SensoDat: Simulation-based Sensor Dataset of Self-driving Cars

Jupyter Notebook Code: https://huggingface.co/datasets/it4lia/sensodat/blob/main/analysis.ipynb

Latest Release on Zendodo:

Original Paper Artifact:

SensODAT: Sensor Data of Autonomous Driving Tests

SensoDat is a dataset of self-driving car simulation data (32,580 executed simulations across 14 campaigns). The simulations were generated using state-of-the-art test generators (Frenetic, Frenetic-V, and AmbieGen) and executed in the BeamNG simulation environment.

Dataset Overview

Concretely, the dataset contains:

• Simulation description data in ASAM OpenDRIVE format
• Trajectory logs for each execution
• Sensor data as time series of 81 simulated sensors/properties
• Execution metadata (configuration, duration, validity, and predicted outcomes)
• PASS/FAIL results based on the OOB (Out-of-Bounds) safety metric

The recorded sensor streams include vehicle dynamics and control signals such as RPM, wheel speed, throttle and brake input, brake temperatures, steering signals, transmission states, ABS/ESC activity, and many more.

Scale & Storage

• 32,580 total simulation executions
• 81 sensors per test case
• 3.34 GB of structured data
• Stored in MongoDB for efficient querying and large-scale analysis

All data was collected from high-fidelity soft-body physics simulations, enabling realistic vehicle dynamics representation without requiring researchers to re-execute computationally expensive simulation campaigns.

Research Applications

SensODAT supports research in:

• AI and machine learning for autonomous systems
• Regression testing and test prioritization
• Simulation flakiness analysis
• Safety validation and fault detection
• Reproducibility and benchmarking in simulation-based SDC testing

Reference

• Original paper: https://dl.acm.org/doi/10.1145/3643991.3644891

Associated Paper

@inproceedings{sensodat,
  author       = {Christian Birchler and
                  Cyrill Rohrbach and
                  Timo Kehrer and
                  Sebastiano Panichella},
  title        = {SensoDat: Simulation-based Sensor Dataset of Self-driving Cars},
  booktitle    = {21th {IEEE/ACM} International Conference on Mining Software Repositories,
                  {MSR} 2024, Lisbon, Portugal, April 15-16, 2024},
  year         = {2024},
  doi          = {to appear},
}

@article{sensodat-preprint,
  author       = {Christian Birchler and
                  Cyrill Rohrbach and
                  Timo Kehrer and
                  Sebastiano Panichella},
  title        = {SensoDat: Simulation-based Sensor Dataset of Self-driving Cars},
  journal      = {CoRR},
  volume       = {abs/2401.09808},
  year         = {2024},
  url          = {https://doi.org/10.48550/arXiv.2401.09808},
  doi          = {10.48550/ARXIV.2401.09808},
  eprinttype    = {arXiv},
  eprint       = {2401.09808},
}

Requirements

You need to have Docker installed and running.

NOTE: The following instructions were tested with Windows and Linux with 32GB RAM.

Automatic setup

To set up a MongoDB with the SDC simulation data, ensure Docker is up and running. Then simply execute the setup.sh (Linux/MacOS) or setup.bat (Windows) script.

Automatic clean up

To tear down the database simply execute the cleanup.sh (Linux/MacOS) or cleanup.bat (Windows) script.

Manual setup

Run the following commands in the exact order to setup the database:

Start an uploader and a mongo container:

docker compose -f ./environment/docker-compose.yml up -d --build

Verify the containers are up and running:

docker ps

Copy the data to the uploader container:

docker cp ./data uploader:/app/data

Unzip the data in the uploader container:

docker exec uploader unzip ./data/data.zip -d ./data

Copy the code to upload the data to the mongo container:

docker cp ./code uploader:/app/code

Upload the data to the mongo container:

docker exec -it uploader python ./code/fill_mongodb.py

To tear down the database when you don't need it anymore:

docker compose -f ./environment/docker-compose.yml up down

License

    SensoDat: Simulation-based Sensor Dataset of Self-driving Cars
    Copyright (C) 2024  Christian Birchler & Sebastiano Panichella (AI4I - The Italian Institute of Artificial Intelligence for Industry, sebastiano.panichella@ai4i.it)

    This program is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 3 of the License, or
    (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program.  If not, see <https://www.gnu.org/licenses/>.