Oh wow - I never did it so scientific. I just changed angles a bit or changed roadspeed and looked at the effect. One problem is that even if using faster time - time is not everything - at least for longer routes. But so it seems - angles below 22.5° should really be avoided. However I'm not sure if it is really good to change all angles so that they are 45° at least - even in nature I would guess in general a route will not take too many of them. Maybe dropping roadspeed by -1 for angles 22.5-45° if unavoidable would be better than changing the angle? On 7 September 2015 at 08:20, Gerd Petermann < gpetermann_muenchen@hotmail.com> wrote:
Hi Felix,
I think I found some more rules, but I still need more tests before I can code a new fix. I found at least two errors in the v2 + v3 patch, not sure if they cause harm or if they just prevent good results. One was in the highest speed calculation, the other in the calculation of the needed heading change (some angles were not enlarged, others were enlarged too much)
Here is what I learned so far: 1) The time penalty for a turn depends on the angle and the sum of the road speed values of the arcs which are used. (I thought that I have to look at the maximum (or minimum)) 2) the sum of speeds gives a factor which is mulitplied with a value that depends on the angle. The threshold values for the sum of speeds: 0,1 : 0 2,3 : 1 4,5 : 2 ... 12..13: 6 14: 7 2) penalties (for a car) are steps of 60s for angles below 22.5° (0s, 60 s, 120 s, 180s, ...,420s) steps of 12s for angles > 22.5 and below 45 (0s ,12s , ..., 84s) steps of 6s for angles > 45 and below 67.5 (0s, 6s, ...42s) steps of 3s for angles > 67.5 and below 90 (0s, 3s, ..., 21s)
3) The penalty is always completely added to the travel time of the 2nd arc.
TODO: - The above values are for a right turn in a drive-on-right area. I still have to check left turns. - I also have to measure the effect on bicycle routing and maybe other vehicles - The effect of multiple arcs. They have an effect on the precision of the stored heading values. Not sure how this changes the results. I guess that Garmin will only use an arc with a high speed when that is long enough so that the higher speed weights out the penalty . A few tests show that it typically prefers the slower roads at the junction with the sharp angle, and "jumps" on the faster road at the next node.
If you think that I should also look at other parameters, please let me know. Gerd
------------------------------ Date: Sat, 5 Sep 2015 15:17:17 +0200 From: extremecarver@gmail.com To: mkgmap-dev@lists.mkgmap.org.uk Subject: Re: [mkgmap-dev] [Patch v3] sharp angles
Yes - I think for the start and end of a route - the algo will choose topmost route only. For inbetween sections however not - there it chooses the best - at least according to my tests a few years ago... There is a max of 6 or 7 roads lying on top of each other - if more it will produce a routing error and not go there at all.
And yeah - I also don't know what's best. V2 seemed best to me so far.
On 5 September 2015 at 15:09, GerdP <gpetermann_muenchen@hotmail.com> wrote:
Hi Felix,
okay, I came to similar results with other styles today, so maybe I'll revert the change that makes most angles larger.
Reg- --x-cycle-map: Without it only angles < 45° are changed to 45°. With it and v2, angles were changed to 68° when road speed of an arc is >= 3 and 90° when road_speed >= 5. With it and v3, all angles < 90 are changed at junctions with only 3 directions.
I think the Garmin algo is not really able to handle multiple arcs with that your style creates, at least it doesn't seem to "look" at all of them, esp. the last part of a route seems always to use the "topmost" road, means, the one that was last added in the style (as long as the selected vehicle is allowed). I think we first have to understand how the Garmin algo works before we can try to manipulate the data for better results. This is time consuming and error prone, so I'll need a few more days to work out facts.
Gerd
Felix Hartmann-2 wrote
Well - I guess this will never be without errors - with Patch v3 there is again a little loop - now again at the first place:
(only happens with "Shorter Distance" and one of the driving/motorcycle/Dirtbike and so on profiles..).
Also on some other places still detours - in general "Shorter Distance" seems to be much more problematic than "faster time". So for sharp turns - ironically - shorter distance chooses the detour to avoid the sharp turn.
I'm not really sure patch v3 is better than v2 however. Results are 50/50 better/worse. However the arrival times got a bit slower (even if following exactly the same route). I always tried with --x-cycle-map switch - never without. I'm still not so sure what this option really changes for outcome in the end. So - additional intersection roads would still be king IMHO... (but yes I know - not possible).
On 04.09.2015 19:34, Gerd Petermann wrote:
Hi all,
attached is v3, only small changes: 1) the message "maybe duplicated OSM way " was printed too often 2) with --x-cycle-map, change the headings at junctions with only three different arcs so that angles of 90° or more are created.
If I hear no complains I'll commit this patch on monday, probably without the --x-fix-sharp-angles switch.
I am still trying to work out in what case the Garmin routing algo prefers a detour, so maybe I find more improvements later.
Ciao, Gerd
_______________________________________________ mkgmap-dev mailing list
mkgmap-dev@.org
-- keep on biking and discovering new trails
Felix openmtbmap.org & www.velomap.org
_______________________________________________ mkgmap-dev mailing list
mkgmap-dev@.org
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
ajjdjdgb.png (24K) <http://gis.19327.n5.nabble.com/attachment/5854037/0/ajjdjdgb.png>
-- View this message in context: http://gis.19327.n5.nabble.com/Patch-v3-sharp-angles-tp5853996p5854043.html Sent from the Mkgmap Development mailing list archive at Nabble.com. _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
-- Felix Hartman - Openmtbmap.org & VeloMap.org Floragasse 9/11 1040 Wien Austria - Österreich
_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
-- Felix Hartman - Openmtbmap.org & VeloMap.org Floragasse 9/11 1040 Wien Austria - Österreich