CI-NN: A Model-Driven Deep Learning-Based Constructive Interference Precoding Scheme

Constructive interference (CI) precoding is a promising and efficient interference management scheme. However, the symbol-level operations required for CI precoding make this precoding scheme face a bottleneck of high computational complexity. To solve the above problem and make CI precoding applicable to high data rate transmission scenarios, in this letter, we propose a deep learning (DL)-based precoding design method driven by a CI communication model, and develop a CI neural network (CI-NN). By carefully designing a neural network with our customized loss function, the proposed scheme well meets the requirement of CI. Simultaneously, this scheme can realize user adaptive precoding according to the number of active users. The simulation results show that the proposed CI-NN can reduce time complexity effectively, while ensuring the performance of the communication model.