An improved LLR approximation algorithm for low-complexity MIMO detection towards green communications

As the number of transmit/receive antennas gets large in wireless communication systems, the drastically-increasing complexity in MIMO detection imposes significant challenges in implementing green communications while achieving high spectral efficiency. The winner-path-extension (WPE) K-best algorithm is an efficient detection algorithm in uncoded MIMO systems, known for its stable throughput and excellent symbol-error-rate (SER) and bit-error-rate (BER) performances under relatively low complexity. However, when applying the WPE K-best algorithm into coded MIMO systems, where soft-output information such as log-likelihood ratio (LLR) is required, missing counter-hypotheses issue in LLR calculation often degrades the error performance. To solve this problem, in this paper we propose an improved LLR approximation algorithm, such that WPE K-best algorithm can be well suited to coded MIMO systems. Specifically, when a counter-hypothesis misses, we set a metric threshold for the missing counter-hypothesis by calculating the metric of the bit flipping vector, and then randomly choose a value below the threshold as the approximation. We conduct simulation evaluations for our proposed algorithm in an 8 × 8 MIMO multiplexing system employing 16QAM modulation and Turbo coding. Simulation results show that compared with other existing LLR approximation schemes, our proposed approach can effectively improve the block-error-rate (BLER) performance as well as reducing the complexity in the tree search of WPE K-best algorithm. Moreover, we use a look-up table method to determine the Schnorr-Euchner (SE) enumeration order, which can further decrease the computational complexity of WPE K-best algorithms.