Reduced-Complexity Robust MIMO Decoders

We propose a robust near maximum-likelihood (ML) decoding metric that is robust to channel estimation errors and is near optimal with respect to symbol error rate (SER). The solution involves an exhaustive search through all possible transmitted signal vectors; this search has exponential complexity, which is undesirable in practical systems. Hence, we also propose a robust sphere decoder to implement the decoding with substantially lower computational complexity. For a real 4 x 4 MIMO system with 256-QAM modulation and at SER of 10(-3), our proposed robust sphere decoder has a coding loss of only 0.5 dB while searching through 2360 nodes (or less) compared to a 65536 node search using the exact ML metric. This translates to up to 228 times fewer real multiplications and additions in the implementation. We derive analytical upper bounds on the pairwise codeword error rate and symbol error rate of our robust sphere decoder and validate these bounds via simulation.