Expectation Propagation-Based Parallel Iterative Detection and Decoding for Massive MIMO.

Expectation propagation (EP) has recently been considered with iterative detection and decoding (IDD) to enhance the bit error rate (BER) performance for coded massive Multiple-input Multiple-output (M-MIMO) system. However, current state-of-the-art (SOA) EP-based IDD schemes like double-EP (DEP) are still far from realization due to unaffordable latency brought by their serial structure and large computational complexity. To relieve this issue, in this article, an efficient EP-based parallel IDD method is proposed for coded M-MIMO. First, by investigating the factor graph (FG)-based message passing in EP-based IDD, a double non-resetting framework named EP-dNRe is proposed to improve the IDD efficiency. Based on this framework, a novel parallel EP-based IDD scheme named EP-based parallel detection and decoding (PDD-EP) is further proposed. After proper initialization, the parallel loops of PDD-EP can execute the EP detection and decoding modules simultaneously, which brings latency reduction and improved performance-complexity trade-off. Simulation results and complexity analysis are presented to confirm the efficiency of the proposed PDD-EP. Particularly, both DEP with double non-resetting (DEP-dNRe) and PDD-EP can greatly reduce the required number of iterations to reach the same performance as the SOA DEP. Furthermore, the proposed PDD-EP can attain the same BER as DEP-dNRe with about 33.1% less complexity, and outperforms DEP-dNRe up to 0.2 dB by similar complexity in various LDPC-coded M-MIMO systems.