Optimal Structure Based Intelligent Precoding Design for Network Massive MIMO Communications

Massive MIMO employs large antenna arrays to support multi-user wireless communications with high data rate. To mitigate the inter-user interference, various precoding schemes are proposed, which usually need iterative calculation and suffer from high complexity, especially for multi-cell scenarios. This paper investigates the intelligent design of network precoding schemes based on the optimal structure. Firstly, we derive an interference avoidance precoding in network massive MIMO communications, and propose an optimal precoding structure. We extract part expression in the optimal structure as parameters and propose a parameterized precoding without iterations. By using the convolutional neural network (CNN), these parameters can be trained and predicted, which significantly reduces the computational complexity. Simulation results present that the proposed precoding scheme can increase the transmission rate by 20% than the RZF scheme.