The suq-controller is a Python package for to run simulation studies in crownet. The sampling is automated and can be parallelized. All results are returned in a convenient format (pandas DataFrame).

The main motivation is to provide a tool to build Surrogate (S) or Uncertainty Quantification (UQ) models.

Note, that the project is a work in progress. Features are only implemented when required and we welcome contributions (feedback, code, documentation, tests, ...). Please interact through issues.

Use Python's package mangement tool and install with

pip3 install --upgrade pip
python3 -m pip install git+https://gitlab.lrz.de/vadere/suq-controller.git@master

Test if the installed package was installed successfully by running:

python3 -c "import suqc; print(suqc.__version__)"

This command should print the installed version number (and potentially infos to set up required folder) in the terminal.

Requirements:

• Python>=3.6
• Other packages listed in requirements.txt .

Run the code from the source directly (without install), please check if you meet the requirements (see requirements.txt file). You can also run pip3 install -r /path/to/requirements.txt. Make also sure that the Python paths are set correctly (possibly add with sys).

See [SRC_PATH]/tutorial

Please only contribute (report bugs, code, ideas, etc.) in the original repository.

https://gitlab.lrz.de/vadere/suq-controller/

To update the suq-controller with the current master run the following git command from Vadere's root source code path:

git subtree pull --prefix Tools/SUQController [email protected]:vadere/suq-controller.git master --squash

A few hints for your Vadere .scenario files:

1. ScenarioChecker
   Before running your scenario automatically on suqc, activate the ScenarioChecker (Project > Activate ScenarioChecker) and run it in the VadereGui. The ScenarioChecker will point out potential problems with your scenario file.
2. Default parameters
   Make sure to set realTimeSimTimeRatio to 0.0. (Values > 0.0 slow down the simulation for the visualisation)
   Another point that may cost a lot of computation time is the optimizationType, consider using DISCRETE (discrete optimization) instead of NELDER_MEAD. Please note, that varyStepDirection should always be activated with discrete optimization.
   Remove attributesCar from the .scenario file if you are not using any vehicles to avoid confusion of attributes.
3. Visual check
   Visually check the results of your simulation, maybe check upper and lower parameter bounds.
4. Clean topography
   Remove elements in your topography that are not used. Sometimes through the interaction with the mouse, tiny obstacles or targets are created unintentionally. Check the elements in your topography, you can take a look at the ElementTree in the Topography creator tab. Remove all elements that are unused, especially focusing on targets.
5. Data processors
   Remove all data processors and output files that you don't use. In particular, remove the overlap processors, they are intended for testing purposes.
6. Reproducibility
   Make sure that your runs are reproducible - work with a fixedSeed by activating useFixedSeed or save all the simulationSeeds that have been used. (Another way is to provide a fixedSeed for each runs with suqc, in this case make sure that useFixedSeed is true.)