If you follow the topographic diagram with your mouse you’ll notice that instead of going to WP1 directly, Kurviger goes around, on the road to the left, then comes to WP1, and then instead of turning around to WP1 it goes in the same circle again
This is of course nothing major. It just seems like the “extra curvy” setting really really REALLY likes the road to the left. If go to “curvy” the problem goes away. But I think it’s funny that this also goes away when you UN-select the “avoid using same road twice” (when you do that you see that “extra curvy” really wants to drive on the western route. When you select “curvy” you see the route that I personally would take, from start directly to WP1 then directly to WP2 -> so it is technically possible, there’s no one-way-street or anything in the way)
Uhm… ok I think we have a forum bug on top of this here LOL? The link works when I edit my post and use it directly. But once the forum converts it to a “box” and you click on the box it gets broken. Try the original form:
That’s an interesting example, thanks for posting this. Internally, the route from the start to WP1 is calculated first. Now Kurviger tries to avoid the already used roads when calculating the route from WP1 to WP2, this won’t change the route form the start to WP1. There are not many options to take different roads in your example, that’s why it looks like that. If you use only the curvy routing, the results look a lot more sensible. I am not sure if I would call it a glitch though.
Thanks for the note on the one box issue, that seems to be an issue with the forum software.
Haha ok that makes a lot of sense, I guess that gets us into CS territory of tree search and backtracking. If the algorithm hadn’t chosen the route from 0 to 1 yet, then 1 to 2 would be a lot easier to calculate.
And I guess that’s the correct decision, because otherwise the cost would become O(n!) if you have to check every previous path after computing the next one, so yeah, not worth it for this edge case