Accurate Gridless Indoor Localization Based on Multiple Bluetooth Beacons and Machine Learning
Konstantinos Kotrotsios, Theofanis Orphanoudakis
Abstract
In this work we present an indoor location method using smartphones as a source of location information. The proposed method uses the Received Signal Strength Indicator (RSSI) value from Bluetooth Low Energy Beacons scattered around interior spaces. We present the results of our model using machine learning, which was developed based on measurements of RSSI values from Beacons inside a lab environment occupying a space of 31m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> . Measurements were fed to the open-source TensorFlow framework to develop an estimator of the distance between the mobile phone and the beacon. Next, based on the cross-sections of peripheral lines having as a center the location of the Beacons and radius the predicted distances we compute the intersection points from all circles and base our position estimation on the Geometric median of intersection points. Through experiments, we show that our system has an average accuracy of 69.58cm and can predict position with an accuracy of less than a meter in 80% of the cases.