[Tagging] Theoretical crosswalks (was Re: Pedestrian access tagging)

Minh Nguyen minh at nguyen.cincinnati.oh.us
Mon Jan 25 10:01:58 UTC 2021


Vào lúc 16:14 2021-01-23, Graeme Fitzpatrick đã viết:
> 
> 
> 
> On Sun, 24 Jan 2021 at 08:04, Brian M. Sperlongano 
> <zelonewolf at gmail.com 
> <mailto:zelonewolf at gmail.com>> wrote:
> 
>       When a user has run on 98% of the roads in a city, and the
>     remaining 2% are roads that aren't runnable for various reasons,
>     they are pretty motivated to fix the map to get that number to 100%
> 
> 
> But if the last 2% aren't runnable, haven't they already achieved 100%? :-)
> 
> In regard to access, does foot=no "stop" (as far as routers are 
> concerned) a pedestrian from walking "across" a road, or only "along" it?

As far as I know, the current crop of OSM-based routers doesn't consider 
anything about the cross street when trying to cross it. Ideally, a 
router wouldn't tell a driver to cross six busy lanes of traffic to 
continue straight along a quiet side street, but that happens today.

Speaking of which, does this discussion about legal restrictions extend 
to the inclusion or omission of crossings? As you put it, a crossing is 
just another way to access and use a road, just in a different 
direction. It stands to reason that we should apply the same standard, 
but this is where I think the strict legal standard breaks down.

Consider California, where there is technically a crosswalk on every 
side of every intersection even if there are no crosswalk markings, 
signals, or signs. It would be jaywalking to cross where a sign says not 
to [1] or to cross midway along a block between two adjacent signalized 
intersections [2]. However, that leaves a lot of potential theoretical 
crossings that would be legal but surprising in practice, perhaps 
including cases where you might have to argue your way out of a 
jaywalking ticket in court.

In San José, as we systematically imported sidewalks as ways, we 
simultaneously identified and mapped crosswalks to connect the sidewalks 
to the routable road network. We mapped many unmarked crosswalks 
throughout the city, especially in residential neighborhoods. However, 
we consciously omitted unmarked crossings across collector and arterial 
streets. For example, in the T-intersection at [3][4], we mapped an 
unmarked crosswalk across the side street (Edwards Avenue) but not 
across the main street (South 1st Street), even though both crossings 
are legal apart from laws on due diligence or negligence.

Our reasoning was that renderers and routers need to reliably 
distinguish between the unmarked crosswalks that are commonly thought of 
as crosswalks (like the one across Edwards) and those that are more or 
less a loophole in the law (across South 1st). A less capable pedestrian 
would be ill-advised to cross South 1st here compared to the safer, 
marked crosswalk just a block or two away. Conversely, a capable, 
desperate enough pedestrian would cross the street anywhere it's safe to 
do so, regardless of the router's suggestion. (They might even cross 
halfway down the block -- which would also be legal in this case.) 
Additionally, if we had mapped all the legal crossings, Osmand would've 
annoyed drivers with excessive warnings about theoretical crosswalks, 
making the warning less effective at real ones.

Though I'm casting the rationale in relative terms, it isn't really a 
subjective decision. As locals, we intuitively knew not to map unmarked 
crosswalks across a road that mostly had signalized intersections, 
because that happens to be a reliable heuristic for an unsafe crosswalk 
in this city. Other heuristics like lane count and speed limit would've 
been unreliable here but might be reliable in a nearby suburb.

It'd be unrealistic to expect a router to infer the safety of a 
theoretical crosswalk, especially at scale across a variety of 
localities, but that seems like the logical conclusion to focusing on 
legal technicalities to the degree that this thread has for the foot=* 
key on roadways. Yet I'm uncomfortable with the notion that I have to 
suppress my local knowledge in favor of a legal analysis that defies 
common sense.

[1] R9-3bP in the MUTCD
[2] California Vehicle Code §21955
[3] https://www.mapillary.com/map/im/tA6zQ8oDotikoPTvZiAMFE
[4] https://www.openstreetmap.org/node/65546220

-- 
minh at nguyen.cincinnati.oh.us




More information about the Tagging mailing list