LI Qianrui

Thesis

Decentralized Transmitter Cooperation and Signal Processing with Limited Feedback and Backhaul

 

Transmitter cooperation is considered a promising tool for dealing with interference in wireless networks with an aggressive reuse policy of spectral resources. Cooperation is meant here as the joint optimization of certain transmission parameters, where such optimization can be carried out over several independent domains by using different techniques such as power control, user selection in time/frequency, antenna selection or beam/precoder design. Although transmitter cooperation comes in many flavors, a recurrent assumption behind proposed methods lies in the need for cooperating devices to (i) acquire, share information pertaining to the propagation channel toward the multiple receivers and (ii) perform cooperation based on the disseminated information in the previous step. This holds true for instance for coordinated beamforming methods and, to an even greater extent, for network-MIMO (Joint Processing coordinated multi-point (JP

CoMP) in the long term evolution (LTE) terminology). As feedback and exchange of channel state information (CSI) come at a price in terms of signaling overhead, there arise two important questions: (i) What information should be fed back or exchanged such that the CSI acquired at each transmitter is most informative to perform cooperation? (ii) Which techniques can reap the benefits of cooperation while living with an imperfect channel representation that varies from transmitter to transmitter?

 

In this thesis, we address both aforementioned questions. We consider first each transmitter acquires an initial imperfect CSI based on limited receivers feedback. The transmitters can then exchange their initial CSI through the limited backhaul, which leads to an imperfect and non-identical final CSI at each transmitter. We optimize first the use of limited backhaul such that the channel estimation at each transmitter is the most accurate. We also consider the problem of optimally allocating the backhaul resource so as to get more accurate CSI estimates or balancing the accuracy and the consistency of the CSI estimate at each transmitter.

 

For the design of efficient cooperation techniques that copes with the imperfect and non-identical CSI configuration at each transmitter, we investigate specifically a regularized zero forcing (RZF) precoder design in large system scenario. With random matrix theory tools, we find the RZF precoder which maximizes the system ergodic sum rate and is robust to the effect of limited feedback and backhaul. Finally, interesting and challenging research directions and open problems are discussed.