Incremental residual learning-based dynamic modeling and stability analysis for multipower underwater vehicles

Abstract

Dynamical modeling is the basis of state prediction, motion control, and performance analysis for underwater vehicles. Existing dynamic modeling methods were derived from physical principles, less considering the modeling reality gap caused by environmental uncertainty. Therefore, an incremental residual learning-based dynamic model is proposed for a novel multipower underwater vehicle with strengths of the water-jet pump and buoyancy drives. First, the multimodal physical dynamic model is derived from multibody systems, including driving forces and hydrodynamics. Then, the continuous residuals caused by the environmental uncertainty are compensated by the incremental residual learning. After that, multimodal stability criteria are established based on the hydrodynamic equilibrium equation. Finally, experiments on Computational Fluid Dynamics and lake trials are conducted to demonstrate the effectiveness of the proposed method.