Massive-MIMO: From Theory to Practice – System Efficiency Under Constraints. Wednesday, June 6, 15:00 – 17:00
Chair: Vladimir Lyashev (Huawei Technologies)
15:00 – 15:10: Vladimir Lyashev Huawei Technologies — Welcome and brief introduction
15:10 – 15:30: Eugene Tyrtyshnikov the Institute of Numerical Mathematics — How to sense less working with large multi-dimensional matrices
Astronomically large data must have never been considered as absolutely arbitrary, otherwise they become intractable. Here the crucial paradigm means parametrization
and structure. Such data should be regarded rather as a function that we can use to obtain its values at some points, but we never do it for all points. Sensing less by far is necessary.
A very general kind of structure is given by low rank representations of matrices and their generalizations to multi-dimensional matrices, also called tensors. In this talk we consider recent essential developments of the ideas of low-rank matrix approximation proposed in “A theory of pseudo-skeleton approximations”. The practical importance of the very approach consists in using only small part of matrix entries that allows one to construct a sufficiently accurate approximation in a fast way for “big data” matrices.
15:30 – 15:50: Mikhail Kirichenko Huawei Technologies — Massive-MIMO Era: From High Performance to Smart Performance
Each year makes wireless networks are more and more dense, the communication links transform from noise-limited to interference-limited; introduction of massive-MIMO techniques and wide bandwidth increases dimensions for processing. Such quick growth communication technologies and corresponding 5G requirements define new challenging problems for layer 1 (PHY) and layer 2 (MAC) algorithms. To achieve higher performance in straightforward way is obvious solution, which is expensive for practical usage. In the talk we are going to discuss smarter ways to develop L1 and L2 algorithms, which almost becomes low-dimension approximation of well-known techniques, but allows to achieve better performance with reasonable cost.
15:50 – 16:10: Alexander Sergienko Saint-Petersburg Electrotechnical University “LETI” — Machine Learning Techniques for Channel Estimation
As mobile communication systems use more and more complex transmission technologies, channel state information (CSI) estimation problem for these systems becomes more and more challenging. One of promising CSI estimation evolution paths lies in the area of machine learning. A survey of these methods will be presented, including channel estimation for multicarrier systems, MIMO and mmWave.
16:10 – 16:30: Martin Kurras the Fraunhofer Heinrich Hertz Institute — Hybrid Precoding in FDD systems under Sum Feedback Constraint
Massive MIMO in FDD is an ongoing challenge in research, especially due to pilot and feedback overhead. To keep pilot overhead limited hybrid beamforming is a good compromise. To limit the feedback overhead, current focus on CSI compression. However, this only solves part of the feedback overhead problem in FDD systems, where the sum feedback scales linear with the number of active users and feedback per user. In our opinion, the scaling with the number of active users is more critical, because costly massive MIMO are likely to be deployed at places with high user density. Thus, the following topic is a critical one and not well investigated so far: How to design feedback in massive MIMO FDD systems with many users under a sum feedback constraint?
We propose to scale down the feedback per user as a function of the total number of active users. This talk discuss some of the trade-offs at hand and presents numerical examples for one way to keep the sum feedback constant.
16:30 – 16:50: Alexander Maltsev the University of Nizhny Novgorod — Massive MIMO antenna techniques in the upcoming millimeter wave IEEE 802.11ay standard
The 5G systems for performance improvement will exploit all possibilities including the straightforward channel bandwidth increase, network densification, massive MIMO and MU-MIMO techniques. The millimeter wave technology is well-suited for all mentioned ways, allowing having a peak rate up to tens of gigabits due to large bandwidth and having a highly-directional, MU-MIMO capable antennas in a small form-factor. The first millimeter wave standard in the license-free 60 GHz band, the IEEEE 802.11ad, also known as WiGig, offers up to 6.76 Gbps data rate on the distances up to tens of meters. Initially it was developed for indoor applications, with peer-to-peer connection, and does not support MU-MIMO, although introduces the directional transmissions and reception using the small phased antenna arrays. But to realize the full potential of the millimeter wave systems as a part of 5G, several important problems should be solved. Firstly, the millimeter wave coverage should be increased up to several hundred meters – and this is possible only with implementation of high-gain steerable antennas and special beamforming algorithms. Secondly, the MU-MIMO should be introduced in the millimeter wave standards, which lead to significant increase of simultaneously served stations and therefore, require proper collision avoidance mechanisms and association algorithms. The overview of these key PHY/MAC techniques in the upcoming millimeter wave IEEE 802.11ay standard will be presented in this report.
16:50 – 17:15: All participants — Free discussion