Controversial railway opinions (without a firm foundation in logic..)

[robert thomas]

Delay Repay should not apply when the incident is beyond the railway's control, as per the Stockport trespass on Saturday, which will have been hugely expensive.

Conversely, forms of transport competing with rail should be subject to identical Delay Repay rules.

seconded

 

[InkyScrolls]

Driverless trains are easier to create than driverless cars, because they don't have to worry about steering.

This is demonstrably false.

 

[Bletchleyite]

This is demonstrably false.

It's not, as they exist and have done since well before driverless cars - DLR etc. They may not be economic for the mainline yet (in terms of needing to e.g. remove all level crossings), which is an entirely different question.

I'm certain they will exist on the mainline at some point, but I doubt you need to worry about your job because it won't be in the near or mid-term future.

 

[Indigo Soup]

It's not, as they exist and have done since well before driverless cars - DLR etc. They may not be economic for the mainline yet (in terms of needing to e.g. remove all level crossings), which is an entirely different question.

I don't even think you need to remove level crossings. Self-driving cars have demonstrated obstacle detection. Determining the level of brake force required to stop in time is a job that's perfect for automation. Interfacing with the signalling system is almost trivially easy. And if there's an obstacle for which the vehicle can't stop in time, there's no choice to be made. For a road

The reason why cars are being pushed isn't that they're easier to automate (they aren't) but that there aren't powerful unions trying to keep drivers in them. Secondarily, because the tech companies pushing autonomy have the standard US car-centric view of the world.

 

[HSTEd]

1. Delay Repay should be awarded on the basis of the Anytime Single fare, not the fare paid. This would provide a stronger financial incentive for rail operators to run to time, as well as a disincentive to extortionate walk-up fares with heavily discounted options for those who travel only in odd-numbered seats under the light of the full moon.

Beyond anything else this would create a huge potential for malfeasance.
People with discounted fares, like staff, could manipulate the system to make money.

Especially as staff may know a train will be delayed before the public does.

 

[Krokodil]

Conversely, forms of transport competing with rail should be subject to identical Delay Repay rules.

Imagine delay attribution on the roads.

Nope. The Beeching take on roads would have been very interesting. You could expect to see, based on things he actually said in interviews:

• Significant restrictions on bringing cars into cities, especially on on-street parking, and potentially a requirement that car owners would have to have off-street parking for their vehicles
• Better integration of bus and rail transport
• A reduction in the indirect subsidies provided to air transport, coach operators, and road hauliers

Broadly speaking, I think that everyone should have cause to curse his name, not just rail enthusiasts!

I've suddenly decided that I rather like the bloke. Are there copies of these interviews anywhere?

I don't even think you need to remove level crossings. Self-driving cars have demonstrated obstacle detection. Determining the level of brake force required to stop in time is a job that's perfect for automation. Interfacing with the signalling system is almost trivially easy. And if there's an obstacle for which the vehicle can't stop in time, there's no choice to be made.

How far ahead can an autonomous vehicle "see" an obstruction?

 

[yorksrob]

Delay Repay should not apply when the incident is beyond the railway's control, as per the Stockport trespass on Saturday, which will have been hugely expensive.

Certainly not. It's not the passengers fault that someone decides to use the OLE as a climbing frame.

If the industry is that worried about the costs, go beck to giving out RTV's so that the money stays in the industry.

 

[LUYMun]

14. All those supermarkets and retail parks on old goods yards should have their rail link reinstated and be required to receive deliveries by rail.

I thought the provision of smaller-scale freight services was obsolete by the 1950s? Why resurrect this now?

 

[Indigo Soup]

I've suddenly decided that I rather like the bloke. Are there copies of these interviews anywhere?

The major one I tend to refer to is The Transport Conflict, published by the Observer. I believe others may exist. His views on parcels service are rather interesting in light of the huge volume of parcels being transported by the modern, hugely fragmented, parcels industry.

He did say somewhere, but I'm not sure where, that he thought that road transport was in need of a Reshaping-type study.

How far ahead can an autonomous vehicle "see" an obstruction?

That very much depends on how far ahead the sensor suite is designed to detect an obstruction. The obstacle-detection systems fitted to autonomous ships have a useful range well in excess of anything the railway might need - if you think a high-speed train takes a long time to stop, try a large container ship!

I thought the provision of smaller-scale freight services was obsolete by the 1950s? Why resurrect this now?

Because I was writing the post while waiting on a station platform looking at the supermarket in the former goods yard, thinking what a nuisance large delivery vehicles are on local streets. Does it make economic sense? No. Not in the slightest.

 

[Krokodil]

That very much depends on how far ahead the sensor suite is designed to detect an obstruction. The obstacle-detection systems fitted to autonomous ships have a useful range well in excess of anything the railway might need - if you think a high-speed train takes a long time to stop, try a large container ship!

Container ships don't tend to operate along Network Rail's linear forest. If you had obstacle detection on trains they'd forever be chucking the brake in.

 

[Indigo Soup]

Container ships don't tend to operate along Network Rail's linear forest. If you had obstacle detection on trains they'd forever be chucking the brake in.

Obviously you wouldn't spec the systems the same. But maritime autonomous systems demonstrate the 'looking a long way ahead' thing. Autonomous road vehicles demonstrate the 'deciding if something just off the line of route is going to get in the way' thing.

I don't think it's anywhere near an acceptable standard of accuracy or reliability at the moment for rail or road use. For maritime use it probably is, because the operating environment is so different. But while rail is a more complex regulatory environment, the technical challenges for autonomy are greatly reduced compared to road.

 

[HSTEd]

I don't think the obstacle-on-the-tracks detection problem is really a showstopper.

Trains regularly operate in conditions that make visual detections of obstacles on the line functionally impossible - at least in time for a driver to take meaningful action.

I'd expect eliminating the driver would result in a net reduction in injuries and fatalities from that cause by removing the most vulnerable to injury person on the train.

 

[Trainguy34]

Make it so that, unless the person is proven to be doing something out of their control, they must pay for at least 20% of the delay repay caused by the incident.

 

[Bald Rick]

Make it so that, unless the person is proven to be doing something out of their control, they must pay for at least 20% of the delay repay caused by the incident.

And what if they don’t have any money?

 

[Trainguy34]

And what if they don’t have any money?

Maybe not 20%, maybe 20% of income?

 

[jfowkes]

I can imagine a far-off future where self-driving mainline trains are the common-place default more so than self-driving cars, but I think both are a long way off from true mass adoption.

My take: the main reason for going self-driving for trains is elimination of the cost of a driver with reducing or eliminating human error from driving a distant second. Drivers are high cost because the mainline UK railway is a complex environment to drive a train in and the public rightly demand a very high standard of safety from the railway. Any self-driving system capable of replacing a driver in all circumstances while maintaining that level of safety would be very expensive to build and maintain, to the point where I suspect it's not worth doing.

 

[Meerkat]

Any self-driving system capable of replacing a driver in all circumstances while maintaining that level of safety would be very expensive to build and maintain, to the point where I suspect it's not worth doing.

How come various systems have already been doing it for years then?

 

[jfowkes]

How come various systems have already been doing it for years then?

Oh, you can do it on isolated networks like the DLR, or have "very close to driverless" operation in controlled environments like Crossrail or the Thameslink core.
Is there an example of driverless operation in the context of a more complex network with mixed-use traffic, inter-operating with trains that still have human drivers, inter-operatiing with legacy signalling systems, in all weather and track conditions? That sort of thing.

I'm sure these things will happen with time, but I don't think it'll be easy.

 

[Krokodil]

How come various systems have already been doing it for years then?

On classic routes with occupation crossings, overgrown trees etc?

 

[Meerkat]

On classic routes with occupation crossings, overgrown trees etc?

Can drivers do much about these, and can’t a computer do it if it can get a taxi through city streets?

 

[Krokodil]

Can drivers do much about these

There is a long list of collisions in recent history where the driver has managed to reduce the impact speed and therefore the severity of the accident.

and can’t a computer do it if it can get a taxi through city streets?

I don't suppose that the taxi is fitted with steel wheels and running on steel rails?

 

[HSTEd]

On classic routes with occupation crossings, overgrown trees etc?

Deployment of obstacle detection at a few thousand level crossings, only a small fraction of which are open, is probably not going to break the bank given the savings driverless trains would provide.

 

[Bald Rick]

Deployment of obstacle detection at a few thousand level crossings, only a small fraction of which are open, is probably not going to break the bank given the savings driverless trains would provide.

It would be somewhat cheaper to close most of them. Don’t forget that the majority of level crossings are not public roads.

 

[Krokodil]

Deployment of obstacle detection at a few thousand level crossings, only a small fraction of which are open, is probably not going to break the bank given the savings driverless trains would provide.

Sticking obstacle detection on all 6,000 crossings in the UK, are you mad? Quite apart from the fact that farm vehicles are often associated with mud.

Then there's the line away from crossings. What about trespass, animals, fallen trees, fallen cement mixers...?

 

[HSTEd]

Then there's the line away from crossings. What about trespass, animals, fallen trees, fallen cement mixers...?

The train hits them, as it would anyway.
The elimination of the cab, with its exposed crewmember, will also mitigate the effects of such impacts.

 

[Krokodil]

The train hits them, as it would anyway.

Except that often the driver does react and reduces the impact speed. Up here many sheep have been spared due to the vigilance of the driver.

 

[InkyScrolls]

Except that often the driver does react and reduces the impact speed. Up here many sheep have been spared due to the vigilance of the driver.

Ditto on the S&C!

 

[Indigo Soup]

Except that often the driver does react and reduces the impact speed. Up here many sheep have been spared due to the vigilance of the driver.

Is your desired outcome for the train to slow down when there's an obstacle on the line, or for it to not slow down when there's an obstacle on the line? Because either can be done.

The sensor suites fitted to autonomous cars aim to tell the difference between an obstacle with which the vehicle may collide, and one merely moving parallel to its course. A similar setup could be fitted to a train. The detection range would need to be increased, but that's a solvable engineering problem. Real-time measurement of visibility and railhead conditions would allow for speed to be moderated to a safe level in a similar way to a human driver.

Likewise, if an autonomous car can tell the difference between a red traffic light and a green traffic light, an autonomous train can tell the difference between a red signal and a green signal. It can even be programmed with the necessary route data to tell it where to look, which isn't possible for cars - and is limited only by the storage available. Semaphore signals might be more of an issue, I'll grant.

Yes, there are still significant problems with obstacle detection. That applies to both rail and road. Silicon Valley tech investors have decided they want to see a return on their money from road vehicles, and have the political influence to force acceptance despite the issues. There isn't so much money riding on autonomous rail vehicles, and rail unions are strong enough to resist the pressure of capital. For now.

 

[Krokodil]

Is your desired outcome for the train to slow down when there's an obstacle on the line, or for it to not slow down when there's an obstacle on the line? Because either can be done.

To slow down when there's an obstacle, and to not slow down when there isn't an obstacle.

If you want autonomous trains with no human oversight, you're going to need to sort out the lineside. You need to eliminate all level crossing risk, securely fence the line from trespassers and livestock, cut back all vegetation properly and maintain structures so that they aren't in the habit of falling down on the track. You'll need to secure the PTI too, so that drunks leaning against trains can't be dragged under. That means platform edge doors. You also need a means of evacuating passengers safely in an emergency where staff aren't immediately on hand. I gather that the Crossrail core has catwalks running through it, that won't be easy in Victorian tunnels. All that on top of the cost of rolling ETCS out to the nation.

All that cost, just to save a few salaries and you still haven't improved the service. The money would be far better spent on electrification which would both save costs and speed up the service.

 

[Bletchleyite]

If you want autonomous trains with no human oversight

I can't see that happening. More likely is "guard only operation" a la DLR, where there would be a member of staff on board whose primary role was dealing with passengers and operating the doors, but they were also trained to drive the train at low speed on sight to a safe location in emergency. Such an approach could do a lot for the economics of rural branch lines, for instance.

As for PEDs, the DLR doesn't have them. They'd be needed for fully automatic operation but not for "GOO".