Underactuated Ship Path Following Control Based On Heuristic Dynamic Programming

The ocean is the treasure house of the earth ’s resources.With the development of science and technology and the pace of human development in unknown fields,the intelligent development of ships has attracted more and more attention.As one of the key technologies of intelligent ships,autonomous navigation technology enables ships to automatically plan navigation paths,avoid obstacles and reach their destinations according to predetermined paths.In order to realize the autonomous navigation of ships,path following technology is indispensable.Due to the characteristics of large inertia,large time delay and strong nonlinearity,it is difficult,complex and time-consuming to establish an accurate mathematical model of ship motion,thus posing a great challenge for the design of ship motion controller.From the perspective of ship path following control.In this thesis,from the perspective of ship path following control,in order to solve the path following problem of a model unknown ship,research is carried out from the following aspects:(1)In order to solve the problem of course control of a non-linear ship with unknown system dynamics,we propose a BPNN-based ship dynamic identification model network,and design a course controller based on a heuristic dynamic programming method,to optimize the course control strategy,we use a dual network structure of actor-critic network,the requirements for the system state parameters in the method are solved by applying the identification of model network results,and realize the purpose of training the approximate optimal course controller of an unknown ship model using ship navigation and steering data.Finally,we use the Mariner’s integral mathematical model to simulate the ship to verify the control effect of the approximate optimal course controller.(2)We use an integral line-of-sight navigation algorithm to solve a non-linear underactuated ship path-following control problem with unknown system dynamics,and add an attenuation factor to the overshoot and oscillation problem that the traditional integration method is easy to cause.We introduce the action-dependent heuristic dynamic programming method and replace the cost function by a mass function to solve the problem that the original method requires model dynamics,and optimize the controller training step from a two-step process that first train the model network offline and then train the actor-critic network online to a fully online training step,which improves the convergence speed and the real-time performance of the algorithm when track errors occur due to the lateral force of the ship under interference.Finally,we use the Mariner ship model for simulation experiments to verify that the controller can effectively track the reference path under disturbance.The research results show that the ship path following controller based on heuristic dynamic programming designed in this thesis has reliable control performance,can stably follow a given straight and curved reference paths under high wind interference.It has certain reference value and practical significance for the improvement of the ship’s autonomous navigation capability.