ASRLab develops and shares research datasets supporting work in autonomous systems, software testing, artificial intelligence, robotic perception, verification and validation, and digital twin engineering. Our datasets are intended to support reproducible research, benchmarking, experimentation, and the development and evaluation of intelligent systems.
Multi-Robot HIL & Digital Twin Dataset
Experimental data for multi-robot Hardware-in-the-Loop and Digital Twin research
This open dataset contains experimental data collected from synchronized physical and simulated robotic systems. It supports research on Digital Twins, trajectory analysis, robotic inspection, anomaly detection, and system verification.
The dataset contains 9.97 GB of experimental data, including:
- Processed robotic data in CSV format
- Reference and executed robot trajectories
- Camera recordings and experiment screencasts
- Real and simulated point clouds
- Robot pose information
- Spatial benchmarking data
The accompanying ROS 2 workspace provides source code, reference plans, and execution trajectories to support reproducibility of the experiments.
Dataset:
https://data.mendeley.com/datasets/xsnpgfz5mb/3
Source Code:
https://github.com/ESOGU-SRLAB/ESOGU-HILTest-DualRobot
ChArIoT Dataset
Dataset for AI-assisted software testing and mutation research
The ChArIoT Dataset supports research on software defect analysis, mutation-based testing, automated program repair, and AI-assisted software engineering.
The dataset was constructed from buggy and fixed source-code pairs and organises software changes according to three main modification types:
- Update
- Delete
- Insert
It includes method-level buggy and fixed code samples, edit actions, abstracted source-code representations, and formatted and unformatted dataset variants.
These resources can be used for machine learning experiments involving code transformation, defect prediction, mutation generation, and automated software testing.
Dataset:
https://tinyurl.com/chariot-dataset
Faulty Image Dataset
Fault-injected image data for robotic perception and verification
The Faulty Image Dataset was created to support the evaluation of camera-based robotic perception systems under abnormal operating conditions.
The dataset contains 10,000 images, consisting of:
- 5,000 normal images
- 5,000 fault-injected images
The fault-injected samples represent different types of camera and image degradation, including:
- Dilation
- Erosion
- Morphological opening
- Morphological closing
- Gradient effects
- Motion blur
- Partial image loss
The dataset can support research on anomaly detection, fault detection, computer vision robustness, AI quality evaluation, and verification of camera-based autonomous systems.
The faulty samples were generated using CamFITool, ASRLab's camera fault injection tool.
Dataset:
https://drive.google.com/drive/folders/1pIqEnIIKFN13Z4m38zEB-XYQp_qb0T6S
Related Research:
https://arxiv.org/abs/2108.13803
Open Research Resources
ASRLab aims to make research outputs reusable whenever possible by sharing datasets, software, experimental configurations, and supporting source code.
Additional open-source software developed through ASRLab research activities is available on our Tools page and GitHub organisation.
ASRLab GitHub:
https://github.com/ESOGU-SRLab