check.


While I think there is no good technical reason why we can't have driverless trains I think there should be personnel 'passenger side'. (If we have to have drivers why to we stick them in the crumple zone? Is that to 'encourage' good behaviour?).

I was also looking at this article - It’s Not a Lack of Technology That’s Keeping Trains From Going Driverless and wondered if some of the factors mentioned (line of sight, and the massive distances trains need to identify and react to obstacles and bring a few hundred tons of steel and cargo to a stop safely) are still issues that we cannot resolve?


I am going to add a link to the Railway TPN for this discussion as well. 

Lynsay,

Thanks for that. Maybe I need to put on a hard hat now!

Sighting and stopping distances just don't apply to the situation that I have tried to describe, the existing signalling system looks after that.


Essentially we 'solved' the safe route problem once we added track circuiting to the interlocking. The Achilles Heel is that we rely on the driver to observe the signals and line speed limits. The aproach in the UK has been to overlay devices such as AWS, which has never achieved 100% coverage and doesn't address the speed limit problem except in very rare cases. Newer systems like ERTMS ultimately require a total refit of the railway. But the information is already there! We just have to read and act on the existing signals. The route profile and speed limits are also known, just put it on the train.


Drivers could still be retained at stations and depots where trains are prepared for use. They would be able to work fixed hours and have a fixed workplace.

I wonder how driverless trains would have coped with the issues we had yesterday after 'Storm Doris' though?


Just within our own region in Hertfordshire, I understand that we had a tree down on the track towards Cambridge and a roof from a warehouse on the track just outside Sandy on the mainline track from London up to Glasgow. Therefore any train travel in our area was pretty much cancelled for a while! 


Would a driverless train be able to identify these obstructions early enough to come to a safe stop, or ignore them and run the risk of a derailment, potentially costing many lives and causing serious injuries?


I use the driverless DLR quite often (just LOVE Greenwich ) and I feel very safe using that as there are limited options for 'obstruction' on the line and it's only a short service and not high speed. However, not sure how I'd feel about being on a driverless train at speeds of 125mph...

Hi James,


I would just add a couple of caveats to your thoughts.


Firstly, remember the level of safety you are trying to achieve. To satisfy the public acceptance of risk on the railways you need an unsafe equipment failure rate of (typically) one failure in 10^14 years. (As an example, this is what track circuits achieve.) This does not come cheaply - particularly in low volume equipment.


Secondly, unfortunately it is not going to be practical for any "line of sight" system (however automated) to brake a 200 tonne train travelling at 150mph on metal rails in its viewable distance - wheelspin detection is already in place, but to quote a famous engineer "ye canna change the laws of physics captain!" Hence the "block" system to keep trains a safe distance apart, and a huge amount of other activities to keep other obstructions off the line - which is why level crossings are such a problem.


I certainly agree that there need to be lessons shared between autonomous rail and autonomous road, however I suspect the learning process might go the other way to the one you expect. Road vehicles have traditionally been based on safety arguments based around driver intervention, with fully autonomous vehicles this argument changes completely. The rail industry has over 100 years experience of considering safety arguments based around automatic control methods, and there will be a lot of learning of how to approach a safety decision that will be transferrable. (Incidentally, ditto from aviation.) A key part of this - and I suspect, a particular area of upcoming controversy - is going to be transparency. Rail supply companies have to be fully transparent in their IP, no secrets are allowed about how their safety is assured. Once the risks of fully autonomous vehicles become fully apparent then the automotive industry, where extreme product secrecy is consider vital to staying ahead, could be in for a bit of a shock.


Interesting discussion, thank you. You might like to look at the UK "Digital Railway" project digitalrailway.co.uk to see what is coming along for the UK. Cash permitting there are some big changes ahead!


Andy

Hi Lisa,

Would a driverless train be able to identify these obstructions early enough to come to a safe stop, or ignore them and run the risk of a derailment, potentially costing many lives and causing serious injuries?

Unfortunately not, for the reasons I gave above: in general mainline trains cannot be stopped in time to avoid obstructions once they become visible. What drivers are very good at is spotting and reporting issues on adjacent tracks before they cause an incident, so personally I think this gives some argument for keeping the driver.


It's worth bearing in mind the latest available figures (from 2015):

• Road fatalities 1,732 (www.gov.uk/.../reported-road-casualties-in-great-britain-main-results-2015)

• Rail fatalities 297, of which 252 were suicides, and of which none were due to train accidents - for the ninth year in succession (www.rssb.co.uk/.../2016-07-annual-safety-performance-report-2015-2016.pdf)

The biggest risks - in fact the only significant risks - for the non-suicidal public on the UK railways are to people trespassing near the track, standing near the platform edge when trains are moving, or being less than 100% careful on a level crossing. It's not autonomous trains we need - it's autonomous people! Tongue firmly in cheek - but you can see the frustration in the rail industry given the huge efforts they've made to warn people to take care in these situations - over many decades - and people still don't. So the rail industry is investing a lot on these three issues; particularly the latter two, with a lot of work on platform screen doors on new rail lines (e.g. Crossrail) and low cost improvements to lightly used level crossings. But we're still only talking 45 fatalities from these types of causes last year. Even so, low cost innovations in these areas will be hugely appreciated by the UK rail industry. So, for instance, perhaps using the imaging technology from automated vehicles to automatically detect when someone is misusing a level crossing? 


Hope that's reassuring!


Cheers,


Andy

Andy,

Thank you for your comments particularly coming from a 'rail' perspective.

I fully appreciate that sighting distance isn't stopping distance on a mainline railway and that is why what I am suggesting does not replace the signalling system but instead provides a more reliable system of reading it. The fact that the industry has an acronym SPAD suggests that 'wrong side failures' of driver's isn't anywhere near the parts per million zone let alone parts per million million.

Language is important and  'driverless' is a better term than 'autonomous' , which perhaps fits the car situation better as it has to adapt to the environment that it finds itself in. Even then it isn't completely autonomous as it must have some sort of map built in as otherwise it will be like me trying to get through London and knowing that I had to pass through Islington. Unfortunately there were no signs saying 'here is Islington' so I must have passed through its boundaries several time before I swapped to using the setting Sun as a guide!

I think the driverless train would do even better than a human driver at spotting obstructions to other traffic. I can't stress enough that this train will know its route better than any driver, photographically to the metre! It won't be looking the other way, it won't be speaking to train crew, it won't be drowsy.

While I can conceive of having a single driverless train, (it uses existing infrastructure), there is obviously scope to add more features with an expanded fleet. A 'down' train can report precisely to an approaching 'up' train any cautions - cow on line at 10,539 metres from York datum zero etc. and without the distraction that happens with a human driver - 'rubber necking'.


I fully appreciate and applaud the levels of safety that the rail industry in the UK has achieved but I think there is a safety gap that can be closed, 'the last link'? While I titled this topic 'Driverless Trains', knowing full well that for many that would be a leap too far, what stops such a system being trialled/developed as an 'add on'? Initially it needn't have any control input at all - 'mother-in-law' system! I bet there are universities out there that would love to run a 'camera in the cab' system project to see how well it would work and with absolutely no safety consequences.