Autonomous cars and robot taxis - the future of motoring is driverless

[roryer]

Transport is about to get very interesting.

Google’s Self-Driving Car Project could destroy taxis, Uber and Lyft because it knows everything about you

Google's self drive cars have been on the road now for several years and the technology is nearly market ready. BMW, Daimler, Toyota etc. all have their own models under testing.

The most dangerous part of any car is of course the driver, and removing human control will not remove the possibility of some death or injury, but will certainly make our streets much safer.

Most new cars already adopt some aspects of autonomous cars, and with Google going into production on what look like 'pods' with restricted speed, lightweight, with no steering wheel or control pedals, which can be designed to see your Google diary, we are starting to see how this will change our transport system.

Google is in fact one of the main investors in Uber, so I predict autonomous cars to begin to be introduced as taxis within 5 years. It will quickly be adopted by car clubs (Zipcar etc.), and perhaps within 15 years will replace most new private cars.

What does anyone else think?

 

[Bob_the_lost]

We're still miles off autonomous vehicles.

It's possible to do some amazing things with autonomous vehicles, the DARPA challenge was always worth following for example. However those are prototype vehicles, carefully monitored maintained and cost a fortune. Even the Google bubble car will cost a fortune and is restricted to low speeds and simple routes (as far as i know). Making a car capable of handling the absurd number of permutations needed to handle city driving in a non grid (ie. not US) city is really difficult.

I'd expect some things like collision detection to become mainstream in the same way ABS brakes did. They appeared in the 80s and grew steadily to become required kit on new cars by 2007 in the EU. Automatic braking started to crop up around 2002 and there are already suggestions to make it mandatory but i'd be surprised if it were before 2020 (ish) Edit: that might be pesimistic, All Volvos now ships with some basic collision avoidance tech built in but then again that's Volvo. Once you've bolted enough of those sensors to the vehicle you can expand it to provide other behaviour such as 'convoy' driving.

Beyond that I could envisage 'automatic' lanes on Motorways and A roads where you can put the car in full auto and where it'll either hand control back to you or pop into a service station. From there perhaps a white-list approach to rolling out to other roads that have been mapped and tested to be survivable.

Taxis in 5 years? Maybe in some dedicated and simple situations but not a chance in general world-wide use.

 

[Pickman's model]

it depends if google are still trusted.

 

[roryer]

We're still miles off autonomous vehicles.

You could be right, but I'm told that the problem is not technology, which is far better and safer than human drivers in real world conditions, but market acceptance. This is two pronged problem, not only public acceptance, but also the big auto companies are not pushing it for they still market cars on 'sex appeal' and 'power', not as simple transportation and mobility. There are still a number of legislative issues to deal with too.

Finally, obviously going from working prototype to mass market is a big step, but as you point out autonomous driving technology will be increasingly introduced to the average car, and motorway convoys are very likely to come next.

How long will it take for driver controlled cars to be phased out? Add in peak oil and the inevitable financial collapse, we will need to be smarter about transport than the current astounding 91% of car commuter trips being single occupancy, autonomous cars are part of the solution.

 

[Bob_the_lost]

You could be right, but I'm told that the problem is not technology, which is far better and safer than human drivers in real world conditions, but market acceptance. This is two pronged problem, not only public acceptance, but also the big auto companies are not pushing it for they still market cars on 'sex appeal' and 'power', not as simple transportation and mobility. There are still a number of legislative issues to deal with too.

Finally, obviously going from working prototype to mass market is a big step, but as you point out autonomous driving technology will be increasingly introduced to the average car, and motorway convoys are very likely to come next.

How long will it take for driver controlled cars to be phased out? Add in peak oil and the inevitable financial collapse, we will need to be smarter about transport than the current astounding 91% of car commuter trips being single occupancy, autonomous cars are part of the solution.

Safer is an easy thing to do. Not crashing is a doddle, if you detect that there's something ahead of you and you're closing too fast then stop. Doddle, zero injuries.

It's the intelligence part of driving that's difficult. Making room for emergency vehicles, merging into lanes of traffic during rush hour, Identifying gaps to pull out in if someone's letting you out at a junction, handling unexpected lane closures, blind corners. Knowing when to stop is really easy, knowing when to go isn't as clear cut and as far as i'm aware we're well off that point being ready in prototypes.

Which is why we'll be seeing the collision avoidance stuff soon, it's conceptually simple.

 

[Dogsauce]

Bikes will never be autonomous, and will still move about unpredictably. Can't see how the two can coexist on urban roads, although at least it will be better if these things are guaranteed to indicate and change lane or direction in a predictable manner.

 

[Crispy]

Bikes will never be autonomous, and will still move about unpredictably. Can't see how the two can coexist on urban roads, although at least it will be better if these things are guaranteed to indicate and change lane or direction in a predictable manner.

Check out the car's treatment of cyclists from 1:00, both at speed and at an intersection.

These things are already very sophisticated and can handle all sorts of unexpected situations.

 

[Dogsauce]

Hopefully they'll also equip them with autonomous doors that won't door cyclists, in fact they should have these on regular cars too.

 

[HAL9000]

The problem is not so much technology, but the human factor. With modern aircraft the level of automation is so high, pilots are failing to react correctly to system failures (in at least one case, failure to control an aircraft correctly when the ILS auto landing system was unavailable)

* Thomasfly flight, autothrottle failed silently causing air speed to drop, pilots reacted late to the airspeed problem.
(http://www.aaib.gov.uk/cms_resources.cfm?file=/Summary - AAR 3-2009 Boeing 737-3Q8, G-THOF 06-09.pdf) If the pilots had performed a manual approach without autothrottle, its likely they would have watched the airspeed all the way to touch down.
* Air France flight, airspeed sensor froze. Incorrect action of one of the pilots and the aircraft was lost. (http://en.wikipedia.org/wiki/Air_France_Flight_447)

Autonomous car driver cannot be expected to take control of the car in the event of a failure, the car needs to handle failures without driver input.

 

[TodayIsCaturday]

So as a pedestrian, if you know the oncoming car has autobrake you can cross the road whenever knowing you'll be safer than at a zebra crossing facing a person-braked car?

 

[weltweit]

There will be a lot of development around hybrid cars and collision avoidance certainly but I agree with Bob_the_lost there is a lot more to driving than knowing when to brake or stop.

And there will be a slow transition period. I can remember when ABS came out on production cars and people in Germany were saying ABS cars should have big warning lights on their backs because they could stop so much quicker than following non ABS vehicles.

The future is coming, but perhaps not as fast as some expect.

 

[Bob_the_lost]

So as a pedestrian, if you know the oncoming car has autobrake you can cross the road whenever knowing you'll be safer than at a zebra crossing facing a person-braked car?

Yes, if you can identify the model of car, trim level and know that it's fitted with automatic braking you could. In 50 years this may become a sport in the distopian nightmare that our children live in. In the medium term it isn't going to be a major issue.
You'd probably find that buses and HGVs get it first in which case they might be subject to chicken entertainment.

 

[weltweit]

So as a pedestrian, if you know the oncoming car has autobrake you can cross the road whenever knowing you'll be safer than at a zebra crossing facing a person-braked car?

You could still step out closer than the auto braking car could stop and still be hit.

 

[pogofish]

What does anyone else think?

Puzzled:

http://www.urban75.net/forums/threads/googles-self-driving-car-lets-a-blind-man-get-around.291052/

 

[Dogsauce]

So as a pedestrian, if you know the oncoming car has autobrake you can cross the road whenever knowing you'll be safer than at a zebra crossing facing a person-braked car?

You get extra points if you can trigger their airbags.

 

[Bob_the_lost]

Mercedes has released it's optimus prime prototype

http://www.engadget.com/2014/07/06/mercedes-semi-autonomous-truck/#continued

 

[The Boy]

Google cars have been programmed to speed by up to 10mph over the limit if needed to keep up with the flow of traffic.
http://www.bbc.co.uk/news/technology-28851996

 

[hash tag]

Seeng above re: the Auto braking systems, Wont this just encourage "drivers" to read newspapers whilst driving or use their mobiles, facebook or whatever?

http://www.trw.com/integrated_systems/active_passive_systems/automatic_emergency_braking

 

[Bob_the_lost]

Seeng above re: the Auto braking systems, Wont this just encourage "drivers" to read newspapers whilst driving or use their mobiles, facebook or whatever?

http://www.trw.com/integrated_systems/active_passive_systems/automatic_emergency_braking

Just like airbags and seatbelts have encouraged drivers to simply crash into brick walls instead of braking.

Or in other words. No. That braking is very, very harsh and unpleasant. Whilst you might get added risky behaviour it's going to be a net gain.

 

[hash tag]

I sort of feel that as another piece of safety technology comes along so people tend to be less cautious/take more risks.

 

[Bob_the_lost]

I sort of feel that as another piece of safety technology comes along so people tend to be less cautious/take more risks.

Perhaps people do take more risks, but if you can point to a rise in casualties after the introduction of ABS, Seatbelts, Airbags, Lane Assist, Automatic braking or any other technological safety improvement you will be the first person to ever do so. Insurance companies crunch the numbers like no-one else and they like these things.

For example, using made-up numbers, if there's one extra crash because someone decided that auto-braking would protect them another five crashes that would happen for other reasons would be prevented.

 

[Crispy]

Oh. The google cars are considerably more stupid than I thought. They can only drive on routes that have been meticulously mapped and can't interpret the road for themselves at all.

http://www.slate.com/articles/techn..._car_it_may_never_actually_happen.single.html

Chris Urmson, director of the Google car team, told me that if the car came across a traffic signal not on its map, it could potentially run a red light, simply because it wouldn't know to look for the signal.

 

[dylanredefined]

You get extra points if you can trigger their airbags.

Or the government decides that some people are not worth saving and allows a software fix that removes idiots from the gene pool.

 

[Bob_the_lost]

Oh. The google cars are considerably more stupid than I thought. They can only drive on routes that have been meticulously mapped and can't interpret the road for themselves at all.

http://www.slate.com/articles/techn..._car_it_may_never_actually_happen.single.html

Disappointing but not surprising when you think about it.

On the other hand Google Streetview shows that it's logistically possible to map the roads and with modern data storage it isn't an impossible task to map and store every road for a route.

Of course road works and changes are still a potential issue.

 

[Dogsauce]

It'll work by limiting the roads cars can work on, and new procedures being required for when changes to road layouts - temporary or permanent - happen. Temporary lights for roadworks may have to have some sort of transducer, or a requirement and method to register their location. If you come to the end of a mapped section of highway, you'll need to take manual control, that's all. There may be scope for 'training' your car to park on your driveway. There's also scope for 'crowd mapping' routes, where the driving alignment of a vehicle under manual control on a certain section of road is recorded and fed back to a database, with collected data helping automatic cars to 'learn' how to drive through a section. A lot of this will need improved positional data, I know from the wiggly line I get tracking my bike ride on my phone through a built-up area that we have a long way to go on this!

I reckon we'll see it in taxi/shuttle bus type uses a long time before anything else.

 

[hash tag]

There has been talk of building roads for exclusive use of autonomous cars, which could in part work. If not roads, a lane or two on motorways perhaps.

 

[Dogsauce]

Given these cars won't be cheap, that'll just be roads for the wealthy then. I can see problems in that being accepted.

More likely is the motorway network being set up with markers or transponders to assist autonomous vehicles. Thing is, it's a convenience thing rather than a safety thing, motorways are not where attention is needed as far as accidents go.

Could be stuff like HGVs where the costs add up to start with, most distribution depots are just off the motorway network so it would be a fairly easy drive for a machine. Maybe more sensible would be joining them all together and sticking them on rails, imagine that.

 

[SikhWarrioR]

Given these cars won't be cheap, that'll just be roads for the wealthy then. I can see problems in that being accepted.

More likely is the motorway network being set up with markers or transponders to assist autonomous vehicles. Thing is, it's a convenience thing rather than a safety thing, motorways are not where attention is needed as far as accidents go.

Could be stuff like HGVs where the costs add up to start with, most distribution depots are just off the motorway network so it would be a fairly easy drive for a machine. Maybe more sensible would be joining them all together and sticking them on rails, imagine that.

We could also do that with cars and buses join them together and run them over a network of guide tracks built all over the country..........Hang on didnt we used to do that all under one organisation, I seem to remember it was called something like "British Railways" and it was owned by us all

 

[T & P]

We could also do that with cars and buses join them together and run them over a network of guide tracks built all over the country..........Hang on didnt we used to do that all under one organisation, I seem to remember it was called something like "British Railways" and it was owned by us all

How about a real-life Scalextric style system? That way you don't need to build more roads. It works for trams in many cities worldwide.

 

[rover07]

Elon Musk reckons autonomous driving is 5-6 years away.



Latest Tesla Model S has various autopilot features, adaptive cruise control, collision detection, auto lane changing and parking.