OpenCSI: Self-Calibration Layer for Heterogeneous Mesh Wireless Sensor Networks

WiFi sensing models trained in one room usually fail in the next: standard normalization bakes chip- and room-specific artifacts right into the feature space. OpenCSI is a self-calibration layer that hides those artifacts, exposing each link as a dimensionless Z-score against its own quiet-period baseline, learned online from a 120-second empty-room bootstrap and reported with a maturity tag so downstream code can detect drift and abstain. Evaluated across three rooms and three ESP32 generations, it is a practical step toward WiFi sensing that survives hardware swaps and new environments. Presented at the IEEE Conference on Local Computer Networks (LCN) 2026 in Coimbra, Portugal.