Skip to main content

Massive MIMO Localization for 5G Networks

Periodic Reporting for period 1 - MassLOC (Massive MIMO Localization for 5G Networks)

Reporting period: 2017-09-01 to 2019-08-31

Next-generation 5G wireless data networks promise significant gains in terms of offering the ability to accommodate more users at higher data rates with better reliability while consuming less power. To meet such a challenge, massive MIMO systems have been proposed to allow for orders of magnitude improvement in spectral and energy efficiency using relatively simple processing. The basic idea is equipping cellular base stations with rectangular arrays, each of them consists of a very large number of antennas. The extra antennas help to focus energy into ever-smaller regions of space to bring huge improvements in throughput and radiated energy efficiency. Other benefits include reduced latency, simplification of the media access control layer, and robustness against intentional jamming.

An unexplored and unintentional side-effect of using a very large number of antennas combined with high carrier frequencies is the ability to pinpoint the location of the user with high accuracy. This project aims to develop several analytical tools in order to model, design, and analyze massive MIMO-OFDM systems from the localization point of view, and ascertain their validity via experimental datasets. Ultimately, our broad goal is to conceptualize an engineering research idea and then transition it into innovative applications that can be replicated for real-world cellular networks operated by established service providers and mobile manufacturers. In parallel, the project will allow the fellow to achieve several knowledge transfer objectives and increase prominence in his research field.

Robust parameter estimation and localization for massive MIMO systems are complex, and optimizing their overall performance is challenging. To explore and tackle this problem, diverse technical pathways can be regarded, ranging from analytical descriptions to experimental techniques. Our main goal is to analyze and design localization methods for massive MIMO systems. To reach this goal, we put forth the following technical objectives: (i) develop generalized error models for massive antennas to describe the gain and phase error, mutual coupling and antenna position errors. (ii) derive low complexity subspace-based multi-dimensional parameter estimation algorithms. (iii) propose robust localization methods for massive MIMO systems based on the estimated parameters.
Track 1: Channel Modeling for 3D MIMO

We discuss the stochastic nature of dense multipath stemming from rough surfaces, e.g. roughcast as well as balconies. Based on the effective roughness approach, we propose a scattering function which is symmetric with respect to transmitter and receiver positions. The scattering function enables the calculation of a joint angular delay power spectrum. The power spectrum describes dense multipath in a stochastic manner as a function of the geometric setup and the parameters of the rough surface. We analyze the power spectrum in the angular and delay domains and compare our results with measurements reported in the literature. The importance of this work for multipath-assisted positioning is as follows: The results of this track to be useful for the derivation of position error bounds and position estimation algorithms under realistic channel conditions.

Track 2: Estimation of Basic Parameters of the 3D MIMO Channel

2.1 We propose a search-free R-D beamspace tensor-ESPRIT algorithm for mmWave channel estimation. The proposed approach is higher-order singular value decomposition (HOSVD) based and it is the R-D generalization of the beamspace-ESPRIT method. The standard tensor-ESPRIT in element space is achieved by using an identity matrix. Furthermore, the multidimensional parameter association is critical but challenging for both 5G communications and localization. For the proposed approach, the R-D parameters are automatically associated. The performance of the proposed algorithm is evaluated by considering different precoders and combiners. Furthermore, the effect of the size of the precoder and combiner is also investigated by considering both the reflectors and scatters.

2.2 We propose an incomplete iteratively reweighted HOSVD (i-IR-HOSVD) algorithm for robust multi-dimensional channel estimation from partial observation and in impulsive noise environments. Inspired by the tensor completion technique, the key idea is to minimize the ℓp-norm of the residual error instead of ℓ2-norm and recover the low-rank tensor measurements at the same time. i-IR-HOSVD can be applied for robust higher-order tensor decomposition from crossly corrupted and incomplete measurements. It achieves accurate channel estimation performance in impulsive noise environments, even with incomplete measurements.

Track 3: Localization and Orientation Estimation Algorithms

We consider the radio localization problem from a massive MIMO point of view, with a substantial focus on the millimeter-wave (mmWave) regime and provide an overview of different methods for estimating angles and delays with respect to sources in multipath channels and demonstrate how such estimates can be used for localization.
We provide an overview of 5G massive MIMO localization, describing the common channel models and propagation effects, contrasting different channel estimation methods as well as localization techniques. We also presented recent research progress and outlined four promising research directions that involve (i) accurate mmWave propagation modeling, (ii) efficient channel parameter estimation techniques to handle the complicated propagation models and system constraints, (iii) cooperative localization and (iv) artificial intelligence in 5G networks.
Summary for publications