eqasim-java

This repository contains all the Java (and MATSim) related parts of eqasim. Especially, it contains the standardized code to run the simulation and to cut smaller regions from the reference scenarios.

The general workflow to run a reference scenario is as follows:

• Generate a scenario (using the respective scenario pipeline)
• Run the simulation (using this repository)

Currently, there are two scenario pipelines: Switzerland and Île-de-France.

Once a reference scenario exists (and has been run for a sufficient number of iterations, we recommend 40) it is possible to cut out a smaller region. The workflow for that is as follows:

• Run the reference scenario for 40 iterations
• Cut the smaller scenario based on the output population
• Run the smaller simulation

All of these steps can be performed with this repository. A more detailed guide through the steps outlined above is given in the following sections.

How to generate a reference scenario

Currently, there are two pipelines that generate reference scenarios. One is for Switzerland and Île-de-France. At the moment, each of them has a separate repository and a lot of code duplication is going on, which is subject to be improved soon. Nevertheless, the pipelines can be run in almost the same way.

The links to the pipeline repositories are:

• Switzerland: https://gitlab.ethz.ch/ivt-vpl/populations/ch-zh-synpop
• Île-de-France: https://gitlab.ethz.ch/ivt-vpl/populations/idf-paris-synpop

Please follow the instructions from the Switzerland pipeline in both cases. The first step is to set up an environment with Java, Maven, Python, etc. and then to run the scenario synthesis. In each case, you need a pipeline configuration, which should looks as follows.

For Switzerland, such a configuration should be sufficient:

# switzerland.yml

raw_data_path: /nas/ivtmatsim/scenarios/switzerland/data
target_path: /path/to/my/custom/cache/directory

output_path: /path/to/my/custom/output/directory
output_id: switzerland_10pct

input_downsampling: 0.1
use_freight: true

java_memory: 300G
threads: 24
hot_deck_matching_runners: 24

stages:
    - matsim.final

By running python3 run.py switzerland.yml the pipeline should create a new folder called switzerland_10pct in the specified output folder after processing all the data. Note that the paths need to be adjusted accordingly. Also, this example assumes the standard IVT path for the raw input data. In another set up this may need to be changed. Using the output_id the name of the output folder can be specified and the input_downsampling setting defines the sample size (in this case, 10% was chosen). Note that it makes sense to remove all contents of the cache directory if you attempt to create a new sample size, because several stages in the process would need to be generated.

For Île-de-France the process is similar, but the settings file looks slightly different. A template is this one:

# ile_de_france.yml

raw_data_path: /nas/ivtmatsim/scenarios/paris/data

target_path: /path/to/my/custom/cache/directory
output_path: /path/to/my/custom/output/directory
output_id: ile_de_france_10pct

input_downsampling: 0.1
hts: egt

java_memory: 300G
threads: 24
hdm_threads: 24

stages:
    - matsim.final

Again, python3 run.py ile_de_france.yml should generate the scenario.

Running a simulation

Suppose there is a freshly generated scenario. The only thing that is necessary to run the scenario is to use the RunSimulation script in this repository.

The script can either be started directly in Eclipse/IntelliJ, or from a Maven build. To create a Maven build of the repository, you first have to clone it:

git clone https://github.com/eqasim-org/eqasim-java.git

Then you have to enter the directory and package it:

cd eqasim-java
mvn -Pstandalone package

It is important to keep the -standalone flag, because it creates a full jar with all the libraries needed. The eqasim Maven project is a multi-module project. All general purpose classes are located in org.eqasim.core. Additionally, there are two scenario-specific modules called org.eqasim.switzerland and org.eqasim.ile_de_france. For each of them a separate jar will be created. They are located at:

eqasim-java/switzerland/target/switzerland-1.0.0.jar
eqasim-java/ile_de_france/target/ile_de_france-1.0.0.jar

Both of the scenario-specific modules contain the aforementioned RunSimulation class for their respective use cases. Hence, to run the simulation from the command line, one needs to call:

java -Xmx100G -cp /path/to/switzerland-1.0.0.jar org.eqasim.switzerland.RunSimulation --config-path /path/to/switzerland_config.xml

or

java -Xmx100G -cp /path/to/ile_de_france-1.0.0.jar org.eqasim.ile_de_france.RunSimulation --config-path /path/to/ile_de_france_config.xml

This should start the simulation according to what is given in the respective config file.

Cutting a scenario

In order to cut a scenario, it is possible to call the respective script from Eclipse/IntelliJ, or from the command line. In the latter case, please follow the instructions from the section above to package a jar file.

The relevant run script is then org.eqasim.scenario.cutter.RunScenarioCutter. Note that either the switzerland or ile_de_france jar can be used. Since the script is located in the core package, there is no scenario-specific functionality in the cutter. We simply don't create an additional jar for the core package since all the functionality is automatically included in the scenario-specific jars.

The script expects a number of arguments:

java -Xmx100G -cp /path/to/switzerland-1.0.0.jar org.eqasim.core.scenario.cutter.RunScenarioCutter
--config-path /path/to/ile_de_france_config.xml
--output-path /path/to/output/paris_10pct
--extent-path /path/to/paris_shapefile.shp
--prefix paris_
--threads 24

First, the path to the reference scenario must be given by providing its config file. Second, an output path needs to be specified. Third, the path to a shape file must be provided. This shape file determines the extent of the cut region. Make sure that it is a compact shape without holes. The shape file should only contain a single polygon feature. Finally, a prefix can be specified that defines according to which pattern files will be written to the output folder. In this case, we would have a paris_population.xml.gz, paris_facilities.xml.gz, etc.

The generated scenario can then be run. For that, you can use the same scripts as for the reference sceanrio. If you cut a Zurich scenario from Switzerland, the switzerland module can still be used to run the scenario. The same goes for the Paris / Ile-de-France case.

To get started, the gis/ folder in this repository contains a number of common shape files:

• zurich_city.shp The 12 districts of Zurich city
• zurich.shp Zurich city and several high density areas in the surroundings
• zurich_20km.shp Zurich city and surroundings, plus a 20km buffer
• paris.shp City area of Paris (20 arrondissements)
• paris_20km.shp City area of Paris, plus a 20km buffer

Note that the shape files must be given in the correct projection. For Switzerland this is EPSG:2056 and for France it is EPSG:2154.