[Imports] Uploading sidewalks in San Jose, California, US
nbolten at gmail.com
Fri Oct 20 16:05:32 UTC 2017
Even just knowing and seeing where the sidewalks are is a massive
improvement over the typical situation, which is, unfortunately, no data at
But routing is a primary target and the sidewalks were never intended to be
disconnected on the long term. As crossings are added, I'll recommend (and
probably help with) a pass with the JOSM validator to find disconnected,
overlapping ways, a good best practice in any case. With a couple hotkeys
in mind, it takes only a few seconds to connect a block.
But this really is something of a router issue, and seems to be the main
point of contention. Sidewalks should definitely be part of a connected
network (as is the plan in San Jose), but there are all kinds of situations
where incremental mapping will create pedestrian "islands" that remain in
the map for a long time:
- building courtyards when indoor corridors have yet to be added.
- small paths in residential areas, like back yards.
- rooftop features that lack connection to indoor corridors.
- some boardwalks (gotta go over sand)
And of course, eventually, sidewalks and pedestrian infrastructure end and
become disconnected. Recommending walking in the street is a consequence of
the primary forms of routing being car-centric and is a bit of a hack:
there's no real confidence that it's safe or legal.
In the examples provided with graphhopper, there are two things it should
probably be doing:
1. Preventing disconnected subgraphs from being selected in the first
place, or at least providing an alternative route. Real pedestrian
infrastructure has gaps and islands.
2. Given the extremely short distances involved, it should also offer a
path that ignores the map, which is what most routers will do.
The examples that have been given for breaking routing are over extremely
short distances, unlikely to ever be part of a real route outside of
multimodal trip planning where the destination happens to be by a bus stop.
Since that's part of the goal of this import, I'm sure it'll be addressed.
On Fri, Oct 20, 2017, 7:28 AM Andy Townsend <ajt1047 at gmail.com> wrote:
> On 20/10/2017 14:07, Clifford Snow wrote:
> > ... The goal isn't really to have pedestrian routing, but to enable
> > people, especially those with limited mobility, to get to and from
> > their destination.
> Apologies if I'm missing something here, but surely the latter has the
> former as a prerequisite? A router designed for people with limited
> mobility might take into account things like drop kerbs, crossing
> widths, tight turns etc., but as a minimum it still needs to be able to
> find a connected route from A to B, surely?
> Best Regards,
> Imports mailing list
> Imports at openstreetmap.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Imports