AURSAD: Universal Robot Screwdriving Anomaly Detection Dataset

Abstract

Dataset containing normal and anomalous operation measurments from UR-3e robot and OnRobot Screwdriver. The dataset contains 2045 samples in total. The robot was sampled with frequency of 100 Hz. <pre>| Type | Label | Samples | % | |--------------------------|-------|---------|-------| | Normal operation | 0 | 1420 | 69.44 | | Damaged screw | 1 | 221 | 10.81 | | Extra assembly component | 2 | 183 | 8.95 | | Missing screw | 3 | 218 | 10.65 | | Damaged thread samples | 4 | 3 | 0.15 | </pre> Additionally, there are 2049 supplementary samples describing the loosening/screw picking motion, labeled 5. The data is in continuous form with every event having its own label. Events are grouped in samples. For more detail please look into the technical report: https://arxiv.org/abs/2102.01409. The dataset has an accompanying Python library for easier data loading and manipulation: https://pypi.org/project/aursad/.