Home // UBICOMM 2023, The Seventeenth International Conference on Mobile Ubiquitous Computing, Systems, Services and Technologies // View article

WiFi CSI/RSSI Fingerprints Positioning based on Data Augmentation Technique

Authors:
Mohamed Amin Elaoud
Wiem Fekih Hassen

Keywords: IPS, CSI, RSSI, private dataset, Raspberry Pi, GAN, CNN.

Abstract:
In today’s digitally connected world, Indoor Positioning Systems (IPS) are of paramount importance, especially for applications within enclosed spaces such as buildings. Leveraging the widespread deployment of WiFi technology, this paper presents an IPS that hinges on WiFi signal data specifically, Received Signal Strength Indicator (RSSI) and Channel State Information (CSI), and the powerful generative model, Tabular Generative Adversarial Network (TabGAN). The work entails a meticulous data collection process conducted within a controlled laboratory environment at the University of Passau, in Germany. Subsequently, data augmentation through GAN is employed to enrich the dataset. The augmented data is then evaluated using LightGBM and Convolutional Neural Network (CNN) models, with the Root Mean Square Error (RMSE) as the primary metric and Positioning Error for comprehensive evaluation of the IPS’s accuracy and positioning capabilities. The IPS achieved a remarkable result of 0.99 meters for LightGBM and 0.8 meters for CNN, showcasing its high accuracy on unseen data and validating the efficacy of GAN-based data augmentation for enhancing indoor positioning capabilities.

Pages: 40 to 45

Copyright: Copyright (c) IARIA, 2023

Publication date: September 25, 2023

Published in: conference

ISSN: 2308-4278

ISBN: 978-1-68558-106-0

Location: Porto, Portugal

Dates: from September 25, 2023 to September 29, 2023