• OpenStreetMap

"OpenStreetMap is a free editable map of the whole world. It is made by people like you." (from https://www.openstreetmap.org). This page describes the conversion of files containing data from OpenStreetMap to SUMO network files.

There are several ways to download the data from OpenStreetMap to a file. Please read Networks/Import/OpenStreetMapDownload to learn about them. For more information about the file format, see OpenStreetMap file.

3-click scenario generation#

By using the osmWebWizard.py script, a complete scenario can be built quickly and comfortably. The network will be imported with options and typemaps suitable for the selected traffic modes. If more control is needed the options described below can be used.

Importing the Road Network#

netconvert can import OSM-files natively. The corresponding option is called --osm-files <FILE>[,<FILE>]* or --osm <FILE>[,<FILE>]* for short.

The following call to netconvert imports the road network stored in "berlin.osm.xml" and stores the SUMO-network generated from this data into "berlin.net.xml":

netconvert --osm-files berlin.osm.xml -o berlin.net.xml

OSM-data always uses WGS84 geo coordinates, which are automatically transformed to UTM by netconvert (since sumo 0.11.1). Thus you only need explicit projection parameters if a different projection is needed. Refer to the netconvert documentation for other conversion options.

The number of tiles given in both calls must match.

Recommended netconvert Options#

 --geometry.remove --ramps.guess --junctions.join --tls.guess-signals --tls.discard-simple --tls.join --tls.default-type actuated

Rationale:

• --geometry.remove : Simplifies the network (saving space) without changing topology
• --ramps.guess : Acceleration/Deceleration lanes are often not included in OSM data. This option identifies roads that likely have these additional lanes and adds them
• --junctions.join : See #Junctions
• --tls.guess-signals --tls.discard-simple --tls.join : See #Traffic_Lights
• --tls.default-type actuated : Default static traffic lights are defined without knowledge about traffic patterns and may work poorly in high traffic

Countries with left-hand driving#

Recommended Typemaps#

There are multiple degrees of freedom when importing data from OSM. Sometimes, information such as speed limit is missing in the raw data and must be inferred from the abstract type of the road (i.e. motorway). Different simulation scenarios require different modes of traffic and thus different parts of the traffic infrastructure to be imported. The tool for making these choices are typemaps. SUMO provides recommended typemaps in the folder <SUMO_HOME>/data/typemap/. They are explained below.

• osmNetconvert.typ.xml default settings, appropriate for rural and motorway scenarios. This is used in the absence of user-specified types. All other typemaps are intended as patches to this typemap.
• osmNetconvertUrbanDe.typ.xml Changes default speeds to reflect typical urban speed limits (50km/h).
• osmNetconvertPedestrians.typ.xml Adds sidewalks for some edge types and sets permissions appropriate for pedestrian simulation.
• osmNetconvertBicycle.typ.xml Adapts the width of bicycle lanes.
• osmNetconvertShips.typ.xml Imports waterways and ferry routes. This typemap can be combined with any other typemap.
• osmNetconvertRailUsage.typ.xml Imports additional [https://wiki.openstreetmap.org/wiki/Key:usage usage information for railways (main,branch,industrial,...). This typemap only works in combination with other typemaps.
• osmNetconvertBidiRail.typ.xml. Changes the default from uni-directional railroads to bi-directional railroads. This may be useful in some regions of the world where OSM contributors used this style of date representation. The use of this typemap supplants the older option --osm.railway.oneway-default <BOOL>.
• osmNetconvertAirport.typ.xml. Imports aeroways (runway, taxiway, etc.)

Bicycle Traffic#

Bicycle infrastructure can be imported using the option --osm-bike-access. It will evaluate the bicycle=yes/no tags as well as oneway information for bicycles and trigger the addition of separate bike lanes. If you want to modify the width of the bike lanes depending on the street type use the bicycle type map mentioned above.

Pedestrian Traffic#

By default only footpaths (osm ways dedicated for pedestrian use) are imported. To import all sidewalk related information, the option --osm.sidewalks can be set. Alternatively, sidewalks can be added heuristically via typemaps (see above) or guessing-options

Lane-To-Lane Connections#

By default, lane-to-lane connections are guessed by netconvert and only turning restrictions are loaded from OSM to influence connection generation. When setting option --osm.turn-lanes, the turn direction road markings from OSM are evaluated to guide connection generation.

Importing additional Polygons (Buildings, Water, etc.)#

OSM-data not only contains the road network but also a wide range of additional polygons such as buildings and rivers. These polygons can be imported using polyconvert and then added to a sumo-gui-configuration.

To interpret the OSM-data an additional typemap-file is required (the example below is identical to <SUMO_HOME>/data/typemap/osmPolyconvert.typ.xml):

<polygonTypes>
  <polygonType id="waterway"                name="water"       color=".71,.82,.82" layer="-4"/>
  <polygonType id="natural"                 name="natural"     color=".55,.77,.42" layer="-4"/>
  <polygonType id="natural.water"           name="water"       color=".71,.82,.82" layer="-4"/>
  <polygonType id="natural.wetland"         name="water"       color=".71,.82,.82" layer="-4"/>
  <polygonType id="natural.wood"            name="forest"      color=".55,.77,.42" layer="-4"/>
  <polygonType id="natural.land"            name="land"        color=".98,.87,.46" layer="-4"/>

  <polygonType id="landuse"                 name="landuse"     color=".76,.76,.51" layer="-3"/>
  <polygonType id="landuse.forest"          name="forest"      color=".55,.77,.42" layer="-3"/>
  <polygonType id="landuse.park"            name="park"        color=".81,.96,.79" layer="-3"/>
  <polygonType id="landuse.residential"     name="residential" color=".92,.92,.89" layer="-3"/>
  <polygonType id="landuse.commercial"      name="commercial"  color=".82,.82,.80" layer="-3"/>
  <polygonType id="landuse.industrial"      name="industrial"  color=".82,.82,.80" layer="-3"/>
  <polygonType id="landuse.military"        name="military"    color=".60,.60,.36" layer="-3"/>
  <polygonType id="landuse.farm"            name="farm"        color=".95,.95,.80" layer="-3"/>
  <polygonType id="landuse.greenfield"      name="farm"        color=".95,.95,.80" layer="-3"/>
  <polygonType id="landuse.village_green"   name="farm"        color=".95,.95,.80" layer="-3"/>

  <polygonType id="tourism"                 name="tourism"     color=".81,.96,.79" layer="-2"/>
  <polygonType id="military"                name="military"    color=".60,.60,.36" layer="-2"/>
  <polygonType id="sport"                   name="sport"       color=".31,.90,.49" layer="-2"/>
  <polygonType id="leisure"                 name="leisure"     color=".81,.96,.79" layer="-2"/>
  <polygonType id="leisure.park"            name="tourism"     color=".81,.96,.79" layer="-2"/>
  <polygonType id="aeroway"                 name="aeroway"     color=".50,.50,.50" layer="-2"/>
  <polygonType id="aerialway"               name="aerialway"   color=".20,.20,.20" layer="-2"/>

  <polygonType id="shop"                    name="shop"        color=".93,.78,1.0" layer="-1"/>
  <polygonType id="historic"                name="historic"    color=".50,1.0,.50" layer="-1"/>
  <polygonType id="man_made"                name="building"    color="1.0,.90,.90" layer="-1"/>
  <polygonType id="building"                name="building"    color="1.0,.90,.90" layer="-1"/>
  <polygonType id="amenity"                 name="amenity"     color=".93,.78,.78" layer="-1"/>
  <polygonType id="amenity.parking"         name="parking"     color=".72,.72,.70" layer="-1"/>
  <polygonType id="power"                   name="power"       color=".10,.10,.30" layer="-1" discard="true"/>
  <polygonType id="highway"                 name="highway"     color=".10,.10,.10" layer="-1" discard="true"/>
  <polygonType id="railway"                 name="railway"     color=".10,.10,.10" layer="-1" discard="true"/>

  <polygonType id="boundary" name="boundary"    color="1.0,.33,.33" layer="0" fill="false" discard="true"/>
  <polygonType id="admin_level" name="admin_level"    color="1.0,.33,.33" layer="0" fill="false" discard="true"/>
</polygonTypes>

Using the typemap file typemap.xml the following call to polyconvert imports polygons from OSM-data and produces a Sumo-polygon file.

polyconvert --net-file berlin.net.xml --osm-files berlin.osm --type-file typemap.xml -o berlin.poly.xml

The created polygon file berlin.poly.xml can then be added to a sumo-gui configuration:

 <configuration>
     <input>
         <net-file value="berlin.net.xml"/>
         <additional-files value="berlin.poly.xml"/>
     </input>
 </configuration>

Railway-specific Objects#

By using typemap <SUMO_HOME>/data/typemap/osmPolyconvertRail.typ.xml, numerous railway-related POIs such as

• stations
• switches
• stop positions
• platforms
• position markers

can be imported.

Import Scripts#

The helper script osmGet.py allows downloading a large area. The resulting file called "<PREFIX>.osm.xml" can then be imported using the script osmBuild.Py. Both scripts are located in <SUMO_HOME>/tools.

The call is:

osmGet.py --bbox <BOUNDING_BOX> --prefix <NAME>
osmBuild.py --osm-file <NAME>.osm.xml  [--vehicle-classes (all|road|passenger)] [--type-file <TYPEMAP_FILE>] [--netconvert-options <OPT1,OPT2,OPT3>] [--polyconvert-options <OPT1,OPT2,OPT3>]

If "road" is given as parameter, only roads usable by road vehicles are extracted, if "passenger" is given, only those accessible by passenger vehicles are.

When using the option --type-file an additional output file with polygons of rivers and buildings as well as Points of Interest (POIs) will be generated. This can be loaded in sumo-gui for additional visualization. Useful typemap files can be found at <SUMO_HOME>/data/typemap/.

Additional options for netconvert and polyconvert can be supplied using the options --netconvert-options and -polyconvert-options

Note that the scripts also support a secondary syntax for loading even large areas by splitting them into multiple tiles and download requests. In this case the calls look like this:

osmGet.py --bbox <BOUNDING_BOX> --prefix <NAME> --oldapi --tiles <INT>
osmBuild.py --oldapi-prefix <NAME --tiles <INT> [--vehicle-classes (all|road|passenger),ramps,tls] [--type-file <TYEPMAP_FILE>]

Elevation Data#

Incooperating z-coordinates in networks is still experimental so please report anything odd.

SRTM#

When using option --osm.elevation, z-data is imported from tags with key="ele" in OSM nodes. Since this tag is not yet in wide use, tools exist to overlay OSM data with elevation data sources (https://wiki.openstreetmap.org/wiki/Srtm_to_Nodes). When using the osmosis-srtm plugin the option tagName=ele must be used since only the 'ele' tag is evaluated and the plugin would use the 'height' tag by default.

Layer Information#

When using option --osm.layer-elevation, z-data is imported from tags with key="layer" in OSM ways. Since this data does not encode full elevation information, a heuristic is used to interpret the given information. Manual correction may be necessary.

Other data sources#

Further options for importing elevation data are listed at the Elevation overview page.

Further Notes#

Junctions#

In OpenStreetMap roads forming a single street and separated by, for example, a lawn or tram line, are represented by two edges that are parallel to each other. When crossing with another street, they form two junctions instead of one. To merge such junctions into a single junction, one can define which nodes to merge. See Networks/Building Networks from own XML-descriptions#Joining Nodes and netconvert documentation for usage details.

The netconvert option --junctions.join applies a heuristic to join these junction clusters automatically and is used by default when using the osmBuild.py script described above. However, some junction clusters are too complex for the heuristic and should be checked manually (as indicated by the warning messages). To manually specify joins for these junctions, see JoiningNodes Also, sometimes the heuristic wrongly joins some junctions. These can be excluded by giving them as a list to the option --junctions.join-exclude <ID>[,<ID>]*.

When leaving junctions unjoined, there is a high risk of getting low throughput, jams and even deadlocks due to the short intermediate edges and the difficulty in computing proper traffic light plans for the junction clusters.

Traffic Lights#

Interpreting traffic light information in OSM#

netconvert prefers each intersection to be represented by a single node with a single traffic light controller. To achieve the former, see #Junctions. To achieve the latter some extra options are recommended. OSM often uses nodes ahead of an intersection to represent the position of traffic light signals. The actual intersection itself is then not marked as controlled. To interpret these structures the option --tls.guess-signals and --tls.guess-signals.dist <FLOAT> may be used. To cover the cases where this heuristic fails, the options below may be used to computed a joint tls plan for multiple nodes. To identify the guessed signals in a network they all start with the prefix GS_ before the node id.

Joining traffic lights#

OSM does not have the possibility to assign several nodes to a single traffic light. This means that near-by nodes, normally controlled by one traffic light system are controlled by two after the network is imported. It is obvious that traffic collapses in such areas if both traffic lights are not synchronized. Better representation of the reality can often be achieved by joining nearby junctions into a single junction. However, if the junctions should stay separate, it is possible to at least generate a joint controller by setting the option --tls.join. For fine-tuning of joint traffic lights, the attribute tl can be customized for individual nodes. The joined traffic lights get the id joinedS_id0_id1 where id0 and id1 are the junction ids. If there are more nodes in the join than given by --max-join-ids (default 4) the id will be abbreviated to something like joinedS_id0_id1_id2_id3_#5more (for a 9 node cluster).

If you want to let netconvert guess joined traffic lights at node clusters which were not joined previously (e.g. by using --junctions.join) and which do not have traffic lights assigned in the input, you can use the option --tls.guess.joining. This is not recommended in general. The traffic lights joint by this method get the id joinedG_id0_id1 where id0 and id1 are the junction ids and the same abbreviation rules as above apply.

Debugging missing traffic lights#

Occasionally intersections that should be TLS-controlled are set to uncontrolled in the exported .net.xml-file. This may either be due to lack of data in OSM, or due to the invalid interpretation of that data by netconvert. Either of the following steps may be useful to diagnose the problem:

• run netconvert without the options --tls.discard-loaded --tls.discard-simple
• run polyconvert with a type-file that contains <polygonType id="highway" name="highway" and then look at the generated POIs in sumo-gui. They should include all traffic light locations defined in the OSM file.

Overriding the traffic light information#

If the traffic light information embedded in the OSM file does not fit your needs, you can strip it with --osm.discard-tls option in netconvert and then provide your own definition in a separate *.nod.xml file in a second run of netconvert:

# 1. Import the OSM file to SUMO, discarding TLS information.
netconvert --osm-files berlin.osm.xml --output-file berlin-without-tls.net.xml \
   --osm.discard-tls

# 2. Set traffic light information.
netconvert --sumo-net-file berlin-without-tls.net.xml --node-files tls-controlled-nodes.nod.xml \
   --output-file berlin-with-tls.net.xml

where tls-controlled-nodes.nod.xml overwrites the type of node to "traffic_light". If the node already exists (which is usually the case) you don't have to provide any information other than the node's ID and new node type.

Highway On- and Off-Ramps#

OSM networks often lack additional lanes for highway on- and off-ramps. They can be guessed via netconvert using the --guess-ramps option.

Isolated Edges#

When dealing with strictly vehicular scenarios it usually helps to add the option

--remove-edges.isolated

To discard edges which have no predecessor and no successor edge. However, this often causes the removal of railways or waterways which is not desirable for multi-modal scenarios.

Shared Space for Trams and Road Vehicles#

In the OSM database, shared space for tram and road vehicles is often modelled with distinct elements that occupy the same space. When imported directly, this would allow those modes of traffic to ignore each other due to running on different edges. To fix this, the option --edges.join-tram-dist FLOAT may be used. When this option is set (values between 1 and 2 are recommended), overlapping OSM elements will be converted to road lanes with shared permissions (indicated by a dark purple in sumo-gui).

Further way/edge attributes#

When setting option --output.street-names, the 'name' attribute of every edge will be set to the street name defined in the OSM data.

When setting option --osm.all-attributes, all OSM tags of a way are exported as generic params of the edge. If only a specific selection of tags should be imported, this can be set with option --osm.extra-attributes.

Railway signals#

By default, rail signals of category "main" and "combined" are imported. In case the region of interest doesn't provide the necessary level of detail it is possible to customize the import rules using option --osm.railsignals. The following values are supported

• ALL: imports all signals (key="railway" and value="signal")
• KEY=VALUE: imports all nodes with a tag where key is "railway:signal:KEY" and value is VALUE
• KEY=VALUE,KEY2=VALUE2,...: as above with multiple combinations
• KEY=: imports all nodes with a tag where key is "railway:signal:KEY" (also supports giving a list)
• DEFAULT: equivalent to option value "main=,combined="

Editing OSM networks#

JOSM#

From George Dita, on 01.07.2009 JOSM can be used to edit OSM-data (i.e. for trimming a rectangular map and deleting unwanted features). After you delete the part that does not interest you, you have to alter the file using xmlstarlet which actually deletes the nodes.

xmlstarlet can be used like this:

xmlstarlet ed -d "/osm/*[@action='delete']" < input.osm > output.osm

OSMOSIS#

From Christian Klotz, on 01.07.2009, tip by Christoph Sommmer

The java tool osmosis (https://wiki.openstreetmap.org/index.php/Osmosis) can be used to filter out unwanted features from an OSM-file. The following command keeps motorways and motorway links while filtering out everything else:

java -jar osmosis.jar --read-xml file="orginal.osm.xml" --way-key-value \
    keyValueList="highway.motorway,highway.motorway_link" \
    --used-node --write-xml file="filtered.osm.xml"

osmfilter / osmconvert#

osmfilter is a command line tool used to filter OpenStreetMap data files. It can be used to preprocess the data and can save a lot of work for netconvert. With this tool it is possible to import the top-level roads for a whole country within minutes of processing time. It is recommended to use the associated tool osmconvert to convert the data into the '.o5m' format and back to speed up the work on large datasets. The tools can also be used to cut an osm file to a boundary box like this (be aware that the order of coordinates is lon,lat):

osmconvert -b=10.54,52.257,10.545,52.26 fullnet.osm.xml > myplace.osm.xml

netconvert Details#

Road Types#

When importing road networks, netconvert searches for the street type, encoded in OSM as a key/value-pair where the key is either "highway", "railway" or "waterway". Only if such a key occurs in the edge definition, the edge is imported (see also below). The edge's type name is built from the found key/value pair by building a name as: <KEY>.<VALUE>. Using this type name, the edge's attributes are determined using a predefined map of type names to type definitions. It is possible to override the default types with own type definitions. This is documented in the article about the SUMO edge type file.

If no explicit type file is given, the default one located at <SUMO_HOME>/data/typemap/osmNetconvert.typ.xml is used. If you want to change the values to add pedestrian infrastructure or have bidirectional railway edges you may want to load additional type maps from <SUMO_HOME>/data/typemap/.

Explicit Road Attributes#

In case an edge contains the definition about the number of lanes (key="lanes") or the allowed speed (key="maxspeed"), this information is used instead of the according type's value. Also, the per-edge information whether the edge is a one-way edge is read (key="oneway"). In case the edge belongs to a roundabout (key="junction" and value="roundabout"), it is also set as being a one-way edge.

Dismissing unwanted traffic modes#

In most cases, tracks and edges which not may be crossed by motorised traffic are not interesting for road traffic research. It is possible to exclude these edges from being imported using the netconvert-option --remove-edges.by-vclass <STRING>[,<STRING>]*.

For removing all edges which can not be used by passenger vehicles the call must be extended by:

--remove-edges.by-vclass hov,taxi,bus,delivery,transport,lightrail,cityrail, \
  rail_slow,rail_fast,motorcycle,bicycle,pedestrian

For removing all edges which can not be used by road vehicles the call must be extended by:

--remove-edges.by-vclass rail_slow,rail_fast,bicycle,pedestrian

Relationship between OSM ids and SUMO-ids#

For the most part, the relationship between OSM and .net.xml is simple:

• OSM node id "1234" translates into junction id "1234".
• OSM way id "5677" translates into multiple edge ids "5678#0", "5678#1", "5678#2", "-5678#0", "-5678#1", "-5678#2"

OSM ways are mostly bi-directional and cross multiple intersections. In .net.xml they are split at each intersection with a running index #n and the edges with prefix '-' are those in the opposite direction.

By setting the option --output.original-names, each lane in the network will record the element name from the input network that it is derived from in a child element:

<param key="origID" value="...."/>

Some OSM elements may not appear in the created network because they are joined with other elements or converted to (unnamed) geometry points along an edge (due to option --geometry-remove). Thus, a sequence of ways (i.e. "42", "43", "57") without intersection, would be joined into a single SUMO edge using the id of the first way ("42").

Also, option --junctions.join can lead to the removal of some ways and nodes. Multiple nodes ("1", "2") are joined into "cluster_1_2" and the edges that are within the cluster are converted into internal edges named ":cluster_1_2_INDEX".

Warnings during Import#

Several types of warnings and errors with different levels of severity may be issued during OSM import.

Additional warnings are described here.

Importing large Networks#

It is possible to build SUMO-networks for very large areas (i.e. the whole of Scotland) but some precautions should be taken:

• reduce the network size with the following options

--no-internal-links, --keep-edges.by-vclass passenger,
  --remove-edges.by-type highway.track,highway.services,highway.unsurfaced

• if you only need major roads you can reduce the network further by setting the option

--keep-edges.by-type
  highway.motorway,highway.motorway_link,highway.trunk,highway.trunk_link,highway.primary,highway.primary_link

• if your network is larger than ~500MB make sure you use the 64bit version of sumo and use a computer with plenty of RAM

Importing Public Transport Data#

netconvert and further SUMO tools can be used to import public transpot data from OSM. The easiest way to do this is by using the OSM-Web-Wizard tool. Performing the import without the wizard is explained in Tutorials/PT_from_OpenStreetMap.

The following data is imported:

• Public transport stops
• Public transport lines that service these stops

Public transport schedules which are needed to make use of the above data are generated with a user-defined service period based on a simulation of the lines.

Importing Sidewalks#

By default (osmNetconvert.typ.xml), only dedicated edges for pedestrians will be built and all other roads (except motorway) will permit pedestrians on all lanes. For a pedestrian simulation, sidewalks are needed and pedestrians should be forbidden on most road lanes.

OSM data coverage and definition style for sidewalks varies by region which may require different import options to achieve adequate sidewalk coverage.

Sidewalks from typemap#

By adding a typemap such as {{ SUMO_HOME }}/data/osmNetconvertPedestrians.typ.xml, sidewalks of a pre-configured width a added to a specific set of road types and all non-sidewalk lanes are forbidden for pedestrians.

OSM-sidewalk data will only be considered if it explicitly disables sidewalks on an edge. This ensures good sidewalk coverage but may lead to double-sidewalks if OSM modellers have added the "sidewalks" as parallel foot paths edges.

Sidewalks from OSM#

By setting option --osm.sidewalks, all sidewalk data from OSM will be loaded. When combined with a typemap such as {{ SUMO_HOME }}/data/osmNetconvertPedestrians.typ.xml, the typemap will only be used to configure sidewalk widths but no extra sidewalks will be added.

When --osm.sidewalks is set, by setting option --osm.crossings the "highway=crossing" tags will be loaded and crossings will be generated.

This definition style prevents double-sidewalks but may lead to missing sidewalks wherever OSM modellers did not add sidewalk information.

osmWebWizard uses this style beginning with version 1.11.0.

Importing OSM Data via Python/ Overpass API#

Another way to get OSM data is to query via the Overpass API, e.g. with python.

Python Example (Get OSM River Data from OSM in BBox)

   import requests
   import json

   def getData(lsouthern-most latitude, western-most longitude, northern-most latitude, eastern-most longitude):
     query = 'way["waterway"="river"]["ship"="yes"](bbox:%s, %s, %s, %s)' %(southern-most latitude, western-most longitude, northern-most latitude, eastern-most longitude)
     overpass_url = "https://overpass-api.de/api/interpreter"
     overpass_query = """
     [out:json][timeout:25];
     (
       """+query+""";
     );
     out body;
     >;
     out skel qt;
     """
     response = requests.get(overpass_url, params={'data': overpass_query})
     data = response.json()
     return data, query

Caution: If the data volume is too large, the overpass API aborts.

More information about the query syntax and a GUI for testing the queries can be found on the website https://overpass-turbo.eu/.

Importing other formats (pbf, o5m)#

To use a pbf or o5m network in SUMO, you first need to convert it to the OSM format. This can be done using Osmconvert.

Missing Descriptions#

• TLS computation
• computation of lane-2-lane connections
• what is exactly imported (how edge attributes are determined)
• other traffic modes
• Network quality

References#

• https://www.openstreetmap.org/ - the home site
• https://www.openstreetmap.de/ - the German home site
• https://wiki.openstreetmap.org/wiki/Map_features - information about database attributes