Support Relation:connectivity: invalid right turn direction with lanes about osmand

[TheSnoozer]

Just saying there is also a no_right_turn restriction on that intersection: https://www.openstreetmap.org/relation/1775212.

Interestingly the intersection before:
[MUTE] Go ahead (C|TR) onto New England Highway A15;ACW and go 0.3 km. Way 256694082
which is correct.

the
Turn right (C|+TR) onto New England Highway A15;ACW and go 0.0 km. Way 761885309
should be a [MUTE] Go ahead

[yuriiurshuliak]

The bug has been reproduced. While navigating directly from one point to another, both the Next turn widget and the Lanes widget falsely suggest a right turn despite there being no actual turn along the route.

OsmAnd~ 4.7.0#1320m, released: 2024-01-04

video_2024-01-05_16-01-13.mp4

[ivanPyrohivskyi]

C++ version will do after Java review and together with issue #18158 and #19041

[vshcherb]

Not a regression

	<segment id="761885309" oid="48760659793" start="1" end="0" time = "1.0" rtime = "1.0" name = "New England Highway (A15;ACW) " maxspeed = "80" distance = "32.0" trunk turn = "Turn right (C|+TR)" turn_angle = "-0.5658264" lanes = "[2, 11]" start_bearing = "-169.40848" end_bearing = "-169.40848" height = "[]" description = "Turn right (C|+TR) and go 32.77 meters" />

Though mapping needs to be checked - https://www.openstreetmap.org/way/256694081 if this turn right is correct on this way

[mbartlett21]

The mapping is correct. The right turn lane continues both before and after the slip road comes off.

(Image from Southern Downs Regional Council mapping)

[vshcherb]

It's still is better for algorithm I think to have mapping as turn:lanes:backward=left;straight|right.
We will try to fix with this mapping, but I'm not sure cause it is a confusing mapping which could lead to issues in other places

[mbartlett21]

If it checks whether the lanes are the same configuration after the intersection that should eliminate that.

However, with example b below (which I don't know that any real world road would use), it would then give wrong directions if the intersection was mapped as a single point.

EDIT:
Part of the reason this is hard is because OSM doesn't seem to have any way of mapping which lanes are the same across intersections. For example, if we go from lanes:direction=2 to lanes:direction=3, there is no information telling us whether the left-most lane from before is now in the centre or still on the left. see edit3

EDIT2:
If there was a tag something like destination_way:lanes:backward=1234;1235|1235 that would tell us where each lane is directed it would probably be useful to differentiate intersections like this. see edit3

EDIT3:
Found the connectivity relation. I've created the appropriate connectivity relations to this intersection and will test once the map update comes through to OsmAnd.

EDIT4:
Nope. It still says right turn.

[vshcherb]

I didn't know that https://wiki.openstreetmap.org/wiki/Relation:connectivity already approved. No we don't support it yet.

Do you support an idea that turn:lanes is real indication on the ground or turn:lanes indicate real movements by lanes. That basically a question cause ground marking sometimes are missing however have an assumption of allowed movement directions

[mbartlett21]

Do you support an idea that turn:lanes is real indication on the ground or turn:lanes indicate real movements by lanes.

I would generally map it as whatever the indications are on the ground.

[vshcherb]

In this situation I would say we need to support https://wiki.openstreetmap.org/wiki/Relation:connectivity as it's only good choice in many confusing sutations