I was listening to a podcast today discussing self driving cars and how difficult it is for them to work safely in complex urban environments.
I do not understand why work on this technology is starting in the most complex environment.
Why not automate motorway driving first as that must be easier.
When I replace my campervan I want to join the Edinburgh bypass and have the new van take me to Inverness!
Am I missing something obvious here?
This is quite long but I think is worth a watch…
I imagine, in large cities, once people realise self driving cars will stop automatically if a pedestrian wanders into the road not much will move.
Active cruise control is already pretty good, combined with lane assist, although there is a section of the M4, where my van reads the speed limit as 100mph. It also doesn’t understand van speed limits.
If you really wanted to start with the easy bits it would be trains...
Wasn't there an example recently of someone in a Waymo in SF who had some ne'er do well try to break in?
With a human the driver would just floor the accelerator to get away but instead you're at the mercy of whatever the driving software decides, eg just remaining stationary.
I do not understand why work on this technology is starting in the most complex environment.
Why not automate motorway driving first as that must be easier.
I think you need to divide it into what the likes of Weymo are doing in just small areas of san fran/phoenix and what the car manufacturers are doing, which is what we'll get in small increments. In the short term at least we are just looking at one step on ADAS to hands off eyes off and it'll just probably be certified for motorway use. Is it ford and Mercedes who currently have certification in the uk for hands off eyes off on motorways, possibly just hands off. For you and me over the next 10-15yrs it's probably just better ADAS. If you want something that sort of gives the experience right now, then buy a vehicle with lane centring rather than lane keep.
There was some suggestion that pedestrians/cyclists would have to wear beacons so self-driving cars could see them, which wouldn’t be open to abuse in any way, shape or form.
If they have a place, it’s as above, on motorways or equivalent. In urban areas it’s hard to shake the notion that it’s a combo of Musk hype machine and public transport spoiler.
If they have a place, it’s as above, on motorways or equivalent. In urban areas it’s hard to shake the notion that it’s a combo of Musk hype machine and public transport spoiler.
This ^^. It's a bullshit notion that we can all just carry on as we are, no need to minimise driving or stop building roads cos "any day now" we'll have self-driving electric/hydrogen/fairy dust powered cars.
In fact in some respects it takes some of the focus off public transport and active travel, delays implementation of any sort of mass transit because everyone is going "ooh, AI / self-driving / "technology" will solve all our problems..."
Meanwhile the simple cheap Stuff That Works like trams and buses gets ignored.
Watch the video… some of this stuff is covered.
I for one would rather not let the genie out of the bottle… although I fear that it may already have happened. I think self driving cars are going to be as bad as the cars we all drive… which, unfortunately, we can’t un-invent.
Meanwhile the simple cheap Stuff That Works like trams and buses gets ignored
I don't think those things are being ignored, more a case of overlooked as the big financial wins for the transport industry are still centered on personal cars.
The above video is very long but I watched it on my commute and agree with most of the points.
I worked in the industry and I used to think it's great, but watching how lots of other bigtech ideas have turned out I've massively changed my mind.
In answer to the OP, the goal here is to make money. Since they are basically autonomous taxis, there are more people in towns needing lifts so thats where the autonomous car companies (uber, lyft, wayve) are focusing their efforts. They have had billions poured into them by investors who will want to see a return.
It's different for legacy car makers like General Motors etc who are focusing on motorways.
We are bascially handing control of our transport over to a few rich private companies who will lobby governments to bring it in, replace bus lanes with autonomous only car lanes, they'll start blaming pedestrians and cyclists for being hit so they'll be banned from crossing, it'll be very space inefficient, it'll start off cheap, then they'll ramp up the cost once capturing the market and removing the comptition, just like with uber, airbnb, netflix, AI and all the other mass user, high investment services.
It will become a necessary monthly subscription. You pay more money and but don't own anything.
There was some suggestion that pedestrians/cyclists would have to wear beacons so self-driving cars could see them, which wouldn’t be open to abuse in any way, shape or form.
IIRC that was something originally posted on a rabid anti transport/hippie/anti technology blog a few years ago.
IIRC that was something originally posted on a rabid anti transport/hippie/anti technology blog a few years ago.
It may have been posted on such a blog, but the technology is real and being considered. The protocol is called Vehicle-to-everything "V2X". Could be used for pedestrian beacons (V2P), but could also be used for cars sharing information about where they've seen pedestrians.
We are bascially handing control of our transport over to a few rich private companies who will lobby governments
That’s been the case for about 100 years - Ford, GM, etc.
The difference now is that the technology is changing and one or two companies might drop out or get established.
It may have been posted on such a blog, but the technology is real and being considered. The protocol is called Vehicle-to-everything “V2X”. Could be used for pedestrian beacons (V2P), but could also be used for cars sharing information about where they’ve seen pedestrians.
As noted above, autonomous vehicles don't like unpredictable things (like cyclists/pedestrians), they don't always recognise them for what they are and can't effectively anticipate their moves so the options are either everything being equipped with beacons (what could possibly go wrong there...?) or simply banning pedestrians and cyclists. Cos the option to simply run out in front of the car and watch it slam the brakes on is the other outcome and no car company wants that.
Cyclists saying "ooh, I can't wait for autonomous cars, they won't run me over!" are missing the point. You just won't be allowed on the roads with them!
How come Tesla's currently do a pretty good job of spotting pedestrians, cyclists, cones etc? What do you mean they don't like unpredictable things? How do they deal with buses pullng in and out , or other cars (both of which they do)? Some of the comments above are a bit 'Facebook' and detached from the current reality
The protocol is called Vehicle-to-everything “V2X”. Could be used for pedestrian beacons (V2P), but could also be used for cars sharing information about where they’ve seen pedestrians.
Yeah, i work with V2X. The only V2P stuff that i've seen has been exactly that, vehicles sharing data about VRUs they've spotted using onboard sensors, location, speed, direction, ped crossings etc. We already have shared data available (but not for pedestrian movement, yet).
Nothing about "beacons" though, other than a couple of papers about using phones to identify the existence of a VRU, which was fairly roundly dismissed as an evolutionary dead end as you can't rely on something like that. Everything is either not directional enough, or too short range, or easy to obscure (or switch off). Needs a complete technology rethink and some frankly horrific legislation to be put in place first.
Yeah, it (the thing about beacons) did get picked up by various media outlets and run in various different ways.
There's a new Integrated Transport Policy on the way from DfT to replace most of the nonsense from the previous Government, I know some of it is including autonomous considerations.
As noted above, autonomous vehicles don’t like unpredictable things (like cyclists/pedestrians), they don’t always recognise them for what they are and can’t effectively anticipate their moves
You could write exactly the same thing about a lot of drivers in conventional cars (and a small number of people even suggest banning cyclists from the roads as a solution).
Cos the option to simply run out in front of the car and watch it slam the brakes on is the other outcome
Self driving cars cannot circumvent the laws of physics. They will be capable of reacting faster than a human driver and stopping in a shorter distance, but if a pedestrian jumps in front of a self driving car that can't stop in time, they're still going to get hurt. I'm puzzled why some people think pedestrians will start throwing themselves in front of moving vehicles en masse.
I think it's a valid concern. The cars will stop and won't move again until the object dissappears. This is already happening (according to above video) with people placing cones in front of unoccupied autonomous cars.
How come Tesla’s currently do a pretty good job of spotting pedestrians, cyclists, cones etc?How do they deal with buses pullng in and out , or other cars (both of which they do)?
They slow down or slam the brakes on and then you intervene to deal with it, that's not autonomy (or at best levels 1/2) it's just glorified ADAS in your case adaptive cruise control with lane centring and pictures on the screen that show you bins and cones. There's been a few deaths in Tesla s where people have presumably thought it's more than just that, Tesla are currently under investigation in the states for their whole full self drive / city streets claims or whatever it's called. I think this is particularly following a few emergency service vehicles being collided with.
How come Tesla’s currently do a pretty good job of spotting pedestrians, cyclists, cones etc?
They do a reasonable job but still make lots of mistakes. However, it's a tech company, and as such it is all about creating a buzz over new features to increase its profile and keep investors excited. However, where safety is concerned car manufacturers need to be better than 'reasonable', they need to be absolutely spot on. And - surprise! - the last bit going from 'reasonable' to 'spot on' is really very difficult indeed.
Mercedes have taken on this challenge and their system is far more robust because it relies on the entire city having already been mapped in 3D beforehand.
What happens today if a pedestrian wanders onto a road?
I think the point is, a pedestrian is less likely to wander into traffic randomly when there’s a fair chance of the human driver not reacting in time, seeing them or something daft.
If you know the cars will stop, you don’t need a pedestrian crossing, which sounds awesome, but actually as mentioned I’d have thought we’d end up with Jaywalking rules like the US or Australia. Which is worse.
What happens today if a pedestrian wanders onto a road?
The driver winds down their window and tells them to get out of their way – “You got a problem? Do you know who I am?”
But an autonomous vehicle can’t do that.
Why do you think an autonomous vehicle without passengers wouldn't be able to shout at a pedestrian (if that's the solution)? For vehicles with passengers, I suspect windable window technology will still be available. Unless large numbers of pedestrians start standing in the middle of the road, this doesn't sound like it would be a regular problem anyway.
I think the point is, a pedestrian is less likely to wander into traffic randomly when there’s a fair chance of the human driver not reacting in time, seeing them or something daft.
Autonomous vehicles won't be able to stop instantaneously, so wandering randomly into the road in a busy city will still be hazardous, and for that reason, I don't think people are going to start obstructing the roads in large numbers if autonomous vehicles are introduced.
There are many practical issues to overcome before autonomous vehicles become common technology, but I don't think pedestrians creating grid lock is anywhere near the top of the list. We'll adjust our behaviour slightly as the technology changes, and life will carry on. In any case, worrying about pedestrians obstructing cars is looking at the whole thing back to front. Cities where pedestrian traffic has priority would be much nicer places to visit or to live and work in.
If you know the cars will stop, you don’t need a pedestrian crossing, which sounds awesome, but actually as mentioned I’d have thought we’d end up with Jaywalking rules like the US or Australia. Which is worse.
TBH, with a proper integrated transport policy, you'd ban the cars from areas where pedestrians are going to be.
And there seems to be the idea that self driving cars will drive round exactly like humans, 20kph over the speed limit, tapping away on their phones and generally oblivious to whats going on. They won't be.
Some of the comments above are a bit ‘Facebook’ and detached from the current reality
Nope its more you buying into the detached from reality hype. Just look at teslas habit of parking in emergency vehicles.
As with many problems the first 80 % is "easy" but its the last 20% of edge cases which is really hard.
My boss spent last week in San Francisco - every taxi journey he took was via a service called Waymo - autonomous taxis, no driver. Despite initial concerns, he was amazed and convinced with how good an experience it was.
given how far technology has come in the last 30 years I have no reason to doubt self driving cars will be safer than people driven ones in time to come. I welcome it. along with electric vehicle tech it will make personal car ownership obsolete and vastly improve cities. No need to own a car - just call one up on your phone. No more drunk drivers, no more road rage, whats not to like?
no more road rage, whats not to like?
People like road rage through. I think the biggest bar to wide spread adoption of self driving cars is they require you to abandon your sense of entitlement. All those micro aggressions- not letting someone else pull out or change lane, Over taking some one just in time for your both to stop at the traffic lights - all that shit. People love it and the love to be seen to do it.
Self driving cars, all working together with a common goal so we all get where where going easier, safer and quicker - thats practically communism. Thats what fungus does. We're better than that. Well I am. People don't want something thats fair, they want something thats advantageous to them- they want to get where they are going at the expense of someone else, even if it means their own journey taking longer - they still win and are seen to be the winner.
TBH, with a proper integrated transport policy, you’d ban the cars from areas where pedestrians are going to be
Obviously, but on a week day in central London for example, the vehicles are mostly Busses, taxis and service vehicles. Which are essential and there are pedestrians, cyclists etc everywhere. I’d love to see how an autonomous vehicle would get on.
I have the pleasure of driving into London on a regular basis. Public transport would be nice, I could be there in 1 hour vs 2 but the equipment is too bulky.
A self driving company van would be an awesome upgrade. Just like a decent automatic and active cruise control, that only took 30 years.
I’ll probably be retired or dead from the stress of London driving by then.
That’s pretty impressive to be fair.
Interesting take on where it 'could' be going:
Another advantage will be massively reduced congestion once all cars are self driving as they will co operate, not block junctions and improve traffic flow by not concertinaing ( as properly introduced 20mph zones do)
As I say… watch the video. Summary below. It’s only one guy’s view and obviously he has an agenda but on the whole he is spot on in most of his videos. I have been thinking many of the things he raises for a while and was interested when this popped up.
0:00 Intro
1:17 Reasons for optimism & skepticism
2:33 Moving fast and breaking people
8:50 Fatal Uber Crash
10:34 The Real Road Safety Issue
14:27 Promoting car-centric cities
17:03 Cheaper taxis (yay?)
21:07 Traffic congestion will be even worse
25:19 The promised future (past and present)
28:04 How AVs will destroy cities
29:56 Eliminating public transit
32:37 Consuming all streets
34:02 Eliminating pedestrians
36:05 Eliminating speed limits
38:05 Pollution and noise
39:41 Eliminating traffic lights
42:16 Do we actually need AVs?
43:28 Utrecht vs Fake London
48:33 What should we do about it? (Hard infrastructure and time-based road pricing)
51:53 Where to learn more
I know it’s coming and resistance may well be futile but I can only see autonomous vehicles as a step in the wrong direction. The motor industry screwed us all but suppressing public transport and active travel and making everywhere a worse place to be. Self driving cars are the second wave and whilst they will bring some benefits to some people, they will make things even worse in so many ways.
Self-driving cars. Thoughtful emoji. PMSL emoji x 3
Another advantage will be massively reduced congestion once all cars are self driving as they will co operate, not block junctions and improve traffic flow
Some way to go yet....
To me fully autonomous self driving cars feel further away now than when this thread started.
Maybe because both me and Mrs OD have new vehicles with lots of driver aides but it's obviously how massively far these are from door to door full automation - not least how buggy fairly basic (compared with a smart phone) infotainment systems can be
