smif: Simulation Modelling Integration Framework

smif is a framework for handling the creation, management and running of system-of-systems models.

A system-of-systems model is a collection of system simulation models that are coupled through dependencies on data produced by each other.

smif provides a user with the ability to

  • create system-of-systems models

    • add simulation models to a system-of-systems model
    • create dependencies between models by linking model inputs and outputs
    • pick from a library of data adapters which perform common data conversions across dependencies
    • create user-defined data adapters for more special cases
    • add scenario data sources and link those to model inputs within a system-of-systems
  • add a simulation model to a library of models

    • write a simulation model wrapper which allows smif to run the model
    • define multi-dimensional model inputs, outputs and parameters and appropriate metadata
  • run system-of-systems models

    • link concrete scenario data sets to a system-of-systems model
    • define one or more decision modules that operate across the system-of-systems
    • define a narrative to parameterise the contained models
    • persist intermediate data for each model output, and write results to a data store for subsequent analysis

In summary, the framework facilitates the hard coupling of complex systems models into a system-of-systems.

Should I use smif?

There are number of practical limits imposed by the implementation of smif. These are a result of a conscious design decision that stems from the requirements of coupling the infrastructure system models to create the next generation National Infrastructure System Model (NISMOD2).

The discussion below may help you determine whether smif is an appropriate tool for you.

  • smif is not a scheduler, but has been designed to make performing system-of-systems analyses with a scheduler easier

  • Geographical extent is expected to be defined explicitly by a vector geometry

    • smif is not optimised for models which simulate on a grid, though they can be accomodated
    • smif is designed for models that read and write spatial data defined over irregular grids or polygons using any spatial format readable by fiona
  • Inputs and outputs are exchanged at the ‘planning timestep’ resolution

    • smif makes a distinction between simulation of operation, which happens at a model-defined timestep resolution, and application of planning decisions which happens at a timestep which is synchronised between all models
    • smif is not focussed on tight coupling between models which need to exchange data at every simulation timestep (running in lockstep)
    • smif does accomodate individual models with different spatial and temporal (and other dimensional) resolutions, by providing data adaptors to convert from one resolution to another
  • smif has been designed to support the coupling of bottom-up, engineering simulation models built to simulate the operation of a given infrastructure system

    • smif provides a mechanism for passing information from the system-of-systems level (at planning timesteps scale) to the contained models
    • smif is appropriate for coupling large complex models that exchange resources and information at relatively course timesteps
  • smif is not appropriate for

    • discrete event system simulation models (e.g. queuing systems)
    • dynamical system models (e.g. predator/prey)
    • equilibrium models without explicit timesteps (e.g. Land-Use Transport Interaction)
    • for simulating 100s of small actor-scale entities within a system-level environment

Installation and Configuration

smif is written in Python (Python>=3.5) and has a number of dependencies. See requirements.txt for a full list.

Using conda

The recommended installation method is to use conda, which handles packages and virtual environments, along with the conda-forge channel which has a host of pre-built libraries and packages.

Create a conda environment::

conda create --name smif_env python=3.6

Activate it (run each time you switch projects)::

conda activate smif_env

Add the conda-forge channel, which has smif available::

conda config --add channels conda-forge

Finally install smif::

conda install smif

Installing smif with other methods

Once the dependencies are installed on your system, a normal installation of smif can be achieved using pip on the command line::

    pip install smif

Versions under development can be installed from github using pip too::

    pip install git+http://github.com/nismod/smif

To install from the source code in development mode::

    git clone http://github.com/nismod/smif
    cd smif
    python setup.py develop

Spatial libraries

smif optionally depends on fiona and shapely, which depend on the GDAL and GEOS libraries. These add support for reading and writing common spatial file formats and for spatial data conversions.

If not using conda, on Mac or Linux these can be installed with your OS package manager::

# On debian/Ubuntu:
apt-get install gdal-bin libspatialindex-dev libgeos-dev

# or on Mac
brew install gdal
brew install spatialindex
brew install geos

Then to install the python packages, run::

pip install smif[spatial]

Running smif from the command line

Follow the getting started guide to help set up the necessary configuration.

To set up an sample project in the current directory, run::

    $ smif setup

To list available model runs::

    $ smif list
    demo_model_run
    ...

To start the smif app, a user-interface that helps to display, create and edit a configuration, run::

    $ smif app

To run a system-of-systems model run::

    $ smif run demo_model_run
    ...
    Model run complete

By default, results will be stored in a results directory, grouped by model run and simulation model.

To see all options and flags::

    $ smif --help
    usage: smif [-h] [-V] {setup,list,run} ...

    Command line tools for smif

    positional arguments:
    {setup,list,app,run}  available commands
        setup               Setup the project folder
        list                List available model runs
        app                 Open smif app
        run                 Run a model

    optional arguments:
    -h, --help        show this help message and exit
    -V, --version     show the current version of smif

Citation

If you use smif for research, please cite the software directly:

Here’s an example BibTeX entry::

    @misc{smif,
          author       = {Will Usher and Tom Russell and Roald Schoenmakers},
          title        = {nismod/smif vX.Y.Z},
          month        = Aug,
          year         = 2018,
          doi          = {10.5281/zenodo.1309336},
          url          = {https://doi.org/10.5281/zenodo.1309336}
    }

A word from our sponsors

smif was written and developed at the Environmental Change Institute, University of Oxford within the EPSRC sponsored MISTRAL programme, as part of the Infrastructure Transition Research Consortium.

Click links for code, data, and publications associated with this project.
Will Usher
Will Usher
KTH Royal Institute of Technology
Tom Russell
Tom Russell
Senior Research Software Engineer

Tom develops software for infrastructure planning, adaptation and risk assessment.