Indoor Localization Through AI and Smartphone Sensors

Addressing the challenges posed by GPS technology in indoor environments, this thesis investigates the integration of smartphone sensors and WiFi signals with machine and deep learning algorithms to enhance indoor localization accuracy. After analyzing existing technologies and methodologies, the study is based on an approach that combines the data from smartphones' IMU sensors along with infrastructure-based information. Highlighting the potential of AI to process complex sensor data, the study introduces two frameworks: the Accuracy Focused Architecture and the Lightweight Architecture. The first model fuses sensor data with WiFi information, applying convolutional neural networks (CNNs) and recurrent neural networks (RNNs) for movement and position prediction, alongside error correction methods and optimization processes to refine predictions further. The second model streamlines the architectural framework to prioritize efficiency and computational speed, while ensuring accuracy remains uncompromised. Utilizing spline fitting for data interpretation and optimization, the model is particularly adept at supporting real-time applications on smartphones and other devices with limited computing resources.