osrm-backend icon indicating copy to clipboard operation
osrm-backend copied to clipboard

Broken round-about numbering for roundabouts with service roads

Open TheMarex opened this issue 7 years ago • 4 comments

Breaking this out from https://github.com/Project-OSRM/osrm-backend/issues/2113

Both exits here are called the first exit.

service_1 service_2

TheMarex avatar Dec 04 '17 10:12 TheMarex

This issue has been encountered again in https://community.openstreetmap.org/t/directions-say-2nd-exit-reality-is-3rd-exit/103341. It should not be too hard to fix, right?

KronosTheLate avatar Sep 01 '23 11:09 KronosTheLate

The fix does seem simple, but the behaviour is explicitly defined in the code.

https://github.com/Project-OSRM/osrm-backend/blob/14dcf918121fa7a12addda60395cc5c3dee6ad33/src/guidance/roundabout_handler.cpp#L420-L431

The original PR implementation in #4075, seems to think this is the correct behaviour: https://github.com/Project-OSRM/osrm-backend/issues/4075#issuecomment-303412765

Is this expectation still correct? The other routers seem to think not.

I can see there is an obvious problem when you do actually take a service-road exit, such as when your journey ends on the service road.

mjjbell avatar Sep 01 '23 16:09 mjjbell

Fron the example I linked, the exits were of equal size (based on actually driving there). So the type of road the exit leads to may not influence the size of the roundabout flare, and thereby not influence how a driver would see and count the exits from inside the roundabout.

Furthermore, the exit roads could have imperfect classifications. So I would vote heavily to count the exits as long as a car can drive on them, irrespective of classification.

KronosTheLate avatar Sep 01 '23 21:09 KronosTheLate

Here is the actual roundabout object for investigations: https://www.openstreetmap.org/way/40837718

KronosTheLate avatar Sep 04 '23 06:09 KronosTheLate