Traffic Scenarios Generation

Welcome to the project page!
This site documents how to run experiments for generating traffic scenarios.

🚧 Page Under Construction – More content coming soon! 🚧

Script Usage

This repository contains a script for running traffic scenario generation experiments with several configurable parameters.

Command-Line Arguments

  1. num_case (type: s)
    • Description: The case number for the experiment. You can find the different experiment cases in NeuroExperiment.py.
    • Example: --num_case 1

    For more details about the available experiment cases, see the

    num_case documentation

  2. experiment_name_suffix (type: string)
    • Description: Suffix to add to the experiment name for identification. When performing multiple repetitions of an experiment, this string will indicate the folder suffix, followed by the seed number used .
    • Example: --experiment_name_suffix METR16_experiment
  3. main_folder (type: string)
    • Description: The folder path where experiment results will be saved.
    • Example: --main_folder "experiments"
  4. repeat (type: int)
    • Description: The number of times to repeat the experiment with several random seeds.
    • Example: --repeat 5
  5. optimization (type: yes/no)
    • Description: Whether to perform BO hyperparameters optimization.
    • Example: --optimization yes
  6. load_model (type: yes/no)
    • Description: Whether to load a pre-trained model.
    • Example: --load_model yes
  7. train_models (type: yes/no)
    • Description: Whether to train the model again.
    • Example: --train_models yes

      Studies Cases

For more details about the available experiment cases, see the num_case documentation.

Example Usage

To run the script with specific arguments, use the following command format:

python script_name.py --num_case <num_case> --experiment_name_suffix <experiment_name_suffix> --main_folder <main_folder> --repeat <repeat> --optimization <optimization> --load_model <load_model> --train_models <train_models>

Example

python3 test.py --neuroD --num_case 1 --experiment_name_suffix 2024_07_10_METR_16 --main_folder 2024_07_10_METR_16__OPT_split --repeation 5 --optimization yes --load_model no --train_models yes

Help

To see the full list of available options, run:

python3 test.py --h

Publications and Conferences

Carbonera, M., Ciavotta, M., Messina, E. (2024). Variational Autoencoders and Generative Adversarial Networks for Multivariate Scenario Generation. DATA SCIENCE FOR TRANSPORTATION, 6(3) [10.1007/s42421-024-00097-y].
Paper 2024 - DATA SCIENCE FOR TRANSPORTATION
Carbonera, M., Ciavotta, M., Messina, V. (2023). Driving into Uncertainty: An Adversarial Generative Approach for Multivariate Scenario Generation. In Proceedings - 2023 IEEE International Conference on Big Data, BigData 2023 (pp. 2578-2587). IEEE. [10.1109/BigData59044.2023.10386128].
Paper 2023 - IEEE Big Data Conference

Dataset, code and contacts

The code to run is available at:
https://github.com/mikeleikele/TransportScenariosGeneration

Due to the large size of the dataset, it is not currently possible to host it on this page.
If you are interested in accessing the dataset, please contact us at michele.carbonera@unimib.it.

If you have any questions, please contact us at:
michele.carbonera@unimib.it

If you make use of this work, we kindly ask you to cite our related publications.

Acknowledgements

This work has been funded by:

MOST Logo MOST - National Sustainable Mobility Center, part of the European Union’s NextGenerationEU project (MOST - National Sustainable Mobility Center CN00000023, Italian Ministry of University and Research Decree No. 1033-17/06/2022, Spoke 8).
MOST Logo ULTRA OPTYMAL - Urban Logistics and sustainable TRAnsportation: OPtimization under uncertainty and MAchine Learning PRIN2020 project funded by the Italian University and Research Ministry (grant number 20207C8T9M).