Traffic Signal Timing

[mrsman]

Some of the more recent discussion in the FYA thread has lead me to ponder some basics about traffic signal timing.  I'm hoping that the discussion will get technical to help improve my understanding. 

What is the best way to time a traffic signal at a standard intersection between a major street and a minor street? 

For the sake of simplicity, let's say that there are no left or right turn arrows involved at the intersection.  The major street has 3 lanes in each direction and a left turn lane and is 70 feet wide from curb to  curb.  The minor street has 1 lane in each direction and parking lanes on both sides of the street and is 35 feet wide.  The major street runs east/west and has a 35 MPH speed limit and the minor street runs north/south and has a 25 MPH speed limit.  There are push buttons in all directions.

The phases of the signal would be set up as follows:

1) Major Street Green/Walk  [7 seconds, see http://mutcd.fhwa.dot.gov/htm/2009/part4/fig4e_02_longdesc.htm]

2) Major Street Green/FDW+ countdown timer [7 seconds]

3) Major Street Green/Solid Don't Walk [3 seconds minimum, note that the ped clearance time@ 3.5 ft/sec is (2) and (3) combined]

4) Major Street Yellow/Solid Don't Walk [4.1 seconds, see http://www.thenewspaper.com/rlc/docs/2012/nchrp12.pdf]

5) All Red [1 sec]

6) Minor Street Green/Walk [7 seconds]

7) Minor Street Green/FDW+ countdown timer [17 seconds]

8) Minor Street Green/Solid Don't Walk [3 seconds]

9) Minor Street Yellow/Solid Don't Walk [3.4 seconds]

10) All Red [1 second]

When phase 3 is at minimum, the cycle length is a total of 53.5 seconds.  But as I expect phase (3) to generally be a lot longer, cycle length can be extended considerably - while of course keeping track of the timing of adjacent signals.

Now for the above intersection, I envision phase (3) as being the resting phase.  Unless there is any actuation, phase 3 would be displayed.  Phase (3) has a minimum of 3 seconds, but could be displayed a lot longer if there is no actuation by peds or by side street cars.  Phases 1,2, and 3 combined should also be long enough to enable progressive signal timing with other intersections along main street.  I envision that during the busy parts of the day, phase 3 will be at least 20 seconds long.  During the middle of the night, phase 3 could be measured in hours.

(A)  Now if a main street ped actuates the signal during phase 2 or phase 3, and there is no call from the side street, the signal should immediately go back to phase (1), then (2) and then again rest at (3).  The actuation should reset at the beginning of phase (1).

(B)  If there is actuation by a side street car is detected during phase (3), independent of actuation by main street ped, then after phase (3)concludes, the signal should generally follow this sequence: (4), (5), (8*), (9), (10), (1), (2), (3). 

* While phase (8) has a minimum of 3 seconds, it should be displayed long enough to clear the number of cars detected in the intersection, up to a maximum of 26 seconds.  If more than 26 seconds is needed, then instead of going from (5) to (8), the signal should go from (5) to (6), (7), (8), (9), (10), (1), (2), (3).  In other words, green should be displayed for 27 seconds and the pedestrian signals should be displayed, even if there is no pedestrian actuation.  My thinking is that if the signal is going to be green long enough for a pedestrian to cross, the pedestrian signals should be displayed.

(C) If there is actuation by a side street ped, the signal should follow the sequence: (4), (5), (6), (7), (8), (9), (10), (1), (2), (3).  The actuation should reset at teh beginning of phase (6). 


I think phasing like this serves multiple purposes.  First, it will give main street peds a walk signal after the side street phase concludes, independent of whether there is actuation.  So during much of the day, the main street peds will get their signal at the beginning of the green phase without the need to push a button, but in the middle of the night when the only expected traffic is main street cars, peds will have to push a button to get an immediate display to walk.  There is no need to cycle through the side street phases, when there is no call from the side street.

Second, this will reduce delays during the quieter times of the day.  Since the signal rests at (3), if at 2 AM the only car that approaches is a side street car, the signal could quickly change from yellow to red to allow the side street car to go without having to go through a ped clearance interval.

Third, during the absolute busiest times of the day even the side street peds do not need to push a button.  If the side street signal will stay green long enough to allow a relatively large number of cars to cross, then the signal will display the ped signals as well.  But for most of the day, the side street peds will need to push a button to ensure that the signal will stay green long enough to cross the main street.

I hope this leads to more efficient signal timing and more respect for the signals by pedestrians.  Let's tell ped activists to stop complaining about "beg buttons" by designing the signal in a way that helps all users.

[jeffandnicole]

Just a few fixes to your assumptions:

You mentioned the side street is 35 feet wide, 2 lanes, 2 parking lanes.  That's probably too narrow.  A typical lane is 11 or 12 feet wide, so you're talking 22 - 24 feet.  A parking lane is about 8 feet wide.  That's 16 feet for 2 of them.  The side street is probably going to be 38 - 40 feet wide, not 35 feet.  3 - 5 feet doesn't seem like much, but that's another second or two for the flashing don't walk signal.

Ped clearance time is supposed to be entirely contained within the flashing don't walk signal, and shouldn't include any solid don't walk signal.

All-Red phases are generally longer than 1 second.  More like 2 or 3 seconds.

In a city where the traffic timings are fixed, there's little or no reason to have a solid don't walk phase while the light is still green.  That phase is reserved more for a traffic light where the pedestrian gets the walk phase, then after the ped time has ended, the light remains green until there's a presence of traffic or peds on the side street. 

Remember...the minimum time for the ped walk phase is 7 seconds.  You can certainly give more time there, and that often depends on the amount of pedestrian traffic expected, combined with the amount of vehicular traffic they want to move thru the intersection.  In a city, this is generally part of a very large network of signals.

At one, and for some reason only one, intersection in Trenton, NJ, there are stickers on traffic light poles saying "Use Button between 8pm and 8am".  I'm guessing this is to do what you're hoping to do in the above.  I have no clue if the signal actually works in this manner though.

[tradephoric]

Here is an intersection with a similar geometry you are describing outside Atlanta.  If you look on streetview there is no detection on main-street (either inductive loops for vehicles or pushbuttons for pedestrians).  The only detection is along the side street.  This would be referred to as semi-actuated signal.  A fully-actuated signal would have detection everywhere.   

https://www.google.com/maps/@33.8275862,-84.38748,55m/data=!3m1!1e3

IMO, fully-actuated signals are overkill at many mid-block intersections.  Here is the amount of detectors needed in that Atlanta intersection running fully-actuated and semi-actuated (assume there is just one-lane along each side-street approach)

Fully-actuated:  8 vehicle detectors; 8 pushbuttons (16 detectors total)
Semi-actuated:  2 vehicle detectors; 4 pushbuttons (6 detectors total)

You increased your detection cost by 166% by running fully-actuated (16 detectors vs 6).  For the added construction and maintenance cost, there is little benefit having main-street detection at this Atlanta intersection.

I hope this leads to more efficient signal timing and more respect for the signals by pedestrians.  Let's tell ped activists to stop complaining about "beg buttons" by designing the signal in a way that helps all users.

In your initial post you are proposing that the WALK comes up automatically during certain times of day but then require the pedestrian to press the pushbutton during other times.  Is this correct?  How does that help pedestrians respect pushbuttons?  If anything it adds confusion for pedestrians because sometimes the pushbutton does something, sometimes it doesn't. 

[roadman65]

Don't get me started on this one here, as here in Central Florida our engineers need to go back to school and learn the basics in math!  We have so many signals on our arterials that could easily work together and do not.  Driving OBT in Orlando can get you stuck at two traffic signals in a row with 2 signal waits each going through Southchase because our engineers can't figure out how to make the signals at Wal Mart and Deerfield Blvd. work together.

[tradephoric]

While phase (8) has a minimum of 3 seconds, it should be displayed long enough to clear the number of cars detected in the intersection, up to a maximum of 26 seconds.  If more than 26 seconds is needed, then instead of going from (5) to (8), the signal should go from (5) to (6), (7), (8), (9), (10), (1), (2), (3).  In other words, green should be displayed for 27 seconds and the pedestrian signals should be displayed, even if there is no pedestrian actuation.  My thinking is that if the signal is going to be green long enough for a pedestrian to cross, the pedestrian signals should be displayed.

The problem is traffic signals are reactive, not proactive.  When a green light is displayed, the controller really has no idea how many vehicles will pass through the intersection in advance.  It can only count the number of cars that the previous cycle received and adjust the timing accordingly (assuming that the next cycle will somewhat match the previous cycle.. which isn't always the case either).  The point is, I think it's a mistake to bring up the side-street ped automatically when no side-street pedestrians presses the pushbutton.  During rush hour when you have multi-mile backups on main-street, you want to maximize the green time for them and not have to wait for a phantom side-street pedestrian to finish their crossing.

[roadman65]

I would like to know why the latest trend is to have some signals give the protected left turn arrow post straight through over the previous before the straight through?   In Orlando area it depends on the time of day as in the AM it could be pre straight through and in the PM it could be post.

Then some signals like Town Center Blvd. and John Young Parkway do not have consistency.  One day at one specific time it will be post, and the very next day at the same time it will be pre.  Unlike JYP and Deerfield Blvd. which always during the evening rush, SB JYP will give the protected left onto Deerfield post straight through, while the rest of the day before except on weekends where its always pre.

At least the latter has consistency and it has to do with the traffic flow at peak times to warrant the change, but at JYP and Town Center, its a random call made by the controller to decide who goes first and who goes last.  Sometimes at Town Center/ JYP we have more left turn vehicles over straight through and yet the protected lefts have short green times not even allowing the whole queue to advance, while the straight through green is indefinite with no cars anywhere the intersection.   How can that one be making accurate calls on traffic?

[Jet380]

It's normal around these parts for the pedestrian phase to start only at the beginning of the associated car phase: so if the signals are resting on green for the major road and a pedestrian comes along wanting to cross the minor road, pressing the button will first call a green light for the minor road (even if there are no cars there), then give the pedestrian a green when it reverts to the major road again.

This makes sure pedestrians are only introduced into the intersection just as the parallel cars are being given a green light. Combined with a 5-second 'head start' for pedestrians, this goes a bit towards reducing the chance of turning vehicle collisions.

In your initial post you are proposing that the WALK comes up automatically during certain times of day but then require the pedestrian to press the pushbutton during other times.  Is this correct?  How does that help pedestrians respect pushbuttons?  If anything it adds confusion for pedestrians because sometimes the pushbutton does something, sometimes it doesn't.

In Melbourne the pedestrian buttons have a little red 'call recorded' light, and during busy times it stays lit continuously to tell you that it is set to provide a pedestrian phase at every cycle: (it's the small circle shape on the left)

[mrsman]

Here is an intersection with a similar geometry you are describing outside Atlanta.  If you look on streetview there is no detection on main-street (either inductive loops for vehicles or pushbuttons for pedestrians).  The only detection is along the side street.  This would be referred to as semi-actuated signal.  A fully-actuated signal would have detection everywhere.   

https://www.google.com/maps/@33.8275862,-84.38748,55m/data=!3m1!1e3

IMO, fully-actuated signals are overkill at many mid-block intersections.  Here is the amount of detectors needed in that Atlanta intersection running fully-actuated and semi-actuated (assume there is just one-lane along each side-street approach)

Fully-actuated:  8 vehicle detectors; 8 pushbuttons (16 detectors total)
Semi-actuated:  2 vehicle detectors; 4 pushbuttons (6 detectors total)

You increased your detection cost by 166% by running fully-actuated (16 detectors vs 6).  For the added construction and maintenance cost, there is little benefit having main-street detection at this Atlanta intersection.

I hope this leads to more efficient signal timing and more respect for the signals by pedestrians.  Let's tell ped activists to stop complaining about "beg buttons" by designing the signal in a way that helps all users.

In your initial post you are proposing that the WALK comes up automatically during certain times of day but then require the pedestrian to press the pushbutton during other times.  Is this correct?  How does that help pedestrians respect pushbuttons?  If anything it adds confusion for pedestrians because sometimes the pushbutton does something, sometimes it doesn't.

I initially proposed 8 push buttons but only 2 vehicle detectors.  Main street traffic is generally assumed to be a default.  You are correct that this would be more expensive.

The idea with having detection for main street peds is to avoid having to go through a whole signal cycle at 2 a.m.  If the signal rests at main street green and solid don't walk - if a main street ped comes first, we can go directly back to walk, if a side street ped/vehicle comes first we can go directly to yellow without cycling through the pedestrian signals.

Main street walk will always come on after side street signal.

[johndoe]

Well if you really want to get more technical you may enjoy this document (particularly chapter 4): http://ops.fhwa.dot.gov/publications/fhwahop08024/fhwa_hop_08_024.pdf

The problem is traffic signals are reactive, not proactive...It can only count the number of cars that the previous cycle received and adjust the timing accordingly
accordingly

Most traffic signals don't even have that much flexibility.  It sounds like you're describing adaptive signals, which is also what roadman65 asks about (when sometimes a left is leading and sometimes lagging) .  The system has algorithms which prioritize movements based on volumes and desired arrival times.  Engineers can allow certain movements to be called in a different order depending on the coordination plan.

[tradephoric]

The idea with having detection for main street peds is to avoid having to go through a whole signal cycle at 2 a.m.  If the signal rests at main street green and solid don't walk - if a main street ped comes first, we can go directly back to walk, if a side street ped/vehicle comes first we can go directly to yellow without cycling through the pedestrian signals.

This only works if road agencies are willing to bring up the WALK in the middle of the associated vehicular green phase.  Many agencies don't allow this to happen.  Instead they operate the signal as Jet described above:

It's normal around these parts for the pedestrian phase to start only at the beginning of the associated car phase: so if the signals are resting on green for the major road and a pedestrian comes along wanting to cross the minor road, pressing the button will first call a green light for the minor road (even if there are no cars there), then give the pedestrian a green when it reverts to the major road again.

This makes sure pedestrians are only introduced into the intersection just as the parallel cars are being given a green light. Combined with a 5-second 'head start' for pedestrians, this goes a bit towards reducing the chance of turning vehicle collisions.