[Talk-GB] Imaginery footpaths added by user "Gavaasuren"

[SK53]

Hi David,

Most of these problems are issues for a router for interpreting OSM data,
rather than specific problems for the data.

There are plenty of examples of people building routers for people with
restricted mobility using OSM data (for instance wheelchair users, blind
people etc). Most of us will map steps on footways simply because even one
step acts as a barrier to wheelchair users or many older people.
Fortunately barriers for pedestrians are not as common as they used to be.

For places as mapped as areas a simple strategy for a routing engine is to
find the centroid and build virtual paths to places on the edges. Similarly
such areas can be buffered by, say 5 metres, to determine any overlaps with
highways mapped as centre-lines. All such things are relatively simple
post-processing steps on the data which can be easily carried out in
PostGIS and do not need special tagging on OSM.

Many of the members of this list routinely use OSM on a daily basis (we
'eat our own dog food') for pedestrian routing. I have used OSM for this
purpose for 5 and a half years and have never encountered any problems
other than missing data. My main purpose of routing is to have accurate
estimates of time to get to place X (usually a bus stop or railway station)
so as not to miss a public transport connection. It works very well.

Jerry


On 18 August 2014 11:40, David Woolley <forums at david-woolley.me.uk> wrote:

> On 18/08/14 10:59, SomeoneElse wrote:
>
>> Whilst the existance of a highway=pedestrian area that isn't connected
>> is an indication of something, it's usually just an indication of that
>> mapping in a particular area is not complete.
>>
>
> Considering the longer term problems:
>
> 1) There needs to be better guidance to routing software developers on how
> to route when there are parallel features accessible on foot;
>
> 2) There needs to be a lot more mapping of barriers.
>
> Ideally, the routing rule for foot needs to be something like that,
> subject to access and surface quality considerations, if there is no
> barrier between adjacent features, you may cross at any point between
> them.  In this case, there has probably been pressure to make life easier
> for the router.
>
> I think this also came up recently with regard to central reservations on
> non-motorways.
>
> The other difficult situation we have here is that pedestrian areas are
> mapped physically, as the actual area occupied, but most roads are mapped,
> abstractly, as an infinitely narrow line on the centre of the carriageway,
> so you will get a gap between the two and the router has to use some
> heuristics to decide whether that gap is bridgeable on foot.  I have seen
> cases where the pedestrian area was mapped out to the centre of the road,
> but I considered that wrong.  (In fact, mapping roads as areas will
> generally confuse routing software.)
>
> Another variation of this routing problem is that of where is it
> reasonable to cross a road.  Ideally, physical barriers at the centre of
> the road should be mapped, and access restrictions put on any reservations
> that is not supposed to be used by the public, but the main consideration
> tends to be the level and speed of traffic and the visibility of that
> traffic, combined with whether or not there is a designated crossing point
> near enough to be used.
>
> There really isn't enough information mapped to make a decision as to
> whether it will be safe to cross.  Also, a little old lady may not be safe
> crossing at an arbitrary point, whereas it will be no problem for a more
> able bodied person.  Some people may want to avoid pedestrian subways,
> particularly after dark.  Any mapping of crime levels in them is likely to
> be volatile and may even move the crime.
>
> Particularly for residential roads, you might get into the dangerous area
> of mapping actual maximum speeds on rat runs, as, there is a road near  me
> with a 20mph limit, but, apart from speed bumps it is long and straight, so
> vehicles may get up to 40 mph between bumps, with visibility limited by
> parked cars.  The council policy is to only use passive enforcement.
> Mapping that as 40 mph de facto, may encourage people to use it that way,
> but saying it is safe for little old ladies to cross at night, based on the
> 20 mph limit may also be wrong.
>
> Maybe there is a need for a verification tool that renders additional
> random interconnections and crossing points, so that one can see whether
> there is a need to add barriers, and other hints, to prevent such routings.
>
>
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-gb
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20140818/db66fbd7/attachment-0001.html>