Dynamic pedestrian intervals

[tradephoric]

The idea is to terminate the FDW when sensors no longer detect the presence of pedestrians in the crosswalk.  Using a standard 3.5 ft/sec walking speed found in the MUTCD can be restrictive to an intersections operation.

https://www.youtube.com/watch?v=6yCbtgr7rSI
https://www.youtube.com/watch?v=Yc5i9-mVxfM

[riiga]

Interesting. What's the significance of this proposal? Is the duration of green at US crosswalks always the same?

Here at least we have different intervals at crossings depending on whether there are pedestrians crossing or not together with bicycles. Only bicycles = shorter duration of green at crossing, only pedestrians or both = standard duration.

[froggie]

Is the duration of green at US crosswalks always the same?

No, it's highly variable depending on jurisdiction, age of the signal, and the environment of the signal.  There are numerous traffic signals in the US that don't even have pedestrian indicators...and most of those that do require a pedestrian to push a "beg button" in order to get a walk signal to cross.

[roadfro]

Is the duration of green at US crosswalks always the same?

No, it's highly variable depending on jurisdiction, age of the signal, and the environment of the signal.  There are numerous traffic signals in the US that don't even have pedestrian indicators...and most of those that do require a pedestrian to push a "beg button" in order to get a walk signal to cross.

The issue comes from not the walk duration, but the flashing don't walk clearance time. This time is fixed and based on the width of the crossing. So while a walk man can be displayed for 4 seconds or 44 seconds at the same intersection, the flashing don't walk time is always the same. By MUTCD regulation, a 3.5 ft/sec time has been adopted as of 2009 where a 4.0 ft/sec time was used previously. This meant an increase in cross traffic time whenever a pedestrian is present, and has negative affect on signal coordination. The idea presented thus could have the ability to decrease the flashing don't walk time if it can be known that the pedestrian finished crossing earlier.

[myosh_tino]

Back in 2009, our local road guru, Gary Richards (a.k.a. Mr. Roadshow) of the San Jose Mercury News ran a story on the changes to the clearance times for pedestrians.  In the article, he discusses a system that would dynamically adjust the clearing time based on the people crossing...

Santa Clara County is testing a new signal system at Bascom Avenue and Renova Drive near Valley Medical Center in San Jose. Pedestrians still must push a crosswalk button to trigger their light, but sensors on poles at curbs will then detect how many pedestrians there are and how much time they need to clear the crosswalk.

If walkers are fast, the pace could be as high as 4 feet per second. If they are older or disabled or just ambling across, it could be slowed to about 2.8 feet per second.

The extra time provided by pedestrian detection will not show on the countdown signal, but the light for cross traffic will remain red until pedestrians clear the crosswalk.

"With this approach," Prasad said, "the pedestrian timing is only extended when needed rather than extending it every time, with or without pedestrians. Our approach will result in significant time savings."

However, it is costly. Adding the new system could run from $12,000 to $20,000 per intersection, or about $2 million for the expressways.

[Mdcastle]

Having a different pedestrian clearance interval each time is going to royally screw up the countdown modules.

[roadfro]

As quoted, the countdowns won't be changed every time, but the time before cross traffic green starts becomes longer.

My bigger concern about this would be the effect variable pedestrian clearance times would have on major arterial coordination.

[Zmapper]

How failsafe is this proposed system? What happens if the system thinks a nighttime pedestrian dressed in all black has finished crossing and thus changes the signal, but in actuality the pedestrian is still three lanes away from the curb?

[jeffandnicole]

How failsafe is this proposed system? What happens if the system thinks a nighttime pedestrian dressed in all black has finished crossing and thus changes the signal, but in actuality the pedestrian is still three lanes away from the curb?

That's the person in all black's fault for wearing all black while walking outside at night, and jaywalking.  He still has the walk signal and countdown timer.  If he started walking so late into the don't walk phase that he's still 3 lanes away from exiting the crosswalk, that's on him.

And that brings up another point - would, technically, the signal be held up forever if people kept ignoring the Don't Walk phase and kept walking into the road, once they figure out the light will remain red for them?

[Zmapper]

Perhaps it may not be prudent, but is completely legal to wear all black clothing. Additionally, darker complexioned pedestrians may cause a failure, through no fault of their own.

If a detection system failed to recognize black cars, it would be deemed a failure. If a pedestrian detection system fails to consistently detect "black" (clothing, skin color, hair color, etc) pedestrians, it must also be deemed a failure.

When designing a system for the general public, ALL of the general public must be accommodated. If the detection system is reliable enough to consistently detect all of the general public, my concerns are irrelevant. If not, then it is simply unacceptable to implement a flawed and potentially discriminatory system.

[roadfro]

How failsafe is this proposed system? What happens if the system thinks a nighttime pedestrian dressed in all black has finished crossing and thus changes the signal, but in actuality the pedestrian is still three lanes away from the curb?

That's the person in all black's fault for wearing all black while walking outside at night, and jaywalking.  He still has the walk signal and countdown timer.  If he started walking so late into the don't walk phase that he's still 3 lanes away from exiting the crosswalk, that's on him.

[/quote]

It will really depend on the type of detection used. This conversation seems to imply that it would be camera detection, but there are other methodologies (e.g. radar) that could potentially work. Even if camera detection were used, it's not a strict color on color detection method–rather, cameras detect motion and change in contrast within a predefined area, so it's likely that a pedestrian with dark clothing would be detected without too much issue (especially if there is decently adequate intersection lighting).

And that brings up another point - would, technically, the signal be held up forever if people kept ignoring the Don't Walk phase and kept walking into the road, once they figure out the light will remain red for them?

The signal wouldn't be held up forever. All modern signals can be set for a maximum phase duration, especially if there are other calls to service detected.

The greater question is at what point would the detection kick in? Does it start detecting only during the walk interval, and calculate clearance based on the pedestrian speed during that interval (leaving out pedestrians that begin crossing during FDW)? I don't know how that would work, but some kind of safeguard like that would seem reasonable.

[Zmapper]

Roadfro - Are heat sensors reliable enough for public use?

[roadfro]

Roadfro - Are heat sensors reliable enough for public use?

I've never even heard of that as a detection technology for traffic signals...

The main three types of detection I am aware of are inductive loops, microwave detection and video detection.

[spooky]

Thermal imaging is being used in video detection. The camera detects heat instead of motion.

[roadfro]

^ Huh... never heard of this. Gotta wonder if that would be worth it...seems more expensive to deploy on a wide scale.

[tradephoric]

Thermal imaging is becoming more common.  Here's an example of FLIR in action:

https://www.youtube.com/watch?v=_90Skumc_Zw#

[roadfro]

That's a thermal camera? Looks very similar to regular contrast detection cameras...