Development of an indoor localisation system based on ambient Wi-Fi and BLE fingerprinting and machine learning classification

ABSTRACT

Localisation system based on ambient Wi-Fi and BLE (Bluetooth Low Energy) is widely used in recent years. However, performance and localisation accuracy of deployment models are still key problems. In this project, the working demonstrator of localisation system based on ambient Wi-Fi and BLE (Bluetooth Low Energy) fingerprinting is developed to evaluate the system’s performance and compare between methods. For the experimental deployment, the SparkFun ESP32 Wi-Fi and BLE supported boards are used as the terminals to receive Wi-Fi and Bluetooth signals, and simultaneously are as the Bluetooth signal transmitters. In the first step, the system collects ambient Wi-Fi and BLE fingerprinting data for designated reference points in an area of interest. Then, the data is used to train a simple machine learning model (using Matlab classifier and Neural Networks). The training is implemented separately for ambient Wi-Fi and BLE fingerprinting data, and for combined ambient Wi-Fi and BLE. The final step in the project involves evaluation of the system's performance. The experimental results demonstrates the good performance of indoor positioning system Wi-Fi based on Wi-Fi fingerprinting as well as the greater localisation accuracy in combined Wi-Fi and BLE model. 

Quang Loc Ngo_Engineering Minor Thesis.pdf

Download PDF • 4.46MB