Abstract
Indoor localization remains an important task in wireless sensor networks (WSNs), especially in complex multi-room environments. This study proposes a three-dimensional (3D) localization method based on received signal strength indicator (RSSI) fingerprinting and adaptive path loss modeling. The method provides accurate RSSI estimation at unmeasured locations, thereby reducing the need for exhaustive manual data collection. Experimental results on ZigBee-based WSNs show that machine learning models trained on synthetically generated radio maps achieve accuracy levels comparable to those using traditional fingerprints. The multilayer perceptron (MLP) achieved the best result on real data with a root mean square error (RMSE) of 0.83 m, while the Extreme Gradient Boosting (XGBoost) model achieved the best performance on synthetic fingerprints with a RMSE of 1.03 m. These results demonstrate the effectiveness of the proposed method and highlight the generalization capabilities of state-of-the-art machine learning models, especially in scenarios with limited empirical data.