SubCat: Design and Control of a Vectored-Thrust Underwater Robot With Model Predictive Control

Abstract

Small-scale underwater vehicles face significant challenges in rejecting strong environmental disturbances due to restricted individual actuator capacities and the inherent thrust projection losses of conventional fixed-thruster configurations. To overcome these limitations, this letter presents SubCat, an underwater robot featuring a mechanical vectored-thrust architecture driven by four servo-actuated rotatable aquatic thrusters. This design dynamically concentrates propulsion along actuated motion axes, substantially enhancing maximum effective thrust utilization along demanded directions. To handle the resulting highly coupled nonlinear dynamics, this work derives a comprehensive rigid-body model with hydrodynamic parameters identified via Computational Fluid Dynamics (CFD). Furthermore, a real-time Model Predictive Control (MPC) framework is developed to concurrently optimize thrust magnitude allocations and predictive servo orientations, while explicitly accounting for finite servo reorientation-rate constraints. Extensive experimental validations, encompassing sharp disturbance recovery and high-agility trajectory tracking against a standard fully-actuated baseline, demonstrate stable path fidelity, improved thrust utilization, and lower task-level net electrical power demand in the tested high-demand trajectories.