Robust Adaptive Ship Autopilot with Wave Filter and Integral Action

SUMMARY A stable minimum phase transfer function from rudder angle to yaw angle is used to design a globally convergent adaptive ship autopilot. 1st-order wave disturbances in yaw are ltered by applying a notch lter. Integral action is introduced by augmenting the plant. Global convergence is proven for the total system which include the observer, the parameter update law, the feedback controller, the notch lter and the integral part of the controller. The simulation results showed that the performance is excellent, even with no a priori knowledge of the ship parameters. KEY WORDS Adaptive control Ship autopilot Marine systems 1. INTRODUCTION The design of ship autopilots have been analyzed extensively in the literature, and the control laws applied ranges from conventional PID-type controllers to more advanced self-tuning and adaptive algorithms. Adaptive ship autopilots have been discussed by numerous authors, see e.g. SSlid and Jenssen 1 and Holzh uter and Strauch 2. Model Reference Adaptive Control (MRAC) of ship was rst discussed by Van Amerongen and co-authors 3?5 , and this work was based on the assumption of full-state feedback. In this paper, the input error direct adaptive control scheme of Sastry and Bodson 6 is used to derive an adaptive output feedback autopilot. This approach to the adaptive ship autopilot problem was rst presented by Lauvdal and Fossen 7. In order to make an autopilot work properly, there are several things needed to be analyzed and addressed. First, varying ship dynamics due to e.g. changing load and weather conditions and cruise speed have to be addressed. In conventional schemes, speed-scaling is introduced to compensate for the speed dependent change in the dynamics, but variations due to changing load conditions are usually not addressed. This is the motivation for introducing an adaptive control law. Secondly, knowledge about the environmental forces acting on the ship is crucial in order to obtain robustness and satisfactory performance. The robustness issues are particularly important in an adaptive controller, since disturbances may cause parameter drift. Thirdly, robustness issues related to unmodeled dynamics and nonlinearities in the input need to be addressed. The problem of high-frequency rudder motions due to 1st-order wave disturbances in the feedback loop is explicitly addressed. This is usually solved by using a Kalman-lter to estimate the low-frequency motion components of the ship. Wave ltering in terms of the Kalman lter algorithm has been discussed my numerous authors 8?12 the last decades.