Regularization for Cholesky Decomposition in Massive MIMO Detection

In this letter, we propose a new regularization approach to low-precision Cholesky decomposition. This decomposition is employed to inverse an ill-conditioned matrix in the linear detector of the Multi-User Massive Multiple Input, Multiple Output (MU-MIMO) receivers. The proposed method was implemented in half-precision format and tested in realistic scenarios generated by the QuaDRiGa 2.0 channel. Simulations show that our approach demonstrates performance close to the single precision implementation. In high interference scenarios, it outperforms state-of-the-art alternatives, such as regularization based on weighted noise power.