Localization of Radio Frequency Sources based on Received Signal Power

Accurate estimation of the large-scale path-loss coefficient is essential for reliable localization of unknown radio frequency (RF) sources. Traditional empirical path-loss models use fixed coefficients but fail to capture spatial and environmental variability. This work presents a data-driven approach that learns the path-loss coefficient from real-world received power measurements using regression with transmission frequency, geolocation and automatically classified terrain features derived from public available satellite imagery. The predicted coefficients are converted into distance estimates and integrated into a robust 3D multilateration framework to determine RF transmitter\u2019s location. Experimental results show improved localization accuracy and generalization compared to fixed-coefficient models, demonstrating the method\u2019s practicality for field deployment using standard measurement logs.