Iterative SVD for Massive MIMO:- Control Theoretical Approach
Speaker:Prof. Vincent LAU
Professor
Fellow of IEEE
Department of Electrical and Computer Engineering
Hong Kong University of Science and Technology
Date & Time:6 Sep 2013 (Friday) 16:00 - 17:00
Venue:J420
Organized by:Department of Electrical and Computer Engineering

Abstract

Distributive and iterative algorithms play a very important role in wireless resource optimizations, game problems or interference management in wireless networks. One important issue associated with distributed algorithm is on the convergence analysis. Traditional convergence analysis is all based on an important assumption that the wireless channels are static during the iterations. However, this assumption is quite unrealistic in practice if the algorithms involve over-the-air iterations. In this case, the channel fading will change in a similar timescale as the algorithm iteration and the analysis of convergence behavior under such dynamic environment is very challenging. In this talk, we propose a general framework to address the convergence behavior of iterative algorithms under time-varying channels. We shall first propose a systematic framework to analyze the “algorithm trajectory” and the “time varying target” using non-linear control theory. We found that the time-varying channels act like an "external force" in a virtual dynamic system. Based on the framework, we can sometimes improve the convergence of distributed algorithms using compensation method. Finally, as an illustration, we propose a low complexity and robust iterative SVD for massive MIMO applications. Specifically, SVD plays a critical role in MIMO processing. However, it has a complexity of O(n3), which can be very high when the number of antennas n is large. In order to exploit temporal correlation in the channel matrix, we do not need to compute a new SVD every frame. Instead, we compute the SVD recursively. The challenge is the convergence and tracking robustness of the iterative SVD. We show that using the proposed framework, the performance of the iterative SVD is very close to the brute force SVD but at a much lower complexity of O(n2).

Biography

Vincent obtained B.Eng (Distinction 1st Hons) from the University of Hong Kong (1989-1992) and Ph.D. from the Cambridge University (1995-1997). He completed the Ph.D. degree in two years and joined Bell Labs from 1997-2004. He joined the Department of ECE, Hong Kong University of Science and Technology (HKUST) in 2004 and is currently a Professor and the Founding Director of Huawei-HKUST Joint Innovation Lab at HKUST. He is also elected as IEEE Fellow, Croucher Senior Research Fellow and Changjiang Chair Professor. Vincent has published more than 220 IEEE journal and conference papers and has contributed to 32 US patents on various wireless systems. In addition, he is also the key contributor of four IEEE standard contributions to IEEE 802.22 (WRAN / Cognitive Radio). His current research focus includes robust cross layer optimization for MIMO/OFDM wireless systems, interference mitigation techniques for wireless networks, delay-aware cross layer optimizations, network control theory as well as multi-timescale stochastic network optimization. He has obtained three IEEE best paper awards and is currently an area editor of IEEE Transactions on Wireless Communications, area editor of IEEE Signal Processing Letters, EUARSIP Wireless Communications and Networking as well as guest editor of JSAC.