André Pirard A.Pirard.Papou at gmail.com
Thu Aug 29 21:29:22 UTC 2013

```On 2013-08-29 16:57, Colin Smale wrote :
>
> Won't the calculated radius depend on the number of points on the way,
> and the width of the road? If you look closely at the geometry of a
> curved road in OSM it is of course made up of straight segments with a
> certain angle between them. A right angle junction might be a 45 steps
> of 2 degrees each, giving a smooth curve with a certain radius, but it
> can also be represented by a single node with an angle of 90 degrees.
> How do you work out the turning radius from that?
>
I said on the Osmand forum that for the same total angle, the radius can
be long or short.  It does not really depend on the number of points
(1). The most significant parameters are the straight ways with their
points at which the road starts turning. If the bend is as simple as
that, with a constant curvature, then it's simply the radius of the
circumference that's tangent to the straight ways at the said points.
But, of course, the road may not be an arc of circumference, which makes
But, having seen what I saw, I doubt that the mapping of many bends is
precise enough, especially those I moved 20 m or much more back to their
real place, which some sometimes called micro-mapping.
Regarding 90° junctions, their radius is theoretically infinite so that
rotating a car around its vertical axis is unsafe at any speed ;-)  In
reality, it depends on the minimum turning radius which you'll find in
car catalogs. But as the cars stop or almost at such junctions, you
divide a very small speed by that radius and that's why the cars very
rarely skid there ;-)

On 2013-08-29 17:12, Pieren wrote :
> Then, instead of going on site to measure it with a decameter and add
> a new tag on a splitted segment, I would simply add more nodes....
With a new tag precisely_computable_radius=yes and a trig lib in the GPS
to compute it.
With an average of 1 per continent, the turning_radius tag is mapped on
a node, no split, they forget to specify which but logically at the
As to measuring it, you're right about the decameter, the first problem
is finding the center and the second one spinning round with trees in
the way and cars passing on the road ;-)
And well, amusing methods for funny tags, I would use a varying radius
circumference to overlay the road on an ortho, the largest radius so
that no part of the road extends outside the circumference. Even a
single circumference in a zoomable transparent layer does it. No math
even needed with the two zoom factors. Slide the circumference under the
scale of the road layer to measure it, or use a measuring tool in the

Did I say fun?
Drive carefully,

André.

(1) 2+3 points exactly on a circumference make a good bet that it is
one, but more points are needed to make sure there is no very improbable
zig-zag in between. Take some time to read about the Bézier curves
<https://en.wikipedia.org/wiki/B%C3%A9zier_curve> for more complicated
ones if you like.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20130829/8882da7f/attachment.html>
```