Saturday, April 25, 2015

Signal Timing - the Phoenix Experience

One of the most difficult, challenging, hair-pulling, and rewarding experience is to completely re-time a city. Over the years, I have been blessed to be involved in many re-timing projects.  Timing a city provides a great benefit/cost ratio because for a small amount of money, you can gain huge user benefits.

When I say signal timing, I am not simply talking about settings in the controller - walk, don't walk, clearance, all-red, etc.  I am talking about creating cycle lengths and offsets for the safe and efficient movement of traffic along a corridor.  It's one of the most challenging projects a traffic engineer will ever undertake.

Normally, the process goes like this.  First, you collect traffic data along major corridors and at intersections.  Then, you will typically find logical progression corridors or grids.  The next step involves using signal timing software to optimize the timing.  Finally, these timings are loaded into the system.  Fine-tuning is necessary because as you change the system - you will likely modify the behavior of the motorists.  If motorists are suddenly faced with different timing, they might speed up or slow down - depending upon what they see.

I have seen some cities never perform major traffic signal timing projects.  For example, Phoenix is laid out in a grid pattern.  There are typically traffic signals every mile, half-mile and sometims quarter mile.  And it turns out that a 90 second cycle length provides incredible progression in all directions.  That is because it takes about 90 seconds to travel one mile at 40 miles per hour. If the capacity and phasing of the major intersections can be controlled with a 90 second cycle length - then this provides a perfect backbone for signal timing.  Simply time all of the intersections the same and put in a 90 second cycle length - you'll get pretty decent progression.

For years this worked in Phoenix - mostly because for a long time there were few left-turn phases.  But now that exclusive left turn phases are at almost every major intersection - the 90 second timing no longer works.  There isn't enough phase time to accommodate all the phases.  So even Phoenix now should perform signal timing on a frequent basis.

I am not sure how many cities perform signal timing overhauls.  As traffic patterns change and volumes increase, it can have a dramatic effect on the system.  In my opinion, these overhauls should be done every five years at least.  But budgets are tight, the work is difficult, and few people will recognize the difference.  It's not the flashiest project in the world. But the benefits are huge.

Bill Reilly Story

I'll never forget when I met Bill Reilly.    In the late 1980s, he was a big name in the development of the Highway Capacity Manual.  I think he was the chairman of the signalized intersection committee. This was when Dolf May was the overall chairman.  I was a newbie traffic engineer doing work out of Phoenix, Arizona for Lee Engineering.

Anyway, we had been measuring traffic flow rates at intersections in Phoenix.  We had determined that the 1800 Ideal Saturation Flow Rate (The rate at which vehicles move through an intersection) was too low.  But the Saturation Flow Rate was also modified by these factors that were listed in the Highway Capacity Manual.  Each factor lowered the flow rate by a bit.  One factor, I think it was the truck factor, really substantially lowered the flow rate.  But no matter how hard I searched, I couldn't find any research documenting this particular factor.

It turned out that one year, I was to present one of my papers to the Transportation Research Board.  I was excited to travel to Washington, D.C. to present the paper, but I was also excited to attend the Highway Capacity Meetings.  Because I would be able to find out where the research was to back up the factors.

Well, the meeting wasn't really the right place to bring up this issue, and so after the meeting, I boldly went up to Bill Reilly and asked him where they came up with the factor.  His answer surprised me, to say the least.

"We made it up", he said.  "We needed a factor and we didn't really have any research, and so this is what the committee guessed".  I was astonished.  It had huge implications across the United States and there was no data to back it up?  Amazing.

Of course I now understand that much of engineering is educated guesswork.  You can't have a procedure or factor for everything.  That's why experience and grey hair is really quite valuable.  There's nothing that will give you that experience except monitoring and studying traffic for a career.  Bill Reilly certainly did that.

We sent our research into the Signalized Committee and I think this prompted more research into the area.  But there is nothing wrong with the approach that the committee took.  They needed a value and they took their best guess.  It was certainly a better guess than I would have made.

Loop Detection

I remember in college my friend, Steve, lived in a dormitory on the other side of campus.  To get to his dorm room, we had to cross a major street.  There was a pedestrian crossing but it was not actuated.  It simply provided a pedestrian phase in synch with traffic progression along the street.  In other words, it provided a pedestrian crossing every 120 seconds or so.

But as engineering students, we didn't know about traffic engineering yet.  Someone had told my friend that the "traffic" pullbox was really a pressure sensor.  If you wanted to cross the intersection, you simply needed to jump up and down on the box and soon you would get a green phase.

It worked!

 If we jumped up and down on the pullbox, we would get a green phase... about every 2 minutes or so.  Remarkable.  The sad thing is that I believed this until I became a traffic engineer.  I look back and wonder how dumb I must have been.

There are many ways to actuate a signal now, but back then, almost everything was done by loop detection.  A loop detector is a wire coil embedded into the street.  A small current is constantly being passed through the loop creating an inductance field.   When a vehicle passes over the loop detector, it creates a small disturbance in the field which is amplified through a loop amplifier and the signal is then sent to the controller.

The loop detectors are quite noticable at most intersections - as it is usually placed into the intersection by sawcutting the pavement.  This leaves a terrific scar on the roadway.  People know this scar has to do with vehicle activation, but they aren't really sure as to how it works.

Over the years, I have heard some pretty amazing theories.  I have heard that there are magnets in the roadway.  I have heard that they detect weight.  I have heard that they detect temperature, sound, and exhaust.  But nope.  They simply detect the change in the field created by the current in the loop.

I bring this up because of a recent article about loop detectors and bicyclists in Chicago.  The article states "Motorcycles and bicycles often aren’t big enough to trigger magnetic sensors that switch traffic lights from red to green, WBBM Newsradio’s Alex Degman reports."

Actually, they aren't magnetic sensors in the roadway.  The inductance field is a magnetic field, but theren't aren't any magnets in the roadway at all.  But, if you jump up and down really hard...

Saturation Flow Rate

There is a fancy term to describe the maximum number of vehicles that can travel through an intersection.  It's called the "Saturation Flow Rate" or SFR.  In 1985, traffic engineers believed that unimpeded queued vehicles would travel through an intersection about once every 2.0 seconds.  Based upon that, and probably a few studies, the Saturation Flow Rate was set at 1800 vehicles per hour.  (This is obtained by dividing 3600 second/hour by 2.0 seconds/vehicle). It was a nice round number.

There was a problem with this value, however.  When we did traffic studies at very busy intersections, we found that traffic moved through the intersection faster than this value.  Sometimes our calculations left us scratching our head.  We would measure traffic through the intersection greater than the calculated capacity.  What was going on?

The answer is that the 1800 Saturation Flow Rate was too low.  Today it is set at 1900, but in 1985 the rate was 1800.

Now the 1985 Highway Capacity Manual method was correct.  Given all the inputs into a capacity calculation (lost time, cycle length, saturation flow rate), it will fairly accurately calculate average vehicle delay. But the 1800 SFR flow rate was not correct.  It was too low.

In the late 1980s, we performed research of saturation flow rates at major intersections in the Phoenix area.  What we found was very surprising.  At busy intersections, we found that vehicles traveled through the intersection at 1.8 seconds per vehicle.  And the results at some intersections were astonishing.

One major roadway through the Phoenix area is Grand Avenue.  Back in the 1980s this road created several six-legged intersections. Since each roadway needed its own phase, the cycle length at these intersections could be almost three minutes.  So when vehicles received a green light, they went.  And boy did they go!

We measured Saturation flow rates at Grand Avenue and found that they were consistenlty between 1.6 and 1.7 seconds per vehicle.  This translated into a Saturation Flow Rate of 2200 Vehicles per hour.  This was substantially greater than the 1800 Vehicles per hour that was imbedded in the 1985 Highway Capacity Manual.

Today, the 1900 might still be low.  Most agencies don't have the resources to measure Saturation Flow Rate, but they should.  It gives a more-accurate picture of the actual traffic situation.