Driverless cars could revolutionise transport and lead to the decline of Railways

[ExRes]

Of the debate about driverless cars here one thing has been missed. I believe it will actually increase congestion and could gridlock the road network in cities etc.

If we have true driverless cars (call them pods) that are fully automated and don't need a driving licence etc to be used then this is an expected scenario. I go to meet my 5 mates in the centre of town to go on the lash. We all get in our pod as no one will share a pod because why would you. An avg household would have 2.4 pods plus a family pod on the drive if there is a enough space. If pods can be used by the kids to go to meet their mates then, you have gone from the number of cars on the road being limited to age, driving licence, lack of intoxication and income to just about the whole population using a pod once every 24hrs. Imagine every kid over the age of 8 turning up at 08:30 for school in a pod.

If a good proportion are pod taxi's then each journey generates an extra journey for the pod to be used as it needs to get you and 3 journeys if the pod cannot be immediately utilised and has to go to a pod park to await its next duty.

If that scenario was ever to occur I'd have to ask how much damage would be done to the environment making all these pods, not to mention the batteries and charger points required and the electricity to power them along with the road re(building) required and the gridlock you've mentioned, I also wonder how many would have to be built to cope with everyone and their dog wanting to use one at the click of a finger and expecting to get one that second because their need is higher than everyone else

 

[Bletchleyite]

The other thing to point out is that before it even gets to testing, it'll have been extensively developed by hundreds of individuals working at their own, comfortable pace - which increases the chance of errors being picked up and resolved. Compare that to the single person, working on reflexes making their split second decisions..

Indeed. Good developers do lots of "unit testing" as they go along, and indeed design and build for testability, e.g. modularising rather than copy-pasting code. There are some bad developers out there, but if you are writing safety critical code you certainly invest in good developers, and potentially have more than one system doing the same thing from a different supplier, so if they disagree you stop.

If that scenario was ever to occur I'd have to ask how much damage would be done to the environment making all these pods, not to mention the batteries and charger points required and the electricity to power them along with the road re(building) required and the gridlock you've mentioned, I also wonder how many would have to be built to cope with everyone and their dog wanting to use one at the click of a finger and expecting to get one that second because their need is higher than everyone else

I'm unconvinced it would result in many more journeys than cars do at present. Most kids get a lift to school in the car these days.

 

[ExRes]

I'm unconvinced it would result in many more journeys than cars do at present. Most kids get a lift to school in the car these days.

Parents must be rather poor down here then, the main roads are full of buses and coaches shifting 50/70 kids at a time every morning and afternoon, replace a 50 seat coach with a dozen or so of these pods and the roads will be even more chaotic than they are now

 

[Bald Rick]

Of the debate about driverless cars here one thing has been missed. I believe it will actually increase congestion and could gridlock the road network in cities etc.

If we have true driverless cars (call them pods) that are fully automated and don't need a driving licence etc to be used then this is an expected scenario. I go to meet my 5 mates in the centre of town to go on the lash. We all get in our pod as no one will share a pod because why would you. An avg household would have 2.4 pods plus a family pod on the drive if there is a enough space. If pods can be used by the kids to go to meet their mates then, you have gone from the number of cars on the road being limited to age, driving licence, lack of intoxication and income to just about the whole population using a pod once every 24hrs. Imagine every kid over the age of 8 turning up at 08:30 for school in a pod.

If a good proportion are pod taxi's then each journey generates an extra journey for the pod to be used as it needs to get you and 3 journeys if the pod cannot be immediately utilised and has to go to a pod park to await its next duty.

If and when it happens, the whole model of car ownership will change. You will just hire by the journey, like a taxi. At busy times, there will be a clear incentive to share. At very busy times you’ll pay through the nose. I expect it will lead to a significant reduction in the number of vehicles in use, but a significant increase in the productivity of each vehicle (given that the average car in the U.K. spends something like 95% of its time parked).

But this is at least 20 years away.

As @Bletchleyite says, autonomous cars are only really a risk to the regional railway. If a car can automatically take you from (say) Glasgow to Fort William, in comfort and in 2 hours, cheaply, then why would anyone get the train? (The view is better from the A82).

 

[yorksrob]

That's always a question I've wondered. Isn't part of the benefit of motoring having your own personalised space ? If everything becomes a glorified taxi, is it the same ?

(I will ask my Dad as he has genuinely enjoyed motoring/driving for many decades).

 

[Bald Rick]

That's always a question I've wondered. Isn't part of the benefit of motoring having your own personalised space ? If everything becomes a glorified taxi, is it the same ?

(I will ask my Dad as he has genuinely enjoyed motoring/driving for many decades).

I’d say the biggest benefit is the door to door transport, at a time of your own choosing.

Having personal space is very much secondary - and in my view getting less important as technology enables short term personalisation (eg not having to worry about what CDs to take with you, just choose your own playlist from your phone).

 

[yorksrob]

I’d say the biggest benefit is the door to door transport, at a time of your own choosing.

Having personal space is very much secondary - and in my view getting less important as technology enables short term personalisation (eg not having to worry about what CDs to take with you, just choose your own playlist from your phone).

Taking off my railway buff hat, it is genuinely interesting.

With my railway buff hat on, it's no, no, no !!!

 

[biko]

Humans are absolutely useless at doing monotonous jobs which very occasionally need them to react quickly when something goes wrong though

That's why it could be argued that the intermediate phase between non-autonomous and fully autonomous is more dangerous than the present situation. Tesla's "autopilot" makes sure the task is more monotonous than before (just watching out of the window) and invites to do stuff that makes reacting slower.

Of the debate about driverless cars here one thing has been missed. I believe it will actually increase congestion and could gridlock the road network in cities etc.

If we have true driverless cars (call them pods) that are fully automated and don't need a driving licence etc to be used then this is an expected scenario. I go to meet my 5 mates in the centre of town to go on the lash. We all get in our pod as no one will share a pod because why would you. An avg household would have 2.4 pods plus a family pod on the drive if there is a enough space. If pods can be used by the kids to go to meet their mates then, you have gone from the number of cars on the road being limited to age, driving licence, lack of intoxication and income to just about the whole population using a pod once every 24hrs. Imagine every kid over the age of 8 turning up at 08:30 for school in a pod.

If a good proportion are pod taxi's then each journey generates an extra journey for the pod to be used as it needs to get you and 3 journeys if the pod cannot be immediately utilised and has to go to a pod park to await its next duty.

Driverless cars will do both actually. It will indeed encourage usage of cars and possibly reduce PT use, but it will also be able to use road space more efficiently as they can communicate with each other and take plans of other vehicles into account. That means that the gap between cars can become very small.

However, this is really something for the future. Most likely, driverless / autonomous driving will first be allowed on motorways only as those are a closed system. Cities become very complex because of the interactions with pedestrians and cyclists.

 

[PTR 444]

That's always a question I've wondered. Isn't part of the benefit of motoring having your own personalised space ? If everything becomes a glorified taxi, is it the same ?

(I will ask my Dad as he has genuinely enjoyed motoring/driving for many decades).

This is just what I was thinking. Many people see their car as an extension of the home, and hence will fill it up with personal items that can be left inside while the owner is away. Replacing car ownership with an automated taxi service will make certain trips much more of a hassle as you’ll have to lug all of your personal belongings with you during the day. You won’t be able to leave them in the car you just used when it’s highly unlikely that same car will come back to pick you up later in the day.

 

[takno]

That's always a question I've wondered. Isn't part of the benefit of motoring having your own personalised space ? If everything becomes a glorified taxi, is it the same ?

(I will ask my Dad as he has genuinely enjoyed motoring/driving for many decades).

It's hardly a "glorified" taxi if there's nobody to spot and clean up the vomit after the previous occupant. Getting home from the pub could become a bit of an ordeal.

 

[yorksrob]

This is just what I was thinking. Many people see their car as an extension of the home, and hence will fill it up with personal items that can be left inside while the owner is away. Replacing car ownership with an automated taxi service will make certain trips much more of a hassle as you’ll have to lug all of your personal belongings with you during the day. You won’t be able to leave them in the car you just used when it’s highly unlikely that same car will come back to pick you up later in the day.

Quite. As far as I understand it, the whole point of car ownership is its your own space.

It's hardly a "glorified" taxi if there's nobody to spot and clean up the vomit after the previous occupant. Getting home from the pub could become a bit of an ordeal.

That's a good point. No doubt the whole concept will have been designed for clean living Californians, like Musky Oblong (or whatever he's called) Traditional British drinking culture may pose profound challenges !

 

[birchesgreen]

The biggest problem for fully autonomous cars is the sheer unpredictability of the roads: pedestrians, cyclists, animals, pot holes. If the driver has to remain alert just in case this removes one of the key benefits plus as has already been shown with some Tesla incidents, a driver won't stay alert...

 

[The Ham]

I wouldn't trust any human to drive better than a well-tested automatic system. An automatic system does not make mistakes, it does exactly what it is programmed to do.



There is that, I suppose, though I'd say it only threatens branch lines. It would likely increase use of longer-distance and commuter routes, because you'd buy your journey as "mobility as a service" and likely not get a through car service unless you chose a very high price, more likely it'd be an electric car to the station and one to your destination at the other end.

Computer systems like the Post Office computer which lead to people being sent to prison.

OK that's an extreme example, however to ensure no faults that's going to require a lot of testing so some time away.

However therein lays the problem, if automated cars add 10% efficiency to the road network, that's no good if road use has increased be 15%

 

[edwin_m]

I’d say the biggest benefit is the door to door transport, at a time of your own choosing.

Having personal space is very much secondary - and in my view getting less important as technology enables short term personalisation (eg not having to worry about what CDs to take with you, just choose your own playlist from your phone).

If you have to share then the door to door transport will often be via someone else's door.

I've always recoiled from "shared" transport in I way I don't from "public" transport. There are obvious personal security issues, particularly as whoever is sharing the vehicle may also get to find out where you live. So I think sharing will only ever catch on in the sense of one vehicle being used sequentially by different individuals or small groups known to each other.

 

[The Ham]

Pedestrians and cyclists are best segregated from general traffic in urban areas.

Whilst that's true that isn't always practical and certainly not for the last bit through housing areas.

You'd also need to make a lot of the existing traffic signal controlled crossings as crossing points and so you'd lose some of the capacity enhancements, as pedestrians wouldn't want to "walk with traffic" (i.e. cross to a traffic Island when one arm is red and cross the next bit when traffic is stopped from using it) and so you'd likely need all stop for motor traffic.

 

[Bletchleyite]

If you have to share then the door to door transport will often be via someone else's door.

I've always recoiled from "shared" transport in I way I don't from "public" transport. There are obvious personal security issues, particularly as whoever is sharing the vehicle may also get to find out where you live. So I think sharing will only ever catch on in the sense of one vehicle being used sequentially by different individuals or small groups known to each other.

It's notable that Uber Pool, where it operates (I suspect it's suspended for COVID), only combines 2 peoples' journeys, with the implication that one will sit in the front and one in the back so they are decently separated.

 

[big_rig]

The 'debate' around driverless cars is a very simple one. Every year driverless car advocates say 'driverless cars are just around the corner, in the year 201x (202x now) we will see big progress towards them.' All they need to do is change the date when the promised land doesn't come Quite convenient for the cottage industry of consultants 'specialising' in this 'field' as they can re-use the same slides for years on end.

 

[172007]

The concept of cars talking autonomously to each other to plan journeys with least conflicts etc has one very obvious snag once you study the "Travelling Salesman problem". You would probably need a Quantum computer(s) to work it all out. There would be so many permutations and individual vehicles to co-ordinate for say a simple journey from Fulham to Chelsea to require such astronomical computer and algorithm power that a block chain would look as simple as 2+2.

 

[Bald Rick]

The concept of cars talking autonomously to each other to plan journeys with least conflicts etc has one very obvious snag once you study the "Travelling Salesman problem". You would probably need a Quantum computer(s) to work it all out. There would be so many permutations and individual vehicles to co-ordinate for say a simple journey from Fulham to Chelsea to require such astronomical computer and algorithm power that a block chain would look as simple as 2+2.

But that isn’t proposed.

 

[Bletchleyite]

The concept of cars talking autonomously to each other to plan journeys with least conflicts etc has one very obvious snag once you study the "Travelling Salesman problem". You would probably need a Quantum computer(s) to work it all out. There would be so many permutations and individual vehicles to co-ordinate for say a simple journey from Fulham to Chelsea to require such astronomical computer and algorithm power that a block chain would look as simple as 2+2.

The idea of cars communicating with each other isn't about journey planning, that's better done with the likes of "Trafficmaster data" as it is now, plus "crowdsourcing" that data as the likes of Waze does. Cars communicating with one another would be more for collision avoidance and similar, just like TCAS (Traffic Collision Avoidance System, I think) communicates between two aircraft to avoid a mid-air collision by telling one set of pilots to descend and one set to climb.

 

[Domh245]

The idea of cars communicating with each other isn't about journey planning, that's better done with the likes of "Trafficmaster data" as it is now, plus "crowdsourcing" that data as the likes of Waze does. Cars communicating with one another would be more for collision avoidance and similar, just like TCAS (Traffic Collision Avoidance System, I think) communicates between two aircraft to avoid a mid-air collision by telling one set of pilots to descend and one set to climb.

Indeed. The best way to think of it perhaps is like birds flocking - as a rough approximation each individual bird only considers it's separation, alignment and cohesion relative to other birds in the flock and adjusts it's own trajectory accordingly - there isn't a central guiding mind for the flock, only hundreds of fairly dumb ones acting with common rules. We don't need a central computer that calculates and assigns paths, just for each car to be able to tell it's neighbours what it's doing, and for others to respond accordingly. In that sense it's not dissimilar to what cars currently (notionally!) do with flashing lights at each corner of the vehicle - but with far more, and precise, data than "I'm slowing" or "I'm changing direction"

 

[lkpridgeon]

I'm sorry but I'm having a hard time seeing driver-less cars actually coming into fruition as someone whom works in software engineering (more specifically deals a lot with AI). From what I've seen it's a long way off and we can barely get simple systems such as AI based loan approval software working without issues let alone a car carrying passengers without continuous human supervision. As much as I love the idea, the technology doesn't look like it'll mature enough for a long while yet... I see more chance we'll have widespread ATO (a much easier thing to do) heck we might even have full (ish) electrification before we see driver-less cars being both reliable and affordable.

What's needed for driver-less cars is commercially available highly accurate positioning (work in progress), highly accurate maps that are kept up to date (still a rather large challenge despite our best efforts), energy efficient yet powerful hardware for the AI to run on (we're about to hit the limit of what silicon can achieve due to the affect of quantum tunnelling of electrons), and many more technical issues. I also personally see there being issues popping up over safety, security, insurance and regulatory the latter will take a while to solve as governments aren't exactly renowned for moving fast or keeping up with technological developments.

Indeed. Good developers do lots of "unit testing" as they go along, and indeed design and build for testability, e.g. modularising rather than copy-pasting code. There are some bad developers out there, but if you are writing safety critical code you certainly invest in good developers, and potentially have more than one system doing the same thing from a different supplier, so if they disagree you stop.

And even with good coverage of unit tests many an issue still crops up and yes modularity is a good aim. However despite the best intentions of some very good developers this doesn't always plan out the way you'd expect as deliverables need to be met at the end of the day hook, line and sinker. After-all if the company doesn't ship anything, the company can't make money to pay ones wage.

Ideally there needs to be an open consortium dedicated to this issue to do as you've suggested, multiple modules that output in a standardised format from various suppliers that can form part of the overall system allowing for redundancy. However as stated up above we still have the issue of hardware constraints due the amount of localised processing needed for AI as edge and peer to peer computing will only get us so far.

We can't outsource all processing to a centralised server due to latency issues, even if we were to co-locate the server closer to the edge the latency would still be too great for a lot of safety critical functions. Heck kernel latency is a major issue in and of itself to deal with.

Maybe I'm far too jaded when it comes to technology and the software industry in general. However if the railways do decline I'll quite happily move my gaze to other complex systems as I don't really care all that much about the thing that runs on top of the rails. I'm in it as I like the way that multiple complex systems interact/integrate with eachother to form an overall system that seemingly "just about works"

Sorry in advance if my scrawl in illegible.

 

[takno]

I'm sorry but I'm having a hard time seeing driver-less cars actually coming into fruition as someone whom works in software engineering (more specifically deals a lot with AI). From what I've seen it's a long way off and we can barely get simple systems such as AI based loan approval software working without issues let alone a car carrying passengers without continuous human supervision. As much as I love the idea, the technology doesn't look like it'll mature enough for a long while yet... I see more chance we'll have widespread ATO (a much easier thing to do) heck we might even have full (ish) electrification before we see driver-less cars being both reliable and affordable.

What's needed for driver-less cars is commercially available highly accurate positioning (work in progress), highly accurate maps that are kept up to date (still a rather large challenge despite our best efforts), energy efficient yet powerful hardware for the AI to run on (we're about to hit the limit of what silicon can achieve due to the affect of quantum tunnelling of electrons), and many more technical issues. I also personally see there being issues popping up over safety, security, insurance and regulatory the latter will take a while to solve as governments aren't exactly renowned for moving fast or keeping up with technological developments.


And even with good coverage of unit tests many an issue still crops up and yes modularity is a good aim. However despite the best intentions of some very good developers this doesn't always plan out the way you'd expect as deliverables need to be met at the end of the day hook, line and sinker. After-all if the company doesn't ship anything, the company can't make money to pay ones wage.

Ideally there needs to be an open consortium dedicated to this issue to do as you've suggested, multiple modules that output in a standardised format from various suppliers that can form part of the overall system allowing for redundancy. However as stated up above we still have the issue of hardware constraints due the amount of localised processing needed for AI as edge and peer to peer computing will only get us so far.

We can't outsource all processing to a centralised server due to latency issues, even if we were to co-locate the server closer to the edge the latency would still be too great for a lot of safety critical functions. Heck kernel latency is a major issue in and of itself to deal with.

Maybe I'm far too jaded when it comes to technology and the software industry in general. However if the railways do decline I'll quite happily move my gaze to other complex systems as I don't really care all that much about the thing that runs on top of the rails. I'm in it as I like the way that multiple complex systems interact/integrate with eachother to form an overall system that seemingly "just about works"

Sorry in advance if my scrawl in illegible.

I'm not sure there are any smart people working in computers who think wide-scale adoption of driverless cars is happening in the next 20 years. The software just isn't there. When the software is there then we will still have to face up to any number of questions about the space they take up, whether they can safely interact with pedestrians, and the resources used in building them. I've probably only got 50 years left in me, maybe less if people start letting computers drive cars at pensioners, so honestly I can't see the value in worrying about it.

 

[PeterC]

It's hardly a "glorified" taxi if there's nobody to spot and clean up the vomit after the previous occupant. Getting home from the pub could become a bit of an ordeal.

You will be lucky if its just vomit.

 

[37424]

Will self driving car have implications for the Rail Network, I think the answer is yes particularly the Rural Network, obviously for of major cities there is limit as to how many cars they can cope with along with some councils regarding cars as the enemy.

Anybody who has done a delivery job will know that the current accuracy of Sat Nav's for some addresses isn't that great and your self driving taxi might might be waiting half a mile away if its reliant on such as google maps etc.

As for the technology its self, I watch the Tesla Self Driver beta testers on you tube in the US and there is no doubt for Motorways or Freeways in the US it is getting very good but off the Freeways on ordinary roads it still makes mistakes, and hands control back to the the driver for situations it cannot cope with, and that to me looks a fair way off level 5 autonomy. Of course Tesla isn't the only game in town and they are using cheap hardware as opposed to others using Lidar etc, but I'm not convinced level 5 autonomy will be here very quickly.

 

[Grimsby town]

I can forsee many problems with driverless vehicles most that have already been mentioned. Private vehicles as they currently exist are very space inefficient compared to a public transport vehicle and therefore inevitably need more resources to construct. Not really acceptable in a world trying to become more sustainable. Let's shrink the size of the vehicle then? The issue is does this not then start to remove some of the advantages of a car such as the ability to transport goods from door to door such as your shopping. It also adds questions around safety. Small pods might be fine travelling at 20mph in urban areas but remove a lot of the crumple zones on a motorway vehicle means that if something does go wrong you're probably less likely to get out. Does this mean we then have different types of vehicles for different purposes. How would you change between systems? This starting to look a lot more like a public transport system.

Other questions spring to mind. Is door to door travel really desirable for a society that already suffers from health issues related to physical inactivity? Do we go down the shared ownership route where autonomous vehicles are akin to taxis? In which case is there potential that people could misuse the vehicle and therefore put it out of action. What systems would need to be put in place to make the vehicles secure against crime?

Its important to remember that automation will also have benefits for the rail and public transport system. No drivers are currently a big source of cancellation on the system. Drivers are also fairly costly so make the use of smaller more frequent vehicles too expensive to be possible at the moment. The rail system is already incredibly secure (especially compared to a road) has less potential for interaction with complex issues and has elements of automation and connectivity between vehicles built in. If automation is going to increase, it will do so on the railways before the roads.

 

[Irascible]

I would have thought driverless cars are more of a danger to buses than ( most ) train services - and if there's something readily available that's the best combination of a taxi & hire car at the ends, doesn't that make rail *more* attractive? you're not going to do your usual 85mph motorway cruise in one either. No matter how many you can pack in nose to tail on an urban road you're not going to match the passenger density of a two deck bus, either.

With a software engineer hat on, I have immense doubts I'll ever live to see fully driverless roads which is really the only way it'll work reliably. Humans are much better at improvising solutions to problems but also far less precise & likely to cause the sort of out of parameters problems current driverless systems give up at. At sea or in the air there's far less constraints on movement, and for guided transit there's no maneuvering anyway ( and paths are obviously extremely predictable ) - roads are unpredictable enough that humans get it wrong all the time even when they have the training, experience and processing to supposedly deal with whatever.

Some of the roads round my way aren't terribly distinguishable from the neighbouring scenery...

 

[Annetts key]

The idea of cars communicating with each other isn't about journey planning, that's better done with the likes of "Trafficmaster data" as it is now, plus "crowdsourcing" that data as the likes of Waze does. Cars communicating with one another would be more for collision avoidance and similar, just like TCAS (Traffic Collision Avoidance System, I think) communicates between two aircraft to avoid a mid-air collision by telling one set of pilots to descend and one set to climb.

A TCAS warning has been a contributing factor in various incidents. See here.
As usual, humans find ways to fool, bypass, override or ignore safety systems. Or to put them into situations where they operate near, at or beyond their limits.

And as the systems become more and more complex, it becomes harder and harder to maintain fully comprehensive ’all possible situations’ testing. Especially as duplicated computers and duplicated sensors will be needed.

As hinted at above, there are plenty to problems to overcome if autonomous road vehicles are to mix with the existing road users.

 

[birchesgreen]

Other questions spring to mind. Is door to door travel really desirable for a society that already suffers from health issues related to physical inactivity?

Well a lot of this is being driven from the US where people use their cars to cross the street.

 

[Bletchleyite]

A TCAS warning has been a contributing factor in various incidents. See here.
As usual, humans find ways to fool, bypass, override or ignore safety systems.

Indeed. In essence TCAS is always right and must never be ignored.