POpulation DIstribution-based Gtfs Generator
A realistic public transport dataset generator, which is serialized as GTFS.
It is based on five sub-generators:
- Region: A geographical area of cells where each cells contains a population value.
- Stops: Tagging of cells with stop or no stop.
- Edges: Adding transport edges between stops.
- Routes: Routes over one or more edges.
- Connections: Instantiation of routes at times.
If you want more background information about the motivations of this work, be sure to have a look at our article.
This generator is a Node.js application that can be installed by running:
[sudo] npm install -g podiggThe easiest way to run the generator is using the command line tool:
podigg [output folder [path to a JSON config file]]The default output folder is output_data.
This config file contains parameters for the generator, as explained below. Example of a config file:
{
"seed": 1,
"stops:stops": 100,
"connections:connections": 3000
}Alternatively, the generator can also be configured using environment variables, as explained below. In that case, the generator must be called as follows:
podigg-env [output folder]The generator can be included into your application as follows:
const PodiggGenerator = require('podigg');
new PodiggGenerator({
"seed": 1,
"stops:stops": 100,
"connections:connections": 3000
}).generate('output_data');Downloading and running the container from the Docker hub:
docker pull podigg/podigg
docker run --rm -it -v $(pwd)/docker-out:/output_data -e GTFS_GEN_SEED=100 podigg/podiggBuilding and running the container from this repo:
git clone [email protected]:PoDiGG/podigg.git
cd podigg
docker build -t podigg .
docker run --rm -it -v $(pwd)/docker-out:/output_data -e GTFS_GEN_SEED=100 podiggParameters must be passed using environment variables.
All parameters are scoped by their generator name in lower-case, except for the general parameters.
For example, choosing a region's latitude offset is done with the parameter region:lat_offset.
When configuring parameters via environment variables,
parameters should be defined with the prefix GTFS_GEN_,
followed by the generator name + __ (or empty if general) and the parameter name.
The generator and parameter names can either be upper or lower case.
For example, choosing a region's latitude offset is done with the parameter GTFS_GEN_REGION__LAT_OFFSET,
and choosing the seed is done with GTFS_GEN_SEED.
| Name | Default Value | Description |
|---|---|---|
| seed | 1 |
The random seed |
Several region generators exist which are explained hereafter, one of them needs to be selected.
Config prefix: region:
| Name | Default Value | Description |
|---|---|---|
| region_generator | isolated |
Name of a region generator. (isolated, noisy or region) |
| lat_offset | 0 |
The value to add with all generated latitudes |
| lon_offset | 0 |
The value to add with all generated longitudes |
| cells_per_latlon | 100 |
The precision of the cells, how many cells go in 1 latitude or 1 longitude. |
| Name | Default Value | Description |
|---|---|---|
| region_file_path | null |
Path to the cells in csv, this can also be a filename of an internal region file from the data directory, for example region_BE.csv. Expected columns (x:integer, y:integer, lat:float, long:float, density:float) |
A noise-based generator, where population values are influenced by nearby cells.
| Name | Default Value | Description |
|---|---|---|
| size_x | 300 |
The width of the region in number of cells |
| size_y | 300 |
The height of the region in number of cells |
| pop_average | 0 |
The average population value for a cell |
| pop_deviation | 10 |
The standard deviation of the population value for a cell |
A generator that creates a given number of circular clusters of population. The population density is high at the center of the cluster and decreases to zero when going to the border of the cluster.
| Name | Default Value | Description |
|---|---|---|
| size_x | 300 |
The width of the region in number of cells |
| size_y | 300 |
The height of the region in number of cells |
| pop_average | 0 |
The average population value for a cell |
| pop_deviation | 10 |
The standard deviation of the population value for a cell |
| pop_clusters | 50 |
The number of clusters to generate. |
| max_radius | 50 |
The maximum cluster radius. |
The generation of stops
Config prefix: stops:
| Name | Default Value | Description |
|---|---|---|
| stops | 600 |
How many stops should be generated |
| min_station_size | 0.01 |
The minimum population value in a cell for a station to form |
| max_station_size | 30 |
The maximum population value in a cell for a station to form |
| start_stop_choice_power | 4 |
The power for selecting cells with a large population value as stops |
| min_interstop_distance | 1 |
The minimum distance between stops in number of cells |
| factor_stops_post_edges | 0.66 |
The factor of stops that should be generated after edge generation |
| edge_choice_power | 2 |
The power for selecting longer edges to generate stops on |
| stop_around_edge_choice_power | 4 |
The power for selecting cells with a large population value around edges as stops |
| stop_around_edge_radius | 2 |
The radius in number of cells around an edge to select points from |
The generation of edges
Config prefix: edges:
| Name | Default Value | Description |
|---|---|---|
| max_intracluster_distance | 100 |
The maximum distance stops in one cluster can have from each other |
| max_intracluster_distance_growthfactor | 0.1 |
The lower this value, the larger the chance that closer stops will be clustered first before further away stations |
| post_cluster_max_intracluster_distancefactor | 1.5 |
The larger the value, the larger the chance that a stop will be connected to more stops |
| loosestations_neighbourcount | 3 |
The number of neighbours around a loose station that should define its area |
| loosestations_max_range_factor | 0.3 |
The maximum range to check around a loose station relative to the total region size |
| loosestations_max_iterations | 10 |
The max number of iterations to try to connect one loose station |
| loosestations_search_radius_factor | 0.5 |
The number to multiply with the loose station neighbourhood size to get the search radius for each step |
The generation of trips and routes
Config prefix: routes:
| Name | Default Value | Description |
|---|---|---|
| routes | 1000 |
The number of routes to generate |
| largest_stations_fraction | 0.05 |
The fraction of (largest) stops between which routes need to be formed |
| penalize_station_size_area | 10 |
The area in which stop sizes should be penalized |
| max_route_length | 10 |
The maximum number of edges a route can have in the macro-step, the larger, the slower this generator |
| min_route_length | 4 |
The minimum number of edges a route must have in the macro-step |
The generation of connections
Config prefix: connections:
| Name | Default Value | Description |
|---|---|---|
| time_initial | 0 |
The initial timestamp (ms) of trip starting times |
| time_final | 24 * 3600000 * 31 |
The final timestamp (ms) of trip starting times |
| connections | 30000 |
The number of connections to generate |
| stop_wait_min | 60000 |
The minimum waiting time per stop in milliseconds |
| stop_wait_size_factor | 60000 |
The factor in milliseconds of stop waiting time to add depending on the station size |
| route_choice_power | 2 |
The power for selecting longer routes for instantiating connections |
| vehicle_max_speed | 160 |
The maximum speed of a vehicle in km/h, used to calculate the duration of a connection |
| vehicle_speedup | 1000 |
The vehicle speedup in km/(h^2), used to calculate the duration of a connection |
| hourly_weekday_distribution | [0.05,0.01,0.01,0.48,2.46,5.64,7.13,6.23,5.44,5.43,5.41,5.49,5.42,5.41,5.57,6.70,6.96,6.21,5.40,4.95,4.33,3.31,1.56,0.42] |
The chance (percentage) for each hour to have a connection on a weekday |
| hourly_weekend_distribution | [0.09,0.01,0.01,0.08,0.98,3.56,5.23,5.79,5.82,5.89,5.84,5.91,5.88,5.95,5.87,5.95,5.89,5.96,5.92,5.94,5.62,4.61,2.45,0.76] |
The chance (percentage) for each hour to have a connection on a weekend day |
| delay_chance | 0 |
The 0-1 chance that a connection will have a delay, 0 will not produce any delays (default) |
| delay_max | 3600000 |
The maximum delay in milliseconds |
| delay_choice_power | 1 |
Higher values means higher chance on larger delays |
| delay_reasons | { ... } |
Default reasons for having delays with their respective chance. Keys must be prefixed with td: http://purl.org/td/transportdisruption# |
| delay_reduction_duration_fraction | 0.1 |
The maximum fraction of connection duration that can be subtracted when there is a delay |
Optionally, PoDiGG can also generate realistic route planning query sets based on the generated dataset.
For this, the queryset:generate option must be set to true.
Config prefix: queryset:
| Name | Default Value | Description |
|---|---|---|
| start_stop_choice_power | 4 |
Higher values means higher chance on larger stations when selecting starting stations |
| query_count | 100 |
The number of queries that should be generated |
| time_initial | 0 |
The initial timestamp (ms) |
| time_final | 24 * 3600000 |
The final timestamp (ms) |
| max_time_before_departure | 3600000 |
The maximum time in ms that a query for a certain departure time must be queried |
| hourly_weekday_distribution | [0.05,0.01,0.01,0.48,2.46,5.64,7.13,6.23,5.44,5.43,5.41,5.49,5.42,5.41,5.57,6.70,6.96,6.21,5.40,4.95,4.33,3.31,1.56,0.42] |
The chance (percentage) for each hour to have a connection on a weekday |
| hourly_weekend_distribution | [0.09,0.01,0.01,0.08,0.98,3.56,5.23,5.79,5.82,5.89,5.84,5.91,5.88,5.95,5.87,5.95,5.89,5.96,5.92,5.94,5.62,4.61,2.45,0.76] |
The chance (percentage) for each hour to have a connection on a weekend day |
The PoDiGG generator is written by Ruben Taelman.
This code is copyrighted by Ghent University – imec and released under the MIT license.