Flexible Manipulator Modeling And Intelligent Control

The motion modeling and control strategy for flexible manipulator are both critical research focus in robotic fields.A kind of the motion modeling for planar two link flexible manipulators has been proposed in this dissertation: The nonlinear dynamic motion equations of flexible manipulator were deduced based on Lagrange theory. The coupling effect between elastic deformation and the rigid motion was concerned in the dynamic equations. At the same time, the flexible manipulators were assumed as Euler-Bernoulli beams, the elastic deformation of the manipulator was represented by modality truncation equations, both elastic and gravitational potential energy were considered to calculate the total energy for whole system, therefore the finial dynamic modeling is more accurate and simple than others published. The results of simulation validate the conclusion.Artificial neural network has attracted great interests for its learning ability, adaptive capability and nonlinear mapping properties. Hence it can be used as a powerful tool for system identification, especially for nonlinear system identification. But the reasonable construction and parameters of networks are hard task for application. In this dissertation the optimal design approach of NN by means of hierarchical evolutionary programming, which was based on evolutionary programming, was introduced to design an optimal NN for a virtual robotic control.Because the existence of measuring inaccuracy, load changing and some uncertain disturb in engineering application, it is difficult to accurately obtain an analytical dynamic model for the flexible manipulator. So the conventional control method is not efficient for this nonlinear system. It seems rational to think over intelligent strategy for this task.The solution of backward and forward kinematics of flexible manipulator must be worked out while you develop a control system for the flexible manipulators. A control strategy using Mutual Mapping Neural Network (MMNN) to compute the solution of the backward and forward kinematics was presented in this paper. MMNN was consisted of two NN and one modify function, one NN was used for forward kinematics calculation, and another for Jacobi matrix calculation, the modify function was formed based on Lyapunov function. The backward kinematics was obtained from the iterative computation of the forward kinematics. The RBF and BP neural network were used for forward kinematics modeling respectively. For the purpose of optimizing the neural network configuration and parameters, the hierarchical evolutionary programming was introduced. The simulation results showed that the precision of trajectory tracking and iterative learning speed are improved by using the hierarchical evolutionary algorithm. So the approach proposed in the paper is an effective control method to meet the requirement of flexible manipulator controlling.