A Phase Optimization Method for Intelligent Reflective Surfaces Based on Deep Reinforcement Learning

Abstract

The invention discloses a method for optimizing the phase of an intelligent reflective surface based on deep reinforcement learning, which includes the following steps: initializing the action network, evaluation network, intelligent reflective surface phase bias matrix and experience pool in the intelligent reflective surface (intelligent body); according to User channel state information to obtain the initial state of the agent; store the experience pool through the interaction between the intelligent reflective surface and the wireless communication system; randomly sample from the experience pool to train the action network and evaluation network so that the evaluation value output by the evaluation network reaches the maximum, and then Obtain the network model parameters after convergence; output the optimal phase offset matrix coefficient of the smart reflective surface that maximizes the signal-to-noise ratio received by the user under the channel state information. The invention can effectively reduce the time required for optimizing the phase offset matrix and the storage space of training samples, and has better robustness.

Description

technical field [0001] The invention relates to the field of communication technology, in particular to a phase optimization method for an intelligent reflective surface based on deep reinforcement learning. Background technique [0002] In recent years, the birth of a number of key 5G technologies has significantly improved the spectral efficiency and capacity of wireless communication systems. However, in the actual deployment process, there are still practical problems such as high energy consumption, hardware implementation complexity, and signal processing algorithm complexity. With the development of radio frequency micro-electromechanical systems and metamaterials, the application of intelligent reflecting surfaces (Intelligent Reflecting Surface, IRS) with low energy consumption and adaptable to time-varying wireless communication systems becomes possible. IRS generally consists of a large number of passive printed dipole antenna elements, and each passive antenna c...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the high computational complexity of the SDR algorithm and the sample acquisition problem of deep learning. The present invention provides an intelligent reflective surface optimization method based on deep reinforcement learning for the downlink transmission system of the base station using a large-scale uniform linear antenna array , the proposed algorithm can train the network model online according to the samples in the experience pool, saving sample storage space and phase optimization time

Images