MultiPhysio-HRC: Multimodal Physiological Signals Dataset for industrial Human-Robot Collaboration

The MultiPhysio-HRC dataset contains multimodal recordings collected to study stress, cognitive load, and affect during industrial human-robot collaboration (HRC) and related tasks. Data were acquired from 55 participants on the first day and 42 participants on the second day, across resting baselines, cognitive tasks (Stroop, N-Back, Mental Arithmetic, Tower of Hanoi), guided breathing, a virtual reality height-exposure stressor, manual battery disassembly, and collaborative disassembly with a voice-controlled cobot. The dataset includes physiological signals (ECG, EDA, EMG, respiration at 256 Hz, and 12-channel dry EEG), voice recordings, and facial action units, together with ground-truth measures from validated questionnaires (STAI-Y1 for stress, NASA-TLX for workload, SAM for valence/arousal/dominance, and NARS at baseline). Data are synchronized, pseudo-anonymized, and released with derivative feature sets in CSV format to facilitate analysis. The data collection was approved by the SUPSI ethics committee, and all participants provided informed consent. A companion GitHub repository provides loaders, preprocessing pipelines, feature extraction, and baseline machine learning models for reproducibility.