Research On Path Following Control Technology For Unmanned Surface Vehicle Based On Reinforcement Learning

As an important part of the marine unmanned system,unmanned surface vehicle(USV)is an important vehicle in the marine environment,and plays an important role in marine military,civil and scientific research fields.As the key technology of USV motion control,path following control is an important basis for ensuring the quality of autonomous navigation,and has always been a research hotspot.With the rapid development of artificial intelligence technologies such as machine learning in recent years,more and more researchers have applied intelligent control technology to the research of path following control of USV.This article will design and develop a set of remote monitoring system and navigation control system on the basis of the existing hardware system of USV.Aiming at the autonomous navigation function of navigation control system,it will focus on the application of deep reinforcement learning method in heading control and path follwing control,and design the heading controller and path follwing controller based on the deep reinforcement learning method respectively,the specific research content is as follows.By introducing the reference coordinate system of the motion of the unmanned boat,the kinematic model and the dynamic model of the USV are established under the corresponding coordinate system.At the same time,the model is simplified into a three degree of freedom model of the horizontal plane which only considers the yaw,surge and sway.On this basis,the control response model of the unmanned boat is established,and the line of sight guidance method used in the follow-up study is determined Specific forms.Then the remote monitoring system and navigation control system of the USV are designed and developed.The data communication protocol between the systems is designed in detail,and the basic functions of the systems are verified through the field test.Aiming at the problems of poor control accuracy and large overshoot of the fixed parameter PID controller,a parameter regulator is designed by using the deep reinforcement learning method,which can make the controller adjust the control parameters online according to the control error state in the course of heading control.By analyzing the simulation results under different desired heading conditions,it is verified that the adjustment of control parameters by the parameter adjuster can improve the effect of heading control.On the basis of the research on the heading controller,combined with the line of sight guidance method,aiming at the poor adaptability of the fixed parameter path following PID controller,and the basic failure of the controller when dealing with the large heading error,the path following PID controller based on the deep reinforcement learning method is designed,and the control performance is compared with the fixed parameter PID controller through the simulation experiment.In view of the uncertainty of ship mathematical model and the influence of unknown interference on the model parameters,a model free path follwing controller is designed by using the deep reinforcement learning method.The end-to-end control of path follwing is realized,and the effectiveness of the controller is verified by simulation experiments.