A Study on Real-Time Optimal MUSV Network Selection for Swarm Unmanned Surface Vehicle Operations

This study proposes a real-time optimal selection method for Medium-sized Unmanned Surface Vehicle (MUSV) networks within a swarm Unmanned Surface Vehicle (USV) system, utilizing an artificial intelligence (AI)-based reinforcement learning technique, specifically Q-Learning. The MUSV functions as the central node of the USV cluster and must autonomously select the appropriate network topology—centralized, hybrid, or distributed—according to varying communication environments and mission conditions. This selection significantly affects communication stability and operational efficiency. In this work, detection rate, signal strength, and packet loss rate are defined as state variables. A Q-learning algorithm is employed to enable real-time learning of the optimal topology selection policy. Based on MATLAB simulations, the success of USV mission execution was verified by selecting the optimal network topology according to the communication environment These findings suggest that AI-based autonomous network management technologies can serve as a viable foundation for improving the operational effectiveness of clustered maritime unmanned systems.