Why aren't more traffic light systems synchronized? [Traffic flow physics]

I live in a city with 1 million+ fellow citizens. I do 2-3 30 minute commutes each day, and while I benefit from an expressway that takes me most of the way, I do have to travel on some roads with traffic lights.

I know that they are computerized, because they all have those big silver boxes next to the intersection that I sometimes see the police fiddling with. Yet they are all on strict timers. Yes, they have sensors too, but if the volume all ways remains heavy the sensors are useless (since somebody is always waiting in every lane).

Far too often I’ll be waiting at say a left turn arrow, and there’s a big break in the oncoming traffic. You’d think that a dynamic system would recognize the gap and flip the arrow to green, but it never does.

Maybe I am incapable of seeing all ends and what seems inefficient to me is actually the best of all worlds for the majority of drivers…<shrug>

Two lights that occur in a row particularly puzzle me. I’ll have a moderate wait at one (on the main drag), it will finally flip green, we’ll all go on down the road-only to invariably be stopped by the next one. Why the two lights aren’t set up to let said vehicle cohort license to cruise on through the next light is unknown to me. It is at the third light I often see these long gaps while waiting to turn left against said flow…

I know that Google Maps will clearly let you see all of the traffic, via green yellow and red bars. Are there any major cities which use the Google traffic data to optimize traffic flow? And if not is it simply a case of budgets falling way behind the times when it comes to updating and sychronizing all of the said equipment?

An older method of traffic control had engineers or their minions go out and note the actual speeds at which traffic flowed, after which the engineers would simply set the timing of the lights on major corridors to keep the bulk of the traffic moving. Drivers who rushed ahead would simply be caught by each light while drivers who went with the flow would rarely have to change speeds or touch their brake pedals. Metro Detroit favored that approach, (at least when I drove there in the 1960s through the early 1980s). I have driven 22 miles along Telegraph Rd. And at least ten miles along Woodward or Jefferson without braking on numerous occasions with similar, if shorter, runs on Van Dyke, Mound, or Southfield on many occasions.

In contrast, the philosophy in Cleveland has been to make sure that no one ever makes it through two green lights in a row. I do not understand what they are trying to accomplish, but it is deliberate. They talked of actually timing the lights on Mayfield Rd. about 25 years ago, but either they could not find competent engineers to accomplish it or the inner suburbs decided to not let commuters “rush” through their towns, because I never did get a trip down that street without stopping at every light.

As to using sensors to detect blockages and gaps, I doubt that any city has the funds to try that, yet.

8 Mile Rd. used to be that way; you could drive all the way from Grosse Pointe Shores to Farmington if you kept the correct speed. Alas, traffic signals are “maintained” by Detroit Lighting Department and their timing has drifted over the years.

that said, a lot of communities in Macomb county have gone to “adaptive” signals monitored by IR cameras. they don’t always operate intuitively but they can be a blessing in off-peak hours; no more getting stuck for 5 minutes at a red light because you’re on a small cross street while there’s no traffic on the main road.

One pain with synchronised lights is that if the road is blocked,
then they give the green to the traffic that is standing still for ages and ages.
Then when they give the green the other way, only a few vehicles move before the congestion on the synchronised road blocks all flow… The trouble is that there were no gaps in the congestion to allow any intersection to have its exits cleared…

Unsynchronised lights create gaps in congestion, which then allows the intersections to clear , randomly admittedly, but thats better than never.

Except on Clifton Blvd., at least. In Lakewood it’s even better: All of the cross-streets are demand only, and stay red only just long enough for cross traffic to get through. If I see a light turning red 100 feet ahead, I don’t even bother slowing down, because it’ll be green again by the time I get there.

No one’s rushing out to spend a lot of money to upgrade traffic lights. In many areas there are electr-mechanical controllers still in use, with signal heads from the 1950s. Even more modern controllers usually use NEMA controllers, while they’ve been updated with microprocessors and jazzy LCD displays the fundamental technology is from the 1970s. Controllers do sense traffic volume by the number of cars running over sensors.

In all but the most modern controllers you’re dealing with at best 1980s vintage microprocessors (the 68030 is a popular one for traffic controllers) and often built in firmware, unless you want to go with a 2070 type system and write your own custom programs. It’s not the same kind of power, flexibility and data available as a internet connected PC

Controllers are expensive due to the limited, captive market that will pay what it has to as well as liability. There’s not much profit in cheap plastic signal heads that almost anyone can make, so most companies try to make their money on controllers.

It’s not a matter of just flipping the light green for a couple of seconds to let you through, there are certain minimum times. Say you pull up to the left turn bay. You might have:
2 second delay to make sure it’s not a false call, say you drove right over the sensor and ran the red light.
2 second yellow light for oncoming traffic
2 second all red
8 second green arrow for you
2 second yellow arrow
2 second red arrow.
Are there really that many 20 second gaps? In places where there are, say suburban arterials, sometimes the signals will give a green arrow out of sequence.

It gets complicated to coordinate signals. Besides the programming, they have to be physically interconnected somehow, usually with fiber. All this programming and stringing wires costs money of course, and again agencies aren’t in a rush to update signals.

Video of my electro-mechanical controller:

My 3M Opticom “Emergency Vehicle” preemption in action

(Initial build of the transmitter had visable LED until I could buy some IR, so I had to put it right next to the sensor to trigger)

Inside a traffic controller


My trial setup, controller running at 1:18

A challenge with “synchronizing” is that in general you can’t do it, even given infinite desire & resources.

If we assume a rough grid of streets with heavy traffic moving in all 4 directions, it’s simply mathematically impossible to time the lights so everybody moving in all 4 directions can maintain a steady speed and get in sync with a wave of lights turning green for their pulse of cars. For some combinations of speeds and distances between blocks it’ll happen to work for a few blocks; for most not. And the less grid-like the street pattern, the more likely you’ll hit problems.

To be sure, if your local traffic pattern is 90%, say, westbound in the morning & eastbound in the evening with much lighter N/S traffic at all times, then you absolutely can prioritize and synchronize the flows in the one major direction. But there are costs …

At some intersections there are lots of left turners. At those intersections that represents a restriction on the flow in the opposite direction, since they have to be stopped to permit the left turners across in front of them. At other intersections along the same route there are few left turners. For smooth flow of traffic in the predominant direction the duration of the green cycle at each light must be the same. So the need for differing length left turns drives a very big delta in the length of the opposite direction green. But you can only make the opposite direction green so short before it gets to be too short to be useful. IOW, you will eventually end up not providing enough green to support your total left turn demand at the worst case light. Which unserviced traffic then backs up into the through lanes of your predominant traffic flow.

Other issue: To make it easy for the predominant direction to flow, you want a long green light. And the longer it is, the easier it is for a car that’s marginally too early or late to adjust speed to get within the lump of cars traveling in sync with the greens. Essentially the traveling time-slot of green lights is bigger, with more happy middle and less problematic edges.

But the longer that green phase gets, the longer the red phase for cross traffic gets. And there’s always some cross traffic at busy times. So what happens if we make their red light longer than most folks have patience for? They start doing unsocial things like pressing late on their yellow and well into the red. Or turning right then left into side streets or businesses to effect a 2- or 3-step crossing of the main street. And usually doing these moves hurriedly & hence with poor safety. And even if they don’t cause a few accidents, they do inject turbulence into the predominant flow which all this green synchronization is trying to comb into fast smooth non-turbulent flow.

And what happens when the cross traffic green gets so short it doesn’t flow all the traffic on it’s road? It starts to back up, perhaps to the next block where it interferes with the predominant through flow in the primary direction on the next street over.

There are lots of other similar examples of theoretical & practical obstacles.
Bottom line: There are real physics & math obstacles to good sync. There are real behavioral obstacles to good sync. Managers nothing to make smarter light timing is an abdication of responsibility; but drivers expecting huge gains from smarter light timing is unrealistic. A few percent improvement is plausible, but that only offset a few years traffic growth & you’re back in the same situation.

The big flaw in your logic is that your field of view far exceeds that of a traffic controllers. Typically, the controller uses a magnetic loop (or two) near the intersection to detect the presence of waiting vehicles. It cannot detect how many vehicles are waiting, and it can’t tell how much time it’ll be until another vehicle passes over that detection loop.

More modern systems can use cameras and fuzzy logic to watch a few car lengths of road for vehicles, but even those can only detect the first 1-3 vehicles in line at an intersection.

Also, the sensors are usually linked to timers, so the controller thinks “I’ll let this lane go, and check back in 1500 milliseconds to see if it’s empty”. 1.5 seconds later, there’s still traffic (the line is just starting to move), so it allows flow for another 1.5 seconds. After a few cycles of this, traffic has picked up enough speed that 1.5 seconds represents a good-sized gap, but the controller still lets them through because it only knows that there are vehicles there every time it checks. Thankfully, there’s a limit to the number of cycles allotted to each flow, but that can be a long wait if the controller happens to be using long cycles during that time.

What I personally don’t understand is why more streets don’t go to blinking yellow mode late at night when there is little traffic. This was pretty common where I lived in Europe. Where I live here, now, there are streets that are practically deserted in the late hours, with minor cross-streets that go into commercial sections of town (trucking/warehouses/etc) that seem like they can easily be converted to blinking yellow from, say, 10 p.m. to 4 a.m., give or take an hour, without interrupting the rare vehicle that may want to get out of this section of town. Or, heck, even blinking reds where appropriate instead of a 60 second stop that is pointless and wasting gas 99% of the time.

An anecdote to illustrate. In the 1970s, I resided for a while in NYC. In the village in lower Manhattan. I actually had occasion to drive this van for a guy I was working for. Driving in Manhattan,or anywhere in New York is generally pretty insane. During heavy traffic, rush-hour times (heh) it could take 45 minutes to get two or three blocks across midtown. Simply too many cars to manage traffic at all.

I had one job for this guy that had me way out in Queens for an all night assignment. I came home somewhere around 4:30-5:30, a magic time in New York between the bars closing and the morning street people coming out. It was then, anyway. It was winter still, so it didn’t get light too early. I drove across the Queensborough Bridge (the 59th St. Bridge) which enters Manhattan at 60th St. I hooked a left on 2nd Avenue (I think that’s the first downtown oneway) and proceeded to hit EVERY GREEN LIGHT on the avenue all the way down to Houston St. More than 60 green lights in a row. I was getting giddy as I got closer and closer, and when turned right on Houston (that’s HOW-Stun) I had to stop and revel in amazement before I could finish getting back to the apartment.

Absolutely true story. Traffic lights can be synched, but traffic is it’s own monster.

Here in Minnesota, Governor Dayton has just proposed a major statewide plan to synchronize traffic signals, to inprove employment by reducing commute times and also reduce gas use, reduce pollution, and decrease pollution-related health problems.

But it is quite costly – they’re talking about something like a third of a billion dollars to do this. (And it will probably be killed by the Republican House, because most of the areas with significant traffic lights and road congestion are in the metro area, which votes Democratic-Farmer-Labor.)

Why aren’t more traffic light systems synchronized? [Traffic flow physics].
Why? because it would cost very little, save the environment and make people happy.

It’s beyond the grasp of politicians why are too busy flying from place to place to lecture the little people on how they should conduct themselves.

Every day millions of people invent new swear words to throw at traffic backups. It’s a virtual no brainer and yet it’s as if it were a deliberate act to maintain the status quo.

Los Angeles did it two years ago, and someone in the legislature determined that there has been a 16% improvement in travel time, though I don’t know how they figured that out. It seems about right, though, from my personal experience.

there’s something to this… by me, I’ve noticed that the signals at “minor” intersections which used to go “flashing yellow/red” after about 10 p.m. don’t anymore. No idea why.

There are very complex algorithms used to control traffic flow using traffic light timing. Sensors in the road measure traffic flow and the lights are coordinated to maximise rate of flow.

About 80 km/h is optimum, above this the cars need to space out further. Below and it is just slower.

I know this because there was an interesting interview on ABC radio last week of a person who does this for a job.

This report is similar to the interview I heard on the radio.

I had a dream where all of the traffic lights turned blue tomorrow…
And it shined emptiness down upon my bed

Optimum for what? It might be optimum for one particular road, but it can’t be optimum for everything.

I was thinking about bringing up this point. A great many traffic lights could safely be turned to blinking from about 9PM to 6PM - far more than actually are.

In addition to lights that should be blinking, there are just too damn many traffic lights. Where I lived formerly, I drove the same route to work for 12 years. During that 12 years, the number of traffic lights increased from six to 12. Where I live now, the number of lights in my town has increased from three to seven in ten years. FAR too often, I am stuck at a red light when I can see both directions and could easily drive through or turn. I have done exactly that on occasion when I get stuck at a red light that doesn’t turn green in a reasonable amount of time. Yes, I’m risking a ticket, but I only do it when I can see all nearby cars and ascertain that none are popo.