Multi-Sensor Fusion for Indoor Smartphone Localization with Attention Mechanism

Indoor localization has become an essential technology for navigation, asset tracking, and context-aware applications in environments where GPS signals are unavailable or unreliable. In this work, we present a multi-sensor fusion framework for indoor smartphone localization that integrates Inertial Measurement Unit (IMU) data with Wi-Fi Received Signal Strength Indicator (RSSI) measurements. IMU data provides high-frequency motion tracking, capturing short-term user movements, while Wi-Fi RSSI offers absolute positioning references from surrounding access points. To address the limitation of IMU (dead reckoning) and Wi-Fi RSSI (fluctuations and loss), we incorporate an attention mechanism that adaptively assigns higher weights to the most reliable input features. This allows the system to maintain accurate localization even when up to 90% of Wi-Fi measurements are missing. Experimental evaluations in realistic indoor settings demonstrate that our approach significantly reduces localization error compared to traditional sensor fusion.