News:

Thank you for your patience during the Forum downtime while we upgraded the software. Welcome back and see this thread for some new features and other changes to the forum.

Main Menu

"Smart" traffic signals

Started by tradephoric, February 15, 2018, 11:27:38 AM

Previous topic - Next topic

tradephoric

Traffic signal countdown timers (TSCT):  Many countries have traffic signal countdown timers (TSCTs) that displays the time remaining in the signal phase.  One problem with these timers is that drivers can only view the timers once they are very close to the traffic signal.  The other thing is that they only work well along fixed time signals (which is actually one of the "dumbest"  traffic signals you can have).  Many people have argued that having these timers just encourage drivers to "gun it"  to make a green light, potentially making the intersection more dangerous.  However, several studies have found that countdown timers increased the probability that a driver in the dilemma zone would stop (reducing potential red light running crashes) and decreased the average deceleration rates approaching the signal (reducing potential rear-end accidents). 



Signal Phasing and Timing (SPaT):  Similar to traffic signal countdown timers, SPaT data displays the remaining time in the signal phase but the data is broadcast by the traffic signal and received by a vehicle display.  This allows drivers to know what a traffic signal miles up the road is doing and a driver could strategically adjust their speeds so that they could potentially make it through a green light that they otherwise would have got stuck at.  Also, since the traffic signal is "talking"  to the vehicle, crash avoidance warnings and alerts could be broadcast to the driver display if approaching a red light at high speed.  Here is an early test bed of SPaT data along Telegraph Road in metro Detroit.



Since those initial test beds, SPaT data has been introduced in some production vehicles, most notably the latest luxury Audi models.  Audi's Traffic Light Information service is only available in cites that broadcast the signal phasing and timing data and for right now Las Vegas is the largest city with widespread adoption.  Here's a youtube video and article detailing the new Audi Traffic Light information service:


https://www.extremetech.com/extreme/240264-hands-audis-exciting-no-really-traffic-light-countdown-timer

What do you guys think of this traffic light tech?  Useful or hype? 


hotdogPi

It's definitely worth it. It could also work with signals that aren't completely on timers; just leave it blank if the time is unknown.

In addition, counting down when it's red is also useful so that drivers will accelerate immediately once it turns green, reducing reaction time, and if it's over a minute, it might even be beneficial to turn the engine off.
Clinched, minus I-93 (I'm missing a few miles and my file is incorrect)

Traveled, plus US 13, 44, and 50, and several state routes

I will be in Burlington VT for the eclipse.

tradephoric

Quote from: 1 on February 15, 2018, 11:35:38 AM
It's definitely worth it. It could also work with signals that aren't completely on timers; just leave it blank if the time is unknown.

Great point.  At variable timed signals there will be some uncertainty to when the signal will change to green.  An actuated signal with detection may want to gap the side-street early if vehicles have cleared, giving that extra time to the main street.  Either the countdown would remain blank or there may be a "quickie" countdown once the decision point to change to the main-street has been reached.  According to the article, the countdown timer is used at variably timed signals in Las Vegas; but as they say the data needs to be "massaged".

QuoteSome is data on fixed duration signals, meaning the light is always green for, say, 30 seconds, yellow for 4 seconds, and red for 30 seconds. Other data comes from variable-length signals, based on traffic flow, time of day, or an approaching emergency vehicle where the light changes to stop crossing traffic; this is the data that needs to be massaged.

DaBigE

Upcoming signal technology will likely make this display moot in the long run, since part of the autonomous vehicle movement includes converting traffic signal controllers to broadcast signal status. While it's potentially useful information to display, I see it as an unnecessary distraction. What happens for protected movement approaches? Do you provide a countdown for the thru movement only or for every movement? As for an increase in crashes, I'm not so sure. Similar arguments were made when countdown timers were installed with ped signals, but for the most part, there hasn't been a spike in attributable crashes.

Quote from: 1 on February 15, 2018, 11:35:38 AM
It could also work with signals that aren't completely on timers; just leave it blank if the time is unknown.
Technically the time is always "known", what isn't is whether an extension will be granted, phase skipped, or a preemption event (train or emergency vehicle) until the controller timer runs out. Similar to a Ped countdown timer, a time could always be displayed, but it would be one with the potential of resetting itself. The question is whether the countdown to zero/reset create a 'boy who cried wolf' scenario?

Quote from: 1 on February 15, 2018, 11:35:38 AM
In addition, counting down when it's red is also useful so that drivers will accelerate immediately once it turns green, reducing reaction time, and if it's over a minute, it might even be beneficial to turn the engine off.
This is assuming everyone is paying attention. Frequently, drivers use the red time to check their phones and are not focused on what's going on around them. Increasing inclusion of auto start/stop engine technology will gradually reduce the environmental benefit.
"We gotta find this road, it's like Bob's road!" - Rabbit, Twister

7/8

I'm hesitant on in-vehicle displays of a countdown because it forces drivers to take their eyes off the road (i.e. "distracted driving"). In fact, I'm not a fan of these new screens in vehicles for that same reason (they're way too distracting)!

I think pedestrian countdown signals work well enough for this purpose. The only problem I have with them is that sometimes the light won't change to yellow at zero, and instead will wait a couple or several seconds to change. In my mind, this ruins the whole purpose of the countdown, and it's frustrating for both drivers and pedestrians. If a new standard could be enforced that all lights must change to yellow when the countdown reaches zero, then I'd be happy (and I think everyone else would be too) :thumbsup:

tradephoric

I created a quick and dirty excel spreadsheet that tracks SPaT data along Woodward Avenue in Detroit.  While the lights along Woodward are timed pretty well, there are stretches that are uncoordinated making it a crapshoot if drivers will hit a red light.  By using the SPaT spreadsheet, I could predict when a driver would most likely hit a string of green lights.  It's a little different than SpaT data, since it is visually showing the "predictive"  green wave.  Essentially, it's estimating what the signals will be doing by the time the driver reaches the signal assuming they drive a specified speed.   In the case of Woodward, a driver can achieve a good green wave or a poor green wave depending entirely on what cycle the driver is in (everything else is equal).



Here's a video of a 40 mile drive down Woodward without hitting a red light.  There are only a few cycles an hour where a drive like this is possible and I probably never would have achieved it without using that SPaT spreadsheet.  This was my very first crack at it and I probably could have tweaked my speed in places to make it a smoother drive, but it's still pretty good.



DaBigE

Quote from: 7/8 on February 15, 2018, 12:41:18 PM
I'm hesitant on in-vehicle displays of a countdown because it forces drivers to take their eyes off the road (i.e. "distracted driving"). In fact, I'm not a fan of these new screens in vehicles for that same reason (they're way too distracting)!
I totally agree, and I think this is where design policy has a ways to catch-up (talking the same policy/rules that dictate how many square inches of reflectivity must be present on the sides, rear of a vehicle, headlight cutoffs, etc.). I would hope the broadcasting technology would only be integrated into a HUD display as an emergency event, e.g., when a red light is about to be run, rather than a dynamic display of time vs. red, yellow, green display.

Quote from: 7/8 on February 15, 2018, 12:41:18 PM
I think pedestrian countdown signals work well enough for this purpose. The only problem I have with them is that sometimes the light won't change to yellow at zero, and instead will wait a couple or several seconds to change. In my mind, this ruins the whole purpose of the countdown, and it's frustrating for both drivers and pedestrians. If a new standard could be enforced that all lights must change to yellow when the countdown reaches zero, then I'd be happy (and I think everyone else would be too) :thumbsup:
Part of that is due to how the ped signal is activated (ped actuated phase vs pre-timed), ped recall settings, and whether or not a vehicular extension is granted or the signal gaps/maxes out. Some of it is just poor planning/programming. It's not always clear as to which is to blame.
"We gotta find this road, it's like Bob's road!" - Rabbit, Twister

tradephoric

A pedestrian actuation can really limit how "smart"  a traffic signal can be.   You can have a state of the art adaptive traffic control system but you are still limited to how long it takes a pedestrian to cross the street.  As an example take Kirkman Road and Carrier Drive in Orlando...
 
https://www.google.com/maps/@28.45738,-81.4586,19z/data=!3m1!1e3?dg=dbrw&newdg=1

There are 200 foot crosswalks for pedestrians crossing Kirkman Road.  Assuming the city is adhering to the MUTCD standards, the side-street would have to run about 70 seconds to fit the pedestrian times.  Maybe there's only one or two cars on the side-street that could clear in 10 seconds, but it's going to have to run the full 70 seconds to fit the peds.  How "smart"  can a traffic signal be when a pedestrian actuation completely hijacks the intersection?  That's why there needs to be a mix of "smart"  signals and good intersection geometry.  It's not just about how smart a signal system is.

hotdogPi

Quote from: tradephoric on February 15, 2018, 03:10:35 PM
A pedestrian actuation can really limit how "smart"  a traffic signal can be.   You can have a state of the art adaptive traffic control system but you are still limited to how long it takes a pedestrian to cross the street.  As an example take Kirkman Road and Carrier Drive in Orlando...
 
https://www.google.com/maps/@28.45738,-81.4586,19z/data=!3m1!1e3?dg=dbrw&newdg=1

There are 200 foot crosswalks for pedestrians crossing Kirkman Road.  Assuming the city is adhering to the MUTCD standards, the side-street would have to run about 70 seconds to fit the pedestrian times.  Maybe there's only one or two cars on the side-street that could clear in 10 seconds, but it's going to have to run the full 70 seconds to fit the peds.  How "smart"  can a traffic signal be when a pedestrian actuation completely hijacks the intersection?  That's why there needs to be a mix of "smart"  signals and good intersection geometry.  It's not just about how smart a signal system is.

35 if the pedestrians are supposed to stop in the median and take two cycles to cross.
Clinched, minus I-93 (I'm missing a few miles and my file is incorrect)

Traveled, plus US 13, 44, and 50, and several state routes

I will be in Burlington VT for the eclipse.

tradephoric

Quote from: 1 on February 15, 2018, 03:16:22 PM
Quote from: tradephoric on February 15, 2018, 03:10:35 PM
A pedestrian actuation can really limit how "smart"  a traffic signal can be.   You can have a state of the art adaptive traffic control system but you are still limited to how long it takes a pedestrian to cross the street.  As an example take Kirkman Road and Carrier Drive in Orlando...
 
https://www.google.com/maps/@28.45738,-81.4586,19z/data=!3m1!1e3?dg=dbrw&newdg=1

There are 200 foot crosswalks for pedestrians crossing Kirkman Road.  Assuming the city is adhering to the MUTCD standards, the side-street would have to run about 70 seconds to fit the pedestrian times.  Maybe there's only one or two cars on the side-street that could clear in 10 seconds, but it's going to have to run the full 70 seconds to fit the peds.  How "smart"  can a traffic signal be when a pedestrian actuation completely hijacks the intersection?  That's why there needs to be a mix of "smart"  signals and good intersection geometry.  It's not just about how smart a signal system is.

35 if the pedestrians are supposed to stop in the median and take two cycles to cross.

I see pushbuttons on streetview for crossing Kirkman.  If pedestrians are suppose to stop in the median then there should be pushbuttons and pedestrian signal heads in the median (which there aren't).  The way the intersection is designed the pedestrians are meant to cross the entire 200' crosswalk in one go.

7/8

Quote from: DaBigE
Quote from: 7/8 on February 15, 2018, 12:41:18 PM
I think pedestrian countdown signals work well enough for this purpose. The only problem I have with them is that sometimes the light won't change to yellow at zero, and instead will wait a couple or several seconds to change. In my mind, this ruins the whole purpose of the countdown, and it's frustrating for both drivers and pedestrians. If a new standard could be enforced that all lights must change to yellow when the countdown reaches zero, then I'd be happy (and I think everyone else would be too) :thumbsup:
Part of that is due to how the ped signal is activated (ped actuated phase vs pre-timed), ped recall settings, and whether or not a vehicular extension is granted or the signal gaps/maxes out. Some of it is just poor planning/programming. It's not always clear as to which is to blame.

Maybe I'm just ignorant of the technical details, but I can tell you specific intersections in Kitchener-Waterloo consistently have this problem, which leads me to believe it's intended. It recently popped up at my University, probably to discourage pedestrians from crossing last-minute. But now everyone knows about it, so people just cross after it reaches zero :-D. This problem is also at a different intersection with red light cameras (seems suspicious :eyebrow:).

I also noticed tonight that the countdowns were off by about 5 seconds on two sides of an intersection (and there is no lagging left to account for this). I thought the light would be turning yellow, but then I noticed the other countdown still gave me a few more seconds.

DaBigE

Quote from: 7/8 on February 15, 2018, 11:50:02 PM
I also noticed tonight that the countdowns were off by about 5 seconds on two sides of an intersection (and there is no lagging left to account for this). I thought the light would be turning yellow, but then I noticed the other countdown still gave me a few more seconds.
I've seen a number of situations like that around here, although usually by more than 5 seconds. Typically, they're caused by a walk phase timing concurrently with a leading protected turn. The one that ran concurrent times out before the one that had to wait for the protected turn to end. It could also be to different crosswalk lengths. But like I said earlier, it could also be sloppy programming. We did the timing plans for one intersection recently, and the contractor screwed up part of the programming. Ped phases were supposed to be actuated by the button, but instead got programmed as auto recalling, resulting in upwards of 20 seconds potentially wasted per cycle. Since the signal still functions "properly", the contractor is going to take their lazy-ass time getting back out to fix it.
"We gotta find this road, it's like Bob's road!" - Rabbit, Twister

jeffandnicole

Quote from: DaBigE on February 16, 2018, 01:39:33 AM
Quote from: 7/8 on February 15, 2018, 11:50:02 PM
I also noticed tonight that the countdowns were off by about 5 seconds on two sides of an intersection (and there is no lagging left to account for this). I thought the light would be turning yellow, but then I noticed the other countdown still gave me a few more seconds.
I've seen a number of situations like that around here, although usually by more than 5 seconds. Typically, they're caused by a walk phase timing concurrently with a leading protected turn. The one that ran concurrent times out before the one that had to wait for the protected turn to end. It could also be to different crosswalk lengths. But like I said earlier, it could also be sloppy programming. We did the timing plans for one intersection recently, and the contractor screwed up part of the programming. Ped phases were supposed to be actuated by the button, but instead got programmed as auto recalling, resulting in upwards of 20 seconds potentially wasted per cycle. Since the signal still functions "properly", the contractor is going to take their lazy-ass time getting back out to fix it.

Yep. Walk phases can differ on both sides of the intersection when programmed to do so. Along with differing left turn phases, this also occurs on traffic lights that can be programmed to change based on expected downstream flow. The ped phase can terminate well before the green light.

UCFKnights

Quote from: tradephoric on February 15, 2018, 03:26:54 PM
Quote from: 1 on February 15, 2018, 03:16:22 PM
Quote from: tradephoric on February 15, 2018, 03:10:35 PM
A pedestrian actuation can really limit how "smart"  a traffic signal can be.   You can have a state of the art adaptive traffic control system but you are still limited to how long it takes a pedestrian to cross the street.  As an example take Kirkman Road and Carrier Drive in Orlando...
 
https://www.google.com/maps/@28.45738,-81.4586,19z/data=!3m1!1e3?dg=dbrw&newdg=1

There are 200 foot crosswalks for pedestrians crossing Kirkman Road.  Assuming the city is adhering to the MUTCD standards, the side-street would have to run about 70 seconds to fit the pedestrian times.  Maybe there's only one or two cars on the side-street that could clear in 10 seconds, but it's going to have to run the full 70 seconds to fit the peds.  How "smart"  can a traffic signal be when a pedestrian actuation completely hijacks the intersection?  That's why there needs to be a mix of "smart"  signals and good intersection geometry.  It's not just about how smart a signal system is.

35 if the pedestrians are supposed to stop in the median and take two cycles to cross.

I see pushbuttons on streetview for crossing Kirkman.  If pedestrians are suppose to stop in the median then there should be pushbuttons and pedestrian signal heads in the median (which there aren't).  The way the intersection is designed the pedestrians are meant to cross the entire 200' crosswalk in one go.
I don't drive by that intersection too much, but from looking at it and knowing the area, I imagine its pretty rare for someone to actually use that crosswalk. There isn't even sidewalks in parts of it. Then again, its Orlando, and they don't have sidewalks even in areas where there is significant pedestrian traffic, I remember lots of missing ones in East Orlando when I went to UCF where people would still walk all the time.

It wouldn't necessarily be required to make them wait the entire 2 cycles to cross the street if implemented with median signals/buttons. A no U turn restriction could be placed and then giving a walk signal on only half the road while the appropriate left turn signal is green. I could think of a few places where that'd be useful but I've never see that done.

UCFKnights

Quote from: DaBigE on February 16, 2018, 01:39:33 AM
Quote from: 7/8 on February 15, 2018, 11:50:02 PM
I also noticed tonight that the countdowns were off by about 5 seconds on two sides of an intersection (and there is no lagging left to account for this). I thought the light would be turning yellow, but then I noticed the other countdown still gave me a few more seconds.
I've seen a number of situations like that around here, although usually by more than 5 seconds. Typically, they're caused by a walk phase timing concurrently with a leading protected turn. The one that ran concurrent times out before the one that had to wait for the protected turn to end. It could also be to different crosswalk lengths. But like I said earlier, it could also be sloppy programming. We did the timing plans for one intersection recently, and the contractor screwed up part of the programming. Ped phases were supposed to be actuated by the button, but instead got programmed as auto recalling, resulting in upwards of 20 seconds potentially wasted per cycle. Since the signal still functions "properly", the contractor is going to take their lazy-ass time getting back out to fix it.
Yeah, lots of sloppy programming in Orlando. UCF's signals are all terribly timed and extremely "inefficient" for pedestrians. A few come to mind as being especially bad:
https://www.google.com/maps/@28.6063113,-81.1986991,3a,56y,66.79h,92.55t/data=!3m6!1e1!3m4!1svdPD_0y5PWnQkkWQZ1TSZQ!2e0!7i13312!8i6656
This signal is programmed with auto-recall along the main road. On the right side, the single lane being crossed has a real short don't walk clearing time, like 9 seconds if I remember right. The other side is quite a bit longer. The other side's activation also is often delayed by the permissive left so the longer side often gets even more of a delayed start. At a minimum, its total crossing time is half of what it could be on every cycle. If the permissive left phase is activated, it is frequently 75% less then what it should be. The other direction at this intersection (and all of them in the area) is button activated, but even though there is no left phases at the 2 sides of the crosswalk are the same length, they are independently activated. Many people forget to push the button, or cross the other way after pushing the button as they are crossing diagonially, so they miss the activation on their side even though there is time for it. You sometimes even see people trying to look at the other sides walk signal to see if they have time.

Another bad one was this one:
https://www.google.com/maps/@28.5963017,-81.1992277,3a,50.1y,300.86h,94.98t/data=!3m6!1e1!3m4!1swyMjlfjdvKPla1llVYCwFw!2e0!7i13312!8i6656
This one is configured as an all red when people are crossing over the main 4 lane road. Even though the side street is much shorter, it receives a don't walk the entire time of the all red phase.

There is one near me that is configured as an all red stop during crossing during school hours only. It has a "bug" during school hours if someone pressed the walk button during the all red phase, it gives them the normal walk signal to cross the minor road. Its a long street with a 60 second cross time, so when kids hit the button after they've crossed the street, it makes traffic went an entire extra phase before going, resulting in a full 2 minute delay with no cars going when it probably only takes the kids 20 seconds to cross.

tradephoric

Another thing that can ruin a "smart" adaptive traffic control system is poor detection.  I think many here would be surprised how inaccurate video detection can be at detecting vehicles.  Tried and true inductive loops are great, but bad pavement can render them useless.  There are new vehicle detection technologies coming out but it always seems the vendor's promises never match reality.  Accurate detection might not be all that important if it's just being used to gap out a side-street, but for completely adaptive traffic control systems good detection is key.

In the future V2I technology takes over and vehicles on the road will be equipped to communicate with the traffic controllers.  At that point the vehicle itself becomes the detector.  A traffic controller will know that a vehicle approaching 1/2 mile down the road is going to make a left turn at the signal.  The vehicles will be in the network and the controller algorithms can figure out the best signal timing based on the "future" conditions (ie. the controller knows a vehicle 1/2 mile away is going to make a left turn at the signal)... and not on "past" traffic conditions like it's done now.



Opinions expressed here on belong solely to the poster and do not represent or reflect the opinions or beliefs of AARoads, its creators and/or associates.