Multi-Agent Reinforcement Learning-Based Routing Protocol for Underwater Wireless Sensor Networks With Value of Information

Efficient data transmission plays a crucial role in the applications of underwater wireless sensor networks (UWSNs). In this article, by considering the differences in transmission requirements for data of varying importance degrees in UWSNs, a multi-agent reinforcement learning-based routing protocol with value of information (MARV) is proposed. First, to distinguish the difference of transmission requirements, we introduce the value of information (VoI) to characterize the importance degree of data to reflect the requirement for the real-time characteristic. Moreover, to ensure the efficient routing for different importance degree of data, we establish a multi-agent reinforcement learning (MARL)-based framework by enabling nodes to learn from the environment and interact with neighbors and elaborately design a reward function by considering the timeliness and energy efficiency of transmission. In addition, to improve the transmission efficiency, we design a packet holding mechanism by designing a priority list and variable holding interval according to transmission requirements. The simulation results show that the proposed protocol performs well for the transmission of different data.