Assessing the impact of multi-channel BLE beacons on fingerprint-based positioning

Bluetooth Low Energy (BLE) beacons are attractive for indoor location due to their ease of deployment, wide support on consumer devices and low cost. Co-ordinate location can be estimated using radio fingerprinting techniques applied to Received Signal Strength (RSS) values from BLE beacons. Earlier work on BLE fingerprinting has noted that the separate BLE advertising channels can exhibit different RSS values, although modern consumer devices do not provide channel information and therefore report a composite RSS with artificially inflated variance that can lead to reduced positioning accuracy. In this work we quantify how often this channel dispersion occurs in typical environments, what the extent of the dispersion is, and what impact it has on positioning. Furthermore we explore how to obtain channel information when using today's consumer devices. We find that the dispersion is both common and significant and leads to significantly reduced positioning accuracy; in our experiments only 10% of composite variances were within 1 dBm of an individual channel's variance, and 40% of composite variances were at least 4 dBm greater. We also show that the behaviour of the channels is sufficiently distinct that a fingerprinting scheme that uses a signal map for each achieves significantly increased positioning accuracy (up to 3 m).