Abstract
Conventional building occupancy modeling is open loop where occupancy is detected based on sensor data and predefined rules/thresholds and ambient controls for illumination, temperature, or other settings are adjusted accordingly without taking into account human feedback regarding the level of comfort. Such an open-loop occupancy model can result in suboptimal ambience in certain situations. This disclosure describes techniques that incorporate user-initiated tweaks to ambience controls to train occupancy models. Per the techniques, modifications to ambience settings made by users are encoded into a preference signal for reinforcement learning through human feedback (RLHF). Model weights are closed-loop tuned based on both sensor signals and crowdsourced manual signals, and converge to a customized occupancy model that is tailored to user preferences.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Shin, D, "High Accuracy Closed-Loop Tuning of Occupancy Models for Ambience Control", Technical Disclosure Commons, (August 01, 2024)
https://www.tdcommons.org/dpubs_series/7252