This library provides a simulator of transportation system. It is intended to be used for experimenting with different ways of designing systems-of-systems (SoS). Through the simulation, the effects of different design decisions can be analyzed.
The package can be used either in interactive mode or in batch mode.
In interactive mode, the simulator runs in the web browser using pyscript. The browser fetches the necessary files from a web server.
To run the interactive mode locally run python dev_server.py.
This starts a local web server from which the files can be fetched to a browser running on the same machine.
This server is primarily intended for development, and therefore also makes sure that generated files are updated, invoking build steps if necessary.
Then, open the link http://127.0.0.1:8000/app/sossim.html in the browser.
When files have changed, simply reload the page from the browser.
A Dockerfile is provided for running the software without installing Python or any libraries.
To use, first build a Docker container using docker build -t sossim ..
Then, run the interactive mode using docker run --name sossim -d -p 8000:8000 sossim.
To stop the container, use docker kill sossim followed by docker rm sossim to remove the container.
In batch mode, the simulator can be run locally from the command line.
First, ensure that all required libraries are installed using pip install -r requirements.txt.
Then, run the command python -m sossim.
This command can take various command line options, which can be enlisted using python -m sossim -h.
The latest commit to the Github respository is automatically published on Github pages. It can be accessed as a live demonstration. This does not require any local installations whatsoever.
In this section, some more details are provided on how the software is implemented.
The core of the simulation is an agent-based model developed using the Mesa library. The Mesa library has some dependencies that are not possible to use in the browser. Therefore, a copy of the core modules of Mesa has been included in this repository.
The simulation takes place on a grid map onto which a randomly generated road network is placed. The road network is implemented as a directed graph using the NetworkX library.
The interactive simulation is designed using the model-view-controller pattern.
The interface components are created by manipulating the DOM elements of a web page, and the graphics are created in SVG.
The DOM manipulation is done using the domed package, which provides a Pythonic approach inspired by the Dominate library.
However, domed works directly on the DOM elements since it executes in the browser, whereas Dominate generates HTML as text and is intended for server side usage.
For development purposes, there is also a Python REPL accessible in the user interface, through the View menu.
Through this, it is possible to access the user interface object using the variable ui.
This in turn gives access to all parts of the system.
As an example, to get the list of all agents, evaluate the expression ui.model.agents().
The models and simulation have a large number of different configuration parameters. This are accessible both as command line arguments when running in batch mode, and as an interactive dialogue in the user interface. The configurations can also be saved and restored, in order to recreate a certain simulation. To handle the configurations in a smooth way, a Configuration class is provided which gathers all parameters. In that way, a single configuration object can be passed around to various functions, rather than individual specific parameters. Special syntax is provided to make the specification of parameters relevant to a certain class as readable as possible.
For development tasks, there is a utility called build.py that automates recurring development tasks, such as type checking, generating documentation, etc.
Its features can be checked using python build.py --help.
During the development of the interactive browser app, the development server should be used.
It is invoked by python dev_server.py.
It serves the app files to the browser for testing.
It also ensures that generated app files are updated before being served, invoking necessary build steps.
To install the dependencies required for using development utilities, run pip install -r requirements_development.txt.
Some documentation is available to browse and search. Those modules that contain non-trivial class structures also come with partial UML class diagrams. These illustrate some of the key relations and attributes of the classes.