From sandors39 at gmail.com  Mon Feb 20 20:37:18 2017
From: sandors39 at gmail.com (Sandor Seres)
Date: Mon, 20 Feb 2017 21:37:18 +0100
Subject: [Routing] Ordinary roads as roundabouts and the contrary
Message-ID: <004501d28bb9$212e2a90$638a7fb0$@gmail.com>

First of all, any suggestion/advice could help me to (further) improve my
data-preparation-tools related to the OSM roads/streets data.

At the same time the issue may be of certain interest for developers working
with rooting/navigation systems.

In the preparation, while removing millions of redundant nodes, vectors,
poly-lines in the source data (related to the 5 road and 2 street classes)
we meet a special roundabout issue. Namely, there are over 140K road objects
(ordinary ways/not tagged as roundabouts) that perfectly match the (6
statistical) parameters of the safe roundabouts in the same class. This
number includes the turning circles as well. Testing on a large sample (from
all over the World, a tedious and difficult job) still shows that most of
the object are roundabouts but missing the proper tag. So, my
dilemma/question is - what is worse (more dangerous) in a routing/navigation
system:

1. Having large number of real world roundabouts as ordinary roads (road
segments) in the network, or

2. Having some ordinary circular (usually one-direction) roads as
roundabouts in the network.

Of course, for a simple rendering the issue is irrelevant but for the GIS
systems and especially for the turn-by-turn car navigation might be
essential.

The mentioned parameters detected on the set of safe roundabouts I use for
the georec (geometry recognition), error detection and reparation. In my
case, the small circular objects have essential role in the
data-generalisation and vector-scale levels.

Finally, it is worth noting that even in case of roundabout tagged ways in
many cases (thousands) the roundabout is "unsafe" (missing segments, mixing
road types, more than one loop, self-crossings, zigzag edges and so on) and
a robust automatic recognition and reconstruction is needed. The old fashion
"contact the mapper" does not work at least for two reasons: one, there are
too many of the cases and two, the cases are there in years and the mappers
are probably not active anymore. For illustration let me take some examples
where the rendering, the visual impression is correct but there is no even a
segment of a roundabout (see e.g. here  <http://osm.org/go/T87HKcHrN-->
http://osm.org/go/T87HKcHrN-- ). One model to reconstruct the correct
objects is to smash the whole shape into none-crossing vectors, so connect
them into simple poly-lines, connect these (if possible) into a
circular/polygonal line as roundabout (using the mentioned parameters) and
the remaining poly-lines move to the set of corresponding ordinary road
segments. 

Regards, Sandor.

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/routing/attachments/20170220/b3d79bb1/attachment.html>

From dieterdreist at gmail.com  Tue Feb 21 00:02:00 2017
From: dieterdreist at gmail.com (Martin Koppenhoefer)
Date: Tue, 21 Feb 2017 01:02:00 +0100
Subject: [Routing] Ordinary roads as roundabouts and the contrary
In-Reply-To: <004501d28bb9$212e2a90$638a7fb0$@gmail.com>
References: <004501d28bb9$212e2a90$638a7fb0$@gmail.com>
Message-ID: <70DD79DE-BD46-4A5F-9BCB-646A799321B4@gmail.com>



sent from a phone

> On 20 Feb 2017, at 21:37, Sandor Seres <sandors39 at gmail.com> wrote:
> 
> mixing road types


different road types in the same roundabout are not necessarily "wrong". E.g. roundabouts are often added to existing crossroads of major and minor roads and with typical dual carriageways (near the crossing) there remain less important sections of the roundabout (in some cases almost nobody uses them) between the links for the "major road".


cheers,
Martin