| The main purpose of this work has been to derive a comprehensive 6 degree of freedom (DOF) nonlinear underwater vehicle model and investigate the feasibility of nonlinear control system techniques for the autopilot design for underwater vehicles.; The main motivation for using nonlinear models and control theory is that remotely operated vehicles (ROVs) are known to operate over a large number of operating points with no specific speed dominating. Thus a linear control design will be quite laborious, since linearization and gain scheduling techniques must be applied to each of the vehicle's operating points. A nonlinear model is also superior its linear counterpart if the vehicle is allowed to perform coupled manoeuvres at some speed. By applying nonlinear control systems techniques nonlinear kinematics, thruster forces and hydrodynamic forces due to quadratic drag, Coriolis, centripetal and added mass coupling terms etc. can be compensated for in a systematic manner. Consequently a comprehensive nonlinear mathematical model of the ROV must be derived.; It is further emphasized on exploiting the robotic system properties to simplify the dynamic description of the ROV equations of motion. Consequently, the nonlinear ROV equations of motion are written in a compact form nearly similar to the representation used in robot manipulator control. This representation is highly advantageous when designing nonlinear controllers since well known matrix properties like positiveness and skew-symmetry can be exploited in the nonlinear control system design. Indeed, advanced nonlinear control design was found often to be simpler and more intuitive than its linear counterpart. Nonlinear autopilots based on feedback linearization, sliding control and passivity based control are discussed in depth. For these schemes, globally stable autopilot schemes are derived in both the earth-fixed and vehicle-fixed reference frames by applying Barbalat's Lyapunov-like lemma for non-autonomous systems. Finally, model imperfectness due to parametric uncertainties is compensated for by deriving adaptive and self-tuning versions of the control schemes. |
