Quantum Machine Learning For Wifi-Based Indoor Localization

Abstract

The limitations of Global Positioning Systems (GPS) in indoor environments have led to the development of a number of indoor localization systems. One potential system is based on WiFi signals. Wifi-based Localization systems leverage Wireless Local Area Network (WLAN) fingerprinting to estimate user positions. This thesis explores a hybrid quantum-classical machine learning approach to im- prove wifi-based indoor localization systems, where quantum circuits extract features while classical models handle classification. The proposed quantum-classical hybrid system achieved 96.83% accuracy in noise-free conditions. The results are further compared with classical machine learning algorithms, including K-Nearest Neighbors (KNN), Support Vector Machines (SVM), and Random Forest (RF), to benchmark our hybrid model. Our model outperformed all these models, where KNN, SVM, and Ran- dom Forest achieved accuracy rates of 74.2%, 92.0%, and 95.4%, respectively. We also included other popular metrics and obtained promising results, where our model achieved precision of 97%, recall of 96.5%, and F1 score of 96.8%, outperforming all the classical models we benchmarked with. However, the hybrid model outperformed them. However, real-world quantum systems face noise-related challenges, such as de- polarizing errors, bit-flip errors, and decoherence, which degrade performance. There- fore, we adopted a number of popular noise-models and evaluated our proposed model accordingly. Our model showed resistance to noise, especially in low to moderate noise rates. The research demonstrates how quantum computing can surpass machine learning, opening new avenues for indoor navigation, human motion detection, and intrusion detection

Date of Award	2025
Original language	American English
Awarding Institution	HBKU College of Science and Engineering