pymsmile / genet

Manipulate MATSim networks via a Python API.

Network Scenario Generator (GeNet)

This package provides tools to represent and work with a multi-modal transport network with public transport (PT) services. It is based on MATSim's representation of such networks. The underlying network available to PT services (roads, railways, but also ferry/flight connections) uses a networkx.MultiDiGraph with additional methods for 'links' which are unique in genet.Network (networkx.MultiDiGraph accepts multiple edges between the same from and to node pair; referring to an edge in networkx.MultiDiGraph and genet.Network has the same effects, i.e. the result is a dictionary indexed by the multi edge index). The PT services are represented through genet.Schedule class which relies on other genet classes: the Schedule relies on a list of genet.Service's, which in turns consists of a list of genet.Route's. Each Route class object has an attribute stops which consists of genet.Stops objects. The Stops carry spatial information for the PT stop.

The goal of GeNet is to:

• Provide a formalised in-memory data structure for representing a multi-modal network with a PT service
• Enable using the data structure for tasks such as generating auxiliary MATSim files e.g. Road Pricing
• Simplify the process of modifying a network and provide simple change log to track the differences between the input and output networks.
• Provide validation methods to check for simple errors such as: whether a Route has more than one Stop or that the underlying graph doesn't have any dead-ends or sources (a place which you can leave but cannot get back to).

Setup

To run pre-baked scripts that use genet in a number of different scenarios you can use docker:

Using Docker

Build the image

docker build -t "genet" .

Running a container with a pre-baked script

docker run genet reproject_network.py -h

usage: reproject_network.py [-h] -n NETWORK [-s SCHEDULE] [-v VEHICLES] -cp
                            CURRENT_PROJECTION -np NEW_PROJECTION
                            [-p PROCESSES] -od OUTPUT_DIR

Reproject a MATSim network

optional arguments:
  -h, --help            show this help message and exit
  -n NETWORK, --network NETWORK
                        Location of the network.xml file
  -s SCHEDULE, --schedule SCHEDULE
                        Location of the schedule.xml file
  -v VEHICLES, --vehicles VEHCILES
  							  Location of the vehicles.xml file
  -cp CURRENT_PROJECTION, --current_projection CURRENT_PROJECTION
                        The projection network is currently in, eg.
                        "epsg:27700"
  -np NEW_PROJECTION, --new_projection NEW_PROJECTION
                        The projection desired, eg. "epsg:27700"
  -p PROCESSES, --processes PROCESSES
                        The number of processes to split computation across
  -od OUTPUT_DIR, --output_dir OUTPUT_DIR
                        Output directory for the reprojected network

Otherwise, you can install genet as a python package, in your base installation of python or a virtual environment. Run the pre-baked scripts, write your own scripts or use IPython shell or Jupyter Notebook to load up a network, inspect or change it and save it out to file. Check out the wiki pages and example jupyter notebooks for usage examples.

Install dev prereqs (use equivalent linux or windows package management)

brew install python3.7
brew install virtualenv

Installation

Create and activate virtual environment

virtualenv -p python3.7 venv
source venv/bin/activate

before installing dependencies you may need to install a dependency of rtree: libspatialindex-dev, the command for linux:

sudo apt-get install -y libspatialindex-dev

Finally install GeNet's Python dependencies

pip install -e .

Install Kepler dependencies

Please follow kepler's installation instructions to be able to use the visualisation methods. To see the maps in a jupyter notebook, make sure you enable widgets.

jupyter nbextension enable --py widgetsnbextension

Install Mathematical Solver

Note: The default CBC solver is installed inside GeNet's Docker image.

To use methods which snap public transit to the graph, you will also need to install a mathematical solver. Methods default to CBC, an open source solver which can be found here. Another good open source choice is GLPK. The solver needs to support MILP - mixed integer linear programming.

Unit Testing

Run the unit tests (from root dir)

python -m pytest -vv tests

Generate a unit test code coverage report

To generate XML & HTML coverage reports to reports/coverage:

./bash_scripts/code-coverage.sh