Adaptive Expectation Propagation for MIMO Signal Detection
Zeliang Ou, Hongwen Yang · 2024
This paper studied adaptive Expectation Propagation (EP) for Multiple Input Multiple Output (MIMO) signal detection. For conventional EP, the calculation of the Minimum Mean Squared Error (MMSE) output in each iteration involves matrix inversion operation and occupies major computational complexity. This paper proposed a low-complexity adaptive EP algorithm based on estimated variance descent. The proposed method utilizes variance descent during iterations as the judgment basis for symbol cancellation, and the computational complexity of the calculation of the MMSE output decreases dramatically after symbol cancellation. Furthermore, the proposed method can also avoid useless iterations in low signal-to-noise ratio (SNR) region in terms of variance descent. Simulations showed that the proposed method kept the same bit error rate performance as conventional EP, while the computational complexity of the proposed method reduced about 25 % ~ 50 % at different levels of SNR compared with that of conventional EP. The proposed methods worked well even when Nrt, where Nrand Ntare the number of receive antennas and transmit antennas in a MIMO system respectively.