Robust Fixed-Complexity Sphere Decoders for Rank-Deficient MIMO Systems

Sphere decoder (SD) has variable complexity, and the traditional fixed-complexity sphere decoder (FSD) is not applicable for rank-deficient (NT > NR) multiple input multiple output (MIMO) systems (i.e. the down link detection in MIMO systems, or systems with highly correlated MIMO channels). To overcome these difficulties, in this paper, robust fixed-complexity sphere decoders (RFSD-s) based on new preprocessing algorithms are proposed for rank-deficient MIMO systems. With respect to the cases without and with information on the level of noise, RFSD using zero-forcing technique (RFSD-ZF) and FSD using minimum mean square error technique (FSD-MMSE) are proposed respectively. To reduce the computational complexity of RFSD-ZF, a simplified RFSD-ZF (SRFSD-ZF) with little performance loss is also introduced, the theoretical proof is given to support the feasibility of SRFSD-ZF. Simulation results show that, besides better performance than the traditional FSD, the proposed techniques are robust to the configuration of MIMO antennas (both NT ≤ NR and NT > NR), which is another advantage over the traditional FSD.