Automatic Decision-Making for Unmanned Surface Vehicle Operation Based on CNN-RF Hybrid Model

Automatic decision-making is crucial for the autonomous operation of unmanned surface vehicles, especially in effectively utilizing multi-source information for dynamic strategy selection. In this paper, we propose a CNN-RF hybrid model for unmanned surface vehicle operations. This model integrates CNN’s feature learning with RF’s nonlinear classification for comprehensive multi-source data modeling. Our approach includes meteorological data alongside state and obstacle information, enhancing decision-making reliability. Comparative simulations with traditional methods demonstrate that our model accurately predicts collision risks and performs effective avoidance, ensuring efficient navigation in complex environments.