Motion Control ofUnderwater Vehicles BasedonRobust Neural Network

Aimingatlowresponse speedandsensitization to external disturbance inmotion control ofunderwater vehicles by adopting neural network, astable robust learning algorithm was presented basedonvariable structure control theory anderror backpropagation algorithm, andtheglobal stability conditions werediscussed indetail. Finally, simulation experiments were carried outonGeneral Detection Remotely Operated Vehicle. Theresults showthatithasgoodrobustness toexternal noises andchanging oflearning-ratio, whichreduces theabrasion ofthe mechanically-driven systemgreatly. Itkeepslearning ofneural networkfastandstable, whichmeetstherequirement ofreal- timecontrol andhastheoretical andpractical value. IndexTerms- underwater vehicle, neural network control, robust learning algorithm, variable structure control theory, global stability.