Channel Simulation and Neural Network Research at 100 GHz in an Office Scenario

Terahertz (THz) channel modeling is seen as the key theoretical basis of sixth generation (6G) wireless communication. Machine learning, as a powerful tool to characterize channel features, has been widely used. In this paper, an indoor THz channel characterization at 100 GHz is presented. The simulation is carried out by the ray tracing method. The path loss characteristics and FI model are researched. Moreover, Back Propagation (BP) neural network and BP neural network based on particle swarm optimization (PSO-BP) are applied to assist prediction of channel characteristics. The results of evaluation index show that the accuracy of PSO-BP model is higher than BP model.