Applications employing Global Navigation Satellite Systems (GNSS) to provide accurate positioning are subject to drastic degradation not only due to electromagnetic interference, but also due to multipath components caused by reflections and refractions in the environment. Typically, the higher the correlation between the line-of-sight (LOS) component and the remaining multipath components, the more inaccurate is positioning estimation. In the case of safety-critical systems that depend on positioning, such as autonomous driving and civil aviation, accurate positioning is essential. State-of-the-art tensor based approaches for antenna array-based GNSS receivers assume that the components are not highly correlated, implying that the measured data is a tensor whose factor matrices are full-rank. In the case of scenarios with highly correlated (clustered) multipath components, the measured data tensor has a rank-deficient factor matrix. In this paper we propose a tensor-based scheme utilizing the multilinear rank-$(L_r,L_r,1)$ term decomposition via generalized eigenvalue decomposition (GEVD) in order to improve the time-delay estimation of the LOS component in challenging scenarios with highly correlated multipath components by exploiting the data model resulting from NLOS component clustering.

使用全球导航卫星系统（GNSS）提供精确定位的应用不仅会受到电磁干扰的严重破坏，而且还会受到环境反射和折射引起的多径分量的影响。通常，视线（LOS）分量与其余多径分量之间的相关性越高，定位估计就越不准确。对于依赖于定位的安全关键型系统，例如自动驾驶和民航，准确的定位至关重要。基于天线阵列的GNSS接收器的基于最新张量的方法假定组件之间的相关性不高，这意味着所测量的数据是其因子矩阵为全秩的张量。在具有高度相关（聚类）多径分量的场景中，测得的数据张量具有秩不足因子矩​​阵。在本文中，我们提出了一种利用多线性秩-$（{\mathrm{大号}}_{\mathrm{[R}}，{\mathrm{大号}}_{\mathrm{[R}}，\mathrm{1个}）$ 利用广义特征值分解（GEVD）进行术语分解，以便通过利用NLOS分量聚类产生的数据模型，在具有高度相关的多径分量的挑战性场景中改善LOS分量的时延估计。