An Efficient Linear Detection Scheme Based on L-BFGS Method for Massive MIMO Systems

For massive multiple-input multiple-output (MIMO) systems, minimum mean square error (MMSE) detection is near-optimal, but requires high-complexity matrix inversion. To avoid matrix inversion, we formulate MMSE detection as a strictly convex quadratic optimization problem, which can be solved iteratively by the recognized most efficient Broyden-Fletcher-Goldfarb-Shanno (BFGS) quasi-Newton method. According to special properties of massive MIMO systems, we propose a novel limited-memory BFGS (L-BFGS) scheme for MMSE detection with one correction search, unit step length, and simplified initialization, which can greatly reduce the storage and computation cost compared to BFGS method. Simulation results finally verify the effectiveness of the proposed scheme.

Automated summary

A novel limited-memory BFGS (L-BFGS) scheme was proposed for MMSE detection in massive MIMO systems, which significantly reduces storage and computation cost compared to the BFGS method. Simulation results confirmed the effectiveness of the proposed scheme.