Real-time regional density detection in outdoor spaces using a low-cost IoT network
Can you estimate crowd density and movement outdoors without expensive cameras or GSM data? Low-cost ESP32 nodes passively observe Wi-Fi signal patterns in the environment. Outdoor RSSI is badly behaved — it reflects off buildings, gets blocked by people and vehicles, and shifts constantly — so rather than chasing exact positions, the problem is framed as regional classification: readings from several reference sensors form a distinct fingerprint for each zone. A machine learning model learns those fingerprints, and Kalman filtering suppresses noise enough to expose movement trends over time.
The result estimates density and flow direction cheaply, at an anonymous and aggregate level — no one is identified or tracked. First field data was collected in Taksim Square.
Project announcement →