[mkgmap-dev] Problem with --adjust-turn-headings

Fri Feb 5 12:47:18 GMT 2010

Hi Steve,

> Please look at 
> http://www.openstreetmap.org/?mlat=51.44827&mlon=-3.49581&zoom=16&layers=B000FTF
> 
> If you approach the marked junction on the B4270 from the south, 
> intending to continue northwards on the B4268 then you get told "turn 
> left" as you approach the junction with 300m or so to go. Trouble is, 
> there is a minor road junctioning with the B4270 at about the same point 
> on the road, and the geometry of the combined junction most certainly 
> does not agree with a "turn left" for the B4268!
> 
> The portion of the B4270 that continues to the north-east makes a 
> junction with the B4268/B4270 as a "give way" on the road. The 
> transition from the B4270 to the B4268 is seamless on the road. The 
> minor road going west does of course join the B4270 as a "give way" on 
> the road. You can see the physical details on Google Earth if you need 
> more detail.
> 
> But to hear "turn left in 300m" is most certainly a surprise! I'd expect 
> no verbal instruction in this case.

Yes, that's not ideal is it?

Tricky case this one. The proximity of the minor road on the left to
the bigger junction does confuse the issue. However, even if that
wasn't there, you would still get the "turn left" message because the
turn heading adjustment code has adjusted the heading of the turn onto
the B4268 to make sure that you know that you have to leave the B4270.
The GPS doesn't know that the road continues in a straight line
and that the turn to the NE is in fact the side road.

Now, if the OSM data could actually express the fact that the straight
on direction was the "main road" (even though the road to the NE has
the same ref as the incoming road), then we could avoid this problem.

Perhaps we should introduce a relation type (prefixed with mkgmap:,
perhaps) that could be used in cases like this to express that two ways
are "the same road" even though their refs are different.

Actually, I believe a lot of routing direction issues would be
solved/improved if there was some way of expressing the fact that road
segments were related. At the moment, as far as I can tell, the GPS has
no idea that at any given junction, which roads are related. Either
Garmin doesn't support that concept or they do and we haven't (yet)
discovered how to express it. Thing is, I have not seen any data in
real Garmin maps that looks like that so my belief is that it's not
implemented.

Mark