Application of Data Driven control in Dynamic Positioning

Abstract

This master thesis examines the motivation of studying adaptive control, presents the framework of Unfalsified Adaptive Control (UAC) and its theoretical background. Detailed explanation is given for key features that make up UAC. Important contributions are summarized in a literature review. Theory is presented for a switched PID controller with a controller bank containing candidates for the switching algorithm. A reset mechanism is developed to ensure bumpless switching between controllers with slow integrator contributions. Switching in terms of hysteresis switching and increasing cost level algorithm is presented for both time independent and varying cases. The thesis describes a marine simulator built in Matlab/Simulink for applying UAC, as a data driven control approach, to Dynamic Positioning (DP) of surface ships. Results from simulations are presented that validate the applicability of UAC to the DP problem when the controller set contain stabilizing controllers. An alternative switching algorithm that combine increasing cost level and hysteresis switching is presented as a possible solution when the bank of controllers contains candidates that destabilize the plant.

Part of the thesis is submitted to the 20th World Congress of the International Federation of Automatic Control awaiting peer review.