6 ms·

Complexity Scientist Beats Traffic Jams Through Adaptation

112 points by theafh 6y ago

ChuckMcM 6y ago
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.
- ampdepolymerase 6y ago
  Can you open source it?
  - ChuckMcM 6y ago
    Like most things I do it will end up on my github (https://github.com/ChuckM 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.
gverrilla 6y ago
most people I know know very little about complexity, and I'll be showing them this article, thanks for sharing!
thepangolino 6y ago
Sad thing is this probably won’t work in the real world due to nimby types.
- bobthepanda 6y ago
  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.
  - toast0 6y ago
    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.
    - 6y ago
tshanmu 6y ago
This is how we do it in India...just that the drivers optimise instead of the cars talking to the signals for a local optimisation. Edit: originally commented too early after reading the selfish driver part.
- ksd482 6y ago
  Are you kidding me? Traffic in India is a colossal disaster. The drivers taking things into their own hands make it 10 times worse than it has to be.
  - jgust 6y ago
    Real life example of living inside a local maxima.
bmsleight_ 6y ago
MOVA, Needs to be more than 50m or just switch to UTC. Sigh - Traffic Engineer.
iggldiggl 6y ago
I was going to say that this sounds suspiciously like something someone experimented with in Germany as well (http://stefanlaemmer.de/en/ http://stefanlaemmer.de/en/), but I've noticed that he's mentioned that example himself as well.
sandworm101 6y ago
>> 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."
- paganel 6y ago
  > 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.
- leetcrew 6y ago
  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).
- wenc 6y ago
  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). [1] https://en.wikipedia.org/wiki/Traffic_light_control_and_coordination#Coordinated_control https://en.wikipedia.org/wiki/Traffic_light_control_and_coor...
cseleborg 6y ago
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.
darepublic 6y ago
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
sova 6y ago
>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.

112 points by theafh 6y ago

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.

ampdepolymerase 6y ago
Can you open source it?
- ChuckMcM 6y ago
  Like most things I do it will end up on my github (https://github.com/ChuckM 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.

gverrilla 6y ago

most people I know know very little about complexity, and I'll be showing them this article, thanks for sharing!

thepangolino 6y ago

Sad thing is this probably won’t work in the real world due to nimby types.

bobthepanda 6y ago
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.
- toast0 6y ago
  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.
  - 6y ago

tshanmu 6y ago

This is how we do it in India...just that the drivers optimise instead of the cars talking to the signals for a local optimisation. Edit: originally commented too early after reading the selfish driver part.

ksd482 6y ago
Are you kidding me? Traffic in India is a colossal disaster. The drivers taking things into their own hands make it 10 times worse than it has to be.
- jgust 6y ago
  Real life example of living inside a local maxima.

bmsleight_ 6y ago

MOVA, Needs to be more than 50m or just switch to UTC. Sigh - Traffic Engineer.

iggldiggl 6y ago

I was going to say that this sounds suspiciously like something someone experimented with in Germany as well (http://stefanlaemmer.de/en/ http://stefanlaemmer.de/en/), but I've noticed that he's mentioned that example himself as well.

sandworm101 6y ago

>> 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."

paganel 6y ago
> 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.
leetcrew 6y ago
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).
wenc 6y ago
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). [1] https://en.wikipedia.org/wiki/Traffic_light_control_and_coordination#Coordinated_control https://en.wikipedia.org/wiki/Traffic_light_control_and_coor...

cseleborg 6y ago

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.

darepublic 6y ago

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

sova 6y ago

>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.