[OSM-talk] Roundabouts - why is a separate segment required?

Colin Smale colin.smale at xs4all.nl
Thu Feb 15 11:22:11 UTC 2018

Just to throw another concept into the mix... so-called flare roads,
where a road joining a roundabout (or other junction for that matter)
splits into two short one-way segments which go either side of an
obstacle. Mkgmap tries to recognise them by seeing if they come together
within X metres. Why is this important? It's about how you present the
turn-off angle to the user. If you look at the pure geometric angles of
the OSM ways up to the next node, taking the first exit can easily get
rounded to "straight on". By looking at where the exit road is after
(say) 50m the you have a more accurate impression about whether it is a
right turn or whatever. Sharing nodes between the ingress from one road
and the egress to the next road makes this all more difficult.

As I said before, this is not about routing in a mathematical sense
(traversing the routing graphs) but about how the resulting list of
nodes/edges is presented to the user. It's great if you can seen the
calculated route superimposed on a base map, but if the written/spoken
instructions don't correspond to your perception as you approach the
junction, then the system has failed. 

On 2018-02-15 11:05, Maarten Deen wrote:

> On 2018-02-14 19:39, Dave F wrote: On 14/02/2018 18:23, Johan C wrote:
> No, they are not. Roundabouts are special types of intersections.  Which is another type of intersection.
> They have a way on which you can drive round. And round. And round.
> And they have other ways leading to and from this round way.
> Whenever you enter the roundabout you drive on this round way, even
> if it's just for a metre. And then you exit this round way on to a
> different way.
> The present tagging (used since 2005 or so, and all around the
> globe) is fine. 
> To repeat myself. You can determine if you need to "drive on this
> round way" from a single node. No need for a section between entrance

You can not determine that from a single node. You need to load the
whole way that connects to that node and than make a judgement call
which roads connected to that node you will traverse (which you don't
know, because from a topology standpoint you are not traversing that

It is like Matej's example.
Suppose it is mapped with the entry and exit road connected to one node.
Yes, you can see if there also connects a roundabout to that node and
you can make the determination that in that case you need to traverse
the roundabout in the correct direction.
But suppose there is not a roundabout connected but a (circular) way
with a oneway direction. Then you also need to make the decision that
you have to traverse that way.
But suppose the way is not circular (making you cross or touch a oneway
street), than you can not do that.
When is a road circular? Most roads are circular from a topology
standpoint, as in: you can reach a node on that way going in either
direction. So you can not determine from a topology standpoint if a road
is a circular road or a roundabout.

Your method of mapping makes for very elaborate edge cases and unwieldy
routing engines.


talk mailing list
talk at openstreetmap.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20180215/1d47a5e2/attachment.html>

More information about the talk mailing list