Prediction Equilibrium for Dynamic Network Flows

Getting Started

To start working:

• Install Python 3.11 (if not already)
• Install poetry (if not already)
• Run poetry config virtualenvs.in-project true
• Depending on your system, run poetry env use python3.11
• Run poetry install
• Run poetry shell to activate the created virtual environment

Disable Assertions

Assertions are used extensively throughout the code and therefore slow down the computation significantly. Often, we use the environment variable PYTHONOPTIMIZE to deactivate assert statements. To disable assertions in your current terminal session, run

• for Bash: export PYTHONOPTIMIZE=TRUE
• for PowerShell: $Env:PYTHONOPTIMIZE="TRUE"
• for cmd.exe: set PYTHONOPTIMIZE=TRUE

Set PYTHONPATH variable

As the code is split into multiple modules, the PYTHONPATH environment variable declares where Python should look for the files. Use the following command to set the variable:

• for Bash: export PYTHONPATH=./src
• for PowerShell: $Env:PYTHONPATH=".\src"
• for cmd.exe: set PYTHONPATH=".\src"

Clone the TransportationNetworks repository

Please clone the TransportationNetworks repository from github.com/bstabler/TransportationNetworks locally. By default, we assume you cloned the folder to ~/git/TransportationNetworks, where ~ is your home directory. However, you can also specify a custom directory by setting the TNPATH environment variable.

Running an experiment

For each experiment described in the manuscript, there is a Python file in the src/scenarios folder. You can run an experiment by simply executing the corresponding file, e.g. python src/scenarios/sample_scenario.py.

Upon execution of an experiment, the following steps are executed:

1. Generation of sample flows that use the constant predictor only
2. Generation of training data for the ML predictors (extracting samples of queues and edge loads)
3. Training the ML predictors
4. Evaluating all predictors by using them side by side in 20 simulation runs

As the computation of most steps use only a single core, you can speed up the computation by running the experiment command multiple times in parallel (e.g. number of cores of your processor). The processes will communicate with each other and only perform tasks that have not been taken by other processes. All of these steps save their output to the ./out folder. Once all steps are executed, the results of step 4 (e.g. tikz plots, visualization files, etc.) are generated into the out/{experiment}/eval folder. You can open files ending in .vis.json in a dynamic flow visualization tool available at https://arbeitsgruppetobiasharks.github.io/dynamic-flow-visualization.

Development

Formatting

We use black as our Python code formatter together with isort to sort imports. In a poetry shell, run

black src
isort src --profile black

To format files automatically on commit, install the pre-commit hook:

ln -s ../../pre-commit.sh .git/hooks/pre-commit

Linting

We use pylint to lint our code. However, because there is currently a lot of noise, we only report errors:

pylint src --errors-only