[mkgmap-dev] special cases with short arcs

Tue Sep 24 08:59:13 BST 2013

Hi,

I am trying to find a better algo for the handling of short arcs.
(Garmins routing algo refuses to route over short arcs)
The typical case of a short arc is a crossing like here:
http://www.openstreetmap.org/browse/node/1449682318
We don't lose much information if we merge the node with
node 1303933872, so merging is a good idea here.

On the other hand I stumbled across a few cases 
were I am not sure what to do. I many cases I'd 
prefer to move one of the two nodes so that the 
arc gets longer.
See e.g.
http://www.openstreetmap.org/browse/way/47138169
The first two points of the way build a short arc (2.8 m),
while the next arc to point 267185122 is > 33m.
The merging of these two nodes will produce a map
which implies that you can turn right directly at the
end of the bridge
http://www.openstreetmap.org/browse/way/47138168
I think it would be better to move the node 627157116
a little bit closer to the next point 267185122.

A similar case: 
The bridge at http://www.openstreetmap.org/browse/way/131860005
has a length of ~ 4.99 m. With the default value 5m, this is considered 
to be a short arc and short-arc-removal will remove the bridge by 
merging the two end points. 

On the other hand, a (service) road with just two points, one point 
connected to a road, the other not connected to a road should
be removed if it is a short arc instead of eventually moving
the point connected to the road (as this produces the zig zag lines as 
posted by popej (road.png):
http://gis.19327.n5.nabble.com/Distorted-lines-tp5775761.html
(this one is easy to detect and I can create a simple patch)

I don't know yet how to implement all this, so I'd like to make sure
that I am not investing much time into a wrong solution.

Comments?

Gerd

--
View this message in context: http://gis.19327.n5.nabble.com/special-cases-with-short-arcs-tp5778708.html
Sent from the Mkgmap Development mailing list archive at Nabble.com.