So self driving car brands are currently in the process of offering subscription based self driving features.
Soon these subscriptions will be cheap enough for everybody to abandon public transport and get into their self driving cars instead.
This means that everybody will be in their cars and it will create the biggest traffic jam in human history!
And finally, self-driving cars will be totally safe and able to drive completely unsupervised! While standing still in traffic going an average of 5 km/h…
It will be beautiful (/s)
There is an argument that a connected Borg collective of self driving cars would be capable of running vehicles with smaller gaps in between them, thus reducing congestion and improving efficiency in traffic.
Hi, traffic engineer here.
That’s never going to happen. It’s nothing but a tech-bro bullshit fantasy.
Why? Because cyclists and pedestrians exist. In order to make it possible for the kinds of gains you’re talking about to happen, every road user has to be an autonomous vehicle, but (aside from freeways) streets simply do not work that way and never will.
(Oh, and also: even at the limit, the best it can ever accomplish is to be an inferior approximation of a train.)
Maybe we should work on building out more trains?
I suppose self-driving cars could drive like regular cars on the small roads, but then go to big specialized highways where they can go full speed.
But that doesn’t make sense anyway, because that would not save you that much time anymore. And in that case you could just have regular cars to drive to a station with parking and then take the train to your destination
We already have those big specialized highways; that’s what freeways are. The trouble is, even if autonomous driving were capable of doubling the lane capacity (and IIRC the theoretical best case is actually less than that, closer to a 50% improvement), induced demand is still a thing. It maybe buys you a reprieve for a decade or so, but after that you’re right back to “just one more lane, bro!”
Since you stated that you are a traffic engineer, quick question for you. Are more lanes all together or more lanes separate better?
What I mean is are you better off with 1 highway that has 4 lanes in each direction or 2 highways that each have 2 lanes in each direct. Same total lanes, but split up differently? It seems like adding more lanes to a congested road just makes a bigger mess; a problem in any lane seems to propagate through all lanes.
I know it’s a lot more complicated than just building a parallel road, but (in my area at least) the lack of alternate routes is frustrating. The entire highway turns into a parking lot if there’s a crash. I also see a lot of near misses due to lane flanking: car in lane A and car in lane C both want in lane B at the same time. More lanes = more likely for this to happen.
But at what point would it just be better to just have a self-driving bus, or even put a railway instead?
At every point. Public transport is more efficient at transporting people than private cars will ever be. If we have enough of it, it’s not even going to be crammed during rush hours.
If all cars were driven with good autonomous features (we’re a long way off) there would be no traffic. Traffic is solely due to human error
No, that’s not how traffic works. That’s like saying a pipe can flow an infinite amount of fluid when you used a liquid instead of a gas because you got rid of the empty space between particles.
Even with theoretically-perfect timing and control, the road still has a finite capacity because cars take up a certain amount of space, both stationary and moving (following distance is still a thing even with computer control because of the mechanical limitations of brake performance). Moreover, it isn’t that much higher than we can manage with humans driving the vehicles already.
The only ways to exceed that limit are to make the vehicles smaller (e.g. bikes) or pack more people into them (e.g. buses or trains).
Traffic is massively impacted by people breaking after ending up too close to another vehicle. There are great demonstration videos in the late 90s / early 2000s demonstrating this. I don’t anticipate me stating this fact to gain any support, because, you know, Lemmy says AI bad no matter what.
You’re right: when traffic is at its critical density, that’s often what triggers the shift from the free-flow regime to the congested regime. But just because you make computers drive the cars – even assuming they did it perfectly without randomly braking, which they don’t – that doesn’t mean it eliminates that flipping between regimes. At best, it might get you a little bit more capacity before hitting that critical threshold, but eventually it’s still going to, and then something – a squirrel darting into the road, a sunbeam glinting off something the wrong way and momentarily confusing the AI, a bump that disturbs the car just enough to make it slow down a fraction of a MPH, etc. – is going to trigger that shift to the congested regime anyway.
This is simply not true on its face.
Unlike the thought expirement you’re running in your head, the real world does not have the luxury of avoiding the possibilities of sudden mechanical failure, a soccer ball being kicked into the road, or unsecured debris falling out of the bed of a truck. These are privately owned and maintained vehicles after all, under many more engineering constraints and usage pressures than just achieving maximum reliablilty in all conditions.
These things require some degree of margin for error. In general, any car should be able to fully stop before entering the space currently occupied by the car in front of it in order to account for unexpected disturbances. However, as car speeds increase, so do the tolerances required to maintain that safe margin, exponentially in fact.
Removing the human from the driver seat doesn’t mean we get to start running bumper to bumper at 70mph; at best it means that cars can get away with slightly smaller follow distances than humans need to account for their comparably slow reaction times.
If you instead believe that we should exert enough control over self driving cars such that we could actually realistically prevent catastrophe while running high speed bumper to bumper traffic, then I have great news for you! We already have that, and it’s called a train.
It’s not a thought experiment in my head. It’s a highly studied and demonstrated fact. And it doesn’t require bumper to bumper cars, it requires cars consistently traveling with safe breaking distance, something humans didn’t do.
I don’t refute that. I agree that far too many humans are far too comfortable with unsafe following distances.That’s not the claim you made though:
And as I said, it’s just simply not true. Congestion happens because car traffic does not scale or recover from disturbances efficiently. Human behavior exacerbates this, sure, but cars are just fundamentally bad at moving people when a road is at capacity. Chalking traffic it up to “human error” alone misses the forest for the trees.