Research On Motion Control Of Underwater Vehicle Based On RBF Neural Network Compensator

Because the underwater Vehicle's motion system is characterized by nonlinearity,strong coupling,time-varying and uncertainty,its actual dynamic model cannot be accurately obtained.At the same time,the operating environment of underwater robots is complex and unpredictable,making real-time control of underwater robots very difficult.Aiming at the above two control difficulties,this paper introduces the Radial Basis Function(RBF)neural network control which can arbitrarily approximate the nonlinear function,and uses its self-learning ability to estimate and compensate the model error and external disturbance,and then improve the motion control performance of the autonomous underwater vehicle.The specific research content includes the following aspects:(1)Kinematics and dynamics modeling of underwater Vehicles.Firstly,two reference coordinate systems are established,and the mutual conversion relationship between the motion parameters between the two reference coordinate systems is given.On this basis,the dynamic mathematical model of underwater Vehicle is established by Lagrangian method.The mathematical model is simulated and verified.which is the underwater Vehicle in the following chapter.The location control system design laid the foundation.(2)An underwater compensation motion control method for overall compensation is developed.The method consists of a sliding mode feedback controller.a sliding mode controller and a compensation controller based on the RBF neural network.The sliding mode controller is used for nonlinear configuration,and the unknown function is estimated and compensated by the RBF neural network controller,which improves the system stability and motion performance of the underwater Vehicle.The convergence of the system is verified based on Lyapunov theory.The simulation results show the effectiveness of the proposed method.(3)A RBF neural network block compensation control method based on nominal model is developed.This method introduces the nominal model,and uses the RBF neural network to adaptively estimate and compensate the three model errors and one external disturbance respectively,and introduces the robust sliding module term to further overcome the estimation error of the neural network and ensure the system within a limited time.Arnived in a convergence state.The convergence of the control algorithm is proved by the Lyapunov stability criterion.The simulation results verify that the control algorithm has strong adaptability and better dynamic performance.(4)An adaptive fuzzy neural network control method is constructed.This method is based on online training and adjustment algorithm to compensate modeling errors and deal with external disturbances in real time,and improve the learning and self-adaptive ability of fuzzy neural network control.The asymptotic stability of the control system is proved by the Lyapunov method.The simulation results verify the flexibility,adaptability and motion performance of the adaptive control system.