Adaptive Identification and Control for Underwater Vehicles: Theory and Comparative Experimental Evaluations
Embargo until
Date
2013-12-20
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
This Thesis reports several novel algorithms for state observation, parameter identification, and control of second-order plants. A stability proof for each novel result is included. The primary contributions are adaptive algorithms for underwater vehicle (UV) plant parameter identification and model-based control. Where possible, comparative experimental evaluations of the novel UV algorithms were conducted using the Johns Hopkins University Hydrodynamic Test Facility.
The UV adaptive identification (AID) algorithms reported herein estimate the plant parameters (hydrodynamic mass, quadratic drag, gravitational force, and buoyancy parameters) of second-order rigid-body UV plants under the influence of actuator forces and torques. Previous adaptive parameter identification methods have focused on model-based adaptive tracking controllers; however, these approaches are not applicable when the plant is either uncontrolled, under open-loop control, or using any control law other than a specific adaptive tracking controller. The UV AID algorithms reported herein do not require simultaneous reference trajectory-tracking control, nor do they require instrumentation of linear acceleration or angular acceleration. Thus, these results are applicable in the commonly occurring cases of uncontrolled vehicles, vehicles under open-loop control, vehicles using control methods prescribed to meet other application-specific considerations, and vehicles not instrumented to measure angular acceleration. In comparative experimental evaluations, adaptively identified plant models (AIDPMs) were shown to accurately model experimentally measured UV performance.
The UV model-based control (MBC) and adaptive model-based control (AMBC) algorithms reported herein provide asymptotically exact trajectory-tracking for fully coupled second-order rigid-body UV plants. In addition, the AMBC algorithm estimates the plant parameters (hydrodynamic mass, quadratic drag, gravitational force, and buoyancy parameters) for this class of plants. A two-step AMBC algorithm is also reported which first identifies gravitational plant parameters to be used in a separate AMBC algorithm for trajectory-tracking. We report a comparative experimental analysis of proportional derivative control (PDC) and AMBC during simultaneous motion in all degrees-of-freedom. This analysis shows AMBC (i.e. simultaneous adaptation of all plant parameter estimates) can be unstable in the presence of unmodeled thruster dynamics; two-step AMBC is robust to the presence of unmodeled thruster dynamics; and two-step AMBC provides 30% better position tracking performance and 8% worse velocity tracking performance over PDC. To the best of our knowledge, the reported comparative experimental evaluation of AMBC and PDC is the first to consider trajectory-tracking performance during simultaneous motion in all degrees-of-freedom.
Description
Keywords
state estimation, parameter identification, underwater vehicle control