The Motion Control Of The Ground Cross-scale Moving Basement

In this thesis, basing on the national973project “The frontier mathematicalproblems of cross-scale from the motion reproduction to the space cooperationtargets”, study the model uncertainty of the moving basement,which produced by the“cross-scale”,and adverse effect of the unknown disturbance caused by the strongcoupling. Design the control strategy to the moving basement, in order to realize theideal motion control of the moving basement.Against the motion control problem of the moving basement, the maincontributions of the paper are as follows:Firstly, introduce the concept of the “cross-scale”. Analyze the reason of the“cross-scale” problems,such as the difference of movement environment, thedifference of motion planning strategy, the different way of information acquisition.In this paper, the preliminary examination of the “cross-scale” factor is mainlyproduced by different motion planning strategy.Secondly, analyze the integrity constraints of the moving basement. Thenonholonomic constraints of the moving basement are essentially caused by itsnonlinear, the nonholonomic constraints are multiplicative, it makes the control of themoving basement complicated. In order to solve this problem, do the research to thecontrol of the moving basement.Thirdly, in view of two main problem, the model uncertainties of the movingbasement and unknown disturbance cased by the strong coupling, using the nonlinearapproximation ability, fault tolerance ability and good self-learning ability of theneural network,design a neural network adaptive control strategy. Through the studyof the stability analysis of the Lyapunov function, get the adaptive control law. Usingthe Matlab simulation verify the feasibility of the control strategy. Through the analysis of the simulation results, find the deficiency of the control strategy, needinglong time to achieve ideal control. By analysis, consider it is caused by neuralnetwork ‘s slow learning efficiency.Finally, in view of the unideal effect caused by the neural network slow learning,algorithm add EKF(Extended Kalman Filter) to improve the control strategy. EKF isextended by the Kalman Filter,which can be applied to nonlinear system. EKF cansolve the deficiency of neural network, such as local extremum, slow learningefficiency,requiring the initial condition. I Design an EKF module which uses themodel outputs and neural networkidentification model outputs to be the input signals.Use the EKF module to train the network weights. As a result,the learning efficiencyof neural network change better. The simulation results show that the improvedalgorithm achieves greatly reduces the time of the ideal control,the control effect isbetter.