Unitary Beamforming Multi-User MIMO System with Efficient User Scheduling Algorithm

Abstract

This letter proposes a new practical multiuser MIMO (MU-MIMO) scheme, which is an evolution of the well-known Per User Unitary beamforming Rate Control (PU 2 RC) proposed for 3GPP-LTE and IEEE802.16m standards. The proposed scheme includes an efficient user scheduling algorithm which alleviates the major weakness of the conventional PU 2 RC. Numerical results show that the proposed scheme provides notable performance improvement especially with small and medium user pool since it effectively exploits the benefit from large codebook size.

Dynamic Channel Feedback Control for Limited-Feedback Multi-User MIMO Systems

“…The problem formulation in (13) belongs to the non-linear integer programming (IP) problems. Unfortunately, no general solution for the IP problem exists.…”

“…The expression in (17) is the optimal real-number solution only for the modified problem in (14), and cannot be directly applied to the original IP problem in (13). However, the integer solution can be found by searching the neighbors of the optimal real-number solution; it can be used as a initial point for searching algorithm.…”

“…On the contrary, orthogonal beamforming scheme such as per user unitary beamforming rate control (PU 2 RC) proposed in [5]- [7] is robust against CSI inaccuracy 1 This work was supported by Samsung Advanced Institute of Technology (SAIT). since the channel feedback, user scheduling, and beamforming are jointly designed by its nature [12], [13]. The PU 2 RC scheme is preferable in low-rate feedback systems.…”

“…Without large user pool, large codebook rather decrease the sum rate of the PU 2 RC scheme[12],[13]. Hence, codebook size of N = 16 instead of N = 256 is used in the numerical results, which is the proper size for small and medium user pool (M = 4).…”

Performance improvement of the PU2RC transmission schemes

“…We refer to the proposed method as enhanced PU2RC (EPU2RC). The same idea was also proposed in [16]. The main difference between our scheme and the method introduced in [16] is in the algorithm proposed for SINR estimation.…”