Non parametric and robust statistics for indoor distance estimation through BLE

Indoor positioning through Smart Bluetooth (Bluetooth Low Energy or BLE) sensors is a promising new field, where noisy data and outliers make challenging even the simplest distance estimates. The power of the BLE signal is known to be highly unstable even when measurement conditions remain unchanged and statistics on repeated measurements are required in order to have a good confidence in the obtained short-range distance estimates. This work proposes a stack of corrections based on non-parametric and robust statistics as a preprocessing step on the measured data, such that both the calibration and the range estimation processes improve their accuracy. According to experiments, robust and non-parametric statistics are able to handle effectively the severe noise involved in RSSI measurements, reaching most of the times a sub-meter precision.