Machine Intelligence Technique for Blockage Effects in Next-Generation Heterogeneous Networks

Abstract

Millimeter wave (mmWave) links such as 28 GHz and 60 GHz propose high data rates and capacity needed in 5G Heterogeneous network (Hetnet) real-time system. The key factors in network planning of Hetnet are the locations and links of base stations, and their coverage, transmitted power, antenna angle, orientation etc. How-ever, large-scale blockages like static buildings, human etc. affect the performance of urban Hetnets especially at mmWave frequencies. A mathematical framework to model dynamic blockages is adapted and their impact on cellular network performance is analyzed. A machine learning approach based on Q-learning with Epsilon-Greedy algo¬rithm is proposed to solve the blockage problem in such complex networks. The proposed results are evident and show the positive effect of increasing the base station den¬sity linearly with the blockage density to maintain the net¬work connectivity. The performance of the proposed Epsi¬lon-Greedy algorithm is compared with Epsilon-Soft algo-rithm. The performances of above said mmWave links are compared in terms of their coverage probability and throughput. The results show that an Epsilon-Greedy algo¬rithm outperforms an Epsilon-Soft algorithm.