This is awesome, I built a toy ML project which ran in a simulated traffic system as a way to explore ML. Like the author, if the individual traffic lights modulate to traffic density "waves" emerge in the simulation of vehicles going through intersections. And, as noted by the author, you only have to do this at the light, it doesn't need any other inputs except an ability to recognize cars coming toward it.
One of the surprising thing is its "disaster" response. Which is to say to totally wipe out like 6 intersections to through traffic. The lights with a way around begin cycling more rapidly, this makes the flows that the lights downstream "see" have more openings in the direction of the exit so they start cycling more vehicles that way, and if the simulation is set up so that vehicles begin to diverge to open roads if enough time has passed with them sitting at an intersection. All of the cars begin to flow around the obstruction.
I couldn't get my "car model" to do an illegal u-turn (as I've seen many cars do in real life when stuck at a traffic light where they could go back the other way) so there were always vehicles that remained stuck trying to get through to impenetrable divide but it was a nice example of emergent behavior from a complex system with automaton type rulesets.
Like most things I do it will end up on my github (https://github.com/ChuckM) eventually. :-) At one time I thought this might have some commercial appeal to city traffic engineers but talking with the folks in Sunnyvale was disappointing to say the least.
>> Coordinating all these programmed traffic lights to keep vehicles moving is a problem.
Is that the goal of traffic light timings? I'm not sure. I think timings are more often used specifically to slow traffic (ie "traffic calming") or to prioritize one form or direction of movement over another.
I remember watching a BBC doc about traffic control in london. The operators all talked about helping vehicles move through the cities, but the policy makers spoke only of reducing traffic by essentially making travel more painful: "The easiest trip to manage is the one that doesn't happen."
I think congestion reduction is probably the more common goal [1] from the point of view of the environment, economics and road capacity utilization.
Traffic calming is probably a localized goal in more pedestrian heavy areas or where there is a vulnerable population (children, disabled, etc.).
Traffic calming also benefits retail stores. I once lived in a city with an artery with unsynchronized lights. Getting to the other side of town was an exercise in frustration, but I can imagine the brick and mortar stores along it benefited from patrons making impulse stops. (I could be remembering wrongly but I believe they lobbied for the lights to be unsynchronized).
> Is that the goal of traffic light timings? I'm not sure. I think timings are more often used specifically to slow traffic
Exactly. There's also the issue of pedestrians (of which most of the time I am part of), i.e. we, as pedestrians, don't really care if the traffic in front of us is fluid or not, we just don't want to wait 5 minutes or more at an intersection in order for the pedestrian green-light to be on.
probably differs city to city. in my city the main north-south and east-west arteries are one way. the greens are synchronized so that a vehicle traveling at or slightly below the speed limit will never hit a red light until it leaves the coordinated downtown area (with a couple exceptions when crossing other major arteries perpendicularly).
The argument that ultimately, politics is what prevents real progress here, rings depressingly true. I heard a journalist explaining that in Berlin, it took 57 administrative steps to put a new traffic light at an intersection (Berliners have a very low opinion of their administration). I guess it's indirectly a political problem because since, as the article says, nobody can agree on how to solve a problem, there isn't the necessary political momentum to reshape the administration to do what needs to be done quickly.
I was going to say that this sounds suspiciously like something someone experimented with in Germany as well (http://stefanlaemmer.de/en/), but I've noticed that he's mentioned that example himself as well.
>If you can predict when cars will cross intersections, then you can prevent them from crashing even if they never stop; you just kind of decrease their speed. They almost never crashed in my simulation, and of course it was much more efficient
That's exactly right, the brake is the enemy of efficient driving and a system that can shirk this (by say, not accelerating to let other cars pass far ahead) would decrease fuel exhaustion by a helpful coefficient.
As much as I like this is the complexity science actually beating the traffic jams or just discussing his unfulfilled desire to do so which is currently blocked by politics
I mean this wouldn't really work because it really optimizes one mode of travel (the car) at the expense of nearly everyone else. People still need time to cross the street on foot or on a bicycle.
It's hard to tell exactly what the article is arguing for. I don't know what the status quo on traffic control is for Mexico City.
I know some big cities have all their lights on a set schedule. Some big cities have a fully integrated system where intersections communicate and predict arrivals to one intersection based on exits from another.
I think the researcher is suggesting that prediction is too difficult to make good results from fixed scheduling or communicating intersections, and that instead, each intersection should adapt to its local sensors. That can include bike detectors and pedestrian buttons.
I think this idea generally works. I've seen lots of traffic signals that respond to cars, bikes, and pedestrians (via button, sadly, I've never seen an inductive loop detector in a sidewalk, would have to be sensitive to count a person). You get better results if you can put sensors a bit back from the intersection to determine queue length, although that adds some expense. It's also nice if traffic controls can be placed into an event mode to help traffic disperse from things like sporting events, commuter trains, and ferries; at least for a few intersections around the site, adaptation may not be enough if there's steady cross traffic, during the event, it may be more efficient to make cross traffic wait longer than normal reasonable limits, or add a cycle for all direction walking, so pedestrians can make diagnol crossings more easily, etc.
Cars idling at a stop light when there's no conflicting traffic (including bikes and peds) is bad for everyone, not just cars. Reducing that is good for everyone, at least in the small --- as the article mentions, improved traffic conditions may encourage more driving, which may not be beneficial for bikes and peds.
I agree intelligent traffic lights that prevent needless waiting would be awesome. I guess in part it becomes a liability thing.. what if the sensors don't detect a child and there is an accident. The city probably doesn't want that liability
If I remember correctly, some types of pedestrian crossings in the UK actually use sensors to automatically detect people waiting near the lights (and also detect if they're still around, or have actually crossed/otherwise disappeared before the pedestrian phase [1]).
If you want to do more than that, i.e. detect pedestrians before they actually reach the crossing, the problem is that pedestrians don't have turning lanes, so it's hard to tell which way people approaching the crossing actually want to turn.
[1] If I'm not severely misremembering things, one thing I like about UK traffic regulations is that it's not actually illegal to cross against a red light as a pedestrian if you're not obstructing anybody. Obviously if there's an accident you're going to be liable if you crossed against a red light, but if the road is clear and nothing happens you cannot be fined.
>one thing I like about UK traffic regulations is that it's not actually illegal to cross against a red light as a pedestrian if you're not obstructing anybody.
Correct "jaywalking" is not a thing here and I think there'd be an uprising if it was.
>Obviously if there's an accident you're going to be liable if you crossed against a red light
Even this is not blankly the case, quite often it would return a split liability result [1], some recent examples [2] are infuriating.
you sound like you have some expertise on this, so I would like your opinion on the following: on my street they have put up signs that display how many seconds it is to the light shift at the cross walk, ostensibly for bikes, but totally readable by cars - it's about 4 houses to the light. It struck me as being idiotic and wasteful for the environment to have made this sign and to use the electricity to power it, but also I would think there would be side effects to the traffic flow that would actually make the sign counterproductive - do you know what the usefulness of these signs is?
I'm just an observant amatuer, no professional traffic experience, although, someone linked a manual for LA's traffic management system (it's a commercial system, but the manual is customized for LA), and reading through some of that confirmed the observations of capability I've made over time.
I haven't seen a time until green? (or is time until yellow/red?) sign like that personally, so I'm not sure about usefulness. I'd imagine electricity cost is pretty negligible, assuming LED lighting etc.
I used to drive frequently in areas with countdown timers for pedestrian cycles; if the vehicle cycle was also ending, the vehicle light would turn yellow when the pedestrian signal hit zero. As a driver in the opposing direction, seeing the countdown helped me decide if I wanted to wait for the signal or make a right turn on red (if conditions permitted). As a driver in the direction of the countdown, it let me plan for a yellow, and I think I was able to make better decisions about stopping or going on yellow; I imagine some drivers might take the countdowns as a queue to speedup to make the light, anecdotally I didn't see that, but you'd need real numbers to decide.
Anyway, if the approach to the intersection is a downhill, as a cyclist, it's nice to know earlier if I need to brake or not, or if I could make it through if I pedal a bit. If there's enough bike traffic to support that, it sounds nice. I don't think you could signal to bikes and not cars --- at least where I've riden, bicycles at reasonable speed are vehicles and should travel in vehicular lanes unless a restricted lane is available and appropriate (bike lanes are restricted to bikes, but bikes are not restricted to bike lanes). Given that, I don't know how you'd signal to cyclists and not car drivers.
Ok, but it really looks like it is placed for the benefit of cyclists coming down the hill of the street. Also it is far enough back that when you are waiting for it to turn green you can't see it.
One of the surprising thing is its "disaster" response. Which is to say to totally wipe out like 6 intersections to through traffic. The lights with a way around begin cycling more rapidly, this makes the flows that the lights downstream "see" have more openings in the direction of the exit so they start cycling more vehicles that way, and if the simulation is set up so that vehicles begin to diverge to open roads if enough time has passed with them sitting at an intersection. All of the cars begin to flow around the obstruction.
I couldn't get my "car model" to do an illegal u-turn (as I've seen many cars do in real life when stuck at a traffic light where they could go back the other way) so there were always vehicles that remained stuck trying to get through to impenetrable divide but it was a nice example of emergent behavior from a complex system with automaton type rulesets.