Reinforcement Learning-Based Optimization of Relay Selection and Transmission Scheduling for UAV-Aided mmWave Vehicular Networks

Abstract

Millimeter-wave (mmWave) communications offer abundant bandwidth for vehicular networks, however it is prone to blockages due to buildings, topology and other environmental factors. To address these challenges, we propose a novel unmanned aerial vehicle (UAV)-aided two-way relaying system to enhance vehicular connectivity and coverage. We formulate a joint optimization problem for relay selection and transmission scheduling to minimize transmission time while ensuring throughput requirements. Proximal policy optimization, deep Qnetwork, and constraint programming models are employed to solve the optimization problem. Extensive evaluations reveal that the proximal policy optimization model achieves close to 100% accuracy. © 2024 IEEE.