Smart traffic lights

[kalvado]

Maybe stupid question - how smart are traffic light controllers these days, can they become smarter?
My main concern is traffic detectors - a failure can go unnoticed for weeks, and glitch in detecting a vehicle can easily cause a mini-jam (e.g. in protected turn lane, with green only on vehicle detection). Looks like controller cannot detect deviation from the usual pattern and respond - although that seems a relatively simple task for single-digit priced industrial controller these days (such controller may be more powerful than my first computer anyway). Calling home with statistics and potential problems is another no-brainer option - although may be a bit more expensive.
DO these things actually exist but didn't make it to remote state capitals yet, or there is an issue with implementing such "advanced" features?

[jeffandnicole]

It's probably not an issue with not being unnoticed - the police no doubt go thru the intersection on a daily basis, and most likely the road crews with jurisdiction on that road go thru there daily or weekly as well - but rather an issue of getting a crew out there to fix it.  It also depends what the problem is.  A simple fix may just take minutes.  A software/hardware/loop/camera/wiring issue can take much longer.  There should be a default option the road crew can select to put everything into a max time cycle so that no movement is missed.

Of course, it always helps to make phone calls or an email yourself.  If it is an issue, someone has to mention it, and most people think everyone else should be the one mentioning it. 

Signals that communicate with the regional or state DOT office do exist, but they tend to be part of a coordinated system.

[cbeach40]

Any actuated timing plan will have a fixed-time plan as a backup in the event of a fault. Usually recognizes it within a handful of cycles. As jeffandnicole said, road crews and other personnel go through on a regular basis and will notice if something's wrong. Actual inspections are supposed to occur weekly (which is the biggest part of the reason why signals operating costs are so insanely high - go roundabouts!). And as jeffandnicole also said, because there's a fault doesn't mean that its easily solved. A software issue could be fixed in minutes, something like a damaged induction loop would cost thousands of dollars to fix, so securing funding to repair it could be quick in a high profile location, or could take years at an intersection of lesser importance.

[kalvado]

It's probably not an issue with not being unnoticed - the police no doubt go thru the intersection on a daily basis, and most likely the road crews with jurisdiction on that road go thru there daily or weekly as well - but rather an issue of getting a crew out there to fix it. 

Broken detectors are basically bread and butter for local newspaper traffic guy. Usually letter starts with "for past  few weeks... "

Of course, it always helps to make phone calls or an email yourself.  If it is an issue, someone has to mention it, and most people think everyone else should be the one mentioning it. 

Signals that communicate with the regional or state DOT office do exist, but they tend to be part of a coordinated system.

I may be a bit more curious about these things than a regular driver, but I still have no clue where to call. To make things worse, I have no idea which roads are state/county/town maintained, so no proper contact point

Any actuated timing plan will have a fixed-time plan as a backup in the event of a fault. Usually recognizes it within a handful of cycles.

Once upon a time, I had to run a red left arrow (red arrow = "no go" here) at the 8th cycle. Thing happily skipped me (and probably 20 cars behind me) for good 15 minutes.  Usual traffic pattern was getting a green arrow in something like 3 out of 4, if not 9 out of 10 cycles  daytime. This is the type of error which seem to go undetected.

As jeffandnicole said, road crews and other personnel go through on a regular basis and will notice if something's wrong. Actual inspections are supposed to occur weekly (which is the biggest part of the reason why signals operating costs are so insanely high - go roundabouts!). And as jeffandnicole also said, because there's a fault doesn't mean that its easily solved. A software issue could be fixed in minutes, something like a damaged induction loop would cost thousands of dollars to fix, so securing funding to repair it could be quick in a high profile location, or could take years at an intersection of lesser importance.

Well, switching to fixed timing is irritating, but workable. And I guess some states save on these inspections..  What I am talking about is effectively self-inspecting controller - checking detection circuitry; add current draw measurement - and burned bulbs can be detected as well. Not very useful with LEDs, but still may detect broken wires...

[jakeroot]

I don't quite understand why we still use induction loops at the intersections. Seems much more cost-effective to use cameras. I can see using loops approx. 100-200 yards away from the intersection to supplement the cameras (to recognize a gap in the traffic that may not be visible to the cameras). But, at the intersection itself, it seems to me that induction loops are far more difficult to replace or repair, are far more susceptible to damage, and must be placed along each approach lane (whereas cameras can watch all approach lanes), which I would guess adds to the construction cost significantly.

Nearly all new signals near me use video detection, and though it's hardly scientific, I very seldom, if ever, see a signal acting up (nor can I say that I've had a signal skip me before).

[tradephoric]

I don't quite understand why we still use induction loops at the intersections.

Inductive loops are pretty damn accurate.  That's why they are still used. 

Inductive loop detectors are still the primary means of detection at traffic signals and elsewhere, and if properly installed and maintained are still the most accurate. However, due to the well-documented problems associated with loops, many jurisdictions are replacing intersection loops with video imaging vehicle detection systems (VIVDS) as loops fail, or even before they fail. While VIVDS have overcome some of the problems with loops such as traffic disruption and pavement degradation, they have not been as accurate in all weather and light conditions as originally anticipated.
http://d2dtl5nnlpfr0r.cloudfront.net/tti.tamu.edu/documents/0-5845-1.pdf

[Thing 342]

Ideally, you shouldn't have to call anyone to report a broken detector loop. At my job, I work on software that coordinates and monitors these traffic monitoring and display devices that is used by several state DOTs. One component of this software monitors the heartbeat signals that these devices send out periodically (usually once or twice per minute), and notifies via email if these devices don't check in after a specified timeout period, so that repair crews can be dispatched as soon as possible.  Usually, the problem with these type devices is the connection between the detector and the controller, rather than the actual components themselves.

The intelligently-managed signal cycles that Baloo talked about are definitely possible (the system can automatically display congestion messages on VMS when traffic speed drops below a certain level), though I don't know of any DOTs that are willing to pony up the money to have such a system put into place for individual signals.

(Disclaimer: my opinions are my own, and not necessarily that of my employers)

[kalvado]

Ideally, you shouldn't have to call anyone to report a broken detector loop. At my job, I work on software that coordinates and monitors these traffic monitoring and display devices that is used by several state DOTs. One component of this software monitors the heartbeat signals that these devices send out periodically (usually once or twice per minute), and notifies via email if these devices don't check in after a specified timeout period, so that repair crews can be dispatched as soon as possible.  Usually, the problem with these type devices is the connection between the detector and the controller, rather than the actual components themselves.

Is it too much to ask about your company name? Maybe - so that it is neither too identifying, nor advertisement - can you list a few companies in that business; like yours and 2-3 competitors?

[roadfro]

Going back to some things in the OP:

Typically when a signal has loop detection (especially for presence detection at the stop line), the controller sends a "heartbeat" signal regularly. If there is a detector failure registered, the controller settings are usually configured to register a "constant call" for the associated signal phase. This means the the controller assumes vehicles are always present on the broken detector. The constant call typically prompts the controller to run the maximum green tinge for the phase. So there shouldn't be delays for cars not getting served, but rather delays for other phases because of extra green to a phase that may not have demand.

As to some of the other thoughts about signal smarts: Detection at signals are typically not "smart" in the sense of collecting data. Signal detectors usually just look for presence of vehicles, and don't collect other statistics like traffic count data. This is why agencies still have to go out and periodically collect traffic count and turning movement data at intersections in preparation for retiming signals. I believe it is possible to modify signal detection to keep count data, but may require additional equipment and modification.

Most signal controllers have the ability to communicate with a central hub. Typically, only larger agencies that employ signal coordination strategies have these hubs and communication networks. I know traffic engineers can upload timing plans to a controller in the field, but I'm not sure if the average signal controller can communicate to the central hub about a detection failure or other fault (apparently the technology does exist). But if a failure is discovered by engineers/techs or the public, engineers can remotely adjust timing from the hub.

LG-D850

[roadfro]

Signals that communicate with the regional or state DOT office do exist, but they tend to be part of a coordinated system.

I may be a bit more curious about these things than a regular driver, but I still have no clue where to call. To make things worse, I have no idea which roads are state/county/town maintained, so no proper contact point

The majority of people on these boards are quite familiar with the proper jurisdiction, so it's not too often someone says here they wouldn't know who maintains which roads.  In general, if there's a state road at an intersection, the traffic light would be maintained by the state.  If it's only county roads, the light is probably county maintained.  The state's transportation website may have a phone number to call - if it's not under their jurisdiction, they can probably tell you to call the county.  The more helpful people will give you the phone number.  And if that doesn't work, you can always call the police department, major, elective representatives, etc.

The point is, there are numerous people one can call.  You may not succeed on the first try, but you'll eventually figure it out.

Regarding the bolded statement, this is not universal either. For example: Nevada DOT will install traffic signals as part of state projects, but NDOT does not maintain any traffic signals whatsoever–that responsibility is handled by the local jurisdiction (city, or the county if not in an incorporated community).

I would say that the best thing to do is to start with city or county's public works or transportation department. If that's not the right spot, they should refer you to the proper place (or report the issue themselves to the right area).

[Jardine]

I've noted this before, on a busy city street with some traffic signals for some side streets, frequently when someone on the busy street turns off onto the side street they will swing wide and trip the inductive sensor for the traffic signals.

One near a place I formerly lived at was tripped far more often for folks swinging wide than for folks coming from the neighborhood and needing the signal to get on the busy street.

If inductive sensors could detect and ignore folks going the wrong way that signal would have probably seen a 75% drop in activations.

Also, traffic on the busy street would not be continuously disrupted by the unnecessary cycling of the signals.


Or we could have a marksman with a rifle shooting the idjits tripping the damn light . . .

[cl94]

Signals that communicate with the regional or state DOT office do exist, but they tend to be part of a coordinated system.

I may be a bit more curious about these things than a regular driver, but I still have no clue where to call. To make things worse, I have no idea which roads are state/county/town maintained, so no proper contact point

The majority of people on these boards are quite familiar with the proper jurisdiction, so it's not too often someone says here they wouldn't know who maintains which roads.  In general, if there's a state road at an intersection, the traffic light would be maintained by the state.  If it's only county roads, the light is probably county maintained.  The state's transportation website may have a phone number to call - if it's not under their jurisdiction, they can probably tell you to call the county.  The more helpful people will give you the phone number.  And if that doesn't work, you can always call the police department, major, elective representatives, etc.

The point is, there are numerous people one can call.  You may not succeed on the first try, but you'll eventually figure it out.

Regarding the bolded statement, this is not universal either. For example: Nevada DOT will install traffic signals as part of state projects, but NDOT does not maintain any traffic signals whatsoever–that responsibility is handled by the local jurisdiction (city, or the county if not in an incorporated community).

I would say that the best thing to do is to start with city or county's public works or transportation department. If that's not the right spot, they should refer you to the proper place (or report the issue themselves to the right area).

This is correct. Things are further complicated in places like New York, where some cities maintain almost every signal inside the city limits even on state-maintained roads (Buffalo and Troy, for example), or parts of Ontario, where certain counties/regions (notably Niagara) maintain every signal inside their boundaries and MTO maintains some stuff elsewhere.

[tradephoric]

That's why video detection is becoming the preferred method, because they can detect all types of vehicles.  IMO though, they aren't as accurate and have more issues than the loops.  Sunglare can affect their visibility sometimes.  If the camera is a bit off the mark, it may wind up looking at the wrong lane.  There's one near me where, if I'm walking in the crosswalk, I interfere with its line of sight and will cycle the signal thinking I'm a vehicle!

This is the problem agencies face.  Video detection is able to detect all types of vehicles (ie. bicycles) but the accuracy of video detection leaves a lot to be desired.  During transitional light periods shadows can lead to extremely high false calls.  Manufacturers will assure you that they have fixed the shadow issues, but every video detection camera I have ever worked with will fault under the right lighting conditions.  I will believe it when I see it.  And the shadow problem has been an issue for the past 25 years (I've read studies from the early 90s that deal with the same problems we see today).  There are other problems with video detection that Jeffandnicole touched on (wind blowing camera outside detection zone, pedestrian walking past a vehicle detector and placing a false-call). 

Should we ignore all the accuracy issues of video detection just because they can detect bikes?

[tckma]

One of the things I was taught in the class to get my motorcycle license is that motorcycles are usually not large/heavy enough to trigger induction loops, and that you could be waiting at that light for a very long time until a car comes up behind you... and if you blow the red, sometimes there is a cop waiting.  The instructor implied that you can maybe use the induction loop not triggering to contest a ticket in court.  It never happened to me (I always had cars come up behind me in fairly short order), and I'm not sure you'd be able to get out of a ticket that way.

Video detection seems like the way to go.  I worked on some computer vision software for my Master's thesis/project, and while it's difficult to do, it's not impossible using even the most basic of video cameras.

[cl94]

One of the things I was taught in the class to get my motorcycle license is that motorcycles are usually not large/heavy enough to trigger induction loops, and that you could be waiting at that light for a very long time until a car comes up behind you... and if you blow the red, sometimes there is a cop waiting.  The instructor implied that you can maybe use the induction loop not triggering to contest a ticket in court.  It never happened to me (I always had cars come up behind me in fairly short order), and I'm not sure you'd be able to get out of a ticket that way.

Newer induction loop installations tend to use tighter coils that are able to detect smaller vehicles. The real problem comes with the metal detectors New York used to use. Don't drive over the 6x6 metal square and the signal is never changing. Good luck doing that in a motorcycle without wiping out, now that most of these detectors are at the bottom of depressions in the surface.

[kalvado]

One of the things I was taught in the class to get my motorcycle license is that motorcycles are usually not large/heavy enough to trigger induction loops, and that you could be waiting at that light for a very long time until a car comes up behind you... and if you blow the red, sometimes there is a cop waiting.  The instructor implied that you can maybe use the induction loop not triggering to contest a ticket in court.  It never happened to me (I always had cars come up behind me in fairly short order), and I'm not sure you'd be able to get out of a ticket that way.

Video detection seems like the way to go.  I worked on some computer vision software for my Master's thesis/project, and while it's difficult to do, it's not impossible using even the most basic of video cameras.

I heard a few times that positioning smaller vehicle on the corner of the loop (if footprint is visible) may improve things.
And optics is likely less than ideal option as well - fog-haze on the street, water (not so clean you know) sprayed by traffic.. Try to wipe a light pole next to an intersection... camera will have hard time sing through that. Then, sun reflection from glass/metal may mess things up quite a bit.

When I think about it, 3 types of detectors working each direction seem a good option.. until I think about the cost and complexity of such installation

[cbeach40]

This is correct. Things are further complicated in places like New York, where some cities maintain almost every signal inside the city limits even on state-maintained roads (Buffalo and Troy, for example), or parts of Ontario, where certain counties/regions (notably Niagara) maintain every signal inside their boundaries and MTO maintains some stuff elsewhere.

In terms of the latter that would not complicate things at all. It would just change who does the repairs, not to whom you'd complain.
Realistically, complain to one of the agencies in the area (whatever level of municipal or higher) and they'll flip it to the responsible agency. It's not that complicated.

[jeffandnicole]

That's why, way above, I said 'Generally', and then if you don't get the correct department on the first try, eventually you'll figure it out. 

The problem is, in most cases no one is going to make that call to begin with.

[tckma]

That's why, way above, I said 'Generally', and then if you don't get the correct department on the first try, eventually you'll figure it out. 

The problem is, in most cases no one is going to make that call to begin with.

I called Howard County Police to report a traffic light completely dark once (thinking the police would just report it to the responsible agency).  The dispatcher told me to call the county highway department (hm, I guess not) and gave me the phone number.  The guy who answered the phone for the county highway department told me that that traffic light was the state's responsibility since it was on a state highway (MD-97), and gave me the phone number for the appropriate SHA office.  It was rather frustrating.  Now I know that anything on a state highway is under SHA jurisdiction so I don't need to play phone tag.  At least I was given the phone numbers without having to look them up.

[PColumbus73]

A pet peeve I have with the loop detection is when people intentionally stop on the loops to trigger a green (arrow or ball), especially when the loops are two or three car spaces back from the stop bar.

[cl94]

A pet peeve I have with the loop detection is when people intentionally stop on the loops to trigger a green (arrow or ball), especially when the loops are two or three car spaces back from the stop bar.

Better than the idiots who stop before them. I know of several signals in New York with the only detection within a car length of the stop bar. I've been stuck at lights for several cycles because the idiot in front of me isn't anywhere close to the sensor.

[jeffandnicole]

A pet peeve I have with the loop detection is when people intentionally stop on the loops to trigger a green (arrow or ball), especially when the loops are two or three car spaces back from the stop bar.

I would be surprised these people are even aware of the loops and their locations. Besides, the loops are supposed to be near the stop bar to detect traffic stopped at a light.

[cl94]

A pet peeve I have with the loop detection is when people intentionally stop on the loops to trigger a green (arrow or ball), especially when the loops are two or three car spaces back from the stop bar.

I would be surprised these people are even aware of the loops and their locations. Besides, the loops are supposed to be near the stop bar to detect traffic stopped at a light.

Most modern designs have 2 sets of loops: one at the stop bar and one further back. And everyone's favorite New York metal detectors are often located a couple car lengths back. Cut over to the left turn lane too late and you might miss it.

[roadfro]

A pet peeve I have with the loop detection is when people intentionally stop on the loops to trigger a green (arrow or ball), especially when the loops are two or three car spaces back from the stop bar.

I would be surprised these people are even aware of the loops and their locations. Besides, the loops are supposed to be near the stop bar to detect traffic stopped at a light.

Most modern designs have 2 sets of loops: one at the stop bar and one further back. And everyone's favorite New York metal detectors are often located a couple car lengths back. Cut over to the left turn lane too late and you might miss it.

What is typically done in Nevada is to have two sets of loops in turn bays (covers two typical car lengths at the stop line) and one set of loops for through lanes (covering one typical car length at the stop line). Recalling that the loops are for presence detection, it's not really necessary to have much more than this.

For actuated signal phases, there is often an additional single loop upstream of the stop bar in the through lanes. This is for passage gap detection ("passage gap", aka "vehicle extension", being the slight extension [~4 seconds] of the phase's green time beyond minimum green value for approaching vehicles detected upstream, in order to allow the approaching vehicles to go through), assuming the controller has a non-zero setting for the unit extension.


Can anybody provide a link to a New York (or other area) metal detector square? I'm curious to see what this looks like as I've never seen this in person.  Inductive loops basically function as a metal detector embedded in the pavement, so I'm curious to see how this is done in other ways.

[tradephoric]

It's three in the morning at this Florida intersection and a pedestrian just hit a pushbutton to cross the 225 foot crosswalk.  It would take 75 seconds to fit this pedestrian assuming the MUTCD standards are being followed.  The intersection is "smart"  and knows that current traffic volumes only require 10-15 seconds to clear out the vehicles along main-street.   I know this is an extreme example, but it shows you the limitations of "smart"  traffic signals since they need to fit required pedestrian times and safety minimums.