Reproducible inter-tile routing error
I have found a reproducible routing error, which occurs when I split and compile the Italy extract from Geofabrik.de. - The error occurs when routing on the A1 from Modena to Parma. (I have only tested this in MapSource so far.) If I route in the opposite direction (South-East), the problem does not occur. - As soon the route crosses the tile boundary, the route starts jumping back and forth over a very large area. - I have attached a screen shot of the routing problem, and a gpx file with the route reduced to two points on either side of the boundary. (If you use the .gpx file, you have to explicitly recalculate the route to reproduce the problem.) Here are the steps I use to recreate the map with the problem: wget -N http://download.geofabrik.de/osm/europe/italy.osm.bz2 java -Xmx1024m -jar ..\..\..\svn\mkgmap-splitter\dist\splitter.jar --max-nodes=1600000 italy.osm.bz2 java -Xmx1024m -jar ..\..\..\svn\mkgmap-trunk\dist\mkgmap.jar --product-id=1 --family-id=503 --series-name=it-2009-03-31 --country-name=Italy --country-abbr=ITA --region-name=Italy --region-abbr=ITA --latin1 --route --tdbfile --tdb-v4 --add-pois-to-areas --road-name-pois --frig-roundabouts --location-autofill=2 --net 63*.osm.gz (The download is about 57MB) - Could this be related to bad data in the highway or relation? Cheers.
Hi Clinton,
I have found a reproducible routing error, which occurs when I split and compile the Italy extract from Geofabrik.de.
- The error occurs when routing on the A1 from Modena to Parma. (I have only tested this in MapSource so far.) If I route in the opposite direction (South-East), the problem does not occur.
- As soon the route crosses the tile boundary, the route starts jumping back and forth over a very large area.
Can you route successfully if the roads in question are contained in a single tile? I'm just wondering if it is a general routing issue or specifically an inter-tile routing issue. I only wonder because I have an example map that has broken routing so I know that there still exists problems with intra-tile routing. Cheers, Mark
On Thu, Apr 9, 2009 at 6:06 PM, Mark Burton <markb@ordern.com> wrote:
Can you route successfully if the roads in question are contained in a single tile? I'm just wondering if it is a general routing issue or specifically an inter-tile routing issue.
OK, I'll try to get an extract on one tile, and see if the problem still occurs. Cheers.
You're right: The routing problems occur even with everything in one tile. I don't know if this is caused by bad data or just general routing weirdness. What would be the best way to analyze this? Cheers. On Thursday, April 9, 2009, Mark Burton <markb@ordern.com> wrote:
Hi Clinton,
I have found a reproducible routing error, which occurs when I split and compile the Italy extract from Geofabrik.de.
- The error occurs when routing on the A1 from Modena to Parma. (I have only tested this in MapSource so far.) If I route in the opposite direction (South-East), the problem does not occur.
- As soon the route crosses the tile boundary, the route starts jumping back and forth over a very large area.
Can you route successfully if the roads in question are contained in a single tile? I'm just wondering if it is a general routing issue or specifically an inter-tile routing issue.
I only wonder because I have an example map that has broken routing so I know that there still exists problems with intra-tile routing.
Cheers,
Mark _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
Hi Clinton,
You're right: The routing problems occur even with everything in one tile.
I don't know if this is caused by bad data or just general routing weirdness. What would be the best way to analyze this?
It may be the same issue that causes problems with the attached map fragment that someone brought to my attention. The road with ref 90 is very badly behaved. I tried reducing the max number of points to a really low number and it cured some of the oddness but not all. So I am thinking that maybe the problem is in the encoding of the polyline (don't really know why I'm thinking about that but that's where I would start looking if I had the time to work on this, which I don't at the moment). Cheers, Mark PS - if you want to process x.osm you will need to use the --ignore-osm-bounds option because it contains multiple bounds elements.
On Apr 10, 2009, at 12:02 AM, Mark Burton wrote:
It may be the same issue that causes problems with the attached map fragment that someone brought to my attention. The road with ref 90 is very badly behaved. I tried reducing the max number of points to a really low number and it cured some of the oddness but not all.
I'll look at the fragment and see if the problem looks similar to the one I described. In the meantime this error is giving me a lot of grief: - The JOSM routing plugin works fine on the route. - Validator does not return anything significant. - The highway where the error occurs is all one segment. Splitting it does not help. The only thing that corrects the error is to recreate the highway node by node. There must be something in the data which JOSM does not display, which causes the error. I'll try to reduce the osm extract to a reasonablehe size in order to see if anything is noticeable. Thanks for your (and everyone else's) help.
Hi Clinton,
I'll look at the fragment and see if the problem looks similar to the one I described.
Thanks.
In the meantime this error is giving me a lot of grief:
- The JOSM routing plugin works fine on the route. - Validator does not return anything significant. - The highway where the error occurs is all one segment. Splitting it does not help.
My hunch is that the data is OK but we're encoding it wrong for some reason.
The only thing that corrects the error is to recreate the highway node by node. There must be something in the data which JOSM does not display, which causes the error. I'll try to reduce the osm extract to a reasonablehe size in order to see if anything is noticeable.
Thanks for your (and everyone else's) help.
Certainly, a small example that exhibits the problem would be very useful. The reason why I think the encoding is wrong is because when I experimented with using smaller limits on the number of points in a line, some of these problems went away. Cutting the line into smaller segments doesn't change the position of the points but, somehow, it made a difference. Cheers, Mark
The only thing that corrects the error is to recreate the highway node by node. There must be something in the data which JOSM does not display, which causes the error. Just a flashing thought: May it have to do with the ordering of the node numbers? This is a invisible thing you change by recreating the street.
On Apr 11, 2009, at 12:30 PM, Johann Gail wrote:
The only thing that corrects the error is to recreate the highway node by node. There must be something in the data which JOSM does not display, which causes the error.
Just a flashing thought: May it have to do with the ordering of the node numbers? This is a invisible thing you change by recreating the street.
I also thought about this, and tried renumbering the node, but this did not seem to have any effect. I'll try it again just to make sure. Cheers.
On Apr 10, 2009, at 12:02 AM, Mark Burton wrote:
It may be the same issue that causes problems with the attached map fragment that someone brought to my attention.
Thanks, this does seem to be the same problem, or at least the symptoms are very similar to the routing behaviour on the A1 in Italy. I'll keep looking at these files to see if I can find anything, or if I can better reduce the problem. Cheers.
try routing with gosmore, traveling salesman, openrouteservice.org, cloudmade.com, ..... there is also a new routing plugin for josm. wherever these create a scenic route you need to check the details. very common errors are wrong direction motorway segments or exits. broken connections are also very difficult to find. Use josm validator. It can find overlapping segments which are not connected. Mapsource isn't very good to find a detour. A Garmin device is usually better in finding a route. In Mapsource itself insert more and more through points and the segment which is connected straight contains the error. The interesting behavior for my Etrex Vista is it can route a long route via multiple tiles even if there are some errors. But when you are driving near the error it fails, does recalculations all time or gives up. there must be some built in guessing code to bridge errors in the data On Thu, Apr 9, 2009 at 2:38 PM, Clinton Gladstone < clinton.gladstone@googlemail.com> wrote:
You're right: The routing problems occur even with everything in one tile.
I don't know if this is caused by bad data or just general routing weirdness. What would be the best way to analyze this?
Cheers.
On Thursday, April 9, 2009, Mark Burton <markb@ordern.com> wrote:
Hi Clinton,
I have found a reproducible routing error, which occurs when I split and compile the Italy extract from Geofabrik.de.
- The error occurs when routing on the A1 from Modena to Parma. (I have only tested this in MapSource so far.) If I route in the opposite direction (South-East), the problem does not occur.
- As soon the route crosses the tile boundary, the route starts jumping back and forth over a very large area.
Can you route successfully if the roads in question are contained in a single tile? I'm just wondering if it is a general routing issue or specifically an inter-tile routing issue.
I only wonder because I have an example map that has broken routing so I know that there still exists problems with intra-tile routing.
Cheers,
Mark _______________________________________________ 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
That's a good idea: I'll compare with the josm plugin route. I have inspected the route for any obvious visible errors, but could not identify any. In many cases in the past I found reversed highway segments, etc, but this appears to be different. Thanks. On Friday, April 10, 2009, Apollinaris Schoell <aschoell@gmail.com> wrote:
try routing with gosmore, traveling salesman, openrouteservice.org, cloudmade.com, .....there is also a new routing plugin for josm. wherever these create a scenic route you need to check the details. very common errors are wrong direction motorway segments or exits. broken connections are also very difficult to find. Use josm validator. It can find overlapping segments which are not connected. Mapsource isn't very good to find a detour. A Garmin device is usually better in finding a route. In Mapsource itself insert more and more through points and the segment which is connected straight contains the error. The interesting behavior for my Etrex Vista is it can route a long route via multiple tiles even if there are some errors. But when you are driving near the error it fails, does recalculations all time or gives up. there must be some built in guessing code to bridge errors in the data
On Thu, Apr 9, 2009 at 2:38 PM, Clinton Gladstone <clinton.gladstone@googlemail.com> wrote: You're right: The routing problems occur even with everything in one tile.
I don't know if this is caused by bad data or just general routing weirdness. What would be the best way to analyze this?
Cheers.
On Thursday, April 9, 2009, Mark Burton <markb@ordern.com> wrote:
Hi Clinton,
I have found a reproducible routing error, which occurs when I split and compile the Italy extract from Geofabrik.de.
- The error occurs when routing on the A1 from Modena to Parma. (I have only tested this in MapSource so far.) If I route in the opposite direction (South-East), the problem does not occur.
- As soon the route crosses the tile boundary, the route starts jumping back and forth over a very large area.
Can you route successfully if the roads in question are contained in a single tile? I'm just wondering if it is a general routing issue or specifically an inter-tile routing issue.
I only wonder because I have an example map that has broken routing so I know that there still exists problems with intra-tile routing.
Cheers,
Mark _______________________________________________ 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
Some time ago I had problems at inter tile routing too. It tunrde out, that the option 'avoid carpool lines' and 'avoid highways' caused them. Disabling them helped for me. Do you have any avoid options set?
A bit offtopic: From what I'm hearing from other users is that the 'avoid carpool lanes' option causes the Garmin to avoid highway=service roads. Perhaps that classification needs to be looked at as a carpool lane != a service road. Johann Gail wrote:
Some time ago I had problems at inter tile routing too. It tunrde out, that the option 'avoid carpool lines' and 'avoid highways' caused them. Disabling them helped for me.
Do you have any avoid options set?
_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
participants (5)
-
Apollinaris Schoell -
Clinton Gladstone -
Johann Gail -
Lambertus -
Mark Burton