Multiagent Deep Reinforcement Learning for Optimal Resource Allocation in AoI-Aware Energy-Efficient Platoon-Based C-V2X Networks

This paper aims to tackle the complex challenge of channel assignment and joint power-energy allocation within a cellular-vehicle-to-everything (C-V2X) network, which is deployed to manage vehicular dynamics at an urban traffic intersection. The primary function of the C-V2X network is to facilitate the coordination of multiple vehicle platoons formed by closely spaced same-lane vehicles. This coordination involves two critical communication tasks including the timely update of platoon states to a roadside unit (RSU) and the reliable exchange of cooperative awareness messages (CAMs) among vehicles within the same platoon. The main objective of this paper is to minimise the average age of information (AoI) to ensure the timely update between vehicle platoons and the RSU, maximise the CAM delivery probability (CDP) to guarantee the successful exchange of CAMs among vehicles, and promote sustainable, green communication practices through the implementation of our optimal power-energy management strategies. Recognising the intricate and dynamic nature of this challenge, we adopt a multi-agent deep reinforcement learning (MADRL) approach based on the Markov decision process (MDP). Two innovative algorithms based on the multi-agent deep deterministic policy gradient (MADDPG) and twin delayed deep deterministic policy gradient (TD3) algorithms are proposed to address this optimisation problem effectively. Finally, comprehensive simulation results are presented, which demonstrate the remarkable performance of our proposed schemes, particularly in terms of energy efficiency when compared to existing research. Importantly, these gains in energy efficiency are achieved while maintaining competitive algorithm convergence speeds, low AoI levels, and high CDP, showcasing the practical viability of the developed methods.