High Resolution Estimation of AoA, AoD and TdoA from MIMO Channel Sounding Measurements with Virtual Antenna Arrays: Maximum-Likelihood vs. Unitary Tensor-ESPRIT

Abstract

Estimating the parameters of a geometric propagation model from MIMO channel sounding measurements will be considered, which requires the solution of an inverse problem. Thus, a model of the measured data is derived, which incorporates a model of the measurement system as well as the parameters of interest. Based on the data model a maximum-likelihood estimator will be derived to infer the model parameters. Because virtual antenna arrays are considerer, formed by step-wise rotating directive antennas at transmitter and receiver side, the MIMO measurements are conducted in the beam-space. Hence, the data model can be described by a multidimensional convolution of the measurement system and the propagation channel. Based on the convolutional modelling, the parameter estimation problem is transformed into a harmonic retrieval problem, which can be solved by an Unitary Tensor-ESPRIT algorithm. The maximum-likelihood and ESPRIT estimator are compared by Monte-Carlo simulations according to their root-mean-square estimation error.