Enhancing Micro-location Accuracy for Asset Tracking: An Evaluation of Two Fingerprinting Approaches Using Three Machine Learning Algorithms

Micro-location is the ability to track a user’s location or asset at a granular level and is the next evolution in location services. However, this evolution is not proceeding without challenges. Current mobile device GPS systems are not accurate enough to provide the granularity required; especially indoors when there is no line-of-sight with satellites. One promising solution involves using Bluetooth Low Energy beacons. In this paper, we present an indoor location system that uses BLE beacons and iPhones. We present the relevant background of work in this area, the architectural framework that we designed and developed, the mobile apps, the machine learning algorithms employed and report on the findings of two fingerprinting methods using three machine learning algorithms. Our indoor micro-location prediction system using beacons consistently performs at an accuracy of at least 1.7 m spherical radius from the target with convergence in under 2 s and was validated on 2,000 trials.