Quanser Aero 2 Pick-and-Place Dataset (QAPPD)

This suite contains ten independent multivariate time series datasets designed for federated and hierarchical federated anomaly detection in repetitive cyber-physical processes, though equally suitable for centralized and other distributed learning settings. Data was recorded on the Quanser Aero 2, a two-degree-of-freedom helicopter testbed, configured to execute repeating trajectories that emulate the cyclic dynamics of robotic pick-and-place operations. The ten datasets are deliberately independent, allowing each to represent a separate client, site, or edge node in a federated setup. In hierarchical federated learning, they can be partitioned into groups to model multi-tier topologies (e.g., devices → cell controllers → factory aggregator). Each dataset is provided as a train/test pair with point-wise anomaly labels for the test split: training data contains only nominal cycles, while test data contains both nominal cycles and anomalies, supporting methods that learn a model of normal behavior and flag deviations. Contents 10 training files (nominal operation only) 10 test files (nominal + anomalous segments) 10 label files (binary point-wise anomaly annotations aligned to the test files) Intended use Benchmarking anomaly detection algorithms under federated, hierarchical federated, centralized, and other distributed learning paradigms — particularly for periodic, multivariate, sensor-driven processes such as industrial robotics, manufacturing cells, and other repetitive control tasks. The ten independent datasets support per-client evaluation, cross-client generalization studies, and aggregated benchmarking across varied conditions. Format: NumPy arrays serialized as pickle (.pkl) Sampling rate: 100 Hz Channels per series: 14 signals covering electrical, command, state, and actuator measurements: Motor electrical: Current0, Current1, Voltage0, Voltage1 Filtered state: FilteredPitch, FilteredYaw Angular rates: PitchDot, YawDot Measured attitude: Pitch, Yaw, CalcPitch Actuator feedback: MotorSpeed0, MotorSpeed1 System flag: Enable License: Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)