On the Crucial Impact of Antennas and Diversity on BLE RSSI-Based Indoor Localization

Abstract

Due to their low complexity, RSSI-based solutions for indoor localization have become increasingly popular in recent years despite lacking the accuracy of more sophisticated localization solutions. One of the main reasons for this lack of accuracy is the highly fluctuating nature of RSSI values as a result of indoor channel characteristics, hardware imperfections and varying antenna radiation patterns. In this paper, we thus critically analyze the log- normal path loss model and its validity with the focus on indoor localization. We show how chip antennas of typical consumer devices affect the RSSI measurements and clarify in what manner this effect can be incorporated into the log-normal model. In this process, we are also able to quantify the impact of small-scale fading and diversity on RSSI-based distance estimation. We furthermore propose a novel calibration scheme that estimates path loss exponents based on a simple training walk and outperforms linear regression in all our use cases. At last, we combine our findings in an implementation of an indoor localization system for a 14×3 m office corridor. On average, our measurements yield a significantly decreased position RMSE of 1.36 m, which compares to the Cramer- Rao lower bound on the position RMSE of 0.71 m in this environment.