Robust Terrain Following Control of Autonomous Underwater Vehicles with Model Uncertaintities

This paper proposes a robust terrain-following control strategy for an autonomous underwater vehicle (AUV) to support seafloor mapping and underwater search missions. To acquire high-quality sonar data, the vehicle must maintain an appropriate sensing distance over complex seabed geometry. Using beam-range measurements from a Doppler Velocity Log (DVL) and an altimeter, a Gaussian-process-regression (GPR)-based terrain estimation method is employed to reconstruct the seafloor profile. The resulting model accounts for fitting uncertainty and altitude tolerance to generate a safe following tube in the vertical plane, which provides a bounded reference for control. To address model uncertainty, a tube model predictive control (MPC) scheme is developed by combining a nominal MPC for trajectory planning with an ancillary feedback law that enforces constraint satisfaction under disturbances. Numerical simulations demonstrate that the proposed method achieves accurate terrain tracking and improved robustness compared with conventional approaches.