Tensor-Based Subspace Fusion for Direct Position Determination

Yuqian Ren,Hang Zheng,Zhiguo Shi
DOI: https://doi.org/10.1109/iccc62479.2024.10681969
2024-01-01
Abstract:Direct position determination (DPD) can achieve higher localization accuracy compared to two-step methods which require independent parameter estimation for each receiver and parameter matching. Nevertheless, most DPD methods are developed for 2-D scenarios, and the existing 3-D DPD methods follow the matrix processing paradigm, resulting in the loss on structural signal characteristics. In this paper, a tensor-based subspace fusion method for DPD (TSF-DPD) is proposed. Specifically, the received signals of multiple receivers equipped with uniform rectangular arrays are modelled as tensors. For the covariance statistics of each tensor signal, the CANDECOMP/PARAFAC decomposition is adopted to retrieve the orthogonal signal and noise subspaces, leading to the formulation of multiple tensor-based spectra for all receivers. To associate the position information from all spectra for 3-D DPD, a fused tensor-based spectrum corresponding to all receivers is constructed based on the minimized sum of distances from orthogonality. Accurate DPD results can be obtained from searching the fused tensor-based spectrum. Simulation results verify the superiority of the proposed TSF-DPD method over the conventional matrix-based DPD method and two-step methods.
What problem does this paper attempt to address?