Stops vs. Signals

Originally posted on my blog October 9, 2010 as: 85 vehicles and pedestrians in 3:11:

This movie was taken by me, with my iPhone, on the way home from work on August 19, 2010 at 5:22 pm. The five-way stop controlled intersection (Franklin Ave/ East River Road / 27th Ave) seems to have a maximum throughput just over 1600 vehicles per hour. Most of the movements are saturated during the peak. This intersection has been blogged about before.

The downside for a stop controlled intersection is that the allocation of time across legs is “unfair”, i.e. drivers are supposed to take turns (yield to the right). Thus a leg which is just saturated will get just as much access to the critical points of the intersection as a leg that is supersaturated, resulting in much higher delays on the supersaturated movements. I did not measure delay, but it is longer on this day for travelers moving WB on Franklin Ave.

There are several other points to note.

  1. Drivers do not all know the “yield to the right” rule.
  2. This results in “negotiations” between drivers about who should go. Less aggressive drivers clearly lose, but eventually go.
  3. This generally increases throughput compared to obeying rules (do not start until the intersection is cleared is violated, to the benefit of throughput).
  4. The intersection is confusing but safe. Any crashes during peak times would be very low speed.
  5. It is more confusing because of the construction.
  6. The intersection was configured with operating signals in September 2010.

Originally posted on my blog October 16, 2010 as 147 Vehicles and Pedestrians in 4:18:

My favorite five-way intersection (Franklin Avenue/East River Road/27th) has now been signalized for over a month. The video was taken on Oct 11 in the late afternoon (I apologize for the poor angle, but I wanted the same position as before as much as possible, unfortunately the sun did not cooperate (or alternatively the clouds did not obscure the sun), also I reduced the resolution for the Web). Other differences to note are that school is now in session.

The intersection was roughly at capacity (as can be seen), in that most conflicting movements were fully served with approaching cars, though I suspect throughput could be a bit higher.

We observe a throughput of about 1948 vehicles per hour (based on my estimate of 147 vehicles and pedestrians and bicyclists in 4:18), which compares favorable with the more chaotic 5-way stop during reconstruction which served about 1600. There seems a long period of lost time that could perhaps be used to improve capacity/lower delay.

The main difference is the extra capacity due to more systematic parallel movements (yielding more than one critical point). Notice the pedestrians just past the 4 minute mark are still quite confused as to whether to go or not.
As a user, the pedestrian timing is still terrible, and I just go whether or not I have the signal, so long as I am unlikely to be flattened like a pancake by oncoming cars.