Support Vector Machine Control For Autonomous Underwater Vehicle

Autonomous Underwater Vehicles (AUVs) are effective tools in exploring and exploiting the ocean,. In order to accomplish variety of assigned complex tasks, AUVs usually have to perform appropriate position and posture changes according to requirements. The motion controller design is the premise and guarantee for the scheduled mission. However, AUVs are usually complicated nonlinear systems. In addition, during the movement, AUVs are possibly influenced by some uncertain external factors such as wind, wave, current and obstacles etc. It makes challenging to precisely control the AUVs.Support Vector Machine (SVM) is a new approach that is specialized in artificial intelligent learning machines, which aims to deal with small samples. It has good performances of nonlinear function fitting and generalization, which make it to be an effective tool for modeling and control of nonlinear systems. In this paper, SVM is proposed for AUV control, in which efforts are devoted to(1) The mathematical models of AUV motion are established. Two coordinate systems are used to describe the motion of an AUV and then the mathematical models are derived. The forces and moments on an AUV are analyzed and a simplified motion equation is recommended for the course-keeping control.(2) The basic principle of SVM is introduced. By using the Least Squares Support Vector Machine (LSSVM), the dynamic model of an AUV is established, so is its inverse dynamic model. The black-box model between the rudder angle and the course angle is derived. The generalization check shows that SVM has good performances of identification and generalization.(3) The inverse model identified by LSSVM is used as a feedforward controller and connected with the control object. Then, a course-keeping control strategy based on LSSVM inverse system is proposed. In order to increase the precision of the control system, a hybrid inverse control system of LSSVM-PID is constructed by adding a PID controller. The numerical simulation results show that the inverse control system based on LSSVM-PID has good accuracy and performance of suppressing external interferences.(4) An internal model control (IMC) system based on LSSVM is presented. SVM is used to build up the model and inverse model. The forward model and inverse model identified by LSSVM are treated as internal model and internal model controller respectively. Then the influence of the feedback filter on the whole control system is analyzed. The numerical simulation about the hybrid control strategy is conducted to verify the control scheme proposed.