speedway - two lane with different direction

[NaWer]

It's difficult to explain this because it's not a real (technical) bug. An example is better than words :
[1] http://map.project-osrm.org/fr
[2] http://map.project-osrm.org/ft

[1] is the expected result. But on the map, the speedway is divised in two lane, each one is one way road (OSM point of view). On via point when searching road, it should choose the right lane (same name but different direction).

The problem is I don't know how to handle if [2] is the expected result ...

Google map is doing the same :
http://g.co/maps/tmpwa

add the possibility to presice the cape ? and via point should care off the cape between previous and next point to choose the right road ?
An option "do not care the direction" may be safe (gps tracking (precision), bicycle, hiking, kamikaze,...) ?

May be related to #153 and #188

[emiltin]

why not just move the via point to the wanted side of the road?

[NaWer]

Because I can't. It's not a bug on the demo map. Let me explain with an example :
the GPS say "your here, want to go there, direction/instruction ..."
If the GPS show my car on the wrong lane (accuracy) and without taking the cape in consideration, it will tell me a sort of "u-turn". next gps fix is better and say "continue, you'll be there soon", and the next say again u-turn...

[emiltin]

ok, that makes sense.
but how would osrm know that you haven't in fact crossed over to the opposite lane? perhaps it would be possible to pass a direction/velocity hint to OSRM, which it could use to select the correct lane?

[NaWer]

osrm can't know that, I agree.
I don't know how turn by turn gps app handle this case... (maybe they don't: who never heard "u-turn as soon as possible" on the highway ? :)

the direction hint seems to be an answer (for car gps, not for walking in the wrong way)

[NaWer]

direction/velocity/cape is not enough :
http://map.project-osrm.org/fE
(just a few pixel down, it's ok : http://map.project-osrm.org/fF)

In this case, speed can be useful too... or maybe the next point (via point or destination)

[EDITION] http://map.project-osrm.org/fH

[DrVanScott]

While your posted links both show the same route to me, i think i get your issue: it is nearly random wheather the main road or the link is used.

The 'direction feature' is still interesting, but i think osrm is also missing of this:

for a given start/end/via point, osrm should not just pick the nearest way. It should also consider ways in a certain (configurable) range. Then those ways (combinations!) should be picked which led to the shortest/fastest route. Sure: There must be a reasonable limit for the range to reduce the amount of combinations. Maybe only using the nearest n'th ways within a range.

[emiltin]

it's not random, you're just not zoomed in enough :-)

[DrVanScott]

emiltin: Just for the case your last comment was not ment as a joke:

• There are use cases for osrm that work in an automated manner - without human inteligence.
• The main interface to osrm is http based. There's no such thing like zooming.
• Even with gui: The gui should also work for "non-technical-people", which do not know that most problems in computer sience can be solved by just zooming in ;-)

[emiltin]

GPS navigation would probably be a case of semi-automated use.

even though i understand these needs, i think the solutions could easily become too clever. there're nothing more frustrating than when you try to do something simple on a computer, but some automatic feature keeps interfering in some completely weird manner :-)

[NaWer]

I've updated links :
http://map.project-osrm.org/fE
http://map.project-osrm.org/fH
( /fH green point is on the slow down lane)
the 2 points were too near (double precision of the short url service I suppose)

It's the same problem (1st comment), but direction are the same in this case.

[emiltin]

does the ability to intelligently move the start point around belong in osrm?

maybe the locate service of osrm could return all ways within x meters of the requested start point. that way the gui could inform the user that there are other ways very close to the start point, and the user can make the intelligent choice.

[NaWer]

In my humble opinion, osrm has to move the point to the right road, according to the next point and the shortest path, as it does when user put the marker near a road (with a nominatim request in the GUI).
This is what the user expects in most cases.
If the user really want to take the exit, he moves the marker further, on the exit lane.

Be it by the GPS, or user's click, or any other mean (float accuracy...), the produced result will always be weird in this case, due to the lack of accuracy. I don't know yet how osrm engine works, if this solution is good and efficient, or if there is a good and efficient solution, but "find near roads, score them according shortest result and take the better" can resolve this 2 cases. There's maybe counter-example or exception...

I would like to use osrm to display gps track. I've got start, via points and end point. I would like to display a polyline insteed of marker on the map. I'm sure osrm can do the job :-)
I imagine too an open source turn by turn GPS apps...

this problem deserve reflexion

[DrVanScott]

maybe the locate service of osrm could return all ways within x meters of the requested start point. that way the gui could inform the user that there are other ways very close to the start point, and the user can make the intelligent choice.

Sounds good. In addition an api "n'th nearest" would enable developers to move the 'automated intelligence' to their code

[DennisOSRM]

Thanks for filing that issue. I will keep it as a feature request and I think we can have some solution soon on that.

[DennisOSRM]

closing in favor of #665