This book presents multivariate time series methods for the analysis and
optimal control of feedback systems. Although ships' autopilot systems
are considered through the entire book, the methods set forth in this
book can be applied to many other complicated, large, or noisy feedback
control systems for which it is difficult to derive a model of the
entire system based on theory in that subject area. The basic models
used in this method are the multivariate autoregressive model with
exogenous variables (ARX) model and the radial bases function net-type
coefficients ARX model. The noise contribution analysis can then be
performed through the estimated autoregressive (AR) model and various
types of autopilot systems can be designed through the state-space
representation of the models. The marine autopilot systems addressed in
this book include optimal controllers for course-keeping motion, rolling
reduction controllers with rudder motion, engine governor controllers,
noise adaptive autopilots, route-tracking controllers by direct
steering, and the reference course-setting approach. The methods
presented here are exemplified with real data analysis and experiments
on real ships. This book is highly recommended to readers who are
interested in designing optimal or adaptive controllers not only of
ships but also of any other complicated systems under noisy disturbance
conditions.