Application Research Of Support Vector Machine In Ship Heading Control

Design of autopilot for ships and ship control have been an important subject in both academic and industry fields. Good control algorithms and reliable marine autopilot ensure the security and economy of ships in the navigation process, which is of important theoretical and practical significance. With the development of computer science and technology, more and more mature and successful application of intelligent algorithms such as neural networks, ant colony algorithm is used in different fields of engineering. This paper will discuss the support vector machine(SVM), which is used in the application of ship autopilot design.First, this paper discusses the ship motion models and the design of ship motion simulation software. Referencing IEC62065, the models of propulsion system, actuator system, surge, sway and yaw are established. Also, when the ship is running in a real physical environment, it will be affected by interference from waves and currents. So the models of waves and currents are also discussed. At last, a simulation system software for ship motion is built, and simulation experiments show that it can be used to verify performance of marine autopilot.Then, the basic principle of support vector machine is discussed. SVM came by the statistical learning theory. Earliest support vector machine is used for data classification by building a hyperplane between different kinds of data. What’s more, because of the kernel function the SVM can also be used for nonlinear problem. In the field of control, the SVM is usually used for regression. SVM can identify the control model of the goal system to help to compute the control signal. In this paper least square support vector machine(LSSVM) is used for offline and online identification of ship model. There is not much difference between online and offline identification by LSSVM in addition that the online LSSVM only use a little data and needs refresh the data on time. In this paper the experiments show that online LSSVM can identify the ship model very well.There are mainly two ways when SVM is used in the control field. One is that LSSVM is used as optimal control signal, which one should know the model of the goal system and reference state. Another one is that LSSVM is used to identify the model of goal system to modify the control law, like generalized predictive control(GPC), PID control. At last, a ship autopilot based on LSSVM-GPC is designed and the main circuit diagram and flow chart are provided. Then experiments show that the autopilot can realize the ship heading control, which have some practical significance.