Hacker News Reader

Comment by saosebastiao | Hacker News Reader

Traffic waves aren't going away. As long as you have non-zero reaction times and slower acceleration than deceleration, the phantom waves will happen as a pure result of physics. Their discovery and subsequent modeling owe to our understanding of fluid dynamics. Computers aren't magical things that destroy the laws of physics.

Faster reaction times reduce propensity, but both higher speeds and closer following distances ameliorate the benefits of the faster reaction times and increase severity. If anything changes they'll become slightly less common but much worse.

EDIT: just to add to the criticism of this list of overly fantastical fantasies of autonomous cars, the ideas about traffic disappearing when people don't have to park are trivially disprovable. It's almost as if the author has never been to an airport or school pickup zone or any other place where people don't have to park but traffic still backs up for miles.

View on HN

Fluid dynamics is local, and so are the dynamics of human-driven car traffic to a first approximation. Computer-driven cars can react non-locally by communicating with each other via radio even if they don't see each other.

At the very least, waves could be damped by a vehicle that sees it coming from a long way away. But if everyone knows the average speed of the roadway, and doesn't suffer from the moronic impulse to push up against the car in front of them, waves will form less often, if at all.

> and doesn't suffer from the moronic impulse to push up against the car in front of them

But see that's the problem. In this piece, we see direct advocacy of that exact moronic impulse under the misguided assumption that computers can will away the laws of physics.

Let's say you've got 50 miles of freeway where it is completely full of cars driving 80mph two feet apart. A dog runs into the road. Locally, the traffic jam is inevitable and bad. But they transmit this condition via radio such that cars downstream can start braking early. What happens when one car doesn't get the message on time? The car in front of it brakes, and this car brakes much harder in reaction to visual cues in lieu of the radio cues, forcing the car behind it to react to visual cues and so on, causing another shock wave.

Don't even get me started on the idea that we can rely on wireless communication in order to make reaction times fast enough that we can drive closer than we can adequately stop. Even locally that level of reliability is decades beyond current capabilities...non visible distances are a non-starter.

There seems to be a confusion among people that self-driving cars are going to fully self-contained, with no interaction or communication between them. At first they probably will - but the greater gains in safety, efficiency, speed, and utility will only happen when they all can communicate with each other - to inform each other about traffic pattern changes, as well as to coordinate on local issues.

There's no confusion here. It will be possible, but relying on it to stop traffic waves is an untenable fantasy. Even directly connected fiber optic communication has levels of unreliability that would make it untenable to drive closer than you can stop from visual sensory alone...and you want to rely on wireless communication?

At it's best, wireless communication may help traffic miles from an incident, by communicating problems (most likely with some central management system) and routing traffic around it. Visual range sensors will dominate all local interactions, just like they do with swarm coordination systems in both nature and robotics.

Sure, but now we've gone from a discussion of changes we could reasonably expect to see over a couple of decades to much more ambitious structural systematic changes which would take most of a century to accomplish.

Thanks for reading! I wrote several paragraphs discussing the complexities of how changing (not removing) parking would both remove some kinds of traffic and add others, and did not at all suggest it would lead to traffic disappearing. I also discussed the issue of queuing for pickups.

On traffic waves: I think it's... unsafe to assume that a phenomenon based on human behaviours (not, in fact, just reaction time) would not change if cars are controlled by software.

Why would you assume that the phenomenon is based on human behaviors? The phenomenon is a result of a physics constrained response to an impetus. It can be replicated without humans present. In real life, that impetus may be human behavior in some cases, but most definitely not all. Imagine a dog running into the road, or a piece of litter obstructing a lidar sensor. And if you're driving closer together and faster, the phenomenon is worse. You can replicate this behavior using any traffic simulator with those three constraints alone (non-zero reaction times, slower acceleration than deceleration, and an impetus that causes braking).

Either way, I'm not sure how you would come to the conclusion that it would be unsafe to assume that traffic waves won't go away. Maybe you could offer some scenario where we actually would be safer to assume they'll go away.

Reply on news.ycombinator.com