gosmore (r20088) fastest-route is making poor decisions #2738
Comments
Author: David Dean After, a bit of mucking around I've determined that the bug in http://i.imgur.com/W8BEt.png does not exist in r18055 but does exist in r18177. A bit changed in r18177 tho, so there is a bit of work to track down what the actual error might be. Some of the other examples appear to be fixed in more recent revisions so we may be dealing with multiple bugs here. |
Author: nroets[at]gmail.com In hpKAN there are two oneway links that I find very odd, considering it is a left hand traffic country. http://www.openstreetmap.org/?lat=-27.42621&lon=153.01838&zoom=16 Some of these bugs is most likely caused by the feature that avoids crossing a busy road. Considering the number of possible layouts and traffic authorities, this feature may be simply too ambitious in which case it should be disabled. |
Author: daviddean (In [20117]) Fixed bug in loading maxspeeds in an attempt to fix #2738. Not quite there yet tho. |
Author: David Dean I found a bug in loading the maxspeeds when they are tagged on a way that set all maxspeeds to 30km/h, and submitted the fixed version as r20117. This appears to fix the problem in http://imgur.com/tKtXq.png, but the other two examples are still not working. |
Author: David Dean http://i.imgur.com/hpKAN.png appears to be related to rhd being set to TRUE when it really should be FALSE in Australia. If I hard code it to FALSE, http://i.imgur.com/hpKAN.png works correctly. I'm still not sure what is causing http://i.imgur.com/W8BEt.png. |
Author: David Dean Isn't rhdBbox actually lhdBbox given that it contains all of the lhd countries? |
Author: David Dean I've fixed the problems with rhd (I hope) in r20118. http://i.imgur.com/hpKAN.png works now. |
Author: David Dean Commenting out the road-crossing section that you suggested makes http://i.imgur.com/W8BEt.png work as expected. Both of the roads have the same speed, but the 90 metre penalty makes it apparently cheaper to go up the side street an u-turn back. Perhaps a very-tight/u-turn penalty would make this work better? |
Author: daviddean (In [20119]) Commented out road-crossing penalty code for the moment to fix #2738 |
Reporter: David Dean
[Submitted to the original trac issue database at 5.48pm, Monday, 22nd February 2010]
I've recently started using the latest versions of gosmore more than my old r17226 version, but I've noticed that there appears to be some problems with the comparative weights of different road types causing gosmore to make some weird decisions.
These do not appear to be related to turn restrictions, as they are happening at junctions without turn restrictions for the most part.
They do not appear to be related to bad data, as a shortest-route performs properly.
A few examples of where this is happening:
http://imgur.com/hpKAN.png
http://imgur.com/W8BEt.png
http://imgur.com/tKtXq.png
The text was updated successfully, but these errors were encountered: