Deep-Learning-Based Lattice Reduction Preprocessing for Time-Correlated MIMO Systems

The lattice-reduction (LR) preprocessing can effectively improve the performance of multiple-input multiple-output (MIMO) systems especially in the highly correlated channel. This study designed a deep-learning model for the LR preprocessing in the MIMO system. Based on the AlphaGo Zero architecture, this paper proposes a deep learning-based lattice reduction (DLLR) algorithm by using reinforcement learning in the AlphaGo Zero. This work investigated the performances of the DLLR in the time-varying correlated MIMO channel environments. Compare with the traditional LR-aided MIMO algorithm, the simulation results show that the DLLR can greatly improve the orthogonality of the MIMO matrices for the detection and the DLLR-aided MIMO system has better bit-error-rate performance than the LR-aided MIMO system.