HLM-MOCAP: A Motion Capture Dataset of Context-Dependent Human Arm Motion for Human-Like Robot Motion Generation

HLM-MOCAP is a motion capture dataset of human upper-limb movements recorded to support data-driven generation of human-like trajectories for collaborative robots in assembly-like scenarios. Forty healthy adults performed a fixed set of seven single-arm tasks at a standardized table-top workstation. The task set includes point-to-point reaches to multiple targets, sequential reaching along a predefined path, planar tracing of zigzag and circular contours, grasp-and-place movements around an obstacle with and without arm crossing, transport of a weighted cylinder, and a precision placement task involving small screws.

The dominant arm was tracked using an eight-camera Vicon MX T10 infrared motion capture system operating at 200 Hz with retro-reflective markers attached to the finger, wrist, elbow, and shoulder. Recordings were exported from Vicon software and organized into trial-wise CSV files.

The repository provides raw and minimally processed trajectories, time-normalized position data, derived velocity and acceleration profiles, and aggregated averages per task and participant group. In addition, diagnostic plots and Python scripts are included to reproduce the preprocessing pipeline (gap filling, trimming, trimming of non-movement phases, time normalization, and smoothing) and to visualize the data. The dataset is suitable for research on human-like robot motion generation, learning from demonstration, human arm movement analysis, and benchmarking of trajectory generation methods in human–robot collaboration contexts.