Deep Learning-based Indoor Localization for IoT Applications
- Publication Type:
- Thesis
- Issue Date:
- 2023
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
In recent years, localization-based Internet of Things (IoT) applications have been developed and deployed, such as interactive and personalized routing, car localization in underground parking systems and patient emergency localization. However, in indoor environment, Global Positioning System signal is not available because it is very sensitive to occlusion. Many researchers have been focusing on utilizing other technologies such as Wi-Fi (Wireless Fidelity), Radio Frequency Identification, Bluetooth and so on for localization services. Among these technologies, Wi-Fi has been most widely utilized for indoor localization due to its low cost and wide availability. There are various signal measurements for the Wi-Fi-based indoor localization such as Received Signal Strength (RSS), Time of Arrival, Time Difference of Arrival, Round Trip Time, Angle of Arrival, and Channel-State Information. However, the indoor localization methods based on RSS can realize the location of low-cost equipment. However, RSS-based indoor localization possesses many challenges due to multipath effects and noise, environment dynamics, device heterogeneity, limited high-quality data and security. To overcome these challenges, in this thesis, RSS fingerprinting-based indoor localization methods are developed using machine learning methods and deep learning methods. For RSS time-series data, the system of Kalman-DNN exploits the temporal dependency of these data by integrating the Kalman filter with deep neural networks, and experiment results validate effectiveness of the Kalman-DNN system. However, for single RSS readings vector, a system called CapsLoc is proposed, which is an RSS fingerprinting-based indoor localization system based on CapsNet (Capsule Network). The experimental results show that CapsLoc can achieve accurate indoor localization, which outperforms some traditional machine learning methods and existing deep learning methods. Especially for heterogeneous IoT devices, RSS can be affected by superimposed challenges, i.e., device heterogeneity, database problem and energy efficiency. In order to improve localization speed, EdgeLoc is proposed based on CapsNet and edge computing technology. Experiment results show that EdgeLoc outperforms state-of-the-art deep learning methods in performance of the localization accuracy and average positioning speed. Considering security issues in localization where malicious attacks at APs (Access Points) exist, a solution of SE-Loc is proposed for RSS fingerprinting-based indoor localization utilizing the deep learning methods. Extensive experiments show that SE-Loc demonstrates superior performance on secure indoor localization over the baseline methods. To address challenges including the multipath effects and noise, the environment dynamics, the device heterogeneity, data limitation, database problem and even malicious AP attacks, deep learning-based indoor localization methods are proposed.
Please use this identifier to cite or link to this item: