MapFactor Forum

Oldie

http://osm.org/go/0MIkzCNag--

Other software uses the downlink connection on the right side to the subsidiary road.

crocodilefarm

usually a missing or wrong turning restriction in OSM. Did you allready fix it in OSM?

MP1902

Can you post the coordinates please, so I can check the OSM maps?

Oldie

I have set a turning restriction. But as said, other software did not use this solution. It seems that they have written their program such way, that routing follows the way, you would immediately exspect.

Yout find the location using the url in my first post.

riccie

I have noticed the same problem, long ago.

The problem is partially in the OSM data, partially in MapFactor Navigator.
In your case, there is a turn restriction "only_straight_on" from the trunk to the primary road, but the "no_left_turn" to the opposite trunk, which would prevent your 'funny' turn, is missing.

http://wiki.openstreetmap.org/wiki/Relation:restriction

I am mapping in OSM only "no_left" and "no_right" restrictions, but not everyone is doing the same.

MapFactor, it seams that "no_..." restrictions are working, but OSM restrictions "only_..." are not taken into account.
Please verify.

mdx

well
if 'restriction' in attr:
    if attr['restriction'].startswith( "only_"):
      type = 1 # restricted maneuver
    elif attr['restriction'].startswith( "no_"):
      type = 0 # prohibited maneuver
    else:
      type = None
is there for a long time ... I would need some example where it is failing. Also note, that only 2 segment restrictions are currently used.
thanks
   Martin

Its_so_funny

Is it possible that Mapfactor will support three segments for U-turn restrictions in the future? I really wouldn't know how to tag u-turns on crossings like this in two segments: http://www.openstreetmap.org/?lat=52.092353&lon=4.331242&zoom=18&layers=M (i'm referring to the primary/secondary crossing there with three no_u_turns)

mdx

Yes, there should be support for three and more restriction segments later on. Please map it as multi-segment ... here is the problem on our side.
thanks
Martin

Track

Same thing here: http://www.openstreetmap.org/?lat=56.76101&lon=60.77019&zoom=16&layers=M

Route via Navigator free: http://img845.imageshack.us/img845/5021/91957282.png

and same route via Marble: http://img850.imageshack.us/img850/9994/87038273.png

tvo

There are also a lot of examples like that in Denmark and in other countries too. I know they can be solved by adding turn restrictions in OSM on each road with this problem, but I can see, that some other navigation programs using the same OSM maps, have solved this problem, and guides you on the right route. One way this problem could be solved, I think, is to make a restriction in the Mapfactor program, which not allow a sharp turning left or right between 150 to 177 degrees or so. The angel 177 to 180 degrees is for still having the possibility to make a U-turn. This turn restriction shall not be active for bicycle and pedestrian. I can´t see this solution should give problems in other situations, but maybe I am wrong.

Track

Okay, but this junction already has turn restrictions: http://img521.imageshack.us/img521/4266/21683080.png

mdx

Do you really have Car profile (and not bike or pedestrian)? I tested it on the latest data (russian_federation_osm_130304.mca) and it routes as expected. The http://www.openstreetmap.org/browse/relation/397839 is relatively old, so it should not be issue of the old data ...
thanks
Martin

Track

Oh, the problem is deeper than I thought. This screencast from my phone describes it better: http://www.youtube.com/watch?v=7aUogdECGkA

tomas

I think that mdx is right, I also cannot reproduce this, I think you changed settings of your vehicle, you are not using default car

Track

Yeah, I've changed speed limits for russian realities, nothing more. I also can reproduce this on clean installation with 100% default settings, tested both Android and Windows (Linux+wine, on 2 different PCs) versions.

P.S. tested clean installation on Win7 under VirtualBox, same thing. Short route from one road to another via this junction calculated correctly, long route (Moscow - Baikal) caused this...

Local russian bug? =)

And sorry for my English :">

tomas

it works correctly with default settings, so I think you made a mistake with your settings, routing engine is clearly not at fault
route from Moscow to Bajkal with default car settings goes through Chelyabinsk, not Ekaterinburgh.

Track

I really don't know what's wrong with my settings because I'm testing it with clean installation. Launching Navigator by clicking "PC Navigator12 Free (default settings)" will reset all custom settings, including vehicle profiles and routing modes, alright? Please, look at this video (new clean installation, default settings): http://www.youtube.com/watch?v=Su5x3Qn3Y1s

mdx

@Track - thanks for report and video :). I can confirm, that it looks like for _long_ routes the restriction is ignored (this route is 5310km) - I have seen it now in PC Navigator. I will check the processing of transit network.
thank you
Martin

mmm

I've found another example of "funny routing".

Coords: 50.14777, 19.42874

Well, in my opinion it should not navigate like this even if no restrictions would be set...

tomas

why not if there are no restriction and the route is faster or shorter?

mdx

@Track - I can confirm problem in the transit network. Proper fix won't be simple, so it now depends on priority list ... thanks again for your report
Martin

mmm

tomas, for sure it won't be faster taking into account that you would have make such maneuver - even if it would be legal you would have to give way to all vehicles on orange road.

tomas

yes, but if routing attributes do not specify this, then the routing engine cannot guess it - what I am saying is that routing is as good as the input data

Oldie

But there is software, that knows how to avoid such obviously wrong navigation, even if the input data does not force correct routing. When I started this thread, Osmand routed the right way with the same data, which were "wrong" for Mapfactor. The same goes for Marble as "track" reports in his post on March 5th.

There are many reasons to prefere MF rather then Osmand. I just would like MF to be better in this point ;-)

tomas

hi oldie, I am not rying to defend MF, I am just raising teorethical points
once you start applying 'intelligence' it usually goes wrong somewhere else

mmm

Well, it's not matter of intelligence. In my opinion algorithm should consider such maneuver (almost 180 degree left turn) so time-consuming, dangerous and unusual that it get so much penalty that it won't be used in such case. Algorithms in another software mentioned above deal correctly with that. That is among other things such as preferring major roads over minor roads, right turns over left turns, U-turn on the roundabout instead of 100 meters before it (despite it is possible or legal) etc., which I suppose are correctly handled by Navigator.

tomas

I am very sorry, I am a little confused, I was referring to another thread :-(
in this case it has nthing to do with intelligence, there is a restriction on this link, so I do not know why is this restriction ignored

mdx

@mmm thanks for your report and comments. I wanted to check the original data, but in that area there were several edits so even wayID of entering road changed (?). Since "Thu, 28 Feb 2013 02:36:26 +0000" there is new restriction http://www.openstreetmap.org/browse/relation/2790300 (maybe you added it?), so in March 2013 data is should be OK (if not please let us know).

Regarding sharp angles on crossings - you are right and we should integrate something similar to route time estimation, but at the moment it is not used.
thank you
Martin

Track

@mdx - thank you, Martin, for response and great product! I can also help you with RU translation, it's a bit incorrect and incomplete, which causes some difficulties for users who do not know English.

MapFactor Forum

Howdy, Stranger!

Categories

In this Discussion

Tagged