YOURS/weird routes

From OpenStreetMap Wiki
Jump to: navigation, search

On this page weird and sub-optimal routes are gathered. Having a list of these routes will hopefully lead to better route calculation weightings, OSM data fixes or the discovery of bugs in the routing algorithm.


  • If you add a route please do so with the shortest possible route that identifies the problem.
  • There are two categories:
  1. Generic routing problems using the mapnik or cycle layer.
  2. Cycleroute specific problems for problems with the cycle networks layer only.
  • You can leave the 'cause' and 'solution' fields empty if appropriate.

There's no need to add more weird routes as there are enough reports to provide work on the route planner for quite some time to come. Thanks for all the reports!

Generic routing
route description cause solution
Routing does not seem to work for Iquitos Return value: 0
fixed Route against oneway on roundabout Gosmore does not recognizes 'true' values Added support for 'true' values in Gosmore
pending No routing over car park Areas like amenity=parking are not routable Add a routing rule to Gosmore to allow routing using areas
But car parks shouldn't be routable, I have a car park close to where I live where there are flowers in the middle to explicitly avoid people driving through it. Walking or cycling from one side to the other is possible though. We should add service roads that route through the car park if we want the routing to be able to do so. A heavy penalty in the routing weights should be applied though... Pov 14:20, 21 January 2009 (UTC)
access=destination, then, if through traffic is not possible, or split into two areas with cycleway inbetween. Alv 14:46, 21 January 2009 (UTC)
fixed Routing against oneway on cycleway Probably no Gosmore rule for oneway/cycleways combo Add a routing rule to Gosmore for oneway/cycleways combo
fixed Routing against oneway on cycleway Probably no Gosmore rule for oneway/cycleways combo

Seems to also route bicycles the wrong way up oneway streets. Stevek 09:57, 30 September 2008 (UTC)

Add a routing rule to Gosmore for oneway/cycleways combo

Add routing rule to obey oneway for bicycles except when cycleway=opposite. Stevek 09:57, 30 September 2008 (UTC)


refusing to route over cyclebridge that works for bikes and has foot=yes? For some reason it doesn't like that portion of cycleway, it never selects the cycleway as part of the route even if you are starting west of the bridge and heading west. See [[1]] Is it related to the fact it is a NCN relation? Mungewell
fixed Does not take the obvious route According to Map Features motorway_link roads are not oneway by default. Gosmore thinks otherwise.
But the Map Features DO state that oneway=yes is implied (see green area on the right side of that page). I did not know that before, but it is written there. RalfZ
Discussion started on talk ML to solve this issue.
Removed default oneway="1" rule from elemstyles.xml
fixed Routing engine does not want to use the bridge to go onto the highway. Instead it takes a 25km detour. Don't understand why. RalfZ Motorway_link way continuing south of the bridge not marked oneway=no. See line above. Need to add oneway=no to the way connecting to South West end of the bridge.
fixed Ignoring a UK_public_rights_of_way#Byway open to all traffic as a potential route for cyclists or motor vehicles (flip the example to motorcar for a different, non-BOAT route). Routing engine does not seem to have highway=byway configured into it as a routeable way for any traffic. --achadwick 14:33, 2 October 2008 (UTC) Added byway to Gosmore's list of highways. So this is hopefully fixed in the next database update.
fixed YOURS crashes with the error message "Status: An unexpected error occured in Gosmore: " and does not render a route. Potentially related to the example above? Note that this way has an explicit bicycle=yes. --achadwick 14:36, 2 October 2008 (UTC) Added byway to Gosmore's list of highways. So this is hopefully fixed in the next database update.
fixed Route does not want to use byway. Welshie Byway is not a feature that Gosmore recognises.

highway=byway with surface=paved, or tracktype=grade1 would be suitable for a normal car route also. Usually byways would not though. Possibly a new issue. Ben 16:43, 28 December 2008 (UTC)

Added byway to Gosmore's list of highways. So this is hopefully fixed in the next database update.
fixed Gosmore doesn't want to use pedestrian bridge --Zorko 15:47, 9 October 2008 (UTC) Gosmore currently does not allow pedestrians on trunk_link roads. Shouldn't the bridge be tagged as highway=footway instead of pedestrian? Tagging changed.
new Area's should be crossed in a straigt line, not allong the borders... (Does a permalink work for the shortest way?) GercoKees 06:19, 10 October 2008 (UTC) Gosmore doesn't support crossing areas like this. Andrewpmk 08:48, 10 October 2008 (UTC)
Areas are very difficult, e.g. what to do when it contains a hole...? --Lambertus 14:53, 3 November 2008 (UTC)
Connect each node of an area with the other node of the area. Delete every connection which crosses the border. Next delete every connection which goes outside the area. Then use remaining connections for routing. GercoKees 20:03, 10 November 2008 (UTC)
Is that necessary? Can't it just be every node on the area 'which has a route linking to it' be connected rather than just 'every node'. The rest makes sense though. Ben 16:43, 28 December 2008 (UTC)
That could cut through concave corners, so no. Alv 14:46, 21 January 2009 (UTC)
new Route is sub-optimal. Result should be this one. Without the two points to aid it, Google routes M4->M5->M42. Chriscf 21:37, 15 October 2008 (UTC) Probably a data problem, but can't identify the specifics as is.
Presumably the same default speed (110/70 mph) used for motorways and trunk roads, so far, so shorter route chosen.
fixed (in test version) Route Hamburg to Bagdad. The first part of the route from Hamburg is missing. Anyway. Google can't caculcate that at all! Gosmore quits prematurely while calculating the route. This is no data error. Fix Gosmore behavior for long routes.
new Route Edinburgh->Alnwick should start from Edinburgh but starts from just outside Lauder instead (about 40km away) Gosmore cannot find a complete route. Change Gosmore's code to allow a larger area to be searched for a possible route (at the expense of slower route calculation).
solved Route Hamburg to Kapstadt (Cap Towm). It starts somewhere in the nowhere in africa and goes to Cape Town. May have something to do with a distinct lack of data. Chriscf 11:12, 17 October 2008 (UTC) Add and/or fix more data in Africa.
new From Hamburg to 'Neustift im Stubaital' by bike is about 300 km shorter than by car.
new From Hamburg to 'Neustift im Stubaital' by foot is about 300 km shorter than by car.
fixed Cannot save the route to gpx because the request uri is too large. Smsm1 08:26, 12 October 2008 (UTC)
This is in the YOURS#Known_issues list --Lambertus 08:51, 12 October 2008 (UTC)
pending Unwanted u-turns from and onto single carriageway road access ramps Every ramp on a single carriageway road requires a (possibly non signposted) turn restriction from the "wrong" direction. Reversing the route shows an obeyed no-u-turn relation. Add a hefty penalty for turns of almost 180 degrees?

Interesting idea, but how to tag the case where a 180 degree turn is allowed ?

Gosmore will work on the B455 if you tag it restriction=no_u_turn or with restriction=only_straight_on. -- Nic 21:52, 29 November 2008 (UTC) For someone tagging such a restriction, it is hard to decide if the situation is a no_u_turn or a no_left_turn. In the description of the restriction it is written that a routing engine should only look for "no_" or "only_" in the restriction tag. Why does Gosmore handle no_left_turn differently than a no_u_turn? This does not make sense to me as the restriction is clearly defined through from/via/to and the "no_" restriction. --RalfZ 23:45, 29 November 2008 (UTC)

pending Taking the wrong (shorter way) primary_link to enter the B455 road. Then makes a U-turn on the primary road. Routing engine assumes that you can just make a U-turn on a primary road. Routing engine also not observing the relation set at this point to disallow this u-turn. RalfZ
[2] Routing contrary to a Relation:restriction. This one does have all the necessary members. In reality, this is a 7-lane tidal flow single carriageway, so it's a brave individual that attempts this manoeuvre. Chriscf 13:48, 17 October 2008 (UTC) Gosmore considers (so I've read) a no_right_turn to mean only turns of 45 to 135 degrees. For right hand traffic countries a no_left_turn should often imply a no_u_turn, also (by legal definitions), and for left hand drive a no_right_turn likewise? That'd be a problem, or two relations for most no_right/left_turn cases, which many are going to forget to enter.
I would suggest that the restriction=* value of a relation is not useful for deciding which movements are prohibited when we're already giving enough information to work this out. Chriscf 09:10, 22 October 2008 (UTC)
Try restriction=no_u_turn or restriction=only_straight_on -- Nic 22:33, 29 November 2008 (UTC)
new Excessively long way when a short dismounted stretch on foot is better for the average user. The problem way is way 23004483, which is tagged access=private/foot=permissive. This in theory allows a bike rider to get off and push for a few metres, saving a few Km of riding, but Gosmore seems to be a bit obsessed about pruning access=no branches: it should consider modes of transport :) In general, cyclists would be OK with dismounting for short distances if the win is big, less so if the dismounts are frequent and the gains not great. This combination of tags is too difficult to process by Gosmore. Probably needs fixing by Gosmore's author.

Please edit your route to show the problem with the shortest route possible (as requested on the top of this page). -- Updated. See the first few hundred meters there we are on a track which even has the tag motorcar=no.

Routing for cars over tracks Gosmore considers tracks as legal ways for cars. I suggest to assume the default tag motorcar=no for tracks. Rainer Dorsch 12:52, 29 October 2008 (UTC)
The example even has a tag motorcar=no and gosmore ignores it. Rainer Dorsch 20:08, 4 November 2008 (UTC)
Likely the access=agricultural is not understood - shouldn't that be agricultural=yes? Gosmore doesn't yet parse combinations of access tags, i.e. access=yes seems to leave motorcar=no unparsed.Alv 21:28, 4 November 2008 (UTC)
modified the access tag as you proposed. Makes a lot of sense from the map feature definition in the wiki. I am not sure who and why the access=agricultural was added. Rainer Dorsch 22:58, 5 November 2008 (UTC)
what's the need for agricultural:anything or access:agricultural? in the routing, or even the osm database? It doesn't mean much in any sense. If a track is used only by agricultural vehicles, that wouldn't be becuase there the only ones who may, it would be becuase there the only ones who need to, and maybe can (phisically), and that's no justification for stating that it's limited to just them to a route planner. I dought anywhere is legally restricted, so that only agricultural vehicles can use it. I imagine you can only use the routes for accessing your land, and you can do so in whatever vehicle you please. Therefore it's private, therefore not a highway. Ben 16:43, 28 December 2008 (UTC)
(Not that this belongs here but I don't yet know where to move it:) Both do exist in other countries: tracks open (besides pedestrians and possibly cyclists) only to all agricultural uses regardless of the vehicle or owner, and roads that allow only agricultural vehicles (tractors, crop harvesters etc.) - both may be exceptionally well physically suitable for bicycles and motorcars. Alv 08:11, 29 December 2008 (UTC)
Tracks can be navigated by cars in many countries so motorcar=no is not an option, I suggest to add access=* restrictions where applicable. I have lowered the average speed on cars so the routing will use it less often. --Lambertus 14:11, 29 October 2008 (UTC)

Accessing a track by bicycles/foot is perfectly legal, only motorvehicles are prohibited. I thought motorcar=no is the tag for these cases (?). If that is the case I think at least for Germany (probably also Austria and Switzerland) assuming motorcar=no is a good assumption for highway=track (if not specified otherwise) Rainer Dorsch 14:45, 4 November 2008 (UTC)

On what basis? It's public if it's legal, but illegal if it's not. If it has "highway=track" then it should public by default as the highway= tag is being used, and this can't be used with another highway tag. If it doesn't then it's private by default. If tracks aren't public by default (i.e no right to rome) where people map then it shouldn't have the highway=track tag additional to physical tags. Why over complicate? Lambertus's point makes sence. Ben 16:43, 28 December 2008 (UTC)
[4] Should have chosen the ferry across the fjord, which will take 25minutes, instead a detour around the fjord which will take hours is chosen. Bernt 14:40, 29 October 2008 (UTC) The average speed of the ferry link is probably too low (3km/h) Average speed set to 10km/h.
new Please use the shortest possible route next time. Route AND_nosr_r=15184978 has a barrier: motorcars=no. Gosmore leads a car through this barrier. --Hugo H 20:07, 8 October 2008 (UTC) Gosmore does not treat a node with motorcar = no as something it can't route.
new Route Danziger Str., Ellerbek to Burgwedelkamp, Hamburg. The provided route recommends to drive through a barrier in the street Moordamm (near crossing Moordamm / Königsberger Straße, Ellerbeker Moordamm). Gosmore should not provide a route through a that can not be passed due to a barrier (node with barrier=gate).
Sorry, I beg to differ here. Gates are designed to be opened - otherwise it would just be a fence. Any restriction on passing should perhaps be shown by access tags on the way beyond the gate. There are plenty of routes I can think of, particularly whilst walking or cycling etc., where one must open a gate to continue along a route. Richard B 22:53, 10 November 2008 (UTC)
A public street where cars are allowed to drive is on both sides of this gate. This gate here is not a gate against sheep, but a gate against cars passing without permission by local admistration. Only some public services have a key for that gate. But, as it can be passed by foot and by bike and can opened by public services it is a gate and not a fence. There are additional access tags to be used in conjunction with a gate to have precise permission information. Such as car=no, bike=yes and foot=yes. restrictions. Thus the router should consider this gate in conjunction with its access tag. Josy
fixed Suboptimal micro level routing for a pedestrian. Sidewalks drawn as footways should be preferred to the parallel road. Same cost used for walking along residentials and footways. Something has been changed.
new Ferry does not connect in Calais? I haven't checked the OSM data yet, but I will.
new new highway=primary_link & highway=trunk_link ignored?
new cycle routing against oneway on highway=residential, and also on highway=pedestrian
new cycle routing on footpath Pushing your bike on foot is allowed and in this case presumed faster than a detour, since trunk roads are not considered suitable for cycling, for now.
This is a difficult topic. I have an example here where a bicycle route is routed over a footway. In this context this does not make any sense, because it is a hiking path, where you would have to carry your bike(!) -- RainerDorsch 21:09, 08 November 2008 (UTC) .
Maybe the default should be to route across footpaths at a walking pace - sure, there are a few cases where you can't do this e.g. really mountainous terrain, but in the majority of cases it would work (I could certainly push my bike up some steep paths and lift it over gates etc.). You can tag the serious exceptions with bicycle=no. But if the default is to assume pushing at a walking pace on footpaths, then if there is a better/faster route it will be preferred. If not, we get off and push (or more likely, try to cycle it anyway). ChrisB 14:17, 27 November 2008 (UTC)
Of course the real reason for this to happen at all is because Gosmore does not allow cycle routing on highway=trunk - which needs to change - especially if there is no other suitable alternative that doesn't have you pushing your bike along a footpath. Richard B 00:24, 30 November 2008 (UTC)
fixed Routing refuses to take morway, might be a dataerror... Found it! GercoKees 20:11, 10 November 2008 (UTC)
new foot routing does not include highway=path; foot=designated; bicycle=designated
new Does not take the obvious route. The algorithm uses road sections of the type highway=primary_link because it understands that this is the shortest way. The logical choice is to continue on the road without "strange" dodge, concatenating lanes of acceleration or deceleration in a few meters if there is another alternative. Probably giving a major resistance to the movement for this type of routes? Or maybe discriminating these complex routes by others more simple calculating the sinuosity index of each one of them? --Tony Rotondas 15:17, 15 November 2008 (UTC)
fixed Walking route (Edinburgh->Musselburgh) stops after 2.4km - should end at destination and be around 10km. Cycle route works ok. Appears to be fixed now
new Car Route stops after a few meter, should be 36 km gosmore routing engine problem (Jump-Error)
This route and the four just below all seem to be symptoms of the same issue: Gosmore stops without a complete route when a significant deviation from the straight line from start to destination is required early in the routing of a car route. All four of these work for walking and cycles. Djlynch 17:24, 25 November 2008 (UTC)
Do you know the line of the "straight line"-check in libgosm.cpp? It try to fix this after Line 338 and 526 with no luck. --M0nty 17:49, 26 November 2008 (UTC)
new 2. Car Route stops after a few meter, should be 14 km gosmore routing engine problem (Jump-Error)
new 3. Car Route stops after a few meter, should be 5 km gosmore routing engine problem (Jump-Error)
new Route stops early, reverse works
as above Route stops early, reverse works. Moving the from just a bit closer to destination works.
new I don't know why it routes through the (longer) service road instead of the main road. Is it because of the sharp turn? It's not that sharp in real life. The problem is caused by misinterpreting the bicycle=designated tag This cannot be fixed in elemstyles.xml. I suggest to remove the designated tag, as it seems a bit silly here.
Ah, that's too bad. I got the advice to tag it as such . But I will have to think of something else. Norpan 13:11, 26 November 2008 (UTC)
Actually, Gosmore is going to translate bicycle=designated to bicycle=yes, which should work the same for a router. --Lambertus 10:09, 1 December 2008 (UTC)
new An over 10 km detour on on a 4 km route, regardless of transport mode or fastest/shortest. Moving either endpoint closer chooses the straight route.
Not really a route problem, but a user interface problem, so possibly this is the wrong place to report it (where else?). In "from" field entering "newcastle" selects Newcastle in Australia for a route search, rather than the one in the UK. There should be some way of choosing from the various "Newcastle"s, preferably sorted by proximity to the source IP eg. for an Australian user, "newcastle" would default to Newcastle in Australia, for someone in the UK it would default to Newcastle-upon-Tyne in the UK. Even trying to force a UK route by using the full name "newcastle-upon-tyne" to "hexham" results in a route through Australia!
New Not a routing problem but when I set the lat, long and zoom to get the view I want (in Suffolk,UK) it shows it for a moment and then always jumps to show some route for a place called Sneek in Holland.PeterIto 22:53, 30 November 2008 (UTC) This happens when a route calculation is requested without any lat/lon parameters. I'm not sure what you are doing to trigger this behaviour. --Lambertus 10:15, 1 December 2008 (UTC)
new Routing from pier-to-pier over the ferry route works as expected, but when endpoints are moved a short ways up the connecting roads, it goes wonky. --Cohort 00:32, 6 December 2008 (UTC)
new Routing engine uses a road tagged surface=unpaved instead of paved roads.

--Farlokko 13:46, 9 December 2008 (UTC)

new yours expects teleportation around ferries? Depending on the start point position, when the route goes over a ferry, only the ferry route and the final part of the route is shown. Much of the beginning part of the route is just missing - incomplete. Moving the start point closer to the ferry, the route is shown closer to what it should be better (marks the entire segment)
new Middle of route changes when end moves farther away. Begin and end of the route stay the same. See this to compare. Both routes use the same ways at their begin and end, but in the middle, a different way is chosen.
new Route is not taking the flyover (marked as highway=trunk_link) but instead takes the bottom route through the intersection which the flyover is meant to bypass. It seems that the router prefers the combination of highway=primary and highway=trunk over a shorter highway=trunk_link route.
Gosmore uses average speeds to determine the fastest route.
Tweak the average speeds and/or add penalties for sharp turns.
new Depending on the placement of both the start and end point, it chooses to add large additional chunks to the route. A diagram can be seen here. 1 is correct, but the final destination has to be moved short of target to do so. 2 has the target almost correct, but an addition stroll over a golf corse seems to have been added. 3. has start and finish in the correct place, and takes a long de tore. It in fact misses a longer route than needed, and takes a yet longer route than needed. 4. has the correct location, but the start point is moved closure, and it takes the correct route. The points at which it goes wrong are not at any divisions between separate ways, and nothing links to/from the way as far as I can tell. The data seems correct on inspection, and also this route works fine on Ben 18:38, 28 December 2008 (UTC)
new Route is going straight instead of taking the U-turn to the north. router is ignoring nodes tagged with barrier=block and possibly any barrier=* router should consider that nodes tagged with barrier=* as access=no by default.
answered Route ignores new slip road from south to west, exiting A404 and joining M40. Route also ignores new slip road from east to south, exiting M40 and joining A404. The route using the trunk road has a higher average speed according to Gosmore because it has mostly trunk and motorway and the slipway is all motorway_link. Are there any traffic lights that are not mapped yet?
The roundabout itself, which is quite intricate, is filled with traffic lights, none of which are entered in OSM. Do I assume that the addition of those traffic lights will slow the route currently taken, to give preference to the motorway slip road, which has none? If that does not work, do I then alter speeds by Key and Value in JOSM?
Adding the traffic lights is a good thing, but Gosmore does not include traffic lights in it's route calculations yet. --Lambertus 21:13, 26 January 2009 (UTC)
Finetune the average speeds, or implement better average speed tags/algo's.
Is there a list of highway attributes and speeds in Gosmore that I can access for general reference?
There is one in the TRAC code repository, but tinkering with average speeds will not solve all these problems. The routing engine needs to become smarter.--Lambertus 21:13, 26 January 2009 (UTC)
Lambertus, many thanks for your responses. Subject closed.
new new new only_straight_on -restrictions are ignored
new "Fast" route is slower and less direct than the Short route in reality. Kinks N at George Street, and S at Botley Road down slow cycle tracks; wastes time at traffic lights and junctions. Excessive weight appears to be given to cycleways in this case. At the George Street bit, Gosmore has a choice of a short direct route with no cycle link, or a convoluted one involving two very stretches of cycleway. It chooses the latter.

The Botley Road setup is more excusable :) Botley Road in that direction is a mess of on-road bus lanes which bikes can use and off-road parallel tracks shared with pedestrians. I don't think it's fully tagged up, so ignore. --achadwick 10:16, 30 January 2009 (UTC)

new routing against oneway direction - the street 1 decembrie 1918 is part of the route, no matter the direction
new Takes a 100 km detour to avoid a motorway_link. Uses it only if given no other choice, as in 1
The data looks quite suspicious. Shouldn't there be some bridges involved? --Spartanischer Esel 13:29, 16 March 2009 (UTC)
I have corrected the data and will test again next time YOURS updates. bielawski 20:29, 3 April 2009 (UTC)
Cycle networks layer problems only
route description cause solution
Open YOURS takes a long detour here. Changing the destination just a few meters to the south (on the same cycleway!) [5] the route becomes 2.5km shorter. The cycleway is correctly mapped as is it used for routing here: [6] Now the route is even worse than before. But the cause is clearer: gosmore seems to consider the highway=secondary faster than the highway=cycleway which is usually wrong.
suboptimal Leaving the cycle network leads to going up a steep hill. Route via the cycle network would be about 1.9 km vs. 1.6 for the hilly one chosen. No knowledge of the effective hindrance of an incline.
new yours makes a giant U-turn just above Amsterdam. Moving startingpoint along the route, yours comes up with a different (shorter) route, here and the u-turn dissapears. However, both routes use the Houtman kade for starters... GercoKees 11:27, 28 November 2008 (UTC)
Another thing is when you click the reverse button the route is correct and will be almost 3 km shorter GercoKees 10:36, 12 December 2008 (UTC)
This issue might have to do something with the ferry-issues above..? GercoKees 07:46, 16 December 2008 (UTC)
general issue Why does YOURS ignore one way streets for bicycle routing? At least here in Canada, it is illegal to go the wrong way on a one way street on a bicycle (unless there is a contraflow bike lane). Andrewpmk 06:24, 26 December 2008 (UTC)
new YOURS using wormholes to Antarctica and back. --Spartanischer Esel 01:04, 25 July 2009 (UTC)