Web-Router doing strange things

[Jakob Mühldorfer]

Hi,

I have observed that the web-router can calculate two different routes, depending on the starting point, even when the second starting point is on the route it calculates for the first one.

To understand, please watch the video attached.
I have attached two profiles and according videos. "happening.brf/mkv", where the bug occurs, and "nothappening.brf/mkv" where the behaviour is as expected.
It is reproducible on both the old and new webinterface.
To repdroduce, these are the starting and end point I used (for the webinterface search mask):

from 51.062577,13.753930 to 51.076748,13.757508

Could you please look into it?
It makes me especially uncomfortable because it seems to calculate a route with higher cost than theoretically possible (happening2.mkv), despite me using pass1/2coefficient that should mean it brute forces all routes?
I think there might be a problem with the heuristics, because problem does also not occur when you change the pass1/2coefficients in the "happening.brf" from 1;-1 to 1.8;0.

Thank you!

[abrensch]

On Tuesday, April 17, 2018 at 12:58:40 PM UTC+2, Jakob Mühldorfer wrote:

I think there might be a problem with the heuristics, because problem does also not occur when you change the pass1/2coefficients in the "happening.brf" from 1;-1 to 1.8;0.

Hi Jakob,

I could not really reproduce, but I could watch your video.

Things like that are not bugs in any case for a profile with elevation costs, because the algorithm is such that a "disturbance" can be carried over long distances if the terrain ha about slope of your "uphillcutoff".

I would only call it a bug if it happens without elevation costs, or if the cost difference is larger than the cost-equivilant of the elevation buffer.

[Jakob Mühldorfer]

Thank you for checking Arndt!
Now I understand why it happend.
Somehow I can also not reproduce it any more myself, seems like new map updates somehow solved it as side effect