Machine learning-based data-driven modeling of deep-sea mining vehicles

This study proposes a data-driven modeling method for mining vehicle motion that aims to address the limitations of traditional methods and the challenges of adaptability to real-world environments. The proposed approach utilizes the APSO-XGBoost enhanced fusion algorithm, training a data-driven model with experimental data from the RecurDyn simulation. The model is optimized using data enhancement techniques, highlighting the differential speed data features and integrating the APSO algorithm. A comparison of the proposed method with traditional approaches reveals its superiority in predicting the speed and angular velocity of the mining vehicle. This provides a novel foundation for the intelligent control of deep-sea mining vehicles.