[mkgmap-dev] unpaved roads

Wed Feb 22 00:17:06 GMT 2017

Gerd Petermann <GPetermann_muenchen at hotmail.com> writes:

> I tried to create a patch for the rules which set mkgmap:unpaved using
> the wiki and taginfo:
> https://taginfo.openstreetmap.org/keys/surface#values
>
> One probem with surface is that we have so many values (taginfo lists
> 4844 different today), many of them typos or combinations like
> "ground;grass" or nonsense like "paved;unpaved" I guess many of the
> latter were produced by "connect ways" functions in OSM editors, so
> not fully intended.
>
> Anyhow, my impression is that it would be better to change the rule so
> that it checks the known paved surfaces and assumes that all others
> mean unpaved.  The current rules are quite old, it was introduced with
> r1425 and last changed with r1489.

[I know I'm late to replying; I left the thread unread until I had time
to read it and think a bit.]

One question is what should "unpaved" mean.  That probably depends on
car vs bicycle, and it seems the real issue is that the Garmin format
isn't expressive enough to allow routing to decide later what it wants
to use.  Thus we are having to map each road to a paved yes/no, road
class, and speed class.

Another way to look at this is to ask why we are representing unpaved
roads, and what the routing policy is we are trying to achieve by that.
In my case, for car, generally what I want is to not use rough roads
(dirt, and even cobblestone) unless it's more or less necessary, which
I'd define as being the only way or saving lots of time, even if I drive
10 km/h on the rough road.

So I would suggest that we think of the Garmin unpaved flag as a "this
is a road I want to avoid, as opposed to a road I don't want to avoid"
bit, and not be so fixated on what paved means.  That may mean different
people build images with different rules.

Then, I would think about what an optimal route is, and how to get the
Garmin unit to compute one when it thinks it is doing shortest time and
avoiding unpaved.  So I would use the unpaved flag for roads that you
really don't want to use, even if they save time, and use lower speed
classes for roads that you have a mild preference for avoiding.

Of course, I am not really sure how this would work.  And I realize it's
trickier with bicycle, where you care about pleasant/safe and time is
not so much the point. But I think the only way is to try to map some
utility function into inverse speed and let the Garmin unit try to
minimize time, since apparently there's no way to compute other utility
functions directly.  So basically if you'd rather ride 5 miles on road A
than 1 mile on road B, as an alternative to get someplace, set A's speed
to 5x B's speed, and then min time will do the right thing, even if you
get bad time estimates.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 162 bytes
Desc: not available
URL: <http://www.mkgmap.org.uk/pipermail/mkgmap-dev/attachments/20170221/24d2d967/attachment.sig>