|
Can anyone explain this rather exotic routing decision? https://www.openstreetmap.org/#map=14/50.8579/-1.1657 I have checked for ways that are not correctly joined, but I cannot find any. The route that I am trying to create is from 50.8507, -1.1538, to 50.8515, -1.1882 by car. Eventually I will rediscover how to attach an image to this question, and I will be able to show everyone what route the computer chose.
showing 5 of 9
show 4 more comments
|
|
when I move the start marker, I notice that for OSRM the problem is located at the traffic signal on Cams Hill (near the bus stop 'Delme Arms Ph') . Exclude that from the route, and it goes over the roundabout. Include it, and it makes a u-turn at the traffic signal. After moving the end marker, it seems the OSRM does not "like' the 2 traffic signals: https://www.openstreetmap.org/directions?engine=osrm_car&route=50.85216%2C-1.16498%3B50.85311%2C-1.16852#map=18/50.85266/-1.16673 include both and it goes wrong, include only 1 and the route is OK. |
|
According to https://github.com/Project-OSRM/osrm-backend/issues/4754 this is an upstream bug caused by an invalidly applied U-Turn penalty. It will be fixed in the next OSRM release. |
... which I think is:
https://www.openstreetmap.org/directions?engine=osrm_car&route=50.8507%2C-1.1535%3B50.8521%2C-1.1884#map=14/50.8579/-1.1506
Thank you, Someone Else. That certainly illustrates the behaviour, although I had intended the journey to stop somewhere around the railway station.
Madryn it seems mapzen and graph hopper seem to work as you would expect. Only osrm takes the longer distance route, perhaps osrm's timings for the junctions cause the odd route. I assume from SomeoneElse's comment you now know that you can just cut and paste the url of the map with the route on it. https://www.openstreetmap.org/directions?engine=graphhopper_car&route=50.8510%2C-1.1554%3B50.8520%2C-1.1910#map=14/50.8510/-1.1673
Just an idea: To test an odd routing result i usually just push red and green flags along in short sections until it reveals the problem, it also gives the distance and the time, that may give a clue why osrm does what it does. no need to open an editor until the "error" has been located, maybe?
Thanks for the suggestion, andy mackey, but I tried that before I asked the question. The routing system behaved as though the Eastern Way sliproad leading south west from the Delme Roundabout was not connected to the main dual carriageway. However, when I looked, there seemed to be nothing wrong with that junction. It seems that the router really wants to avoid the roundabout, but will use it if left with no alternative.
At the risk of stating the obvious, it's perhaps worth mentioning that some routers use older data (sometimes quite a lot older). A problem with one router doesn't mean a problem with OSM data.
Did you check the surface=* tag on the roads? Sometimes it might effect routing.
@SomeoneElse, I checked the history of the Cams Hill, it's from 2013, one of the traffic signals causing the problem from 2014. I think it's unlikely that data is the reason. I think it's a glitch in their algorithm.
You can report potential OSRM bugs at https://github.com/Project-OSRM/osrm-backend/issues. I've created a report for you: https://github.com/Project-OSRM/osrm-backend/issues/4754