Trilateration, fingerprinting, and centroid: Taking indoor positioning with Bluetooth LE to the wild

Bluetooth Low Energy (BLE) beacons raised high expectations when introduced as a cheap, easy to handle indoor positioning solution that is widely supported by smartphones. Experiments soon showed that RSSI values measured by devices vary significantly even when conducted under identical setups. As a result, BLE beacons can be used for what they are designed, i.e., proximity detection. However, setting up indoor positioning systems that cover complete floors and buildings with acceptable accuracy and out of the box is far from being straightforward. This paper evaluates setups for BLE indoor positioning at an inhouse fair to position dedicated smartphones among visitors and exhibition stands. It investigates how established positioning approaches, namely trilateration, fingerprinting, and centroid, respond to the impact of the crowd. Experiments show for the.95 percentile that the weighted centroid technique WCWCL achieves the highest position accuracy of 5.7 meters in the crowded space. Fingerprinting, on the other hand, is heavily impacted by the crowd. While most accurate in an empty/uncrowded space, its accuracy decreases to 11 meters in the crowded space. Further, the paper shows that for a setup with randomly distributed beacons a density of one beacon per 40 square meter is a sweet spot beyond which accuracy does not further increase.