RSSI-based Indoor Localization using Point Mass Filter on a Discretized Grid Map with Pedestrian Dead Reckoning Integration

Abstract

Accurate indoor positioning remains challenging due to the absence of global navigation satellite system (GNSS) signals and the complexity of indoor environments. While foot-mounted inertial measurement unit (IMU) offer infrastructure-free operation, they inevitably accumulate drift over time without external corrections. In this work, we introduce a hybrid localization approach combining a foot-mounted IMU with received signal strength indicator (RSSI) measurements from ambient Wi-Fi access points (APs). Unlike traditional fingerprinting methods, which typically require labor-intensive manual labeling, our approach autonomously constructs a probabilistic radio map during normal pedestrian movement by associating precise pedestrian dead reckoning (PDR) positions with Wi-Fi scans in real-time. The collected RSSI data is discretized onto a two-dimensional spatial grid and statistically normalized to account for variations in signal strength across different devices. During the online phase, a point mass filter (PMF) compares live RSSI measurements against the previously constructed radio map, enabling accurate position estimation without relying on inertial sensors or prior knowledge of AP locations. We evaluated our system extensively in a real-world office environment, testing various hardware devices and multiple users. Experimental results confirm that the method achieves meter accuracy under realistic conditions and exhibits strong generalization capabilities across different devices, underscoring its potential as a lightweight and scalable indoor localization solution.