Processing occupancy and indoor localisation data: data science pipelines for valid and robust databases
Supervisors: Julien Nembrini, Denis Lalanne
Student: Michael Papinutto
Project status: Finished
Year: 2024
Understanding occupants’ behaviour is essential to optimise energy use, enhancing occupants’ experiences, and supporting work environments in buildings, where smart technologies can collect data on space use and human building interactions. This work presents the prepa ration and analysis of two distinct datasets allowing to assess building behaviour acquired through Bluetooth Beacons and OptiTrack motion capture technologies. The Bluetooth Beacons dataset captures room-level location data to analyse interactions and routines building-wide, whereas the OptiTrack dataset offers detailed behaviour insights in two refurbished rooms for collaboration and focus. To ensure the anonymity, transparency, and reproducibility of these datasets’ preparation, data science pipelines aligned with the most recent best practices, such as the Swiss Confederation’s guidelines for Human-Centered and Trustworthy Data Science, were meticulously implemented. Furthermore, the two datasets were enriched with room conditions, weather and energy data to enhance the datasets diversity for richer research questions. Preliminary results are promising: the Bluetooth Beacons dataset shows potential for training a machine learning model to predict room occupancy, while the OptiTrack dataset offers insights into occupant behaviour patterns and interactions within the dedicated rooms. This innovative integration of diverse data sources lays the groundwork for more nuanced and actionable insights into smart building environments.
Document: report.pdf