"Self Driving" Traffic Lights?

[ERD50]

We've had many discussions on Self Driving Cars (SDC) over the past few years, and I'm interested in a parallel path - Smart Traffic Lights (SMT).

It just seems that traffic lights with a small subset and expense of the capabilities of the current crop of SDC would be helpful, save lives, injuries, and accident $$$, save time and fuel, and be far easier to implement than SDC. A lot of intersections already have cameras.

A traffic Light does not need to deal with anywhere near the unknowns of an SDC. The layout is known and static. The software would be far simpler, and it seems extremely unlikely to have any dangerous adverse affect (I'm sure present traffic lights have a 'fail safe' mode so that you can't have all GREEN for example).

For example, at busy intersections near us, the most dangerous situation I see is people trying to scramble to get through a short green arrow for a left turn. Safety and traffic would be greatly improved, if that green arrow was intelligent, and matched the time to the number of cars in the left turn lane (up to a limit, but it's silly to let 4 cars go, and stop the 5th - let all 5 go - or be timed for 4 cars, when only 1 is waiting to turn). So the left turners keep going until that yellow light is red, the oncoming traffic hesitates to go even though they have a green, and that frustrates the cars behind them. And then those cars are rushing to get through before they get a yellow. And if the oncoming traffic proceeds, you could have an accident or near miss (and then maybe they get rear-ended).

Or if the SMT saw a car approaching a RED too fast, it could start flashing alarms (audio and visual) and stop the cross traffic.

And just smooth flow to conserve time, gas and brake wear. How many times have you approached a green light, just to have it turn as you got closer, but there was no cross traffic, and no reason for you to stop? And of course, Murphy assures that some cross traffic appears just in time for the light to change and make them stop. What a waste!

And what they learn, and the mass production of these systems might help improve the SDC/assistance tech on cars as well. They can all work together.

The only downside I can see is - will people start playing 'chicken', and just keep going, thinking the light will safely let them proceed? I imagine the camera will catch that, and they can be ticketed. There would be no need for "Right on RED after stopping to yield" (the current complaints of the Red light cameras), if it's safe to go, you will get a GREEN.

What say you?

-ERD50

 

[Alan]

Are you sure they don’t exist now?

We’ve had smart traffic lights in the UK since the 1980’s. I learned a whole about them in 1984 when I worked on a big project automating all the fault reporting on traffic lights in Greater London (2,300 junctions).

Even back then cells of up to 50 junctions worked together using SCOOT algorithms to dynamically change the timings on lights to speed up traffic flow. (SCOOT algorithms were developed in 1973)
SCOOT = Split Cycle Offset Optimization Technique


http://ocw.nctu.edu.tw/course/sc011/2012-08-22.pdf

Traffic Flow and Control

By the time you've finished this course, you'll be able to:

interpret basic traffic measurements to calculate speed flow and density
select and design appropriate traffic management systems
calculate basic timings for an isolated signal installation
design a coordinated signal installation
estimate emissions from SCOOT algorithms

 

[CaptTom]

We've had many discussions on Self Driving Cars (SDC) over the past few years, and I'm interested in a parallel path - Smart Traffic Lights (SMT)....

What say you?

I say it's long overdue. They just changed the light at the end of my street to wait about 2.5 minutes before changing green for exiting traffic - even if there are no vehicles at all on the main road!

I also say it'll never happen. Politically, there's too much invested in the existing system.

Every time there's a problem, the only solution our illustrious highway department engineers seem to comprehend is a new traffic light. My guess is the municipality or state just writes up a grant request and someone else (Fed or State government) pays. And frankly, I don't think career advancement in the Highway Department requires an above-average IQ or a willingness to embrace new technology. I suspect it's better to take the easy way out and don't make waves.

When the only tool you have is a hammer, every problem looks like a nail.

[Edit: Alan beat me to the punch. I was basing my reply on what I experience here in the US. Glad to hear there are places where a broader view is taken!]

 

[JoeWras]

What I say is that there is a lot of room for improvement in our infrastructure. My city supposedly made a huge investment in syncing lights. The net effect seems to be 3 minute wait times for side street exiting, which can be very frustrating to non-thru traffic. Seems like a fail to me.
[EDIT: CaptTom was writing his reply same time as me. I see he has the same side street exit issue when they "improved" our traffic signals. Maybe the same city.]

We also see a lot of failures of sensors for the lights. Utility crews constantly cut the magnetic sensors. We did have a set of camera based vision sensors rolled out about 20 years ago. These all lasted about 5 years before NCDMV abandoned them. Too many problems.

Then there are the cities that have the special strobe sensor to allow emergency vehicles through. Many of those have been hacked.

Problems, problems, problems. But yeah, I like your thinking. Ultimately, a real curb to curb SDC is going to have to be systematic.

 

[Alan]

Speaking only about the UK, smart traffic lights also react automatically to emergency vehicles and public transport such as buses. When I worked on lights in the 80’s it was transponders carried on the vehicles that communicated with the lights as they approached and they were experimenting with image recognition to see the bus routes displayed on an approaching bus to change the lights in the correct direction of the bus. (This was to save the cost and programming of transponders on every bus). Emergency vehicles would change the lights to all red when they approached with blue lights flashing and transponder emitting the right code.

 

[Ronstar]

Lots of variables out there. Some areas have the green arrow before the green light, some after.

I'm not aware of much real-time automation of traffic signals aside from those controlled by emergency vehicles.

From what I've seen, the timing is up to the engineering group that designs the traffic lights at a particular intersection. I'm not sure that there is a standard design.

 

[easysurfer]

I think countdown traffic lights should be mandatory. So helpful to know I have 10 seconds and counting before the light changes. Both as a driver and pedestrian.

 

[Walt34]

I also say it'll never happen. Politically, there's too much invested in the existing system.

Every time there's a problem, the only solution our illustrious highway department engineers seem to comprehend is a new traffic light.

I think it depends on the jurisdiction. Where I worked they did implement a traffic light control system that purportedly did adjust timing depending on traffic volume and direction (i.e., morning timing was different from evening). From the driver's perspective though, I couldn't see that it made a bit of difference because no matter what the traffic engineers did, traffic volume was simply too high for the existing roadways.

I had a chance to talk with some of the people who designed the roadways, intersections, traffic light control systems, etc. and that was very enlightening. According to them, they could often foresee 10+ years out that some section of roadway or intersection was going to be a huge problem, and they could often resolve it, but if, and only if, they could get the money to make the correct changes.

In many cases though, the volumes of cars were such that even throwing gobs of money at the problem wouldn't solve it because of people's behavior. If you make it easier to drive, then more people drive because that's easier than taking a bus or subway, which in turn drives up traffic volume, thus negating the value of improvements made to the roadway or intersection. So the average driver doesn't see any benefit to the improvements because net, to him, there isn't any.

The book Traffic expounds on this a lot and is an interesting read.

Back to the original topic, yes, I think smart traffic lights will help but they won't make a dramatic difference.

 

[GrayHare]

Around here intersections that were newly signalled within the past 10 to 20 years operate with some intelligence via sensors in the pavement. As for older intersections, some have been retrofitted. Wider retrofitting might be discouraged by better models of this tech expected Real Soon Now.

 

[Ronstar]

+1. Or at least standard length yellow lights. It's nothing but a trick that some communities have much shorter yellow lights purely to generate ticket revenue from out-of-town motorists.

Exactly. I was pulled over for running a red light on a left turn arrow to an interstate ramp. The arrow was green when I started my turn, but turned yellow and red before I finished my turn. ridiculous.

I explained to the officer that the yellow light length was too short compared to standards. He checked, found out I was right, and didn't give me a ticket. I vaguely remember hearing of expressway engineers controlling traffic on their roadways by throttling traffic coming on to their ramps.

 

[Midpack]

For example, at busy intersections near us, the most dangerous situation I see is people trying to scramble to get through a short green arrow for a left turn. Safety and traffic would be greatly improved, if that green arrow was intelligent, and matched the time to the number of cars in the left turn lane (up to a limit, but it's silly to let 4 cars go, and stop the 5th - let all 5 go - or be timed for 4 cars, when only 1 is waiting to turn). So the left turners keep going until that yellow light is red, the oncoming traffic hesitates to go even though they have a green, and that frustrates the cars behind them. And then those cars are rushing to get through before they get a yellow. And if the oncoming traffic proceeds, you could have an accident or near miss (and then maybe they get rear-ended).

Or if the SMT saw a car approaching a RED too fast, it could start flashing alarms (audio and visual) and stop the cross traffic.

The root cause is bad human judgement, not the lack of driver tools. Inductive loops which have been in use for many, many years can vary the duration of turn arrows based on number of cars, they’re not all on dumb timers. And in your example of 5 cars waiting to turn, all that tech and software won’t prevent a 6th, 7th, 8th car entering the turn lane just after the turn light sequence begins from trying to beat the yellow light - if left to humans.

Again 90% of accidents/ fatalities are human error. My money is on vehicles designed to fully protect us from ourselves, and taking judgement away from drivers, more than driver assistance or added external technology. There will still be some accidents/fatalities, but I’d hope anyone would take a 90% (or some large number) reduction over the status quo.

But as you noted, we’ve ‘d this...

 

[ERD50]

Just have a minute to reply, before we head out...

I know that pavement sensors are pretty common, but to be honest, I just haven't seen very much intelligence applied in their use.

And I've heard of the systems to time lights through a city to improve flow, but this doesn't seem to address the things like a better timing of left turns etc, and that could be used at many lights that really don't need syncing with others so much. It's just one application, and there could be so many more.

-ERD50

 

[Gumby]

Midpack is right; we need better drivers. There is mall close by and there is constantly a traffic jam in front of it because people run the light in the left turn lanes. Sometimes this is a result of the guy in front who can't look up from his phone to move when the green arrow comes on. The turners are often unable to completely clear the box after they run the light, so the oncoming or cross traffic then can't move. As people get more frustrated, they edge right up to the guy in front and end up blocking the box in every direction. And it all starts because some people can't just go when they're supposed to go and stop when they're supposed to stop. If I were the police chief, I'd be handing out tickets non-stop at that intersection. I'm pretty sure there is no technological solution.

 

[jollystomper]

Near our house a couple of recent traffic lights seem to have smart technology. Every time I seem to approach them when they are red and there is no cross traffic, they turn green. If there is no traffic behind me I notice that within seconds of me going through they are back to red. One of them is an entrance to a shopping center, at it seems as soon as 3-4 cars are waiting to exit the light will change green to allow the traffic to exit.

But I agree the real issue is the drivers. About a year ago a friend of our youngest child was killed when she slowed down in a left turn lane because the light had turned yellow and she didn't want to risk running a red light. A driver well exceeding the speed limit thought she could weave around the traffic ahead of her, veered into the lane, and rear ended her. The drivers excuse was "I thought she would go through the light and I would have space to get back into the through lane".

 

[Music Lover]

Again 90% of accidents/ fatalities are human error.

It's probably greater than 99%. Other than wildlife jumping in front of a vehicle and and other similar incidents, I can't think of very many collisions that are not the fault of at least one of the drivers. Heavy rain...drive for the conditions. Ice...drive for the conditions.

Some of the newer lights around here are "smart lights". If you're the only vehicle around and are approaching a red light, it will change to green when you get close. They also seem to have with better timing when it's busy, for example, a turning light stays on long enough to clear the turn lane and no lanes are allowed to back up too much.

 

[sengsational]

I don't think throwing technology at this problem will help much.

When traffic is heavy, then the intersection is highly utilized, no matter how incorrectly tuned the traffic light is. If the intersection is highly utilized, no amount of technology is going to improve the throughput. It might make things more "fair", but fairness is for wimps.

When traffic is light, we all notice the intersection is empty, highly underutilized. That's when I make right, make a u-turn, and another right. Problem solved.

 

[Midpack]

It's probably greater than 99%. Other than wildlife jumping in front of a vehicle and and other similar incidents, I can't think of very many collisions that are not the fault of at least one of the drivers. Heavy rain...drive for the conditions. Ice...drive for the conditions.

I didn’t pull 90% out of thin air, that’s what several credible studies have concluded over the past few decades. I don’t know about Canada, but the latest NHTSA stats have been 93-94%, probably up with distracted driving which has clearly gone up. I’ve never seen a study that shows 99%, but 90% is bad enough, and enough reason full SDC’s can’t come too soon.

 

[ERD50]

Near our house a couple of recent traffic lights seem to have smart technology. Every time I seem to approach them when they are red and there is no cross traffic, they turn green. If there is no traffic behind me I notice that within seconds of me going through they are back to red. One of them is an entrance to a shopping center, at it seems as soon as 3-4 cars are waiting to exit the light will change green to allow the traffic to exit. ...

I just haven't noticed any traffic lights near me that actually appear to use much/any intelligence, other than maybe some very rough go/no-go decision on a left hand green arrow - skip it if no cars are in the turn lane. But I have not noticed them dynamically changing the timing based on how many cars are in the left lane, and weighing that against other traffic flow.

At least from what I can tell, dynamic lights like you described seem to be rare. Within a 10 mile radius of me, and I'm sort of suburban/semi-rural, with two medium sized towns in that radius ( 15-20,000 population each), and I'd estimate several dozen intersections that could be made safer and have better flow if they were 'smart'.

... But I agree the real issue is the drivers. About a year ago a friend of our youngest child was killed when she slowed down in a left turn lane because the light had turned yellow and she didn't want to risk running a red light. A driver well exceeding the speed limit thought she could weave around the traffic ahead of her, veered into the lane, and rear ended her. The drivers excuse was "I thought she would go through the light and I would have space to get back into the through lane".

Smart Traffic Lights would not be a 100% solution against bad drivers, but then we get into the "perfect is the enemy of good" viewpoint. I still think a little technology could go a long way towards reducing accidents. Even in this case, that bad driver might not have been thinking that the other driver would have gone through the light, as that would be far less common with more sensible, dynamic response by the Smart Traffic Light.

Smart Traffic Light = STL. I have no idea why I used SMT as shorthand for that in my OP

...and enough reason full SDC’s can’t come too soon.

But that's an issue - SDC's won't be here "soon enough", the tech and wide scale adoption is far off (though that's a better discussion for the SDC thread).

But relevant to this thread, I don't see it as either/or. Seems like everyone could benefit very soon with smarter traffic lights, while we wait for SDC to become common. It would not be dependent on new cars on the road at some future date. I wonder if part of the problem is STL isn't as "sexy" as SDC? And in a way, SDC may be draining attention and resources away from STL?

I would think Elon Musk would be thrilled at getting his tech and brand name into many thousands and thousands of traffic lights across the country ( "This intersection guided by TESLA Technology"). It would be a way to leverage their SDC tech, and diversify their income stream.

-ERD50

 

[dixonge]

A few thoughts on this...

One thing that seems to me to be utterly logical is AEB (Automatic Emergency Braking). After having driven in an AEB-equipped Subaru a couple of years ago I'm all-in on this tech. But many automakers have been kicking this can down the road. Profit over safety, it would seem.

https://www.consumerreports.org/car...-stand-on-automatic-emergency-braking-pledge/

Ford, Chrysler and GM lead the way on providing the *least* amount of safety consciousness, evidently.

If AEB were implemented, combined with smart traffic lights, it would save a lot of injury and death. Throw Self-driving cars into the mix, and AEB just continues to make sense, logically. It would increase safety an additional level or three.

An additional factor is driverless mass transit. In the USA it would appear that only airports have adopted this tech. Europe is *way* ahead of us. Add this to the other tech (traffic lights and SDC's) and it gets real interesting. And safe.

 

[Alan]

A few thoughts on this...

One thing that seems to me to be utterly logical is AEB (Automatic Emergency Braking). After having driven in an AEB-equipped Subaru a couple of years ago I'm all-in on this tech. But many automakers have been kicking this can down the road. Profit over safety, it would seem.

https://www.consumerreports.org/car...-stand-on-automatic-emergency-braking-pledge/

Ford, Chrysler and GM lead the way on providing the *least* amount of safety consciousness, evidently.

If AEB were implemented, combined with smart traffic lights, it would save a lot of injury and death. Throw Self-driving cars into the mix, and AEB just continues to make sense, logically. It would increase safety an additional level or three.

An additional factor is driverless mass transit. In the USA it would appear that only airports have adopted this tech. Europe is *way* ahead of us. Add this to the other tech (traffic lights and SDC's) and it gets real interesting. And safe.

The Toyota Prius we bought last year has AEB and I love it. It has deployed in earnest a couple of times beating my foot to the pedal when someone darted out in front of me. One time I braked when all the traffic on the motorway stopped ahead of me and the system decided I wasn’t slowing quickly enough and increased the pressure on the brakes. The car only just fits into our garage length wise and when I get within a foot of the back wall the brakes come on and I stop dead, which is exactly where I need to be.

 

[NW-Bound]

AEB is something I would love to have in a new car.

The theory of operation may sound simple, but the actual implementation is not. Imagine a malfunction that stomps on the brake accidentally while you are cruising down a congested freeway. Not good!

I am sure car makers test their system very very carefully before releasing to the public.

 

[Midpack]

But relevant to this thread, I don't see it as either/or. Seems like everyone could benefit very soon with smarter traffic lights, while we wait for SDC to become common. It would not be dependent on new cars on the road at some future date. I wonder if part of the problem is STL isn't as "sexy" as SDC? And in a way, SDC may be draining attention and resources away from STL?

How many accidents/fatalities are due to turn lane collisions (it’s a small number, look at NHTSA reports) to begin with?

What would STL’s cost? And I’d question if the situation you’re trying to solve would be changed much at all, as I asked earlier in red. You’ll still have people trying to beat yellow lights. How much more effective would your lights be than existing inductive loops, there are many where I live and I’m in a small town, about 29K?

But carry on, I’m done with this thread.

The root cause is bad human judgement, not the lack of driver tools. Inductive loops which have been in use for many, many years can vary the duration of turn arrows based on number of cars, they’re not all on dumb timers. And in your example of 5 cars waiting to turn, all that tech and software won’t prevent a 6th, 7th, 8th car entering the turn lane just after the turn light sequence begins from trying to beat the yellow light - if left to humans.

 

[ERD50]

Originally Posted by Midpack
The root cause is bad human judgement, not the lack of driver tools. Inductive loops which have been in use for many, many years can vary the duration of turn arrows based on number of cars, they’re not all on dumb timers. And in your example of 5 cars waiting to turn, all that tech and software won’t prevent a 6th, 7th, 8th car entering the turn lane just after the turn light sequence begins from trying to beat the yellow light - if left to humans.

Simple arithmetic. At least from my observations, the green arrow is often shorter than the line. If the green arrow left the full turn lane through 90% of the time, that would be 90% fewer times anyone would try to rush through. I see it shut them off at 5, when it makes sense to let 6, 7, or 8 through. It's only a few more seconds.

I'm not claiming it's any kind of cure-all. But it seems like compared to the massive amount of tech and AI in an SDC, this is child's play, and perhaps (though I think likely), it would be very cost effective. And we could do it today, not some undefined point in the future when SDC has to be so good that we remove driver decisions 100%.

You keep using the "good versus perfect" argument - that's approriate, so why not get some 'good' from STL while we wait for SDC?

I'll repeat, I just don't see the inductive loops driving "smart" timing. If it is do-able, it's not being done very often from what I can see.

If you have a source that says these inductive loops are routinely being used to smartly adjust the turn arrow and can tell how many cars are in line (not just if the lane was empty versus one car), please share. I'll do some searching as well, but my eyes say this isn't being done routinely.

-ERD50

 

[HFWR]

https://www.fhwa.dot.gov/advancedresearch/pubs/SmartTrafficSignal.pdf

 

[Scrapr]

I'll repeat, I just don't see the inductive loops driving "smart" timing. If it is do-able, it's not being done very often from what I can see.

If you have a source that says these inductive loops are routinely being used to smartly adjust the turn arrow and can tell how many cars are in line (not just if the lane was empty versus one car), please share. I'll do some searching as well, but my eyes say this isn't being done routinely.

-ERD50

IMO...what I think the loops are doing is measuring distance between cars. So 1st car goes, 2nd car right on the tail, 3rd car is a bit back as they are not watching the light..but still make the light. 4th car is watching a you tube video and is back farther still. They are the one that runs the red light. And complains about a short light. Now if all 4 left evenly spaced (not possible but whatever) I think all 4 make it with maybe 1-2 more as well. Utopian science in a perfect world. But also a perfect set up for autonomous cars.

my credentials? I went to college with some Civils. They stayed in a did differential equations. I went out drinking