Adaptive Control of Surviving Branches for Fixed-Complexity Sphere Decoder

The fixed-complexity sphere decoder (FSD) has been proposed to attain the near-optimal performance and achieves the same diversity order as the maximum-likelihood decoder (MLD) recently. However, its equally handling all the branches obtained via the full expansion (FE) stages till reaching the final minimum Euclidean distance (MED) decision is unnecessary. In this paper, we conduct an adaptive control of the number of surviving branches for each of the single expansion (SE) stages in the original FSD (AC-FSD). In this controlling method, for each stage, the surviving branches are selected by a threshold value which is calculated from the minimum partial accumulated branch metric and the noise level. We compare the AC-FSD with the original FSD and a recently proposed simplified FSD (SFSD), and simulation results demonstrate that the AC-FSD with proper threshold can attain near-optimal performance with a much lower cost than that of the FSD and the SFSD when the system works in mid and high signal-to-noise ratio (SNR) regimes in practice.