General Recursive Least Square Algorithm for Distributed Detection in Massive MIMO

In this paper, a general recursive least square (GRLS) detection algorithm is proposed for the uplink of distributed massive multiple-input multiple-output (MIMO) to alleviate the bottlenecks in both computational complexity and data bandwidth for interconnection. Different from the existing recursive least square (RLS) detection algorithm which only supports a single antenna in each distributed unit (DU), the proposed GRLS allows for multiple antennas in each DU, rendering it adaptable to a variety of practical scenarios. Moreover, among the total
$C$
DUs and with an integer parameter
$C_{0}$
, the computational complexity of
$C-C_{0}$
DUs in GRLS can be significantly reduced by leveraging the channel hardening property. Through analysis, we demonstrate that the convergence of the GRLS algorithm is guaranteed if
$C_{0} \geq \lfloor {(\sqrt{{{B}}/{2}}+\sqrt{K})^{2}}/{B} \rfloor$
holds, where
$K$
and
$B$
denote the numbers of antennas at the user side and each DU, respectively. Furthermore, based on the daisy-chain architecture, the proposed GRLS algorithm also enjoys excellent scalability, which can be easily extended with extra DUs for further improvement. Finally, the detection complexity and data bandwidth analysis are also provided to unveil the superiority of GRLS compared to other distributed detection schemes for massive MIMO.