Bagging to Improve the Calibration of RSSI Signals in Bluetooth Low Energy (BLE) Indoor Distance Estimation

Abstract

Originally conceived as proximity sensors, smart Bluetooth (Bluetooth Low Energy or BLE) beacons have been quickly adopted as inexpensive means to estimate distance of the transmitter from the receiver. Unfortunately the Received Signal Strength in unstable and produces such oscillations that right beyond a couple of meters the accurate estimation of distances becomes extremely challenging. In this paper, starting from a preprocessed RSSI vector of measurements, a Bootstrap Aggregating procedure is proposed to improve the calibration of RSSI signals. The proposed method, in combination with robust and non parametric statistics, reaches a sub-meter precision up to 6 meters of distance.