A Survey on Multi-Agent Reinforcement Learning Applications in the Internet of Vehicles

The development of the Internet of Vehicles (IoV) and autonomous vehicles plays a significant role in intelligent transportation systems (ITS) that are empowered by vehicular networks. However, the dynamic nature of these networks presents challenges that need to be addressed. Reinforcement learning (RL) has emerged as an effective technique for strengthening vehicular networks. The use of standard single-agent RL and deep reinforcement learning (DRL) has recently been demonstrated to enable each network entity as a decision-making agent to adapt to unknown environments by learning an optimal decision-making policy. However, in the complex and dynamic environments of vehicular networks, the limitations of single-agent approaches become apparent. Multi-agent reinforcement learning (MARL) offers a compelling alternative, enabling net-work entities to learn their optimal policies by observing the environment as well as the policies of other network entities. Due to this, MARL has recently been used to solve various problems in IoV by improving its learning efficiency. In this paper, we review the applications of MARL in IoV networks. Following the review, four main application areas for MARL in IoV were identified, namely: resource management, task offloading, trust management, and privacy preservation. Furthermore, the MARL-based approaches in IoV were classified into three main categories: fully centralized, fully decentralized, and centralized training with decentralized execution (CTDE) depending on the MARL architecture employed. Finally, we discuss the challenges, open issues, and future directions related to the applications of MARL in the IoV.