Driverless trains - why limited progress on the national rail network?

[455driver]

Yeah, I remember a documentary a few years back following a BMI 737 from Manchester (I think) to Malaga. Not long after gear up, a minute max, they turned on the autopilot. and it was literally down to inputting altitudes and directions provided by ATC until the captain took control when the runway at Malaga was in sight, and I mean very clearly in sight and pretty close.

I would have fancied my chances at doing what the flight crew did on that flight, if anything went wrong though I wouldn't have a chance at flying it manually. Brown trousers and a heart attack for me I fear.

There is no need for the pilots to land the plane either, the Trident carried out the first full auto landing in 1968,the only thing the pilots had to do was taxi the plane off the runway,the plane did everything before that including the lineup, descent, flair, touchdown, reverse thrust and stop.

I think there is one (Japanese?) Company that has their pilots let the planes land themselves and it was that which was a major factor in one of their planes crashing on landing, the crew put all their faith in the auto systems until it was too late and then they didnt have the experience to recover it in time.

 

[Dave1987]

There is no need for the pilots to land the plane either, the Trident carried out the first full auto landing in 1968,the only thing the pilots had to do was taxi the plane off the runway,the plane did everything before that including the lineup, descent, flair, touchdown, reverse thrust and stop.

I think there is one (Japanese?) Company that has their pilots let the planes land themselves and it was that which was a major factor in one of their planes crashing on landing, the crew put all their faith in the auto systems until it was too late and then they didnt have the experience to recover it in time.

The Air Asia crash at KSFO (San Francisco) is the one you are referring to I believe. The situation with that was that the pilots involved were inexperienced for their roles. They were also very used to doing ILS instrument approaches. On that day they were given what is known as the 'Quiet bridge visual'. Its quite uncommon to be given that approach but perfectly legal to do so. Because they were unfamiliar with the approach they got behind the aircraft and weren't used to doing a manual visual approach. They were used to the aircraft having speed protection system in place so the aircraft would not go below the critical low speed. But they had managed to put the auto throttle system into a mode that didn't provide that protection and did not realise it. They did not completely understand the auto throttle system and that in certain modes the speed protection wasn't available.

And to do a full Cat 3b autoland the aircraft has to be equipped with 3 autopilots, a radio altimeter, the airport has to be suitably equipped, aircraft on the ground have to be further away from the runway to protect the localiser signal, and the pilots all have to be trained and certified to perform that approach. Doing anything other than a Cat 3 landing the auto pilot has to be disengaged at the MSA (Minimum safe altitude) or DH (Decision Height), and even during an auto land the pilots are required to ensure all the appropriate modes arm at the relevant time or be prepared to go around. Even with an auto land the human is still very much involved.

 

[GB]

Autoland is of no use if the crosswind, tailwind or headwinds are outside of operating range...which gets even more stringent in the event of and engine out scenario.

Automation was also of little use during the Air Canada Flight 143 and BA Flight 9 incidents.

Automations has its place, as do pilots. Both of which have contributed to successful and unsuccessful incidents in the past.

 

[Dave1987]

Autoland is of no use if the crosswind, tailwind or headwinds are outside of operating range...which gets even more stringent in the event of and engine out scenario.

Automation was also of little use during the Air Canada Flight 143 and BA Flight 9 incidents.

Automations has its place, as do pilots. Both of which have contributed to successful and unsuccessful incidents in the past.

Indeed at a lot of airlines only the captain is authorised to land the aircraft to is full cross wind limit, many airlines only allow first officers to land to half the cross wind component the aircraft is capable of handling.

Automation does indeed have its place but where human lives are at stake having human perception there is vital, you just need to ensure that the human is suitably trained and capable.

 

[GB]

Automation does indeed have its place but where human lives are at stake having human perception there is vital, you just need to ensure that the human is suitably trained and capable.

Fully agreed with that.

 

[455driver]

The Air Asia crash at KSFO (San Francisco) is the one you are referring to I believe.

You have the right crash but it was actually Air Asiana, thanks.

 

[455driver]

You have the right crash but it was actually Air Asiana, thanks.

Here is a video if it-
https://www.youtube.com/watch?v=zTXDalv7kNQ

 

[LdnNiko]

The aircraft was in trimmed straight and level flight at the moment the automation dropped out. Everything that happened subsequently was down to the crew's actions. As you said, they had reliable pitch and power information and no reason to doubt its accuracy, that was all they needed to keep the plane out of the water.

And yet when the automation failed, the aircraft was no longer in straight and level flight. See again the text from the report below.

When the autopilot disconnected, the roll angle increased in two seconds from 0 to +8.4 degrees without any inputs on the sidesticks.

Clearly, everything worked as planned until such time as it didn't - it was then necessary for human operators to intervene. Leaving the aircraft to do it's own thing with the autopilot having disconnected in stormy weather, crossing the turbulent ITCZ, would have still resulted in the aeroplane taking an undesired flight path.

The whole point of this thread is to argue what is delaying further automation in transport, specifically rail.

As long as the automatics on transport control systems continue to decide to trip out and subsequently perform an unanticipated action or actions, or fail to continue performing the desired tasks, then automation will be limited.

 

[najaB]

And yet when the automation failed, the aircraft was no longer in straight and level flight. See again the text from the report below.

If it was roll without yaw then it would still have been in basically straight and level flight. 8.4° is a slight bank, and would have produced a very slight rate of turn - the aircraft was not in imminent danger of falling out of the sky. Certainly it was nothing that should cause a problem for professional pilots as was proved by the other flights which experienced pretty much exactly the same failure with no ill effect.

Clearly, everything worked as planned until such time as it didn't - it was then necessary for human operators to intervene. Leaving the aircraft to do it's own thing with the autopilot having disconnected in stormy weather, crossing the turbulent ITCZ, would have still resulted in the aeroplane taking an undesired flight path.

After some time, yes. But there was no need for immediate action.

As long as the automatics on transport control systems continue to decide to trip out and subsequently perform an unanticipated action or actions, or fail to continue performing the desired tasks, then automation will be limited.

It's pretty much impossible to design automation that can handle every possible situation so it will be a long time before we can eliminate a skilled professional. That said, it is important that those professionals understand fully the principles behind how the automation works and its limitations.

 

[NotATrainspott]

Sorry but the auto flight system just like every bit of automation relies on accurate data. In this case because some of the information being fed to the ADIRU was inaccurate the auto flight system was unable to establish what data was correct and what wasn't so it kicked out the auto pilot and handed control back to the human. They should have assessed the situation and worked out how to solve it, they failed to do the most basic task in aviation, fly the plane! Every pilot is taught to fly the plane first and foremost. The pilots still had their backup instruments and should have easily dealt with the situation. If anything the whole thing shows that automation is very very fallible if it isn't fed accurate data. That's why when people's lives are at stake you need human perception there to take over when the data fed to the automation is inaccurate.

You don't understand my point. The plane systems have far, far more data at their disposal than could ever be presented to a human, let alone understood and acted upon in an unexpected emergency situation where instincts take over. Even when the pitot tubes failed, the plane still had its gyroscopes and GPS to give it enough understanding of where it was and what it was doing. Providing this information to pilots at all times would make their jobs harder and could lead to more incidents due to confusion and information overload. Presenting this info only when the plane realises that something is wrong won't work either, because in these situations pilots need even more clarity. The only thing that you could do would be to work out with existing AI techniques which information needs to be presented in a situation. However, at that point the AI is already in effective control of the aircraft, as the pilots would simply follow the instrument data mechanically at that point and would have to trust that the AI was doing it right. If pilots are no longer required to make high-level problem solving decisions when in the cockpit, what exactly is their role? Planes can already fly themselves when everything is normal.

Of course, this AI technology is not yet fitted to aircraft even when all the sensor readings are available. There don't appear to be any real plans to fit it either, since airlines are loath to fit expensive and power-hungry equipment which doesn't lead to higher profits (as two pilots would initially still be required anyway, so there would be no labour cost savings). However, what we are seeing fitted to aircraft for commercial and safety reasons are high-bandwidth permanent satellite internet connections. For an airline, the ability to read sensor data from aircraft when in the air is very valuable as they can then identify issues and have maintenance crews and replacement parts in place to fix it immediately upon landing, rather than needing to take the plane out of service to diagnose. These satellite systems are going to track and keep planes in contact regardless of where they are on earth, so the current problems of oceanic radar and communications blackspots will cease. In the event of an AF447 situation, the plane could immediately send out a mayday signal when things start to go wrong and begin broadcasting even more data than would normally be required. Ground control somewhere on earth would be able to act as a co-cockpit, focussing on diagnosing the underlying issue rather than having to fly the plane. Why would Airbus or Boeing want to have a staff of dozens of specialists on site 24/7 for extremely rare aircraft accidents? Most of their jobs could be done by ground-based AI using all that data, leaving only a few people, if any, needing to be on call out. The AI systems don't need to be sitting there in an Airbus/Boeing data centre wasting energy and space when they have nothing to do. Instead, upon receiving a distress call an automated system can just spin up hundreds or thousands of cloud computing instances on Amazon EC2 or suchlike. Using orders of magnitude more flight-hour information than the most experienced human pilot in history, plus huge amounts of data from real and simulated aircraft incidents (including the full text of every single aircraft accident report ever produced) such a system would be in an exceptional position to identify the root cause of flight troubles. That information can be relayed to pilots, or the ground systems could even take over control themselves. Elon Musk's SpaceX is planning to cover the globe in low-latency gigabit satellite internet, so the control latency between ground control and a plane will be comparable to standard human reflexes. Unless you're about to hit the ground, a second of latency isn't likely to be the root cause of an accident anyway.

At this point, why would you need a human pilot again? You can have fancy AI systems on board, and then if something does goes wrong you can have even more AI processing power available on the ground. Which airline would dare remove the pilot first? A cargo airline. If you have no humans on board, then there is no need for a human cost in the unlikely event of an accident. Rather than having human pilots act with futility as they try to land a striken cargo plane near human habitation, a damaged autonomous cargo plane could simply be written off when in the air and commanded to crash in a convenient field or body of water. The economic incentive for cargo airlines to remove their pilots is bigger than just labour costs though. Automated systems could run through flight checklists in a fraction of the time - they're only there to ensure that humans check every possible thing before take-off - meaning more time in the air making money and less on the ground taking up valuable tarmac. Additionally, without the need for a cockpit at the front your cargo plane can have a full-size front-loading door without the need for an unaerodynamic 747/Beluga-style hump. An autonomous 777 with a nosecose that could lift up to expose the full fuselage width and height would be one that you could unload and load in seconds by simply the entire contents in and out in one go, rather than having to take loads out one-by-one through a side cargo door. Moving stuff around the world is big business with tiny, tiny margins and that auto-777 would simply crush the non-autonomous competition.

So 25,000 train drivers will suddenly be qualified to fix robots, but whart about the people already fixing robots?

If you replace as many humans as possible with machines what do the humans do for work?
In a factory 1 machine replaces (lets say) 20 humans and needs 1 human to look after it, what happens to the other 19 humans?

Anyway I expect to be replace by a ZX81 next month so I am learning the phrases I will need for my new job-

Do you want fries with that mate
Do you want fries with that love

There learnt them already.

This isnt being derogatory to people who work in fast food establishments but merely pointing out that my skills (train driver) dont really transfer over to any other job, and at my age (I aint seeing 50 again) there isnt much chance of me getting a job anywhere, especially in a fast food establishment simply because there are a lot of younger people after the same jobs!

So if you want to spend £billions making me and my sort redundant (how much would it also cost in benefits to keep us all at home?) then fine but by the time you weigh up all the costs of getting rid of us is there any actually real world savings to be had?

Then there is the question of obsolescence, how long can you keep your new fangled auto train set running before it becomes obsolete and needs replacing, so yet more (ongoing) costs, where-as the redundant humans do tend to be quite proficient at learning new things during a phased changeover, something a replacement fully auto system probably wouldnt be able to cope with.

Would you be happy if the whole system shut down for a month every 15 to 20 years so the whole auto system could be replaced with new kit?

Unless of course you make the new stuff backward compatible and do a phased replacement, I mean that always works with simple signalling so what could possibly go wrong in something as complicated as full auto train operation?

You won't be able to get a job fixing robots or flipping burgers or doing pretty much anything. A significant proportion, most likely the majority, of the working-age population will be simply unemployable. Even soft service jobs like hotel service will disappear. If you're an employer, your human costs are far more than just the cost of wages. You're liable for all of the things that could happen to your employees in the course of their work, as well as all of the things that your employees may do. No robot is going to sue for compensation in the event of an industrial accident or cause a customer to sue for discrimination or other costly reasons. Meanwhile, all of the people left under- and unemployed will have less buying power, so they'll have to choose the most affordable option when buying goods and services, which will universally be the option with the most automation. Automation is going to cause a destructive spiral of despair for basically everyone except the top levels of software engineering.

 

[455driver]

Automation is going to cause a destructive spiral of despair for basically everyone except the top levels of software engineering.

And Civil Servants, they will ensure they are okay, eff you Jack I am OK springs to mind.

Was it the Labour Party who, trying to keep to one of their manifesto promises, started inventing Civil Service non jobs to try and lower unemployment numbers, without realising how expensive it would actually be.

But now these staff are laughing all the way to the (pension) bank.

 

[NotATrainspott]

And Civil Servants, they will ensure they are okay, eff you Jack I am OK springs to mind.

Was it the Labour Party who, trying to keep to one of their manifesto promises, started inventing Civil Service non jobs to try and lower unemployment numbers, without realising how expensive it would actually be.

But now these staff are laughing all the way to the (pension) bank.

Not necessarily. The only solution to the automation spiral of despair is the Universal Basic Income, but that policy in itself represents a massive shift towards automation. Instead of having legions of benefits assessors and the other assorted back-room staff that the DWP need to do anything, the welfare system will consist of a computer system running through a list of citizens and sending out BACS payments to each on a set schedule.

HMRC won't have any staff either when automation makes all existing forms of revenue collection obsolete. The only form of tax that can survive the automation revolution is land value taxation, as all others can and will be evaded or become irrelevant due to technological change. Half of HMRC revenues come from income tax and NI, of which both will be pointless when the robots make everyone unemployed. LVT management consists of a GIS database of all land ownership in the country and a list of all of the payments received from landowners. No one can evade it, as ultimately the computer system can just produce a list of non-payers which can then be handed to the courts to have their land ownership rights revoked without compensation. Calculation of land values can be done by interpolating from surrounding land valuations, using AI to work out that land values increase by X amount when such and such an amenity is available and go down by Y amount when there's a problem e.g. soft soil or a landfill nearby.

 

[najaB]

Half of HMRC revenues come from income tax and NI, of which both will be pointless when the robots make everyone unemployed.

That is a very dystopian view of the future. I don't subscribe to the same vision for the future, if for no other reason than fertility rates will continue the downwards trend of the last 50 years meaning that there will be fewer and fewer working-aged people.

 

[NotATrainspott]

That is a very dystopian view of the future. I don't subscribe to the same vision for the future, if for no other reason than fertility rates will continue the downwards trend of the last 50 years meaning that there will be fewer and fewer working-aged people.

The amount of work to be done will be dropping faster than the supply of working-age people.

The income tax system will collapse well before the end of employment. Today it relies upon normal people and their employers not finding it worthwhile to either reduce their working hours or find some way to evade their tax. When people start to drop out of economic work, the shortfall will need to be addressed somehow. This is especially important if you want to have a UBI good enough for people to actually benefit from the post-work society. When you increase the marginal tax rate, you make it more worthwhile to evade and less worthwhile to work in the first place. While you might not have millions of people giving up work immediately, all you need is a gradual reduction in the number of contributors for the effect to snowball and bring the whole system down. That wouldn't be a bad thing anyway, since income taxes are inherently awful and inferior in every possible way to land value tax. Only the poor and the middle class pay income tax; the rich get their income from returns on investments, not selling their own labour. This absurdity has been able to survive up until now but the technological tsunami will end it, just like it will end so many current aspects of our societal expectations.

 

[LAX54]

A driver doesn't necessarily either.

FWIW at Manchester Airport there are "trip lines" above the track which are intended to cause all signals to go to red in the event of an aircraft fouling them. Something like that could deal with the tree issue - line broken? All trains stop. It could cause disruption, but then you could allow for DLR-style manual driving by the "guard" on sight at low speed to avoid the system stopping entirely.
--- old post above --- --- new post below ---


This is true. HS2, OTOH, is new-build - is there any reason that should not be fully automated from day one?



I'd expect to see DLR-style operation, i.e. "guard-only operation", in time. The ability to do this would revolutionise the cost base of branch lines while retaining the customer service role that people like to have and lose on DOO.

If it's possible to make a self-driving car, it's possible to make a self-driving train that would by definition be safer than a self-driving car and probably just as safe as a manually driven train. It's inevitable - there are far fewer variables with a train, even a branch-line train on Victorian infrastructure.

Anyone for giving it a go with the Stourbridge shuttle, perhaps?

No, because there will not be any such ting as a guard by then...they are trying to make them obsolete !

As for trip wires, fine by an airport for a couple hundred yards, not so good mile after mile after mile !

 

[LAX54]

How so? Can you give a situation where an AI system would not perform as well as a person? I can't.

Is it a little robot that climbs out of the cab to check the wheels when a train has been reported with dragging brakes or 'smoke' seen from the train ?

Will it run at reduced speed when one of the horns fail ? If there is a light out on a TSR will it report it ?

Not sure how it would examine the line, The field of vision for the front facing camera would have to be so wide, it would pick up everthing, now where a real person can sort out the bad from the good, a computer based system, its clear or its not, there are no maybes,

 

[455driver]

Is it a little robot that climbs out of the cab to check the wheels when a train has been reported with dragging brakes or 'smoke' seen from the train ?

Will it run at reduced speed when one of the horns fail ? If there is a light out on a TSR will it report it ?

Not sure how it would examine the line, The field of vision for the front facing camera would have to be so wide, it would pick up everthing, now where a real person can sort out the bad from the good, a computer based system, its clear or its not, there are no maybes,

Will it be able to untangle the pushbike or shopping trolley from the front bogie?

 

[LAX54]

I would imagine everything fails safe if the 5G network goes down, which means stopping. Which means chaos.

We are still going to be people, with legs and arms who will seek to find ways to keep moving. So what happens then? People deciding to jump off trains (who will be there to stop them?) and people walking along roads (again, who will stop them)?

All real issues that technology can be aware of, but can't necessarily cope with in a satisfactory way. Well, unless we change laws so if you get near to a driverless car/train and it hits you, tough luck. Wait until there's a chance for a driverless vehicle to come sweep up the body and things carry on.

or we can have a Driver on the train, and keep going even if the radio system 'blips' lol

 

[LAX54]

Will it be able to untangle the pushbike or shopping trolley from the front bogie?

No it would stop, collect said shopping trolley, identify the supermarket and electronically advise said supermarket where they can collect their trolley.

 

[LAX54]

There is more and more automation on the Railways now, Computer controlled this, Computer controlled that, but at the same time, there seems to be more and more major job stopping failures than ever I have known before, there has to be a connection.

 

[najaB]

The amount of work to be done will be dropping faster than the supply of working-age people.

In some countries in Europe the supply of working-age people is already lagging well behind current demand.

 

[455driver]

In some countries in Europe the supply of working-age people is already lagging well behind current demand.

So why are so many of them trying to get over here?

 

[miami]

There is more and more automation on the Railways now, Computer controlled this, Computer controlled that, but at the same time, there seems to be more and more major job stopping failures than ever I have known before, there has to be a connection.

Perhaps it's just that the railway is more intensively used than ever, and automation and technology allows us to run more trains at a lower cost.

100 years ago a small branch line terminus would have half a dozen full time members of staff. How many people work at Looe now?

The arguments about "what will people do" existed then, just as they existed in the industrial revolution. 90% of us don't work to produce food nowadays, that's a good thing.

I hope NotATrainspott's utopian view of a future where nobody is forced to work comes about, I'm not sure society will allow it

 

[najaB]

So why are so many of them trying to get over here?

Stop reading the Daily Mail, there isn't a horde at the gates.

People travel to find higher-paying jobs, it's always been the case.

 

[miami]

Stop reading the Daily Mail, there isn't a horde at the gates.

People travel to find higher-paying jobs, it's always been the case.

Well before my time, but I believe a chap called Norman Tebbit told people to do just that. Indeed many British builders went to Germany, culminating in a sitcom.

"He believed they should move from areas experiencing severe economic decline - such as the North East and parts of Cumbria - to places where work could be found, or found more easily."

If you go back far enough most people didn't have the chance to travel to extract more wealth from their capabilities. Middlemen, called 'merchants', cropped up instead.

Nowadays the movement of people is stronger than ever. A large number of people from my university days have settled in Australia and New Zealand. Many more colleagues have emigrated elsewhere in Europe - Germany, and Netherlands being popular.

Each to his own, as the saying goes. Go somewhere that meets what you want from life.

 

[NotATrainspott]

Is it a little robot that climbs out of the cab to check the wheels when a train has been reported with dragging brakes or 'smoke' seen from the train ?

Will it run at reduced speed when one of the horns fail ? If there is a light out on a TSR will it report it ?

Not sure how it would examine the line, The field of vision for the front facing camera would have to be so wide, it would pick up everthing, now where a real person can sort out the bad from the good, a computer based system, its clear or its not, there are no maybes,

Every single part of the train will have a sensor reading its performance. Using constant readings it will be able to identify problems well before they become apparent to a human, so things like dragging brakes will simply cease to happen in the way that they do today. New fleets of trains, like the Elizabeth Line 345s, are already being fitted with such sensors so that the train owner/operator can minimise expensive downtime. The cost of a service problem is so high that the minimal cost of fitting absurd numbers of sensors is tiny in comparison, especially when scale in numbers makes them cheaper anyway.

Autonomous cars already exist and can do just as good a job of driving a two-dimensional car in an even less predictable environment as a human can. Then, you can use the absurdly low cost of sensors, wireless networks and solar power + storage to cover the railway network in sensing equipment that will alert trains and the infrastructure operator well before a human would see a problem. If you need a track inspection, then you can fly out a drone immediately rather than having to wait for a human inspection team to get on site.

Perhaps it's just that the railway is more intensively used than ever, and automation and technology allows us to run more trains at a lower cost.

100 years ago a small branch line terminus would have half a dozen full time members of staff. How many people work at Looe now?

The arguments about "what will people do" existed then, just as they existed in the industrial revolution. 90% of us don't work to produce food nowadays, that's a good thing.

I hope NotATrainspott's utopian view of a future where nobody is forced to work comes about, I'm not sure society will allow it

Society won't have a choice because it is driven by the free market, and ultimately the free market will always win. If a country tries to hold back progress to keep old things in place they're going to be overtaken by other countries that don't.

 

[Fincra5]

Every single part of the train will have a sensor reading its performance. Using constant readings it will be able to identify problems well before they become apparent to a human, so things like dragging brakes will simply cease to happen in the way that they do today. New fleets of trains, like the Elizabeth Line 345s, are already being fitted with such sensors so that the train owner/operator can minimise expensive downtime. The cost of a service problem is so high that the minimal cost of fitting absurd numbers of sensors is tiny in comparison, especially when scale in numbers makes them cheaper anyway.

Autonomous cars already exist and can do just as good a job of driving a two-dimensional car in an even less predictable environment as a human can. Then, you can use the absurdly low cost of sensors, wireless networks and solar power + storage to cover the railway network in sensing equipment that will alert trains and the infrastructure operator well before a human would see a problem. If you need a track inspection, then you can fly out a drone immediately rather than having to wait for a human inspection team to get on site.

I dread to think of the pure cost of fitting the entire Brighton Mainline, let alone network with sensors to detect everything... What if the sensors fail? The trains all grind to halt? Then what?

I see so much talk of senors...:roll: but how do they work with footcrossings, P-way staff and so on... There's too many variables to make this system work outside closed areas for now. Hence why it's only used on certain areas (mostly tunnels)..

Maybe it will come in on metro areas to increase train running but they already run fairly close. I don't believe Human Drivers are "that" much slower than AI.

Lets look at it realistically... Network Rail hardly have the finance to fix rail issues, such as kinks or rough rides - just slap a ESR on it for a while.
I can't see Driverless trains on the mainlines anytime soon.

 

[NotATrainspott]

I dread to think of the pure cost of fitting the entire Brighton Mainline, let alone network with sensors to detect everything... What if the sensors fail? The trains all grind to halt? Then what?

I see so much talk of senors...:roll: but how do they work with footcrossings, P-way staff and so on... There's too many variables to make this system work outside closed areas for now. Hence why it's only used on certain areas (mostly tunnels)..

Maybe it will come in on metro areas to increase train running but they already run fairly close. I don't believe Human Drivers are "that" much slower than AI.

Lets look at it realistically... Network Rail hardly have the finance to fix rail issues, such as kinks or rough rides - just slap a ESR on it for a while.
I can't see Driverless trains on the mainlines anytime soon.

You don't get it though. The cost of fitting these sensors is falling through the floor at the same time as the cost of making them. A level crossing CCTV camera, for instance, that was nothing but a post driven into the ground with a battery, solar panel and wireless antenna included would cost a fraction of a traditional installation with the need to run cables along the trackside. If the CCTV post is a self-contained unit then a single piling train could fit dozens a night, as it's far easier to install than an OHLE stanchion.

Who doesn't own a smartphone now? Companies have a huge incentive to equip their staff with smart devices. NR would just give a rugged smartphone (bought off the shelf) to each and every permanent way worker that they had to keep on them at all times. This keeps them in contact with each other and the maintenance crews, which is obviously a good thing, and with the GPS location it can then add them to a centralised map of the railway network. Trains would know about every permanent way worker on their route before they even set off, while the permanent way workers would have the information they need about whether the electrification is live or a track is in use. Literally all that I have described could be written as a smartphone application and deployed today.

NR not having the finances to do things is exactly why all of these things will happen, as they will make the railway much cheaper to run. That mass sensor deployment would pay for itself before too long when you consider the huge sums of money involved in infrastructure problems.

 

[Dave1987]

You don't understand my point. The plane systems have far, far more data at their disposal than could ever be presented to a human, let alone understood and acted upon in an unexpected emergency situation where instincts take over. Even when the pitot tubes failed, the plane still had its gyroscopes and GPS to give it enough understanding of where it was and what it was doing. Providing this information to pilots at all times would make their jobs harder and could lead to more incidents due to confusion and information overload. Presenting this info only when the plane realises that something is wrong won't work either, because in these situations pilots need even more clarity. The only thing that you could do would be to work out with existing AI techniques which information needs to be presented in a situation. However, at that point the AI is already in effective control of the aircraft, as the pilots would simply follow the instrument data mechanically at that point and would have to trust that the AI was doing it right. If pilots are no longer required to make high-level problem solving decisions when in the cockpit, what exactly is their role? Planes can already fly themselves when everything is normal.

Of course, this AI technology is not yet fitted to aircraft even when all the sensor readings are available. There don't appear to be any real plans to fit it either, since airlines are loath to fit expensive and power-hungry equipment which doesn't lead to higher profits (as two pilots would initially still be required anyway, so there would be no labour cost savings). However, what we are seeing fitted to aircraft for commercial and safety reasons are high-bandwidth permanent satellite internet connections. For an airline, the ability to read sensor data from aircraft when in the air is very valuable as they can then identify issues and have maintenance crews and replacement parts in place to fix it immediately upon landing, rather than needing to take the plane out of service to diagnose. These satellite systems are going to track and keep planes in contact regardless of where they are on earth, so the current problems of oceanic radar and communications blackspots will cease. In the event of an AF447 situation, the plane could immediately send out a mayday signal when things start to go wrong and begin broadcasting even more data than would normally be required. Ground control somewhere on earth would be able to act as a co-cockpit, focussing on diagnosing the underlying issue rather than having to fly the plane. Why would Airbus or Boeing want to have a staff of dozens of specialists on site 24/7 for extremely rare aircraft accidents? Most of their jobs could be done by ground-based AI using all that data, leaving only a few people, if any, needing to be on call out. The AI systems don't need to be sitting there in an Airbus/Boeing data centre wasting energy and space when they have nothing to do. Instead, upon receiving a distress call an automated system can just spin up hundreds or thousands of cloud computing instances on Amazon EC2 or suchlike. Using orders of magnitude more flight-hour information than the most experienced human pilot in history, plus huge amounts of data from real and simulated aircraft incidents (including the full text of every single aircraft accident report ever produced) such a system would be in an exceptional position to identify the root cause of flight troubles. That information can be relayed to pilots, or the ground systems could even take over control themselves. Elon Musk's SpaceX is planning to cover the globe in low-latency gigabit satellite internet, so the control latency between ground control and a plane will be comparable to standard human reflexes. Unless you're about to hit the ground, a second of latency isn't likely to be the root cause of an accident anyway.

At this point, why would you need a human pilot again? You can have fancy AI systems on board, and then if something does goes wrong you can have even more AI processing power available on the ground. Which airline would dare remove the pilot first? A cargo airline. If you have no humans on board, then there is no need for a human cost in the unlikely event of an accident. Rather than having human pilots act with futility as they try to land a striken cargo plane near human habitation, a damaged autonomous cargo plane could simply be written off when in the air and commanded to crash in a convenient field or body of water. The economic incentive for cargo airlines to remove their pilots is bigger than just labour costs though. Automated systems could run through flight checklists in a fraction of the time - they're only there to ensure that humans check every possible thing before take-off - meaning more time in the air making money and less on the ground taking up valuable tarmac. Additionally, without the need for a cockpit at the front your cargo plane can have a full-size front-loading door without the need for an unaerodynamic 747/Beluga-style hump. An autonomous 777 with a nosecose that could lift up to expose the full fuselage width and height would be one that you could unload and load in seconds by simply the entire contents in and out in one go, rather than having to take loads out one-by-one through a side cargo door. Moving stuff around the world is big business with tiny, tiny margins and that auto-777 would simply crush the non-autonomous competition.

Reading your response tells me you really don't understand the way auto flight systems work on an aircraft. The ADIRU needs reliable airspeed data to provide flight director guidance to the auto pilot. In the air the RNP of the navigation systems in up to 2 miles. The ANP is usually around 0.3 miles. The airspeed data is crucial which is why there are three of them on commercial airliners. On the PFD and the NF the pilots are given all the information they need to fly the plane constantly. I can assure you that planes are not flying themselves when everything is normal, there is still a lot of data required to be input by the crew. They are still required to double check all waypoints and altitude constraints input into the flight management systems. And at busy airports planes are vectored into sequence by ATC. There is still no radar coverage over the Atlantic and all communications are done via HF, it is usually done via CPDLC now but it is still a case of the aircraft having to report its position regularly to ATC. Seems to me that you are desperately trying to believe that a job can be automated just for the sake of automating it without actually understanding the problems involved.

 

[Dave1987]

You don't get it though. The cost of fitting these sensors is falling through the floor at the same time as the cost of making them. A level crossing CCTV camera, for instance, that was nothing but a post driven into the ground with a battery, solar panel and wireless antenna included would cost a fraction of a traditional installation with the need to run cables along the trackside. If the CCTV post is a self-contained unit then a single piling train could fit dozens a night, as it's far easier to install than an OHLE stanchion.

Who doesn't own a smartphone now? Companies have a huge incentive to equip their staff with smart devices. NR would just give a rugged smartphone (bought off the shelf) to each and every permanent way worker that they had to keep on them at all times. This keeps them in contact with each other and the maintenance crews, which is obviously a good thing, and with the GPS location it can then add them to a centralised map of the railway network. Trains would know about every permanent way worker on their route before they even set off, while the permanent way workers would have the information they need about whether the electrification is live or a track is in use. Literally all that I have described could be written as a smartphone application and deployed today.

NR not having the finances to do things is exactly why all of these things will happen, as they will make the railway much cheaper to run. That mass sensor deployment would pay for itself before too long when you consider the huge sums of money involved in infrastructure problems.

Yes I've heard this whole thing of sensors plastered absolutely everywhere. Everyone knows how reliable sensors are in the real world.