Multilevel sequential monte carlo algorithms for MIMO demodulation

We propose low-complexity sequential Monte Carlo (SMC) algorithms for demodulation in MIMO systems that employ large signal constellations. The proposed algorithms exploit the multi-level or hierarchical nature of the signal constellation to reduce the complexity associated with the generation of Monte Carlo samples of the MIMO symbols. The signal space is partitioned into multiple levels and samples are drawn beginning from the highest level space, down to the lowest level, which corresponds to the original symbol space. At each level, we consider only the subspace associated with the sample drawn at the previous level. The advantage of such a strategy is that instead of searching the whole signal space, we restrict our search to the more promising zones of the space, thus saving significant amount of computations. Both stochastic SMC algorithm and deterministic SMC algorithm are considered under such a multi-level framework. For M-QAM signal constellation, the computational complexity of the proposed algorithms is O(log M) in terms of the constellation size, as compared to the O(M) complexity of the existing SMC MIMO detection algorithms (while keeping the number of samples fixed in both the algorithms). We also demonstrate that the performance of these algorithms improves considerably with optimal ordering. The proposed multi-level SMC algorithms are then extended to cope with the case where the number of transmit antennas is larger than the number of receiver antennas, as well as the case of frequency-selective MIMO channels. Extensive simulation results are provided to illustrate the performance of the proposed new MIMO demodulation algorithms in various scenarios