Research On Behavior Learning And Trajectory Tracking Control And Planning Of Unmanned Sailboats

In recent years,the importance of marine strategy is increasing,and the reconnaissance of the marine environment has become the necessary tasks for coastal countries,thus the development of unmanned surface vehicles is in urgent demand.When performing various tasks such as ocean reconnaissance,hydrographic observation and resource exploration,surface vehicles require the lower power consumption due to their long range and long working time.Sailboats are driven by natural wind and have a huge advantage in energy saving.Therefore,the path planning and motion control of unmanned sailboats have been a hot research topic.However,due to the special structure of sailboats,the dynamics model of sailboats is different from conventional unmanned boats,and some of the model structures and parameters are difficult to obtain directly.Meanwhile,the magnitude and direction of sailboats velocities are difficult to control due to environmental constraints,which brings challenges to the autonomy and unmannedness of sailboats.Aiming at the above problems,this paper adopts the predictive control method based on Gaussian model to control the sailing course,and combines the probability potential field and rapidly exploring random trees method to design the path planning for sailboats.The specific research contents are as follows:In the research of course controller,a controller based on data driven modeling is proposed to solve the problems of complex mathematical model,difficulty in sailboats parameter determination and long adjustment time of traditional controller.By collecting sailing data,the state of the sailboat is predicted using Gaussian process model.Aiming at the problem of prediction error caused by data defects,a model predictive control method with adaptive weight item is proposed,which can automatically adjust the parameter weight according to the course deviation.At the same time,the line of sight guidance method is used to achieve path tracking.The results show that the course controller designed in this paper can effectively control the course when the object parameters are unknown,and has faster response speed and adjustment time compared with other conventional controllers such as PID.In the research of path planning,an improved rapidly exploring random trees(RRT)method based on the probabilistic potential fields is proposed.According to the speed polar curve of the sailboat,the method solves the course constraints and speed restrictions of the sailboat.While having the advantage of RRT randomness search,the method can explore the paths according to the speed characteristics of sailboats,and obtain the effective suboptimal path quickly for sailboats in the non-uniform wind field environment.Through several simulated experiments in real wind fields,the average calculation results show that compared with the conventional A*method and other improved methods of RRT,the RRT path planning method based on probabilistic potential fields has faster calculation speed and shorter planning sailing time.Finally,the corresponding control system is simulated in ROS system.A sailboat model is built in Gazebo and the sensors and actuators are added.Based on the course controller proposed in this paper,a simulation control system is built by collecting sailing data,and the interface of control and data display is designed.The experiments prove that the system can realize the accurate course control of unknown sailboat and display the sailing status and trajectory.And the control system built on ROS also provides the possibility of actual sailing experiments.