[Tagging] Proposal for standardization of sidewalk schema (+ import)

Meg Drouhard meg.drouhard at gmail.com
Mon Aug 8 20:43:14 UTC 2016

Hi all,

We've been drafting a longer response that we wanted to share with everyone
to give more context (for those also following the import list-serve, there
is a separate post there that goes through import-specific challenges).

Firstly, thank you to everyone for engaging with such vigorous discussion!
We’ve learned a huge amount about the weaknesses and strengths of our
current proposal.

We think it would be helpful to talk about our motivations in posting this
schema, to give the points we’re making some context. We have started this
project with the goal of making sidewalks in OSM more useful for the
greater community, particularly for people with limited mobility. With its
principles of openness and inclusion, OSM is uniquely positioned to adopt a
data model that can make a big difference for people who use wheelchairs,
crutches, or otherwise have difficulties walking, while also improving the
data in the map. We also see this as an opportunity to get more people
mapping and more data about pedestrian ways into OSM, as the current state
of sidewalk coverage is very poor (and is inconsistently tagged) outside of
a few notable cases.

One thing that we didn’t make very clear in our initial proposal is that it
is primarily about suggesting standard practices for using existing tags to
describe sidewalks. Almost every tag or feature listed in the Pedestrian
Layer section (http://wiki.openstreetmap.org/wiki/Proposed_features/sidewa
lk_schema#Pedestrian_Layer) is already described in other sections of the
wiki with some description of their semantics, but it is difficult to get a
coherent picture of ‘this is the way to map pedestrian ways’ because they
are distributed in different tagging discussions (sometimes in conflicting
ways). Almost all of the suggested tagging conventions are in use in some
fashion or another. For example, Ilya Zverev pointed out that it’s a common
practice to annotate sidewalks as separate ways in Russia. For another
great example of the potential for impact, the wiki user Species (very
helpful and active on the Talk page!) has mapped sidewalks as separate ways
in Graz, Austria, and has wheelchair routing working with an OSRM profile:
47.075407%2C15.440565. And even in locations making heavy use of sidewalk=*
on streets, you will find separately-mapped sidewalks due to important
discrepancies between street shape and sidewalk shape or in order to map
features that are between the sidewalk and the street. So a point we’d like
to make is that many of the concerns raised don’t just apply to a proposal
for tagging in the future - they apply to pedestrian ways that are already
mapped as well.  It would be interesting to see examples of how different
regions have coped with the issues raised here (like crossing streets
anywhere), as there may be multiple strategies that already handle that

The introduction to the proposal has been reworded to make the scope
clearer. We will also start annotating which tagging recommendations
constitute something new or different from standards found elsewhere.

A big issue that’s been raised is with mapping sidewalks as separate ways.
One of the biggest benefits to this approach is that it is a step closer to
the OSM principle of mapping conditions “on the ground”. It allows an
accurate, elegant tagging scheme for the ground conditions of sidewalks,
whereas current conventions lead to an explosion of tags for even basic
information. Also, by enabling the user to simply add in the location of
intersections and kerb ramps where they see them, it makes it much more
intuitive for the user, particularly new users, to add data correctly. We
hope that making this more intuitive will lead to more contributions and
better coverage of sidewalk data. It also resolves several visualization
concerns (the wiki has been updated to explain what we mean by this).
Finally, treating pedestrian routes as separate ways is more in line with
routing concepts (constituting graph edges). For examples of how complex
tagging would need to be to adequately describe pedestrian ways using only
street lines (and how that graph would need to be derived anyways, likely
with gaps in information), there is another proposal (
linked on our proposal), as well as the work of Nathanael Lang that was
previously mentioned (https://github.com/Nathanael-L/pedro). At first it
seems like the street turn lanes problem, but it has the added complexity
of (1) needing to add several annotations to each lane, (2) the lines may
intersect in non-trivial ways, and (3) there are almost always features
located on or between the lanes (kerbs, ramps, crossings, trees, bollards,
grass, bus stops, trash cans, bicycle paths…).

There are valid concerns that this proposal could affect routing solutions
for those who don't have mobility limitations, and in a related theme, that
mapping sidewalks as separate ways may only be appropriate for specific
geographies and doesn’t scale so well to the rest of the world. The common
challenge can be described as: what happens if I’m in a situation where I’m
able to cross the road wherever if I want to? It would be a pain, and
clearly inefficient if I was routed to the nearest crossing when I was able
to just cross the road.

Adding to that, there have been some concerns that the proposal does not
take European use cases into account - for example, that it would break
routing for users who want to cross the street at any location, or that
mapping sidewalk geometries would be inappropriate for certain geographies.
We would first like to point out that these use cases also apply in North
America, including the U.S., where jaywalking is allowed in many places
(regulations operate at the national, state, and local levels). This
Mapillary photo in Seattle, WA, US is typical of many residential areas,
and has a nearly identical appearance to the example of Cambridge listed in
the talk page of the proposal: https://www.mapillary.com/app/
onWI0tm97JK7Z5s5g. Such a user would naturally cross pretty much anywhere
there, and it’s an absolute necessity to walk in the street in areas that
don’t have sidewalks at all (some suburbs, e.g.). As a result, we think
these are universal concerns and would like to account for them.

In cases where the distance to the suggested crossing is short and it is
visible where the route continues after the crossing, the user could just
choose to exercise their judgement, ignore the router to cross the road
(hopefully avoiding cars) and pick the route back up once you’ve crossed.
For cases where the distance is significant to the nearest crosswalk there
are at least two further options. You could go ahead and directly add a way
connecting the sidewalk to the street (as things are currently mapped, in
fact), or with a bit of wrangling, you could structure the router (or
pre-process the data) to make cross-street connections without needing to
‘invisible’ ways to the OSM data layer. We think it helps to have visuals
for these examples, so we’ll be adding them our wiki shortly, please stick
with us!

It would also be really helpful for us if we could have specific examples
of cases where there is a pedestrian routing solution that currently works
for the user that would break with our proposed solution :). To the best of
our knowledge, all of the large-scale routers rapidly fall back to the
street network (ignoring the sidewalk tag) due to the incompleteness of
annotated sidewalks.

Integrating separately-mapped sidewalk ways with street centerlines (for
cycle routing, etc) and areas (plazas) are open problems that aren’t
actually specific to mapping sidewalks. There are already numerous
pedestrian ways that need to accomplish both tasks, and they seem to work
with automatic routers. For integrating with streets in particular, this
proposal suggests linking at street crossings and lowered kerb locations,
which happens at regular intervals in cities (and is addressed in the last

Associating sidewalks with a street is a particularly interesting point.
This problem is not very different from how buildings associate with
streets using either addr:street=* tags or an associatedStreet relation,
and we don’t yet have a perfect solution in place (it’s been listed as TBD
in the wiki for a while now). Knowing the associated street is not actually
necessary to make pedestrian routing on sidewalks work much better than it
currently does, particularly for people with limited mobility (and blind
users http://stateofthemap.us/2016/osm-lights-up-the-world-for-bli
nd-users-with-sound/), though it could probably be enhanced with a tag or
street relation (see the Talk page of the proposal for some good
discussions on this issue).

We’ve tried to address as many concerns as we can, and look forward to more

Thank you,

Tom, Meg, Jess, Anat, Nick and Kaicheng.

2016-08-04 1:06 GMT-07:00 Marc Gemis <marc.gemis at gmail.com>:

> On Thu, Aug 4, 2016 at 3:35 AM, Mike Thompson <miketho16 at gmail.com> wrote:
> > I tested out the proposed mapping/tagging scheme in my local area
> > (http://www.openstreetmap.org/#map=17/40.49192/-105.05655 - not
> claiming I
> > did it perfectly). I didn't think it was especially difficult. Drawing
> the
> > additional ways took a little more time, but offered the advantage that
> you
> > received a good visual as to whether all the sidewalks had been correctly
> > mapped.
> >>
> >>
> >
> Did you miss the crossing of the Stanton Creek Drive near the
> roundabout  ? Proves the argument that it is very likely that one will
> crossings :-)
> And without those extra ways, a router will probably allow you to
> cross Tortola way while walking allow Basseterre Place on the south
> side. Now it, will have to route you south to the roundabout.
> regards
> m
> _______________________________________________
> Tagging mailing list
> Tagging at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/tagging
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20160808/68fa6723/attachment-0001.html>

More information about the Tagging mailing list