Research On Identification Method Of Robot Fuzzy Model And Its Application In Robotic Horse

The robot system is a very complex variable, strong coupling and nonlinear MIMO systems. In the modeling process of robot system it needs to do a lot of assumptions and approximations processing, and ignore a number of uncertainties and uncertain external disturbance which include the friction between the robot joints, high-frequency characteristics, signal detection errors, etc. So it is impossible to obtain the precise mathematical model. In recent years, fuzzy modeling of nonlinear system has become a research hotspot. Fuzzy identification based on fuzzy theory approach is being widely used in complex systems modeling. For robot system, fuzzy modeling will lay a solid foundation to robot robust adaptive control.In this dissertation, the robotic system with unknown model is regarded as controlled plant. It is focus on two fuzzy modeling methods and modeling the biomimetic robotic horse, specific tasks as follows:Firstly, the thesis gives a modeling approach based on adaptive neuron fuzzy inference system (ANFIS). In order to give an appropriate original state, the subtractive clustering is used to process the input data. All parameters of ANFIS net are adjusted by hybrid algorithm which is the premise parameters adjusted by gradient descent and the conclusion parameters adjusted by least square procedure.Then a modeling method is proposed based on improved fuzzy clustering. Use subtractive clustering to analyze the input data. And fuzzy C-means clustering is used to identify the premise structure and premise parameter of this model. And deduce the conclusion parameter by recursive least square estimation.At last, use ANFIS and an improved fuzzy clustering to model the 6-DOF biomimetic robotic horse. The input and output data is collected by the experiments to model the robotic horse fuzzy model. The simulation results show the model including outside interference and many other uncertainties is simple and accurate.