BLE-Based Indoor Positioning Using Extended Advertisement

For a variety of location-based services, it is important to know the location of mobile devices indoors as well as outdoors. Among the wireless communication-based indoor positioning methods, the way using Bluetooth Low Energy (BLE) has the advantage of being able to build an infrastructure at a low cost and use it in an environment with limited power supply. However, multipath and noise caused by wireless environment may degrade the location measurement accuracy. Machine learning-based methods have been proposed to resolve the problem, but they utilize only Received Signal Strength Indication (RSSI). BLE beacon can use extended advertisement according to user's intention, and the resulting change in PHY affects RSSI. This paper shows the effect of PHY on indoor positioning accuracy and proposes a deep learning-based indoor positioning method. We demonstrate that additional learning of BLE beacon's advertisement reception rate can increase accuracy. The accuracy of the location-prediction model was increased by 8.93% compared with the RSSI-only model, and the mean error distance was reduced by 15.02cm. In addition, even in the locations where advertisements were not collected, the locations were predicted with an error of 133.64cm using the data collected from a nearby location.