Distributed Processing Methods for Extra Large Scale MIMO

SUMMARYMassive MIMO (multiple-input multiple-output) systems are key candidates for the fifth generation (5G) of cellular networks. Having a lot of antenna elements at the base station (BS) is an important enabler to provide a very high spatial resolution. Therefore, systems beyond 5G rely on increasing the number of elements at the BS to support future applications. At very large dimensions, e.g. aperture sizes bigger than 100 wavelengths, a new type of array called extra-large scale MIMO (XL-MIMO) emerges that offers enhanced spectral and energy efficiency.However, practical implementation of such arrays requires overcoming several challenges such as computational complexity, hardware limitations and non-stationary propagation patterns.
This thesis presents several techniques to handle major existing concerns in the XL-MIMO arrays, namely: computational complexity of receiver algorithms, scalability and interconnection overheads. In order to address the complexity issue, different low complexity methods are proposed. One of the main differences between these methods and conventional linear receivers in massive MIMO systems is, that they exploit the information about user energy patterns over the array to operate more effectively. Another approach is to distribute the receiver processing tasks between several nodes and create a hierarchy between processing nodes. The thesis studies different architectures and mostly focuses on a distributed way that uses sub-arrays to obtain local estimates at local nodes. Then, a central node collects all the local data to perform a global decision. Furthermore, the thesis suggests several antenna selection methods to limit the area of the array being processed and control the amount of computations. These methods directly use the received energy patterns at the BS to find the best active antenna sets and turn off the rest of the array to save energy. Moreover, to address the hardware considerations such as scalability and inter-connection overheads, a fully decentralized method is proposed that works without a central node. 
In summary, the main outcome of the thesis is the proposal of signal processing enablers for the XL-MIMO systems. The proposed methods address the aforementioned challenges while providing acceptable performance.