Prediction of Amphibious Vehicle Waterborne Posture Focusing on Extreme Data

Amphibious vehicles are prone to the risk of instability, especially in water. Utilizing artificial intelligence for predicting and monitoring the vehicle's waterborne posture is an effective method to ensure safety. This paper aims to develop a prediction model for the waterborne posture of amphibious vehicles with a focus on extreme data. By oversampling extreme data and constructing an extreme loss function, we enhance the model's predictive capabilities for extreme conditions. We establish a prediction model based on the Transformer architecture to ensure robust predictive performance across the entire dataset while improving training efficiency and convergence. Finally, the effectiveness of the model is validated through real-world experimental data.