THÖR-UWB: Collecting Human Motion Data using Ultra-Wideband Localization

THÖR-UWB Dataset The THÖR-UWB dataset is a novel dataset of accurate human motion in a museum-like indoor environment, building on the THÖR dataset protocol. We will provide 2D position estimates from a UWB positioning system, position and head orientation motion capture data, 3D LiDAR and radar scans, and gaze information. In total, THÖR-UWB captures over 130 minutes of human motion data. This is the first data sample, made available for the 7th Workshop on Long-term Human Motion Prediction at the 2025 IEEE International Conference on Robotics and Automation. The complete data will be published after data curation and cleaning have been finished. 1. Data format The sample data consists of two files: 1) a CSV file containing trajectory data, and 2) a Jupyter notebook that can be used to plot the comparison between UWB and motion capture. 1.1. Trajectory file The trajectory file contains three minutes of motion data of a single person. Four trajectories are given: Tobii glasses worn by the participant, measured by the motion capture system (3D position + quaternion) Google Pixel Smartphone, measured by the motion capture system (3D position + quaternion) Google Pixel Smartphone, measured by onboard UWB system (2D position) Unitree Go2 robot, placed statically in the corner, measured by the motion capture system (3D position + quaternion) The units of the positions and the order of the quaternions are given in the trajectory files' header. Missing data is denoted by "N/A". Temporal indexing is facilitated by the "Time" or "Frame" column, indicating timestamps or frame numbers. The motion capture system records at 100Hz, and the UWB data at 5Hz. 1.2. Plotting notebook The notebook contains basic visualizations comparing both UWB and motion capture measurements. The following Python libraries are required: pandas matplotlib plotly