Abstract In Dynamic Positioning, traditional control methods lack stability, while neural networks and reinforcement learning have become research hotspots. By learning decision-making strategies from environmental observation data, we can adapt to environmental changes in real time and achieve optimal control effects. We also design a reward function to evaluate control performance. Dynamic inversion control is a control strategy that transforms nonlinear systems into linear systems. To address the issue of ship dynamic positioning, we propose a novel controller design method by combining RBF neural networks with dynamic inversion control. Its effectiveness has been verified through simulation experiments, demonstrating its potential and value in practical applications. This method will help improve the positioning accuracy of ships and optimize control performance.