Radio Resource Allocation for Vehicle Platooning Based on Hierarchical Deep Reinforcement Learning

Cellular Vehicle-to-Everything (C-V2X) networks provide critical technique support for Intelligent Transportation Systems (ITS) to guarantee vehicle safety and enhance traffic efficiency. However, the traditional centralized resource allocation methods in C-V2X suffer from a dimension explosion problem. This paper proposes a Hierarchical Deep Reinforcement Learning (H-DRL) framework for resource allocation in C-V2X networks. The proposed framework adopts a dual timescale mode in which the high-level meta-controller is responsible for link clustering on a large timescale, aiming to alleviate link interference. At the same time, the low-level controller optimizes the transmission power for V2V links on a small timescale to ensure high reliability and low latency V2V communications. The simulation results show that our work can significantly improve the transmission rates for Vehicle-to-Infrstructure (V2I) links and guarantee high reliability and low latency for V2V links.