[Talk-us] Complex intersection mapping

[stevea]

>So what are we talking about? Intersections like this one, where one
>or more dual carriageways come together at an at-grade intersection:
>
>https://www.evernote.com/shard/s9/sh/6438c196-bb92-4f66-81dc-9b75186286ba/0e8f07ff527c6a85c0dec426b9b79f1e
>
>One of my colleagues at Telenav has remapped this intersection as follows:
>
>https://www.evernote.com/shard/s9/sh/3491f1fe-6afa-4571-bc43-7cb31c9c2625/9dd47d1445fdcf03d3f0bbd93b8e0f92

Hi Martijn:  one thing "wrong" I do see at this particular 
intersection are extraneous nodes with highway=crossing tags:  two 
extra ones on the (northerly) east-west ped-path and one extra one 
each of the (westerly and easterly) north-south ped-paths.  A fairly 
minor error in the context of your larger question.

>The main difference, and the source of some feedback we have received
>over the past few days, is that the dual carriageway roads are
>straightened out, creating multiple intersection nodes (4 in this
>case) instead of the original single intersection node that connects
>all the incoming and outgoing ways. That technique turns out to yield
>more reliable and correct routing and guidance ('keep left, turn
>right') through these intersections in our testing. But of course,
>that cannot dictate how we map as a community, so let's discuss.

I do this myself on intersections which have complex "two-to-one" 
lane collapses in one direction, or "two-to-three" lane expansions in 
another direction, or even both.  I agree with you that making lanes 
which capture dual carriageway and multiple lanes like this 
accurately represents what is on the ground better, is pleasing to 
the eye both in an OSM editor and on an OSM rendering, AND likely 
results in better routing algorithm results (e.g. for offering turn 
directions).  The wiki entry on Braided Streets notwithstanding.

>Some of the feedback we have received about these edits points to a
>statement on this wiki page:
>https://wiki.openstreetmap.org/wiki/TIGER_fixup#Braided_streets: 'It
>is a reasonable and well-used technique to bring the ways of dual
>carriageways back to a single point at intersections to facilitate and
>simplify the mapping of control devices and turn restrictions.' In my
>mapping across the US, my personal experience has been that this
>technique is in fact used, but the 'after' technique with straightened
>out ways is actually much more common. I personally prefer that
>technique as well - I think it is more pleasing to the eye, represents
>what is on the ground better, and is and easier to read. So my feeling
>was that this mapping practice would not be disputed. It turns out I
>was wrong, so I want to see what the consensus is on mapping
>intersections of this type - or perhaps there is none and we can work
>together to get there?

I don't know what the solution is.  It may be to coexist with BOTH 
types and try to do the best that can be done by "smartening up" 
routing algorithms to cope with EACH type of intersection as well as 
can be technically achieved.  That seems "long-term wise" given that 
there will likely be both types of intersections entered into the 
underlying data.

SteveA
California