turn restrictions

[FilipeCB]

hi,

http://map.project-osrm.org/?hl=en&loc=41.533583,-8.619032&loc=41.533470,-8.618990&z=18&center=41.533535,-8.619102&df=0

this route is technically prohibited, as there is a sign that prohibits transit. realized so that the routing engine only prohibit turning through a via node, so that the prohibition would be from a way to another way through a via way. I think that the engine does not see these prohibitions.

thanks

[emiltin]

not sure i understand you, but looking at osm, there are no restrictions added to the map?

[drummer83]

I also think that there was no turn restriction so far. I created the one you described. Let's check again after data update (tomorrow morning probably).

[FilipeCB]

I asked if OSRM see turn restrictions with multiple via points. For example, one restriction like that:

or like that

[drummer83]

According to the link in your first post, it seems in deed that turn restrictions using a via-way instead of a via-node are ignored.
Maybe DennisOSRM can confirm this?

[mjoris]

Internally in the orsm.restrictions file, a restriction is modeled by

• a via node
• a from node
• a to node
• a bit indicating whether it's a forbidden turn or the only turn allowed

So I guess that "a forbidden U-turn" cannot be represented in this model (at least in case of a dual carriageway), because it needs > 3 nodes.

I suppose these nodes are required to be strictly subsequent (because otherwise you can have multiple trajectories that map to the same sequence). So an OSM via way restriction (e.g. http://wiki.openstreetmap.org/w/images/9/96/Via_way_restriction.png ) is not possible either.

[DennisOSRM]

Via ways for turn restrictions are unsupported.

[DennisOSRM]

Support may/will be added with some later version.

[woodbri]

DennisOSRM, This is the same issue I raised on osrm-talk a couple of weeks ago. maybe this ticket should be reopened as an enhancement request to is stays on the todo list. Update: this seems to be a duplicate of #483 and #602.

[DennisOSRM]

Leaving this closed in favor of #483