Dynamic Positioning by Nonlinear Model Predictive Control
MetadataShow full item record
This thesis discusses the theoretical aspects of the unscented Kalman filter (UKF) and nonlinear model predictive control (NMPC) and try to evaluate their practical value in a dynamic positioning (DP) system. A nonlinear horizontal vessel model is used as the basis for performing state, disturbance, and parameter estimation, and attempts at controling the vessel using NMPC are made. It is shown that the extended Kalman filter (EKF), which is much used in various navigation applications including DP, is outperformed both theoretically and practically in simulations by the UKF. Much of which is due to the UKF's improved approximation of the estimated system's true stochastic properties. An attempt to estimate the current from the hydrodynamical damping forces have been applied and shown to be working when the vessel is not subjected to other slowly-varying drift forces. It is implemented a dual estimation approach to try to estimate hydrodynamic damping, which is a very real problem for actual vessels and DP systems. A theoretical evaluation of NMPC is performed and it is concluded that NMPC schemes could fulfill a need in vessel control and DP. Its combination of model based control, optimization approach to achieving performance and predictive properties are indeed useful also for DP. It is found that NMPC could be a step towards a unified control approach combining low and high speed reference tracking, station-keeping and several other control operations which today are handled by separate control approaches. NMPC provides the control designer with an exceptional amount of freedom when quantifying the performance, that it is impossible not to find some use for NMPC.