Hacker News Reader

Comment by purplelobster | Hacker News Reader

Something that puzzled me when I came to the US was that traffic lights (including a majority of pedestrian lights) just cycles through a certain schedule. Where I come from, traffic lights have sensors in the ground that manages the traffic. Also, all pedestrian crossings have very responsive buttons that truly work. The crossing light won't come on unless someone presses the button. In the US, I can rarely tell if the buttons are just props or if they actually do anything.

There's way more traffic in the US though, so not sure if this would work since there are so many traffic lights, and almost always someone waiting in each direction. Something that frustrates me to no end though are traffic lights (especially for left turns) that were obviously programmed with the belief that not many cars would be waiting, but is in fact heavily used. This results in having to wait 2-3 light cycles just to get through the intersection.

View on HN

This will be highly variable from place to place. I've observed them in many places in the US, but that's anecdotal too, and I was unable to find anything online.

On arterial roads in LA, there will be inductive sensors (identifiable by the circles of cut-and-resealed asphalt near the stop line) at every approach to every stoplight.

Some lights will not change unless you are at the correct place. If you creep too far over the stop line, or don't approach it close enough, you will never get a green light (until someone else comes).

Depending on day-of-week and time-of-day, the lights will be on a different tempo. They will respond to ped signals more slowly at rush hour, but more quickly around schools at dismissal time, for example.

Changing the timing of the lights is barely noticeable to drivers but can have a huge effect on road conditions. There is a freeway offramp near my home that feeds directly onto an arterial street. The geometry encouraged people to barely slow down (60 MPH offramp, 35 MPH street). A carload of people were killed on the road (lost control and slammed into a median tree at ~70 MPH -- the tree won).

There are now inductive loops as you traverse the offramp. They sense if your car is exceeding a given speed, and they trigger a red light at the end of the offramp if you are. Then, after that turns green, the system triggers another red light four blocks down the road to be sure you're not drag racing.

It has been highly effective at enforcing reasonable speeds.

> There are now inductive loops as you traverse the offramp. They sense if your car is exceeding a given speed, and they trigger a red light at the end of the offramp if you are. Then, after that turns green, the system triggers another red light four blocks down the road to be sure you're not drag racing.

Oh that's a wonderful nudge! These sound much more better than speed cameras. No need for a bureaucratic system of enforcement for the simple things.

Just as long as they're used judiciously. I can imagine the rage if the guy in front keeps triggering the red for everybody.

As it happens, you have the choice to sense and respond to traffic, or have synchronized traffic signals in a complex grid like a city center. You can't really have both.

To have synchronized signals in a city-wide two-dimensional grid of intersections, you must ignore the fact that some intersections have many cars and some have none. If you give up on the idea of synchronizing traffic, then you can have traffic sensors and working pedestrian crossing buttons.

> In the US, I can rarely tell if the buttons are just props or if they actually do anything.

It turns out that in many places, especially those that have synchronized signals, the pedestrian buttons are what might be called "placebo buttons". I'm sure you can guess what that means.

It would seem that sensors would be most useful with lower traffic volumes, and synchronization better for rush hour. Perhaps using each system at different times of the day would be theoretically optimal.

Or why not incorporate sensor information into the synchronization, and continuously plan the best light cycles for any given situation?

>It would seem that sensors would be most useful with lower traffic volumes, and synchronization better for rush hour. Perhaps using each system at different times of the day would be theoretically optimal.

Most cities do this. I live in chicago, that's how it works. Note that you can also combine sensors simultaneously with timing in certain cases- for example whether to allow an advance green for a left-turn lane or not (if there are cars in it, advance green, otherwise, normal green for both directions of traffic on that road. No change in when the next red comes)

continuously plan the best light cycles for any given situation

Because there are a lot of "any given situation[s]" that might the scenario overly complex. Take a school for instance - you have students crossing the street that want pedestrian signals. The occurs twice a day (before & after school), 5 days a week, 10 months per year. Except there are holidays when there may be no activities. And there are evenings where there may be activities like sporting events, parent/teacher meetings etc. And the occasional weekend event. and so on.

A city may have civic buildings, commercial plazas, construction sites, etc, where these events crop up. Plus roadways have to accommodate bus and bicycle traffic in addition to automobile traffic. Plus certain intersections may back up according to meteorological conditions (for example, blinding conditions caused by a low-setting sun that happens during certain parts of the year).

Trying to synchronize to all of this, in addition to incorporating sensor information and pedestrian switching becomes very complex very quickly.

The pedestrian buttons are not placebo buttons, they actually work. Generally, the pedestrian signal is set to activate in cycles of X minutes (where X is usually a multiple of 2, 3, or 5) but only if it the pedestrian button has been pressed. If the button is not pressed before the cycle begins, the pedestrian signal does not activate on that cycle but will activate on the next cycle. (For example, see dkl's comment.)

In synchronized systems, the pedestrian light works the same way--this is why some intersections will sometimes have red lights for every direction of traffic even though there are no pedestrians.

> The pedestrian buttons are not placebo buttons, they actually work.

Not if the system is grid-synchronized. In places where the system has more than one mode and adapts to the overall traffic amount, yes, but only when the system isn't in synchronize mode.

The US is a very large place.

Where you were I guess they didn't do that, but they most certainly do use sensors in many cities.

Since I haven't been everywhere yet I can't say "most", "some" or any other words like that. But sensors are used in the US.

Of the 15 or so different states I've been to, this is what I've observed. Of course I can't say that there is no black swan, but seems unlikely. At least there's a tendency towards no sensors, and again, with the traffic volumes it might not make sense to have them in most US cities.

Spent a bit of time playing with traffic and parking. One of the issues with the dominant traffic sensor tech (inductive loops in the ground) is that they can fail in cold weather: frost heaves in pavement sometimes crack the loop wires.

In the south, southwest and the pacific northwest the incidence of loops is much higher than in the northeast. There's also more driving, so the incentive to optimize that part of the experience is much higher than say, in the Boston area, where they really just want you to give up the car and use public transportation.

Cities are more and more doing this (what you specified for pedestrian lights outside the US). My problem: there's a very large intersection that I walk through daily that changed to "only allow walk when the button is pushed" and 9 times out of 10 when I walk up to it and push the button when it normally would have been set to "walk", nothing happens. Then, the next traffic cycle I get the "walk" light. However, I've also noticed that from the time I push the button until when the cross traffic starts, 90% of the time I would have had time to walk across the street. So, I've started ignoring the button and just walking when "the time is right." In this case, it has made me less safe. Twice in a few months I've been caught in the middle of the street when the light turns.

I realize that in this case, the walk portion is probably just not tuned properly, but I wonder how often that happens.

I am not a traffic flow engineer, but I have experienced what you describe in regard to the walk button behavior in many places in western Washington, US. I always reckoned that they were using a simple model in which a button press would cause the "WALK" signal to always be fired at the corresponding light configuration at the _next_ cycle start. This is most likely a safety consideration because, as you noted, there may not be enough time to safely cross.

In some complex intersections I've noticed that the 'cycle start' doesn't necessarily line up with the corresponding configuration for the walk button which was pressed. This may result in having to wait nearly two complete cycles for the walk signal in the worst case (pressing the last configuration's walk button immediately after the cycle start). The resulting behavior could be considered a bug, or perhaps just more consideration payed to vehicle traffic. But it is certainly frustrating as the pedestrian.

At many intersections I cross now the signals are automatic with the traffic flow. Most of them also contain a countdown timer indicating the number of seconds until the light changes. It seems that this behavior would be a good tradeoff for pedestrian efficiency and safety. However it comes with a bit of trust from the gov, that you know you can't walk across 5 lanes in 2 seconds.

Reply on news.ycombinator.com