Operating multi-user transmission for 5G and beyond cellular systems

Abstract

Every decade, a new generation of cellular networks is released to keep up with the ever-growing demand for data and use cases. Traditionally, cellular networks rely on partitioning radio resources into a set of physical resource blocks (PRBs). Each PRB is used by the base-station to transmit exclusively to one user, which is referred to as single-user transmission. Recently, multi-user transmission has been introduced to enable the base-station to simultaneously serve multiple users using the same PRB. While multi-user transmission can be much more efficient than its single-user counterpart, it is significantly more challenging to operate. Thus, in this thesis we study the operation, i.e., the Radio Resource Management (RRM), for two popular multi-user transmission technologies; namely, 1) Non-Orthogonal Multiple Access (NOMA) and 2) Multi-User Multiple-Input Multiple-Output (MU-MIMO).

For NOMA RRM, we study a multi-cell, multi-carrier downlink system. First, we formulate and solve a centralized proportional fair scheduling genie problem that jointly performs user selection, power allocation and power distribution, and Modulation and Coding Scheme (MCS) selection. While such a centralized schedule is practically infeasible, it upper bounds the achievable performance. Then, we propose a simple static coordinated power allocation scheme across all cells for NOMA using a simple power map that is easily calibrated offline. We find that using a simple static coordinated power allocation scheme improves performance by 80% compared to equal power allocation. Finally, we focus on online network operation and study practical schedulers that perform user-selection, power distribution, and MCS selection. We propose a family of practical scheduling algorithms, each of them exhibiting a different trade-off between complexity (i.e., run-time) and performance. The one we selected sacrifices a maximum of 10% performance while reducing the computation time by a factor of 45 with respect to the optimal user scheduler.

For MU-MIMO RRM, we focus on the study of the downlink of an OFDMA massive MU-MIMO single cell assuming ZFT (Zero Forcing Transmission) precoding. An offline study is initiated with the goal of finding the best achievable performance by jointly optimizing user-selection, power distribution and MCS selection. The best performance is analyzed by using both Branch-Reduce-and-Bound (BRB) global optimization technique for upper-bounding the achievable performance and a set of different greedy searches for lower bounding the achievable performance to find good feasible solutions. The results suggest that a specific search strategy referred to as greedy-down-all-the-way (GDAW) with full-drop (FD) is quasi-optimal. Afterwards, we design a simple practical scheduler that achieves 97% of the performance to GDAW with FD and has comparable runtime to that of the state-of-the-art benchmark that selects all users, performs ZFT precoding followed by power distribution using water-filling. The proposed scheme performs a simple round robin grouping to select users, followed by ZFT precoding and joint power distribution and MCS selection via a novel greedy algorithm with a possible additional iteration to take zero-rate users into account. Our solution outperforms the benchmark by 281%.