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

[najaB]

If it determined that the risk of flying over a populated area was too high and that ditching was a less bad option, then it would ditch.

Again, you are ignoring the important point - Sully had information that the computer would not have had. He could see that the water state was calm and that there were potential rescue boats nearby.

Here's another one for you - Gimli Glider. Would an AI know about a closed airfield?

 

[najaB]

Of course, because the folks at NASA haven't got the slightest clue what they're on about

NASA's job is to devise the technology of the aircraft of 20 years plus into the future. Nobody is saying that we will never have fully autonomous commercial aviation, but it's not on the short-term horizon.

By the way, you keep saying things along the lines of (paraphrasing) 'We'll start with cargo aircraft, it doesn't matter that much if they crash' - tell that to the families of the passengers on the airliner that it crashes into, or in the houses it crashes onto.

 

[NotATrainspott]

Again, you are ignoring the important point - Sully had information that the computer would not have had. He could see that the water state was calm and that there were potential rescue boats nearby.

Here's another one for you - Gimli Glider. Would an AI know about a closed airfield?

Yes! Of course it would! Why wouldn't it? Why wouldn't the AI be outfitted with a complete geographic map using which it could identify convenient fields to land in too? These are fundamentally trivial problems. We have satellite images of the entire globe and there are companies planning to use their micro-satellites to photograph the entire surface of the earth every day at medium resolution. A runway is a mind-blowingly trivial thing for an AI to identify in an aerial photo even if it had managed to escape human attention somehow. After identifying a possible runway at a certain set of co-ordinates you just reverse geocode those into a meaningful address, or indeed search for human-generated content (like all those geo-tagged photos available on the internet). Of course, it wouldn't be the onboard AI that would trawl through people's Facebook profiles looking for explanations of this daft runway-looking thing ahead, but that ground-based AI system which can share its information.

This is what I mean about the exponential growth in technological capability. All you need is some bright spark to connect two individual ideas together and boom, you've created something far more valuable than the sum of its parts. Google is already doing the reverse geocode/photo detection stuff as part of its plan to map the entire world. Why couldn't they also take in all the open data from the buoys in waterways across the world to see where the water is calm too?

 

[najaB]

Why couldn't they also take in all the open data from the buoys in waterways across the world to see where the water is calm too?

I'm sorry, but I'm out. You've gone too far now.

 

[NotATrainspott]

NASA's job is to devise the technology of the aircraft of 20 years plus into the future. Nobody is saying that we will never have fully autonomous commercial aviation, but it's not on the short-term horizon.

By the way, you keep saying things along the lines of (paraphrasing) 'We'll start with cargo aircraft, it doesn't matter that much if they crash' - tell that to the families of the passengers on the airliner that it crashes into, or in the houses it crashes onto.

20 years is a heartbeat! A new pilot fresh out of school would be only halfway through their career before this happens, when we've had generations of pilots before them.

You seem to have neglected the second part of that idea, which is that you can tell the autonomous cargo plane to go and deliberately crash into the vast majority of the planet where people don't live. Human habitation covers a minuscule proportion of planet earth. Human pilots will not sacrifice their lives if the risk of not making it back to the runway is too low to pass a cost/benefit analysis.

 

[NotATrainspott]

I'm sorry, but I'm out. You've gone too far now.

Why is that absurd? That data is really useful for shipping and leisure craft operators. It's just data, which is what the information age is all built on. There is no plausible reason why you wouldn't have this information somewhere accessed in a geo-referenced way. That the data might never have been intended for use in working out whether to ditch an aircraft is irrelevant.

 

[Domh245]

You seem to have neglected the second part of that idea, which is that you can tell the autonomous cargo plane to go and deliberately crash into the vast majority of the planet where people don't live. Human habitation covers a minuscule proportion of planet earth. Human pilots will not sacrifice their lives if the risk of not making it back to the runway is too low to pass a cost/benefit analysis.

Particularly handy if you are a cargo plane taking off from (for example) Memphis airport - 2nd busiest airport by cargo - and suffer a failure. Just crash where there aren't people

 

[NotATrainspott]

Particularly handy if you are a cargo plane taking off from (for example) Memphis airport - 2nd busiest airport by cargo - and suffer a failure. Just crash where there aren't people

You only choose to crash if the risk of not making it back to the runway is too high! Instead of trying to loop around and land again, spending more time over populated areas, you just head straight out to the nearest unpopulated area and crash there. If you crash before that, then you would have crashed anyway regardless of what you tried to do, so it's really not the fault of AI.

Of course, it's also worth remembering that AI flight systems can be based at airports which are not in the middle of built-up areas. Heavily automated systems don't need a supply of workers, who currently make it hard to have airports out in the middle of nowhere.

 

[najaB]

Why is that absurd? That data is really useful for shipping and leisure craft operators.

You're making too many assumptions. You're assuming that the data is being produced in the first place. You are assuming that it is made available. You're assuming that it will fine-grained enough to be useful. You're assuming that it will be updated often enough to be up to date. You're assuming that there's an infinite amount of bandwidth available all the time. You're assuming that it can be integrated into the flight-control algorithms.

 

[NotATrainspott]

You're making too many assumptions. You're assuming that the data is being produced in the first place. You are assuming that it is made available. You're assuming that it will fine-grained enough to be useful. You're assuming that it will be updated often enough to be up to date. You're assuming that there's an infinite amount of bandwidth available all the time. You're assuming that it can be integrated into the flight-control algorithms.

I mean, it would be totally absurd for me to be able to go to a webpage and tell you that there is currently a wind gust of 17.0 knots at a bearing of 310 degrees in New York Bay.

Going from hourly data intervals to minute or second-based intervals is a trivial matter when you have omnipresent gigabit 4/5G networks and infinite renewable power from wave energy. A company can start churning out data buoys for a fraction of what they cost today because they'll be producing millions of them in highly-automated factories.

I don't say that water temperatures would be included in the flight control systems of aircraft. I say that the entire flight control AI system, including the ground-based backup and enhancement systems, will be able to draw upon this data when it makes decisions about what to do in exceptional circumstances. The software doesn't exist today but as I have just demonstrated, all you need to do is connect two unrelated data sources together by their geographic location and boom, you've got more value than the two on their own.

This is the information revolution, and it is glorious.

 

[najaB]

I mean, it would be totally absurd for me to be able to go to a webpage and tell you that there is currently a wind gust of 17.0 knots at a bearing of 310 degrees in New York Bay.

And what does that make the wind and wave conditions on the Hudson? (Hence my reference to the data being fine-grained enough to be of use.)

 

[NotATrainspott]

And what does that make the wind and wave conditions on the Hudson? (Hence my reference to the data being fine-grained enough to be of use.)

Why is the idea of having internet-connected buoys on the Hudson River so absurd if they already exist a few kilometres downstream? Having that data will be useful somehow to someone, and it'll cost less and less to get that data as time goes on. Of course the data isn't available right here, right now, but I'm not talking about tomorrow. I'm talking about the future, but a future which will happen much, much sooner than most people think.

It's really quite difficult to wrap your head around this sort of thing. Basically anything that can be measured will be publicly accessible data. Some bright spark will then come along and combine two or more sources of data into something that people thought simply wouldn't be possible.

 

[najaB]

Of course the data isn't available right here, right now, but I'm not talking about tomorrow. I'm talking about the future, but a future which will happen much, much sooner than most people think.

And not nearly as soon as you seem to think. You're ignoring the fact that there's a cost to build these massive sensor networks and a cost to make the data available and no clear reason for anyone to spend that money.

 

[Dave1987]

NotATrainspott reading your thoughts on the future of tech kind shows me why we are having some of the problems we currently are with some of the train fleets right now. There are some fleets that the old engineering principle KISS has gone completely out of the window. You may envisage a world where everything is connected to absolutely everything else but I highly doubt that it would work in the real world or it would be incredibly unreliable. Just look at the new trains for Thameslink, even with extensive testing they are proving very unreliable in real world service. Just goes to show that something that works well on paper or in an engineers head may not work reliably in the real world. Automation is and should be used to assist the human but leave the human to make the decisions based on their training and experience. I've known new things be introduced by tech people at TOCs. The tech people have said that these new things will be fantastic etc etc yet when they are used they are terrible because they were designed by someone who doesn't do the job day in day out, and I the end people just stop using them.

 

[NotATrainspott]

And not nearly as soon as you seem to think. You're ignoring the fact that there's a cost to build these massive sensor networks and a cost to make the data available and no clear reason for anyone to spend that money.

Would Airbus and Boeing fund the cost of installing a water buoy network in New York Habour just because of the infinitesimal chance of a ditching? No. Would the environmental authorities and researchers? Yes. As I continue to point out, the cost of these sensor installations is falling through the floor and will continue to do so, making more and more marginal installations worthwhile.

NotATrainspott reading your thoughts on the future of tech kind shows me why we are having some of the problems we currently are with some of the train fleets right now. There are some fleets that the old engineering principle KISS has gone completely out of the window. You may envisage a world where everything is connected to absolutely everything else but I highly doubt that it would work in the real world or it would be incredibly unreliable. Just look at the new trains for Thameslink, even with extensive testing they are proving very unreliable in real world service. Just goes to show that something that works well on paper or in an engineers head may not work reliably in the real world. Automation is and should be used to assist the human but leave the human to make the decisions based on their training and experience. I've known new things be introduced by tech people at TOCs. The tech people have said that these new things will be fantastic etc etc yet when they are used they are terrible because they were designed by someone who doesn't do the job day in day out, and I the end people just stop using them.

Keep It Simple Stupid means that there's no point putting a full desktop-class operating system in a mobile phone. A phone just needs to make calls and send a few texts, you see. No need for all the complexity and cost of a full multi-user operating system with fancy graphics, networking, sound and input control layers.

Ten years ago, Nokia, Motorola and RIM/BlackBerry were major players in the mobile phone market. Now, they're just brands to be slapped on top of other people's technology. All of the people who went around saying that the iPhone would fail have failed themselves.

There isn't one omniscient person deciding which technologies are worthwhile and which aren't. Instead, products are released to the market and consumers decide which ones they like the most. People really liked the iPhone and didn't really like the BlackBerry, so Apple got lots of money and RIM got none. All of these complex technologies are perceived as better than the alternatives by consumers, who are ultimately the people that matter.

Why would consumer choice have anything to do with aircraft automation? Because consumers want to buy things cheaply, easily and quickly. All three of those things are easier to deliver if you've got an autonomous fleet of delivery aircraft. Amazon will have their autonomous planes before the competition, and then they'll be able to offer lower prices, so more people will buy things from Amazon, leaving other companies and pilot-included cargo airlines behind.

It's just like how low cost airlines and the Gulf carriers have been able to crush their traditional opposition. Ultimately, consumers care mostly about low fares, and any airline that has lower fares is one which they'll go with.

For the railways, if they don't keep up with these sorts of technological improvements then passengers will start to look at alternatives. Autonomous taxi and bus services will be able to compete with rail far more easily than their traditional forms. Did you know that Google knows every single place you ever go with your smartphone? If you've got an Uber app installed there's no reason why they couldn't either - if they need permission, then just offer a slight discount on fares and people will happily trade away that information. If Uber knows of lots of people heading into a city centre from their homes every morning by train it could put on an autonomous bus service that would pick them up directly from their homes and take them into work. If it's cheaper than rail, then a slight increase in journey time might not be the end of the world (especially if it removes the time needed to get to and from the station at either end). That sort of precise timetable and route planning simply hasn't been possible before because the information hasn't been available. Now, with the ubiquity of personal computing devices it very much is.

If the railways can't compete, then they will become less useful as a mode of transport and will lose out financially. If you're a London commuter TOC you really, really don't want Uber to come along with the autonomous bus and wreck your profitability, so you're going to push for anything and everything that can to get costs down. The only way to get costs to go down is to invest in technology. If you don't invest, then you lose out, it's as simple as that.

 

[najaB]

Would Airbus and Boeing fund the cost of installing a water buoy network in New York Habour just because of the infinitesimal chance of a ditching? No. Would the environmental authorities and researchers? Yes.

No, the correct answer is 'Possibly. If they have a specific reason to.' Even as the cost of sensor networks decreases, nobody is going to install and maintain one 'because they can', there still needs to be a specific justification.

In real terms science and technology research funding is staying flat at best in the developed economies, there isn't money sloshing around to fund non-targeted research investment.

 

[NotATrainspott]

No, the correct answer is 'Possibly. If they have a specific reason to.' Even as the cost of sensor networks decreases, nobody is going to install and maintain one 'because they can', there still needs to be a specific justification.

In real terms science and technology research funding is staying flat at best in the developed economies, there isn't money sloshing around to fund non-targeted research investment.

Yes, they'll install them for some specific justification. I don't dispute that at all. I'm saying that the cost of installing these sensor networks is dropping all the time, making more marginal projects worthwhile. The data from these networks can be used for a variety of different things of varying economic value. Weather forecasting, for instance, has immense economic value and can justify all sorts of sensor installations. The more sensors you have out in the field, the better your predictions are and the more valuable your activity is to the rest of the economy.

Some of these sensor installations are going to be effectively cost-free too. Everyone now has a smartphone fitted with an accelerometer and gyroscope and an internet connection. Instead of needing to fund a mass installation of seismic sensors you can now just get data from everyone's smartphones instead. Barometers are now common in smartphones, so you can plausibly get metre-precision air pressure data for free from millions or billions of devices all around the world. All of that data can then be fed into AI learning models to help predict future weather patterns.

This is the information revolution. Sensing and processing of data is now so incredibly cheap and easy that we can do things that no one had ever purposely planned. As I continue to emphasise, all you need is some bright spark in a San Francisco office to take these data points and mash them together and you get something which is extraordinarily economically valuable. New ideas can go from initial concept to market within months or weeks when previously they would have been either impossible or implausibly expensive to implement. New technologies then build on previous ones and can happen even faster because they're standing on the shoulders of giants.

 

[Fincra5]

Yes, they'll install them for some specific justification. I don't dispute that at all. I'm saying that the cost of installing these sensor networks is dropping all the time, making more marginal projects worthwhile. The data from these networks can be used for a variety of different things of varying economic value. Weather forecasting, for instance, has immense economic value and can justify all sorts of sensor installations. The more sensors you have out in the field, the better your predictions are and the more valuable your activity is to the rest of the economy.

Some of these sensor installations are going to be effectively cost-free too. Everyone now has a smartphone fitted with an accelerometer and gyroscope and an internet connection. Instead of needing to fund a mass installation of seismic sensors you can now just get data from everyone's smartphones instead. Barometers are now common in smartphones, so you can plausibly get metre-precision air pressure data for free from millions or billions of devices all around the world. All of that data can then be fed into AI learning models to help predict future weather patterns.

This is the information revolution. Sensing and processing of data is now so incredibly cheap and easy that we can do things that no one had ever purposely planned. As I continue to emphasise, all you need is some bright spark in a San Francisco office to take these data points and mash them together and you get something which is extraordinarily economically valuable. New ideas can go from initial concept to market within months or weeks when previously they would have been either impossible or implausibly expensive to implement. New technologies then build on previous ones and can happen even faster because they're standing on the shoulders of giants.

Yet there is much more to the cost than just the sensors themselves... For starters there's all the wiring and networking needed to be created. And that all depends on how many sensors are required to protect the railways and trains and passengers onboard those trains. You'll also need backups incase some fail..

So I wouldn't think it would be as economically viable as you seem to think. All that cost for actually very little gain.

 

[NotATrainspott]

Yet there is much more to the cost than just the sensors themselves... For starters there's all the wiring and networking needed to be created. And that all depends on how many sensors are required to protect the railways and trains and passengers onboard those trains. You'll also need backups incase some fail..

So I wouldn't think it would be as economically viable as you seem to think. All that cost for actually very little gain.

What wires? You use wireless radios for communication and renewable power plus batteries for energy. Soon enough you'll be able to use wireless power too for some low-power requirements. I wonder whether you could inductively power sensors just from 25kV AC fields alone...

 

[FenMan]

And yet we have hundreds of thousands of able bodied people out of work!

The hard truth is that able bodied they may be, but employers will always choose capable and suitably qualified candidates where they can. They are not charities.

The political bit is, what is to be done about it? But that discussion would be way off beam for this topic!

 

[Dave1987]

What wires? You use wireless radios for communication and renewable power plus batteries for energy. Soon enough you'll be able to use wireless power too for some low-power requirements. I wonder whether you could inductively power sensors just from 25kV AC fields alone...

So you are going to base peoples safety on wireless communications? Dear oh dear I dread the world you wish to live in.

 

[najaB]

So you are going to base peoples safety on wireless communications?

Indeed. The words 'wireless' and 'safety critical' rarely go together.

 

[Fincra5]

What wires? You use wireless radios for communication and renewable power plus batteries for energy. Soon enough you'll be able to use wireless power too for some low-power requirements. I wonder whether you could inductively power sensors just from 25kV AC fields alone...

How does the power reach the sensors from this renewable power source? Could it be wires.. not every sensor would be wireless as thats a laughable concept.. more likely a load would be bulked together to a transmitter (less connection issues that way)..

Btw most of the country isn't 25kv AC...

 

[najaB]

I wonder whether you could inductively power sensors just from 25kV AC fields alone...

I would've thought to get useable amounts of power you'd have to be getting close to the hot zone around the OLE. Which would mean that simple sensor changes might require isolation of the traction current.

 

[NSEFAN]

I would've thought to get useable amounts of power you'd have to be getting close to the hot zone around the OLE. Which would mean that simple sensor changes might require isolation of the traction current.

I guess it'd depend on how you coupled the power to the receivers. Given the frequency is only 50Hz I can't imagine using some kind of resonant receiver would be very practical. In any case, something safety-critical should be powered more reliably that a wireless link where possible.

 

[carriageline]

Give up. It seems notatrainspott will not answer posts that provide genuine obstructions to his almighty plans. I think he's another theorists


Sent from my iPhone using Tapatalk

 

[NotATrainspott]

So you are going to base peoples safety on wireless communications? Dear oh dear I dread the world you wish to live in.

Indeed. The words 'wireless' and 'safety critical' rarely go together.

GSM-R or equivalent wireless systems are about to be responsible for the very most basic of train safety systems: preventing trains from crashing into one another. Of course, they're fail-safe so that if there is a fault, trains can come to a halt without any collisions. That would be equally true for any of these wireless safety systems. Each train would still be fitted with all of the sensors it needed to scan the route ahead and stop if necessary; the trackside wireless CCTV would be to detect problems well before the train encountered them. If the wireless link to a level crossing CCTV camera failed then you would treat it just the same as when a cabled CCTV connection is lost. There's no reason why wireless should be any less reliable than a cabled system anyway, as there would be fewer cables to accidentally cut and you can have self-healing network topologies. If the sensors can form an ad-hoc network then even if you lost the long-distance connection you could still have some amount of communication.

How does the power reach the sensors from this renewable power source? Could it be wires.. not every sensor would be wireless as thats a laughable concept.. more likely a load would be bulked together to a transmitter (less connection issues that way)..

Btw most of the country isn't 25kv AC...

The CCTV poles would be self-contained units including a battery and a photovoltaic array. Smaller sensors would be powered wirelessly, possibly from these CCTV poles.

Quasistatic Cavity Resonance for Ubiquitous Wireless Power Transfer - Disney Research

This is the cutting edge of research into wireless power. Of course, this implementation is for a metal-enclosed room, but it demonstrates that wireless power is something that can actually be used to perform useful work in an environment. Systems which do not require the metal cage will be less efficient, but for low-power sensors this probably will not be a problem.

When things are wireless they are cheap and easy to install. The cost of installing the cables is enough to make the sensor system not cost-effective at scale.

I would've thought to get useable amounts of power you'd have to be getting close to the hot zone around the OLE. Which would mean that simple sensor changes might require isolation of the traction current.

I guess it'd depend on how you coupled the power to the receivers. Given the frequency is only 50Hz I can't imagine using some kind of resonant receiver would be very practical. In any case, something safety-critical should be powered more reliably that a wireless link where possible.

What I'm wondering is whether the immense power that runs through a 25kV system is enough to provide <1 watt-level power to sensors within a few metres.

These sensors do not have to be safety critical, because they currently are too impractical to be installed anyway. What these sensors would mean is that infrastructure faults could be detected well before they become a problem, thus allowing far more cost-efficient fixes and also minimal disruption caused by unforseen failures. Automation can mean not only replacing the need for human labour in some tasks but also the ability to technically surpass anything that human could plausibly do regardless of how much they would be trained or paid. This will only make the railway more efficient.

Give up. It seems notatrainspott will not answer posts that provide genuine obstructions to his almighty plans. I think he's another theorists


Sent from my iPhone using Tapatalk

No, I was busy. I love having this discussion/rant, because ultimately I will be proved absolutely correct. The railway industry is already moving in the direction that I'm describing. Just search for "railway sensor networks" on Google Scholar and you'll see exactly what I mean.

 

[LAX54]

What wires? You use wireless radios for communication and renewable power plus batteries for energy. Soon enough you'll be able to use wireless power too for some low-power requirements. I wonder whether you could inductively power sensors just from 25kV AC fields alone...

and look how reliable that is...not !

GSM-R is not reliable for a perm connection, ETRMS is still quite a few years off being used big time, even in its basic form !

What happens when you switch the OHL off ? which is most nights in many many places

Automation is no the be all and end all, the more automation the more to go wrong and affect a much larger area......Oh it already does ! lol

 

[NSEFAN]

What I'm wondering is whether the immense power that runs through a 25kV system is enough to provide <1 watt-level power to sensors within a few metres.

I know power transfer can be done in the LF-HF bands quite easily using this method with ferrites at the receiver, but I honestly don't know how feasible it would be at 50Hz.

What kind of infrastructure sensing do you have in mind for such a system, given the lower power levels which would be available.

 

[najaB]

GSM-R or equivalent wireless systems are about to be responsible for the very most basic of train safety systems: preventing trains from crashing into one another.

There's a very good reason that no Level 3 ETCS installations exist yet. It works well on paper, but in real life - well, see my signature quote.