Description
Reducing the energy consumption of buildings by driving HVAC, lighting, and ventilation control from real-time and predicted occupancy-estimation. This is the dominant application motivation for the smart-building branch of indoor sensing literature; the technical contribution is rarely "a new sensor" and almost always "a new control loop on top of existing sensing". For the thesis, this application context constrains the latency and reliability requirements that any deployed CSI/BLE system must meet.
Why it's hard
- HVAC plant has long thermal time constants; control must anticipate occupancy, not just react.
- Calibration drift in the underlying occupancy estimator silently degrades energy savings.
- Comfort vs energy is a multi-objective control problem with user preferences.
- Sub-zone occupancy is needed for fine-grained control but most sensing operates at room level.
- Integration with legacy BMS protocols (BACnet, KNX) adds engineering friction.
Common approaches
- Model-predictive HVAC control consuming occupancy forecasts.
- ML-based occupancy prediction models trained on multi-modal sensor history.
- Schedule-blending with calendar data for predictable spaces.
- Green-sensing techniques that minimize sensing energy itself.
Source Papers
- esrafiliannajafabadi2022_1342 ↗ — impact of occupancy-prediction models on HVAC performance.
- khan2024_43e8 ↗ — occupancy prediction in IoT-enabled smart buildings.
- koo2026_a08d ↗ — green wireless sensing: energy-efficient sensing via WiFi CSI.
- chaudhari2024_6efc ↗ — occupancy detection fundamentals for smart buildings.
- zhou2022_09d5 ↗ — Integrated sensing and communication waveform design.