My most recent description of not quite getting killed in front of the NYS DOT Region 8 office evidently did some good. Although I wasn’t informed directly, this happened:
The minimum green time was increased to 10 seconds.
Which is five more seconds more than before, allowing us to get nearly all the way through the intersection before crossing traffic on Rt 55 gets a green light.
As before, the numbers are video frames at 60 fps.
T=0.0 – Burnett signal goes green:

T=2.5 – The trailer ahead of us starts and we’re rolling:

T=8.0 – We reach the Burnett crosswalk. Note the car beside us isn’t making much headway, either:

T=11.93 – Burnett signal goes yellow overhead, so the green phase lasted 12 seconds:

T=16.16 – Rt 55 signal remains red, but will change within a second:

Seen from the rear view camera, the Rt 55 signal went green while we were still in the intersection:

You’ll note the cars on Rt 55 behind us weren’t visible three seconds earlier, so, as far as they can tell, we’re running the red.
Fortunately, we’re almost where we need to be:

The timing still isn’t safe, but after three years, five more seconds counts for a lot!
I’ve been thinking for a while now that the CS students at New Paltz could do a project (they do them every year) for using computer vision to control traffic lights. Your little rig does not trigger the inductive sensors in the road. A vision system could make intelligent decisions about changing the lights, constrained by safety protocols (eg. leave the green on at least long enough for a bike to clear! No green lights in both directions (Duh)). It would give students experience similar to what’s required for an autonomous vehicle, but without the difficulty of a changing frame of reference. It would really be low-hanging fruit compared to a full self-driving car.
NYS DOT deployed CV sensors at a few intersections around here and, much to my surprise, our ‘bents pass as automobiles! The camera on Zack’s Way at Hooker / Rt 376 lets us make the left turn with zero hassle; the green time is a bit short, but at least we do get a green signal!
Oh, and give bonus points to anyone implementing the opposing-green detector circuit with hardware gates, not just a few lines of code!