[SDCBC] Fixing a traffic signal (long)

[Michael Elliott]

The intersection of Carlsbad Village Drive and Monroe (in Carlsbad, natch) 
has been part of my daily ride for several years. The southbound Monroe 
portion wasn't well-planned for cyclists, though. It has a posted and 
striped bike lane but at the top of the hill, where Monroe intersects with 
CVD the lane vanishes, leaving cyclists no clear place to go to get out of 
the right turn only lane.

In addition, the buried signal loops never detected bikes (as per pretty 
normal) who wish to go straight or turn left. There is no pedestrian 
crosswalk across CVD at that side, either, so no button to press, either.

Cyclists had to cross the right turn lane -- the busiest -- and plant 
themselves in the straight or left turn lane, often for some time before a 
car triggered the signal to give a green. Crossing CVD against the red is 
very dangerous due to very short sightlines which prevent a cyclist from 
seeing crossing traffic in enough time to safely get across the 
intersection -- it's uphill and the crossing cars go very fast through 
that portion of CVD.

So I contacted the city street department about this mess. SDCBC's Kathy 
Keehan also wrote a nice letter to them at my request (thank you, Kathy!). 
After a couple of follow up letters the city agreed to restripe the lanes 
and put in video detection for the signal.

The bike lane was realigned at the top to put it between the straight lane 
and the curb on the right side of the right turn only lane. And with the 
video detection working for bicycles, a cyclist had a fighting chance to 
get across CVD alive.

But after a couple of months, the lights no longer detected bicycles in 
the left turn or straight lanes. I contacted the city about it two or 
three times, and they responded after sending crews out, but the signals 
continued to be unresponsive. All a cyclist could do is either wait for 
someone to come into the lane behind them, or run the red -- uphill across 
four lanes of hurrying vehicles with sightlines of less than 100 meters. 
I'm too cowardly to try that one, though I confess I do run reds when it's 
safe to do so.

I wrote the city again week before last and they said that they'd send 
someone out to fix the problem. On the following morning I tried it again 
and didn't find any improvement. I wrote again, and this time I 
volunteered my services as the Test Cyclist. They said that they'd have 
the contractor contact me directly. Last Wednesday the field technician 
from Republic ITS (http://www.republicelectric.com/home/) called and we 
discussed the situation. He said he had personally adjusted and tested the 
system the week before and near as he could tell, it had to "see" a 
cyclist, yet my experience was that it didn't. We made an appointment to 
meet Friday morning.

So at the intersection, he opened up the signal box and showed me how the 
detection system works. A small black and white Grundig video monitor* in 
the box displays the lanes where they meet the limit line as seen by the 
camera atop the signal across the intersection.  He showed how each of the 
four lanes (right lane, bike lane, the straight lane, and the left turn 
lane) is "outlined" onscreen into detection zones. The detection system is 
designed to notice contrast and complexity in those zones, he said.

"Watch," he said, "while I cross the street, you'll see the system detect 
me and each box will light up." He walked across the street in the 
crosswalk, and as he did so the right turn lane lit up, followed by the 
the bike lane. But the straight and left turn boxes didn't notice him. Of 
course they "saw" cars, but without an assistant to cross the lanes while 
he was watching, he'd never have seen the controller miss a smaller target.

He said he knew what the problem was: that the signal from the camera was 
degraded -- too weak -- due to the connector on the camera not being of a 
type that is suitable for outdoor use (BNC connectors which are commonly 
found on test equipment used indoors). A common problem, and like a proper 
technician he griped about the idiocy of using parts like that, about the 
lowest bidder process, and the nitwits that install these systems which he 
later has to clean up.

He told me he'd fix it up and that it should work peachy the next time I 
tried it.

I thanked him for his time and willingness to help, and rode into the 
intersection to go home. He watched the monitor. I normally sit on the 
right edge of a left turn lane when turning left so I can easily swing 
wide when making the turn. This apparently put me between the two 
detection boxes so he asked me to lean to the left. The camera detected 
me. But as soon as I stopped moving, it "lost" me. Like with the dinosaurs 
in Jurassic Park, if you stop moving they decide you don't exist.

He said that in addition to getting a stronger signal to the computer, 
he'd set the zones to "lockout" so that once they detect something in the 
lane, they will change the signals even if they lose that motion. Default 
behavior is for the controller to assume that a small object that is seen 
then lost is some unimportant signal, like a leaf or dog or something, I 
guess.

I tested the adjusted signal yesterday morning on my ride and it worked 
perfectly.

To summarize, the city of Carlsbad was very cool about the whole thing, 
Kathy was a great help in coaching me on how to approach the city, I 
enjoyed participating in the troubleshooting process, and (being a nerd) I 
got a kick out of seeing and understanding what goes on in those signal 
boxes.

======
*Who buys Grundig television monitors, anyway?

-- 
Mike "Rocket J Squirrel" Elliott