OSRM profiles

OSRM supports “profiles”. Profiles representing routing behavior for different transport modes like car, bike and foot. You can also create profiles for variations like a fastest/shortest car profile or fastest/safest/greenest bicycles profile.

A profile describes whether or not it’s possible to route along a particular type of way, whether we can pass a particular node, and how quickly we’ll be traveling when we do. This feeds into the way the routing graph is created and thus influences the output routes.

Available profiles

Out-of-the-box OSRM comes with profiles for car, bicycle and foot. You can easily modify these or create new ones if you like.

Profiles have a ‘lua’ extension, and are placed in ‘profiles’ directory.

When running OSRM preprocessing commands you specify the profile with the —profile (or the shorthand -p) option, for example:

osrm-extract --profile ../profiles/car.lua planet-latest.osm.pbf

Processing flow

It’s important to understand that profiles are used when preprocessing the OSM data, NOT at query time when routes are computed.

This means that after modifying a profile you will need to extract, contract and reload the data again and to see changes in the routing results. See Processing Flow  for more.

Profiles are written in Lua

Profiles are not just configuration files. They are scripts written in the Lua scripting language . The reason for this is that OpenStreetMap data is complex, and it’s not possible to simply define tag mappings. Lua scripting offers a powerful way to handle all the possible tag combinations found in OpenStreetMap nodes and ways.

Basic structure of profiles

A profile will process every node and way in the OSM input data to determine what ways are routable in which direction, at what speed, etc.

A profile will typically:

• Define api version (required)
• Require library files (optional)
• Define setup function (required)
• Define process functions (some are required)
• Return functions table (required)

A profile can also define various local functions it needs.

Looking at car.lua as an example, at the top of the file the api version is defined and then required library files are included.

Then follows the setup function, which is called once when the profile is loaded. It returns a big hash table of configurations, specifying things like what speed to use for different way types. The configurations are used later in the various processing functions. Many adjustments can be done just by modifying this configuration table.

The setup function is also where you can do other setup, like loading an elevation data source if you want to consider that when processing ways.

Then come the process_node and process_way functions, which are called for each OSM node and way when extracting OpenStreetMap data with osrm-extract.

The process_turn function processes every possible turn in the network, and sets a penalty depending on the angle and turn of the movement.

Profiles can also define a process_segment function to handle differences in speed along an OSM way, for example to handle elevation. As you can see, this is not currently used in the car profile.

At the end of the file, a table is returned with references to the setup and processing functions the profile has defined.

Understanding speed, weight and rate

When computing a route from A to B there can be different measures of what is the best route. That’s why there’s a need for different profiles.

Because speeds vary on different types of roads, the shortest and the fastest route are typically different. But there are many other possible preferences. For example a user might prefer a bicycle route that follow parks or other green areas, even though both duration and distance are a bit longer.

To handle this, OSRM doesn’t simply choose the ways with the highest speed. Instead it uses the concepts of weight and rate. The rate is an abstract measure that you can assign to ways as you like to make some ways preferable to others. Routing will prefer ways with high rate.

The weight of a way is normally computed as length / rate. The weight can be thought of as the resistance or cost when passing the way. Routing will prefer ways with low weight.

You can also set the weight of a way to a fixed value. In this case it’s not calculated based on the length or rate, and the rate is ignored.

You should set the speed to your best estimate of the actual speed that will be used on a particular way. This will result in the best estimated travel times.

If you want to prefer certain ways due to other factors than the speed, adjust the rate accordingly. If you adjust the speed, the time estimation will be skewed.

If you set the same rate on all ways, the result will be shortest path routing. If you set rate = speed on all ways, the result will be fastest path routing. If you want to prioritize certain streets, increase the rate on these.

Elements

api_version

A profile should set api_version at the top of your profile. This is done to ensure that older profiles are still supported when the api changes. If api_version is not defined, 0 will be assumed. The current api version is 4.

Library files

The folder profiles/lib/ contains LUA library files for handling many common processing tasks.

They all return a table of functions when you use require to load them. You can either store this table and reference its functions later, or if you need only a single function you can store that directly.

setup()

The setup function is called once when the profile is loaded and must return a table of configurations. It’s also where you can do other global setup, like loading data sources that are used during processing.

Note that processing of data is parallelized and several unconnected LUA interpreters will be running at the same time. The setup function will be called once for each. Each LUA interpreter will have its own set of globals.

The following global properties can be set under properties in the hash you return in the setup function:

The following additional global properties can be set in the hash you return in the setup function:

process_node(profile, node, result, relations)

Process an OSM node to determine whether this node is a barrier or can be passed and whether passing it incurs a delay.

The following attributes can be set on result:

process_way(profile, way, result, relations)

Given an OpenStreetMap way, the process_way function will either return nothing (meaning we are not going to route over this way at all), or it will set up a result hash.

Importantly it will set result.forward_mode and result.backward_mode to indicate the travel mode in each direction, as well as set result.forward_speed and result.backward_speed to integer values representing the speed for traversing the way.

It will also set a number of other attributes on result.

Using the power of the scripting language you wouldn’t typically see something as simple as a result.forward_speed = 20 line within the process_way function. Instead process_way will examine the tag set on the way, process this information in various ways, calling other local functions and referencing the configuration in profile, etc., before arriving at the result.

The following attributes can be set on the result in process_way:

process_segment(profile, segment)

The process_segment function is called for every segment of OSM ways. A segment is a straight line between two OSM nodes.

On OpenStreetMap way cannot have different tags on different parts of a way. Instead you would split the way into several smaller ways. However many ways are long. For example, many ways pass hills without any change in tags.

Processing each segment of an OSM way makes it possible to have different speeds on different parts of a way based on external data like data about elevation, pollution, noise or scenic value and adjust weight and duration of the segment accordingly.

In the process_segment function you don’t have access to OSM tags. Instead you use the geographical location of the start and end point of the way to look up information from another data source, like elevation data. See rasterbot.lua for an example.

The following attributes can be read and set on the result in process_segment:

process_turn(profile, turn)

The process_turn function is called for every possible turn in the network. Based on the angle and type of turn you assign the weight and duration of the movement.

The following attributes can be read and set on the result in process_turn:

roads_on_the_right and roads_on_the_left

The information of roads_on_the_right and roads_on_the_left that can be read are as follows:

The order of the roads in roads_on_the_right and roads_on_the_left are counter clockwise. If the turn is a u turn, all other connected roads will be in roads_on_the_right.

Example

           c   e
           |  /
           | /
    a ---- x ---- b
          /|
         / |
        f  d

When turning from a to b via x,

• roads_on_the_right[1] is the road xf
• roads_on_the_right[2] is the road xd
• roads_on_the_left[1] is the road xe
• roads_on_the_left[2] is the road xc

Note that indices of arrays in lua are 1-based.

highway_turn_classification and access_turn_classification

When setting appropriate turn weights and duration, information about the highway and access tags of roads that are involved in the turn are necessary. The lua turn function process_turn does not have access to the original osrm tags anymore. However, highway_turn_classification and access_turn_classification can be set during setup. The classification set during setup can be later used in process_turn.

Example

In the following example we use highway_turn_classification to set the turn weight to 10 if the turn is on a highway and to 5 if the turn is on a primary.

function setup()
  return {
    highway_turn_classification = {
      ['motorway'] = 2,
      ['primary'] = 1
    }
  }
end

function process_turn(profile, turn) {
  if turn.source_highway_turn_classification == 2 and turn.target_highway_turn_classification == 2 then
    turn.weight = 10
  end
  if turn.source_highway_turn_classification == 1 and turn.target_highway_turn_classification == 1 then
    turn.weight = 5
  end
}

Guidance

The guidance parameters in profiles are currently a work in progress. They can and will change. Please be aware of this when using guidance configuration possibilities.

Guidance uses road classes to decide on when/if to emit specific instructions and to discover which road is obvious when following a route. Classification uses three flags and a priority-category. The flags indicate whether a road is a motorway (required for on/off ramps), a link type (the ramps itself, if also a motorway) and whether a road may be omitted in considerations (is considered purely for connectivity). The priority-category influences the decision which road is considered the obvious choice and which roads can be seen as fork. Forks can be emitted between roads of similar priority category only. Obvious choices follow a major priority road, if the priority difference is large.

Using raster data

OSRM has built-in support for loading an interpolating raster data in ASCII format. This can be used e.g. for factoring in elevation when computing routes.

Use raster:load() in your setup function to load data and store the source in your configuration hash:

function setup()
  return {
    raster_source = raster:load(
      "rastersource.asc",  -- file to load
      0,    -- longitude min
      0.1,  -- longitude max
      0,    -- latitude min
      0.1,  -- latitude max
      5,    -- number of rows
      4     -- number of columns
    )
  }
end

The input data must an ASCII file with rows of integers. e.g.:

0  0  0   0
0  0  0   250
0  0  250 500
0  0  0   250
0  0  0   0

In your segment_function you can then access the raster source and use raster:query() to query to find the nearest data point, or raster:interpolate() to interpolate a value based on nearby data points.

You must check whether the result is valid before use it.

Example:

function process_segment (profile, segment)
  local sourceData = raster:query(profile.raster_source, segment.source.lon, segment.source.lat)
  local targetData = raster:query(profile.raster_source, segment.target.lon, segment.target.lat)
 
  local invalid = sourceData.invalid_data()
  if sourceData.datum ~= invalid and targetData.datum ~= invalid then
      -- use values to adjust weight and duration
    [...]
end

See rasterbot.lua and rasterbotinterp.lua for examples.

Helper functions

There are a few helper functions defined in the global scope that profiles can use:

• durationIsValid
• parseDuration
• trimLaneString
• applyAccessTokens
• canonicalizeStringList