Adaptive Signal Detection Method Based on Model-Driven for Massive MIMO Systems

In this paper, the model-driven deep learning (DL) technology is used to solve the problem of either high complexity or poor performance in traditional massive multiple input multiple output (MIMO) signal detection and the model-driven detection network is proposed: JC-Net. The JC-Net structure is designed by unfolding the damped Jacobi detector and adding three trainable parameters to each layer, which are used to control the residual vector, adjust the relationship between the current layer and the previous layer, and for soft projection. Furthermore, the performance of JC-Net can be further improved by increasing the dimension of the residual vector and the JC-Net-Improved is proposed later. Simulation results show that the proposed model-driven massive MIMO detection networks can significantly improve the performance of the corresponding damped Jacobi detector and achieve superior detection performance with low complexity.