I think the Garmin routing algo is probably too stupid when it comes to
If I got that right the algo tries to get and stay(!) on major roads until
it gets close to the target.
This seems to be true for car and bicycle routing (faster or shorter also
doesn't seem to matter)
If such a route is found it is preferred to a much shorter ferry connection
which involves a switch to class-0-roads, no matter how long it is, even if
thes longer route also contains one or more ferry lines.
So, if you select start and end point of your route close to the ferry
terminals you are probably
routed via the ferry. If your start is somewhere out of town so that you use
a major road to get to the
ferry terminal the ferry is probably not used.
Many ferries in OSM are not connected to major roads, sometimes there are
hw=residential or hw=unclassified and of course hw=service between the ferry
line and the next major road, so I also don't
see how mkgmap could "correct" this without risking unpredictable routing
near ferry terminals.
The only work around that I found so far is to select the ferry route as a
part of the route, but this
of course also means that you have to know that this ferry exits.
Gerd Petermann wrote
> Hi Andrzej,
> popej wrote
>> Mkgmap could include following algorithm:
>> Find all suspicious objects, like highway links, roundabouts, service
>> roads connected to a ferry and analyze other roads connected to this
>> object. If any of these other roads get road_class 3 or 4, then increase
>> road_class of processed object accordingly to 3 or 4.
> This would not always help.
> See for example hw=service way 50330706
> One has to crawl through several service ways to find out that the ferry
> is connected with
> the hw=secondary way 475872377
> I see no easy way to implement an algo which would find out that this
> hw=secondary is the
> one that "delivers" the wanted road class. I fear this kind of mapping
> happens quite often.
> I would expect that the Garmin routing algo is aware of this when it comes
> to ferry lines, but did not
> yet test it.
View this message in context:
Sent from the Mkgmap Development mailing list archive at Nabble.com.
mkgmap-dev mailing list