Parametric Precoding Based on Improved Dynamic Gradient Descent in Multibeam Satellite Communications

Abstract

Precoding and beamforming have been studied in multibeam satellite. In this paper, we propose a precoding method based on dynamic gradient descent, which is suitable for large array antenna and multiuser multibeam satellite environment. This method doesn't need complex vector or matrix design because of combining zero-forcing and maximum ratio transmission. And gradient descent algorithm is used to solve this combining zero-forcing and maximum ratio transmission model. Meanwhile, the convergence of gradient descent is also proved. The simulation results show that the improved gradient descent based on parametric design can get higher sum rate, whether it is in the case of multiple users with massive antennas or small users with a few number of antennas. Meanwhile, for the gradient descent convergence problem, we propose a dynamic adaptive learning rate updating method, which has faster and better convergence performance than traditional gradient descent algorithm.