RI: Small: Robust Model-Based Identification, Navigation, and Control of Underactuated Underwater Vehicles

Johns Hopkins University, Baltimore MD

Investigators

Abstract

Over 70 percent of the Earth's surface is hidden by its oceans, which are believed to be the cradle of life on Earth. Yet, less is known about the sea floor and oceans than about the surface of the Moon. Scientific exploration of the oceans is critical to the understanding of the Earth's climate and life forms. However, most of the oceans are inhospitable to humans due to the immense depth, low temperatures, and the absence of light. Since the 1960's, oceanographers and engineers have worked towards the development of remotely-controlled underwater robots, with the goal of exploring the depths of the oceans. The fact that GPS does not work underwater makes underwater navigation especially challenging for autonomous robots. This project seeks to advance underwater exploration by developing new navigation and control methods to enable earth-science, commercial, and national-security missions that are beyond present capabilities. This project seeks to develop improved navigation and control methods for autonomous underactuated underwater vehicles (UUVs) through the development of novel approaches to model identification, model-based vehicle state estimation, and model-based adaptive control. The goal is to improve the accuracy of present-day underwater vehicle navigation by one to two orders of magnitude, down to sub-meter accuracy. Our approach is to develop novel robust RANSAC Nullspace-based Least-Square (NBLS) algorithms for the experimental identification of dual-model (surfaced and submerged) plant and actuator parameters of 6-DOF plant dynamic models for underactuated UUVs in the presence of sensor noise and outliers, to use these plant models in 6-DOF state estimators for underactuated UUVs to improve vehicle navigation precision, and to investigate switched nonlinear model-based adaptive controllers for underactuated UUVs to improve vehicle control precision. The stability of the proposed algorithms will be evaluated analytically and their performance will be evaluated both in simulation and in experiments with robotic hardware This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.