DRL-Based Multi-Dimensional Resource Scheduling for Intelligent Connected Vehicles in UAV-Assisted VEC Systems

Unmanned Aerial Vehicle (UAV)-assisted Vehicular Edge Computing (VEC) has emerged as a novel paradigm for compute-intensive and latency-sensitive tasks for Intelligent Connected Vehicles (ICVs) by introducing UAVs to the vehicular network. However, in the temporary hotspot scenario with traffic congestion, due to the high-speed mobility of vehicles, effective solutions that support vehicles’ higher Quality of Service (QoS) remain a significant challenge. Unlike previous works, we first investigated the UAV deployment problem of maximizing the transmission rate and proposed a Dense Boundary Prioritized Service (DBPS) algorithm to address it. We then investigated the multi-dimensional resource scheduling problem of minimizing the weighting of system energy consumption and latency. Considering the time computing and communication resource, we proposed a Mixed Noise Hindsight Experience Replay-Deep Deterministic Policy Gradient (MNHER-DDPG) algorithm to address it, which improved the DDPG algorithm in exploring noise and experience replay. Finally, experiment results show that the DBPS algorithm enhances the transmission rate, and the MNHER-DDPG algorithm improves the system’s energy consumption and latency.