A Comparative Analysis of AI-assisted WiFi and BLE Indoor Positioning Systems

Indoor positioning has become an on-going issue in recent years.Artificial Intelligence (AI), especially Machine Learning (ML) and Deep Learning (DL) algorithms assisted WiFi and Bluetooth Low Energy (BLE) fingerprinting indoor positioning systems are regarded as common techniques.This work conducts a comparative analysis of some of the state-of-the-art ML and DL algorithms that can be used to process fingerprinting data, as well as the trade-offs when applying them in different scenarios.This includes supervised and unsupervised machine learning models, and an innovative method combining Deep Neutral Network (DNN) and vector embedding.The result shows that vector embedding using i-vectors conducts the most accurate performance after model adaptation, supervised machine learning techniques like Support Vector Machine (SVM) can be regarded as a balance between accuracy and efficiency, unsupervised machine learning technique Density Based Spatial Clustering of Applications with Noise (DBSCAN) is more adaptable when there is much noise, but become less precise when there exists dynamic objects.