An Adaptive Sensor Selection and Ray-Tracing-Assisted TDOA Method for HF Skywave Source Geolocation With Experimental Validation

High-frequency (HF) skywave signals can propagate across vast distances due to ionospheric refraction, rendering conventional line-of-sight (LOS) geolocation methods inadequate for accurate source localization. To address this challenge, we propose an enhanced geolocation framework based on Time Difference of Arrival (TDOA) and semidefinite programming (SDP) to accurately estimate HF skywave source positions using geographically distributed receivers. First, we construct a three-dimensional (3D) ionospheric plasma density grid matrix model by combining an improved wideband ionospheric channel model with a ray-tracing technique. This model not only generates simulated TDOA data but also provides essential prior ionospheric information for real-world geolocation. Then, we approximate the non-line-of-sight (NLOS) skywave path with two equivalent LOS segments using a quasi-parabolic (QP) single-layer model, facilitating geometric modeling of the ionospheric reflection. Finally, we formulate the HF source geolocation problem as an SDP-based optimization task, where the proposed a sensor-selection and weighting strategy selects the receiver subset and assigns optimal weights to each selected sensor, thereby improving localization performance. Numerical and real-data experiments show that our proposed method attains a geolocation mean-square-error (MSE) that is close to the Cramér-Rao lower bound (CRLB), surpassing the performance of existing methods.