Improved sphere decoding algorithm with low complexity for MIMO systems

Sphere decoding (SD) algorithm, as one of the main vector detection mechanisms in digital communication systems, has been referred to have polynomial complexity over a wide range of signal-to-noise ratios (SNRs), rates, and numbers of antennas. The first part of this paper discusses the expected complexity of the SD algorithm over all input SNRs and numbers of antennas, and derives the upper bound of the expected complexity. The result demonstrates how the complexity is affected by the input SNR and the problem size. Moreover, it shows that the expected complexity grows exponentially with the square root of the problem size in low input SNR, while grows polynomially with the problem size in high input SNR for a wide range of the problem sizes. In the latter part, a new algorithm reducing the searching radius in the SD algorithm is proposed. We show that the computational complexity of the novel algorithm is lower compared to the traditional SD algorithm, while the bit error rate hardly changes. Finally, the simulation results show that the new proposed algorithm outperforms the traditional SD algorithm.