An Efficient Approximate EP-Based Iterative Detection and Decoding for Massive MIMO

In this letter, an efficient approximate expectation propagation (EPA) based iterative detection and decoding (IDD) scheme named EPA-IDD is first proposed for LDPC-coded massive MIMO systems. EPA is applied in IDD to bypass the variance updates in the inner loops, which reduces matrix inversions, simplifies extrinsic messages, and effectively enhance the convergence performance of IDD. In addition, a partial resetting scheme is proposed to efficiently adopt the decoder output into the EPA detector. A Neumann series based approximation called wNSA-EPA-IDD is further developed to reduce complexity. Numerical results show that the proposed EPA-IDD schemes outperform the state-of-the-art (SOA) double EP (DEP) in various LDPC-coded MIMO scenarios. Complexity analysis is presented to validate the improved efficiency of the proposed algorithms.