Multi-Agent Reinforcement Learning for task allocation in the Internet of Vehicles: Exploring benefits and paving the future

Inam Ullah, Sushil Kumar Singh, Deepak Adhikari, Habib Khan, Weiwei Jiang, Xiaoshan Bai

Abstract: The Internet of Vehicles (IoV) and its applications are undergoing massive development, requiring diverse autonomous or self-directed vehicles/agents to fulfill various objective and responsibilities in vehicular technology. Similarly, Multi-Agent Systems (MAS) and multi-agent task allocation are currently the main focus of multiple researchers and scholars, and they play a key role in IoV and the Internet of Things (IoT). The development of the IoV and autonomous vehicles plays a significant role in Intelligent Transportation Systems (ITS), which are empowered by vehicular networks. However, the dynamic nature of these networks presents substantial challenges that need to be addressed. In this regard, we trace the historical evolution of the multi-agent task allocation of IoV, highlight its fundamentals and progress, and discuss the existing survey works. This paper comprehensively reviews various IoV strategies, both multi-agent task allocation strategies and Multi-Agent Reinforcement Learning (MARL), emphasizing the intelligent learning architecture, concepts, and security-related issues. Additionally, we highlight various computing platforms and the diverse applications of multi-agent task allocation in IoV, where task allocation is challenging and presents security concerns of multi-agent task allocation in IoV. Finally, we discuss major open problems regarding multi-agent task allocation scalability, complexity, communication overhead, resource allocation, security, privacy, etc., and potential future perspectives on multi-agent task allocation methods are highlighted.