Dynamical modeling of a micro AUV in hydrodynamic ground effect

Abstract

The need for inspection of subsea infrastructures - ranging from ship hulls to pipelines - have led to development of various underwater vehicles specifically geared towards such missions. One such example is the EVIE (Ellipsoidal Vehicle for Inspection and Exploration) robot developed by the d'Arbeloff Laboratory at MIT - which is a water jet propelled ellipsoidal micro AUV with a flat bottom designed for scanning and moving on submerged surfaces. Near proximity to surfaces is important for using sensors like ultrasound for imaging internal cracks that cannot be done by simple visual inspection. However movement on submerged surfaces without wheels or tether is challenging since they are often rough due to rust or corrosion. The d'Arbeloff Laboratory has previously presented an extensive research on utilizing natural ground effect hydrodynamics during longitudinal motion of the AUV to maintain a self stabilizing equilibrium height very close to the surface. Further, it was also demonstrated that using an active bottom jet we can form a stable fluid bed of thickness from 1/200 to 1/50 of the body length and enable smooth gliding of the robot on rough surface. Previous work presented has focussed on the observations, simulations of the phenomenon using computational fluid dynamic (CFD), analysis of the different regions of the ground effect force and some preliminary experimental results. In the research presented in this paper, we go one more step further and provide a deeper insight to the modeling of motion near a submerged surface and its dependency on design parameters. This is of particular relevance since understanding the system dynamics and response characteristics will enable us to design a control system which can deal with the unusual near surface behavior of this underwater vehicle and help transition the design for more practical applications. The work here also substantially explores linearization techniques for this highly non linear system to enable traditional linear control and demonstrates methodologies of using experimental data for modeling purposes instead of a pure analytic model.