| PhD Viva

Name of the Speaker: Ms. Aswathylakshmi P (EE16D404)
Guide: Prof. Radhakrishna Ganti
Venue: ESB-234 (Malaviya Hall)
Online meeting link: https://meet.google.com/ewc-yeog-jmo
Date/Time: 18th September 2024 (Wednesday), 2:00 PM
Title: Fronthaul Compression and Pilotless Strategies for Efficient Uplink Transmission in Large Multi-Antenna Systems

Abstract :

Massive MIMO opens up attractive possibilities for next generation wireless systems with its large number of antennas offering spatial diversity and multiplexing gain that support multiple users and data streams simultaneously. However, the increased number of users and data streams can create bottlenecks and overheads during the implementation of such systems. This thesis addresses two such implementation issues. The huge amount of data generated at a massive MIMO base station can be cumbersome to handle and process. This can throttle the capacity/speed of the network by creating a bottleneck in the fronthaul link, which connects a massive MIMO Remote Radio Head (RRH) and carries IQ samples to the Baseband Unit (BBU) of the base station. This gives rise to the need for data compression techniques tailored towards massive MIMO signal models, that are scalable to the ever increasing number of antennas and are at the same time also compatible with the hardware limitations of the RRH. In this work, we first focus on devising fronthaul compression techniques for massive MIMO base stations that address the above needs. We use matrix representation as an elegant and convenient way to visualize the orthogonal frequency division multiplexed (OFDM) signals received at a massive MIMO RRH in the uplink. From this, we develop a low rank baseband signal model and propose two different fronthaul compression techniques. The first technique is based on low rank approximation using the QR decomposition, which offers higher compression ratios than existing methods and is easy to implement on RRH hardware. The second technique is an iterative matrix decomposition method using alternating minimization that utilizes the convolution structure of the received signals. This method offers orders of magnitude higher compression ratios than the QR technique, but at the cost of higher complexity. We analyse the performance of these techniques in terms of their compression ratios, error rates, and computational complexity, and identify the trade-offs between these factors. We evaluate the performance of the matrix decomposition method under different practical scenarios and constraints. The second problem considered in this work is the increased pilot overhead in massive MIMO systems, especially in multi-user scenarios, which wastes spectral resources and decreases the throughput. We explore an almost blind demodulation solution for massive MIMO-OFDM signals that uses the above iterative matrix decomposition technique to provide estimates of the user data as well as the channel using a single pilot, irrespective of the size of the OFDM signal. Finally, we evaluate performance of the matrix decomposition method in terms of its algorithm complexity versus time, for both the compression and blind estimation problems under user mobility conditions, where the instantaneous channels are temporally correlated. We discuss how the underlying temporal correlations in the channel can be utilized to decrease the complexity of the matrix decomposition algorithm by reducing the number of iterations required.