Closed joovel closed 7 years ago
@joovel it is an effect of OSRM snapping behavior (the route end point is a projection onto the nearest segment) and a data issue (https://www.openstreetmap.org/way/46788688#map=10/33.4023/-118.0062 is not connected to any footway). "Foot (MapZen)" also provides an unrealistic route for slightly shifted edpoint https://www.openstreetmap.org/directions?engine=mapzen_foot&route=33.46620%2C-117.70250%3B33.46009%2C-117.69174#map=10/33.5562/-118.0062
@oxidase Thanks for the quick and detailed analysis! Will someone fix the OSM data, or is that on me? Also, is there a way to tell OSRM to do a fuzzier search for a footway?
data issues usually fixed by local OSM communities. OSRM demo server http://map.project-osrm.org/?z=14¢er=33.468144%2C-117.670956&hl=en&alt=0 has "Small Components" layer that highlights unconnected or partially connected ways.
Without OSM changes ferries can be disabled as a quick fix by removing handle_ferries
.
Also https://github.com/Project-OSRM/osrm-backend/issues/4006 is in progress and will allow avoiding ferries in query time.
@oxidase : Thanks. I've sketched in a few sidewalks that should be more helpful. Thanks for adding the pointer to handle_ferries
.
Here's a brief script to capture this odd result I'm getting from a routing:
Then I issue this curl:
The result is a crazy route taking the user over two ferries (I think) and then back down the PCH. The polyline shows the rough route.
The two points are only 1.5km apart.
When I run this on openstreetmap.org (http://www.openstreetmap.org/directions?engine=mapzen_foot&route=33.4662%2C-117.7025%3B33.4602%2C-117.6918#map=17/33.46324/-117.69720) using "Foot (MapZen)", I get realistic directions.
I expect to get similar realistic directions from OSRM.
We first experienced this issue with 5.6.0 and 5.6.5.