Research On Nonlinear Control Method Of Autonomous Underwater Vehicle

The underwater vehicle have a great deal of concern as an important tool for the development of marine,which varietied complex operations and safe navigation for the motion control have been a higher requirement.However,as a typical non-linear system control,uncertainties and randomness of the object and system disturbances caused great difficulty for the study of underwater vehicle system,lacking of a unified theoretical guidance.Therefore,the study of nonlinear control for underwater vehicle system problem has important theoretical and practical significance.This paper researched the internal structure of underwater vehicle system based on differential geometry,and discussed the nonlinear control problems of underwater vehicle system.Specific research contents are as follows:This paper depending on differential geometry theory as the guiding ideology,establishing the correspondence between geometry and algebra,construct geometric model of the nonlinear underwater vehicle system.Compared with the mathematical model of the underwater vehicle in coordinates,differential geometric theory express on the system structure,analyze from the viewpoint of geometric objects.With affine connection contact the uncontrolled system into a controlled system,provide accurate underwater vehicle system model in formally,which provide a basis for the analysis of decoupling degrees of freedom underwater vehicle.In the absence of external current disturbances for the tracking control of underwater vehicle system,underwater vehicle error equation is established which is through conversion mapping to the same manifold compared from different manifold vectors under Lie group context.PD+feedforward geometric control method based on RBF neural network is proposed for the underwater vehicle system.By the RBF neural network is self-learning to approximate nonlinear viscous drag,which approximation error caused by the RBF neural network is compensated in the feedforward controller using differential geometry method.This control method can improve the ability of PD controller for nonlinear systems.The simulations show that the controller error is small and high precision.Two methods are proposed in considering the influence of underwater vehicle system hydrodynamic coupling and external current interference.The first method is a sliding mode predictive control based on differential geometry method,which eliminate the chattering of sliding mode control.The organic combination of the differential geometric method and sliding mode predictive control method,better overcomes the impact of hydrodynamic damping coupling effect on the system and the chattering of the system,improves the robustness and stability of the system.The second method is L2 disturbance attenuation robust control based on RBF neural network.The approximation error by RBF neural network in the learning process is produced to replace external interference,on this account establish error equation of AUV system,through the L2 controller to suppress and eliminate.The underwater vehicle accurate trajectory tracking ability is achieved under slowly varying current disturbance conditions.Aiming at the time-delay problem of the underwater vehicle system command transmission process,the predictive control is proposed and design predictive and robust combined controller.On the condion of effectively resisting the flow resistance and external interference,method one use differential geometry method to calculate hydrodynamic,then through the performance index of predictive control calculate the optimal control law.Method two is through designing a high order predictive model,to ensure the accuracy of prediction model.According to the prediction error,estanblish filtering error equation to reduce the system dimension,which guarantee the stability of the system.The synchronous tracking of underwater vehicle trajectory is realized.By the simulation,two kinds of method to forecast the effects of system are good,the calculation process of method one is relatively simple,the prediction of method two is more accurate.