Mathematicians have solved traffic jams, and they're begging cities to listen

yosito 2021-08-17 08:58:46 +0000 UTC [ - ]

This article doesn't really go into details about what the solution is. It basically sounds like they're saying they can control traffic when they control enough of the variables. Traffic engineers have been using digital models for decades. The bigger problem is that the real world has too many pesky autonomous drivers. I'm not convinced that many drivers really want a solution that takes their autonomy away. If they did, they'd just use public transport, which, incidentally, is capable of moving far more people with far less congestion issues.

Arnt 2021-08-17 09:38:41 +0000 UTC [ - ]

Drivers wanting autonomy isn't the only uncontrollable variable. Property owners along the streets have ideas, too. Optimising their street for the ability of drivers to drive quickly through it on their way from elsewhere to elsewhere typically isn't high on their list. Adds noise, depresses property prices.

The whole thing reminds me of some modernist architecture dreams. One of the great architects had an idea of cities as... let me be pointed and unfriendly. Of magnificently designed apartment blocks connected to magnificently designed workplaces by wide fast roads, and nothing else. Fantastic looks, very clean, very elegant in a way, but nothing else. Corporatist, not far from fascist, very far from humane.

axelroze 2021-08-17 10:01:57 +0000 UTC [ - ]

Bad old Le Corbusier.

Who would have thought moving architecture away from beauty and into "efficency" would end up so badly? (i am sarcastic)

Arnt 2021-08-17 10:36:56 +0000 UTC [ - ]

It's easy to be sarcastic about him, but be fair. What he was moving architecture away from wasn't so much beauty as it was random slums for most people and beautiful comfort for the 1%.

It's easy to look at the hundred-year-old apartments and see the beauty, but that suffers from survivor bias.

chopin 2021-08-17 12:26:13 +0000 UTC [ - ]

Two years ago I admired the Edwardian blocks in Edinburgh. I don't think they where built for the 1% and are much more humane.

Arnt 2021-08-17 14:35:40 +0000 UTC [ - ]

Yes. And I don't think Le Corbusier ever wanted to replace the best of the old housing, he wanted to replace the mass of bad building.

And he had rather ahem ideas about how people ought to behave, once they had spacious apartments with hot and cold water. Which is the part I was reminded of when I read the article we're discussing. That view of the city as a set of streets to drive through is like his. Very simplistic, one-sided, optimisable but not humane.

axelroze 2021-08-17 12:28:05 +0000 UTC [ - ]

Actually I agree that commieblocks/modern-architecture is very efficient in the same way as Ford's car factory in producing a lot fast. And just as Ford said that the car can be had in any colour as long as it black, so the architecture of this period is in any colour as it is dull concrete gray.

The dream of equality to the letter is what actually lead to this. We should all be equal => We must all live in ugly soulless copy-paste commieblocks. Would person A be less equal to person B if A lived in a box painted red and B lived in a box painted green? Le Corbusier et al obviously thought the answer is yes. I take a lot of issue with that.

Marble pools, huge gardens and other forms of opulence and luxury are a thing to cling in these arguments to but there are many many worker communities who lived in very beautiful buildings at the turn of the 20th century. Why not continue that strategy but go instead for ugliness? The ego of Le Corbusier and the desire to be modern and free of shackles of the past. WW2 destroyed Europe. What to do? Make it so ugly that no one ever again bothers destroying it.

By the way beauty was not completely lost in the communist countries. The new steel mill processing plant in Krakow is a community built in renaissance style. - https://en.wikipedia.org/wiki/Nowa_Huta

karmakaze 2021-08-17 15:28:44 +0000 UTC [ - ]

Why not put an ML system in with some basic safeguards, like the direction changeover duration and that each direction has to be green for at least x within period y. Sure it would suck at first but then it would get smart pretty quick when a long street has all their lights controlled by that model.

necovek 2021-08-17 07:40:43 +0000 UTC [ - ]

I was hoping for something smarter than "central navigation hub" used by everyone.

With the amount of data Google collects from phones riding in cars, it's probably close, except people do not use navigation for daily routes.

axelroze 2021-08-17 10:03:55 +0000 UTC [ - ]

Here an issue is that if all people started using navigation to go via low traffic routes they would clog them and make them high traffic. There is probably some optimal equilibrium but by the time it is reached the rush hour could well have already passed.

necovek 2021-08-17 10:15:53 +0000 UTC [ - ]

A central navigation hub could decide to direct appropriate proportion of drivers a particular way.

Still, the problem is not which routes are least loaded, but how to get people to most efficiently drive en masse.

It is more efficient for everyone to wait ahead of the intersection than to enter it and possibly clog it, yet when others clog it from a different direction, you get stuck. It takes only a few people not driving in everyone's best interest (including their own) for traffic jams to become worse.

And with speed limits, traffic lights and other, non-car traffic (pedestrian/bicycle crossings with and without traffic lights), the problem quickly becomes hard, even for a central navigation hub (eg. there is a long street here that I can go through entirely in 2-3 minutes if everybody is driving at 10-15km/h faster than the speed limit: at the speed limit, you hit at least two traffic lights and spend 5-10 minutes in it).

Problem is likely mathematically solvable only when such a hub drives the cars too.

axelroze 2021-08-17 12:17:07 +0000 UTC [ - ]

> Problem is likely mathematically solvable only when such a hub drives the cars too.

The great promise of self-driving cars of the future. Personally I am super stoked for these. All the comforts of a private vehicle with (almost) all the benefits of public transit.

I am not sure if we need central planning or the 'market' will solve the problem itself. I guess in a fully coherent world of cars some central hub with fair round robin uniform priorities would be best. In reality it is more likely that multiple competitors will arise (Tesla, Uber, Waymo) and we will not have central planning.

necovek 2021-08-17 20:21:05 +0000 UTC [ - ]

Oh, no disagreement there (though I do enjoy driving, and turning it into an optimization problem too :)).

I was mostly debating the OP's claim of how to solve it mathematically. With smarter, self-driving cars, it's likely enough to solve the problem "locally", for a sufficiently large definition of "local" :)

I am not convinced we'll get to self-driving cars that can navigate the real world obstacles anytime soon, though. Like barrels falling off a truck in front of me on the highway. Or a sideway collision that has almost blocked the road.

Unless the self-driving tech is trained in Mario Kart, it will take a while :D

ReptileMan 2021-08-17 13:13:40 +0000 UTC [ - ]

Ask physicists instead... Traffic flow is literally fluid mechanics. I was thinking yesterday how depending on conditions and load capacity traffic moves from laminar flow to turbulent and even greatly reduces capacity

sitkack 2021-08-18 00:21:43 +0000 UTC [ - ]

It can be modeled as fluid flow in a naive case but it diverges

    * It is compressible
    * It involves intelligent agents

The last part is hard for the 'traffic is water' crowd to understand. Lots of folks will skip a trip or defer to another time. It is the mirror of the 'all roads fill to gridlock' phenomena. You cannot build your way out of bad traffic, you have to design it and it might be less capacity than when you started.

Some components of traffic flow can be modeled as fluid flow, but traffic as a chaotic system cannot be.

ccvannorman 2021-08-17 18:46:57 +0000 UTC [ - ]

"All drivers need to be on the same navigation system. Cars can only be efficiently rerouted...."

Um.. Right, should be easy

perl4ever 2021-08-17 23:16:49 +0000 UTC [ - ]

  1. urban streets are too narrow, so ban parking
  2. urban streets are too wide, give electric cars their own lanes. 
  3. what about the bike lanes???
  4. Profit!

Even if this is a good plan, it sounds like social engineering and not math.

sitkack 2021-08-18 00:25:41 +0000 UTC [ - ]

tl;dr they are saying global optimization works, but all the actors need to get a globally optimal plan.

Retaining choice for menial task (driving from a-b) wasting millions of person years on driving locally optimal but globally optimal (you pick the horizon).

This statement is true for much of human condition. False choice wastes our attention from other areas that should get more focus.

perl4ever 2021-08-18 03:47:32 +0000 UTC [ - ]

Mike Rowe suggested everybody only making right turns to reduce the risk of driving. It was a rhetorical example of the sort of tradeoff we choose not to make, putting other things above total safety.

But I like it. It might increase distance and time traveled slightly, but it should decrease both risk and stress.

It seems like an almost achievable plan, if the government ordered Google Maps and whatever Apple's thing is called, to only produce right turn routes.

llimos 2021-08-17 11:44:59 +0000 UTC [ - ]

Does anyone know if navigation systems like Waze already do load balancing? I've had my suspicions about that for a while.

pdimitar 2021-08-17 10:03:25 +0000 UTC [ - ]

If that's the crux of what a 150 EUR hardcover book's synopsis will be, then I'll definitely pass.

llimos 2021-08-17 11:45:44 +0000 UTC [ - ]

Title should have (2020)