Deep learning-based power control method for user-centric network

Abstract

The invention provides a deep residual error network named as UcnNet, and a weighted minimum mean square error (WMMSE) algorithm of a real number domain under a network (UCN) taking a user as a centeris fitted. Specifically, in order to effectively manage the coupling interference in the UCN, deriving and using a WMMSE-based real number domain power control algorithm under multi-cell cooperationfor generating a training label close to the optimal total capacity of the system; and then inputting a multistage residual structure, training a network of a batch normalization layer, giving input channel information by outputting an activation function meeting power constraints, and predicting the transmitting power of each base station. After UcnNet training is completed, under the condition that global channel information is input, output similar to WMMSE can be generated through few calculations. An experiment simulation result shows that the high fitting capability of UcnNet is shown, the fitting efficiency can reach 97.68%, and meanwhile, the efficiency is improved by more than 100 times than that of a WMMSE iterative algorithm.

Description

technical field [0001] The present invention relates to the field of wireless communication technology, in particular to deep learning technology in machine learning and user-centered interference management in ultra-dense network scenarios in future mobile communication systems Background technique [0002] A well-recognized key technology for future networks is to meet the ever-increasing data rate demands through network-dense deployment. At present, Ultra-Dense Network (UDN) is considered to be the main technical means to meet the demand of mobile data traffic in 2020 and in the future. Ultra-dense networking is based on the small coverage and large capacity of the cellular network. By increasing the deployment density of base stations, it can provide seamless connections while maintaining a high data rate, and achieve a huge increase in capacity and frequency reuse efficiency. UDN brings problems such as serious interference and frequent switching, but with the intensi...

AI Technical Summary

Problems solved by technology

However, the fitting ability of the shallow DNN is limited, and as the problem scale increases, the fitting performance will continue to decrease; for a deeper network, a deep convolutional neural network (Convolutional Neural Network, CNN) to improve performance, but there will also be problems of gradient disappearance and gradient explosion, resulting in less than expected fitting effect

Images