TDOA-Based Direct Position Determination Method With Parametric Dictionary Learning

To address the performance degradation caused by grid mismatch and unknown channel attenuation, we propose a time difference of arrival (TDOA)-based direct position determination (DPD) method with parametric dictionary learning. First, the off-grid error and the model parameter error are uniformly modeled as the dictionary offset problem. Subsequently, the DPD problem under dictionary perturbation is transformed into a joint optimization problem involving parametric dictionary learning and sparse signal reconstruction, which is solved within the sparse Bayesian learning (SBL) framework. Furthermore, the Cramér-Rao bounds (CRB) for TDOA-based DPD are derived, serving as a benchmark for performance evaluation. Simulation results demonstrate that the proposed method effectively localizes off-grid sources under unknown channel attenuation. The positioning accuracy can approach the CRB under high signal-to-noise ratio (SNR) and can be improved by over 20% compared with the existing methods under low SNR.