INVESTIGATION OF METHODS FOR ADAPTIVE PATH CONTROL OF SURFACE SHIPS

This dissertation investigates the feasibility and effectiveness of using an integrated system for the adaptive control of a surface ship along a prescribed path in restricted waters. The controller consists of four major components arranged in two loops: an inner or control loop and an outer or gain update loop. The inner loop consists of a Kalman state estimator and an optimal stochastic control law which would provide effective control when subjected to disturbances and measurement noise. The outer loop estimates the parameters of the system equations of motion in order to enable the controller to adapt itself for significant changes in ship characteristics which take place due to changes in the ship operating condition and environment. This work is concerned primarily with the estimation aspect of the adaptive control problem which consists of estimating the values of states and disturbances for the control loop and parameters for the gain update loop. The research covers three major tasks. The first task deals with the use of various disturbance modeling approaches in the design of the control loop Kalman filter. The last two tasks involve the investigation of two possible on-line parameter estimation algorithms: Weighted Least-Squares (WLS) and Minimum Variance (MVE). Computer simulation is used extensively to evaluate controller performance in all cases.; The Brownian motion process is shown to be an effective model for bias disturbances. Furthermore, a control loop designed with this model can perform very well under all types of disturbances. A design criterion is proposed for the selection of the appropriate value of the diffusion coefficient to achieve the desired character of the modeled bias disturbance. This criterion can also serve as a guide in dealing with the random walk process.; The WLS algorithm is shown to be most useful for offline parameter estimation when there is little noise and disturbance. The MVE algorithm, on the other hand, is more suitable to real-time operation. This algorithm can be programmed on a mini computer. The success of a parameter estimation process depends heavily on the amount of information contained in the data base as compared with the level of noise and disturbance. A dither signal is shown to be an effective approach to increase the information content and to avoid the clasic, difficult problem of identifying open-loop dynamics in a closed-loop system. The parameter estimation and path control are not compatible when disturbances are present. The adaptive path control must, therefore, be performed by carrying out these two operations alternately.