Datasets

EngineAD

EngineAD is a real-world vehicle engine anomaly detection dataset designed for safety-critical transportation applications.

  • Collected from a fleet of 25 commercial vehicles over 6 months.
  • Includes high-resolution telemetry sampled at approximately 1-second intervals from engine-related sensors.
  • Covers 13 raw engine signals (e.g., pressure, temperature, fuel-rate, rotational-speed related channels).
  • Includes expert annotations distinguishing normal operation from early signs of incipient faults.
  • Provides processed segment-level data commonly used for benchmarking (including principal-component-based representations).

The EngineAD dataset is now publicly available — no access request is required. To download it, please visit the dataset page below.

Dataset page: BorealisData - EngineAD (DOI:10.5683/SP3/TX13P1)

Related papers:

Citation policy: If you use EngineAD, please cite both papers below.

@inproceedings{hojjati2026enginead,
  title={EngineAD: A Real-World Vehicle Engine Anomaly Detection
    Dataset},
  author={Hojjati, Hadi and Roth, Christopher and Woods, Rory and
    Sills, Ken and Armanfard, Narges},
  editor={Armanfard, Narges and Hojjati, Hadi and Ho, Thi Kieu
    Khanh},
  booktitle={Automated Spatial and Temporal Anomaly Detection:
    Third International Workshop, ASTAD@AAAI 2026, Held in
    Conjunction with AAAI 2026, Singapore, January 26, 2026,
    Proceedings},
  series={Communications in Computer and Information Science},
  year={2026},
  publisher={Springer Singapore},
  address={Singapore}
}

@inproceedings{hojjati2023multivariate,
  title={Multivariate Time-Series Anomaly Detection with Temporal
    Self-supervision and Graphs: Application to Vehicle Failure
    Prediction},
  author={Hojjati, Hadi and Sadeghi, Mohammadreza and Armanfard,
    Narges},
  booktitle={Machine Learning and Knowledge Discovery in
    Databases: Applied Data Science and Demo Track (ECML PKDD)},
  series={Lecture Notes in Computer Science},
  volume={14175},
  pages={239--254},
  year={2023},
  publisher={Springer},
  doi={10.1007/978-3-031-43430-3_15}
}

MultiTab — Synthetic Multitask Data

Generate synthetic multitask regression datasets with controllable task correlations, polynomial complexity, and noise levels.

Generator: Open the MultiTab generator →

Related paper:

Citation policy: If you use MultiTab or the synthetic generator for experiments, please cite the paper below.

@inproceedings{sinodinos2026multitab,
  title={MultiTab: A Scalable Foundation for Multitask Learning
    on Tabular Data},
  author={Sinodinos, Dimitrios and Wei, Jack Yi and Armanfard,
    Narges},
  booktitle={Proceedings of the AAAI Conference on Artificial
    Intelligence},
  volume={40},
  number={30},
  pages={25499--25507},
  year={2026}
}

Upcoming datasets

CARLA-Collide

Description to be added.

Link: TBD

Real-Collide

Description to be added.

Link: TBD