차세대 이동통신 시스템에서 HetNet 구성과 AI기술을 통한 시스템 성능향상 기법 연구

Abstract

Mobile traffic has been increasing exponentially recently as mobile communication terminals and users' usage patterns have shifted from text data, pictures, and voice to video-oriented streaming services. 5th generation mobile communication technology, which is named IMT-2020 in ITU, is a technology that meets 1000 times data traffic capacity, low latency, high energy efficiency, and cost compared to 4th generation LTE systems. It may be difficult to provide services compared to existing 4G LTE systems due to very high path loss due to high frequency area use. Various technologies are being studied to overcome these shortcomings. In this paper, we introduce Small-Cell and D2D technologies and AI technologies as technologies for improving system performance of 5-generation mobile communication systems. In this thesis analyze the performance by comparing the results of the application of each technique and the application of the proposed algorithm. The analysis results show that using Small-Cells and D2D in 5th generation mobile communication systems can significantly reduce the problem of shading area improvement and path loss, which is a disadvantage of 5G. However, since Macrocells and Small-Cells use limited frequencies, interference problems occur, especially terminals on the cell boundary are heavily affected by interference from neighboring cells, and interference problems between Macrocells can occur within cells. In this thesis, we propose algorithm applications for Small-Cell and D2D technologies and hybrid algorithms of both technologies to improve the performance of cellular communication, and further introduce algorithms such as support vector machines, logistic regression, and decision tree. Simulations show that the performance improvement portion of the proposed Small-Cell algorithm compared to the general application of Small-Cells, and the proposed algorithm to control the amount of Small -Cell interference can be observed 2.86x performance improvement over conventional macro systems. D2D communication also showed higher performance compared to macro-systems in application, and eventually applied Small-Cell, D2D hybrid algorithms to produce better results. Furthermore, we applied the learned AI algorithm based on data from the Small-Cell position, user position, SINR, and capacity of the system, and it was shown that the dual training algorithm of the proposed SVM algorithm has 98.1% accuracy and high system performance compared to the previous 92.8%. Configuration of HetNet such as Small-Cells and D2D communication is considered essential for Next-generation mobile communication, and it was observed that AI technology actively improves system performance along with reducing variables in the actual environment.