Deep Learning-Based Path Loss Prediction for Fifth-Generation New Radio Vehicle Communications

Abstract

Fifth-generation (5G) technology is rapidly spreading to vehicle-to-vehicle (V2V) communication, which requires high reliability, high data transmission rate, and low latency to meet service requirements through a new frequency band called millimeter wave (mmWave). However, mmWave bands are difficult to utilize in a dynamically changing vehicle environment because of the propagation attenuation against obstacles. Various studies are underway to predict the path loss in the mmWave band on roads with many obstacles. However, it is still challenging to accurately predict the path loss in various environments because the existing prediction models either generalize the path loss solely based on measurement data or only use specific parameters. Recently, investigations on artificial intelligence have been conducted using various techniques that are different from the existing heuristic methods. Following this trend, we propose a deep learning-based path loss prediction that considers obstacles on roads and weather conditions in V2V communication using mmWave. To consider the various environments affecting measurement, we constructed a realistic simulation environment and collected data that we used to train our deep learning models. Our proposed deep learning-based approach achieves accurate predictions for path loss.

Keywords

• Path loss prediction
• vehicle-to-vehicle communication (V2V)
• millimeter wave (mmWave)
• deep neural network (DNN)
• weather condition modeling