Research Based On Neural Network For Robotic Manipulator Control

Robot is not only a high complicated time-varied, strong-coupled, nonlinear systems, but also subjected to various kinds of uncertainties, such as measurement errors, fractions, varying load, random disturbances, unmodeled dynamics, and so on. It is difficult to get the system of manipulators with entire dynamic model. The intelligent control schemes have been introduced in the uncertain robotic trajectory control system to eliminate the influence of the uncertainties and get the good tracking performance.Solution to the inverse kinematics of robot manipulators is one of important process to control it. Based on the ability of neural networks approximating a nonlinear function, the feed-forward neural networks are applied to solve the inverse kinematic problem of manipulators. BP network is the typical feed-forward neural network. But conventional BP algorithm has defects on slow convergent speed and easy convergence to a local minimum point of error function. After analyzing the disadvantage of BP learning algorithm, a hybrid learning algorithm is introduced. First the genetic algorithm is used to optimize the parameters of the neural network off-line. Then the BP algorithm is used to adjust the parameters on-line. Simulation experiments show that the network can solve the inverse kinematic problem of manipulators, and it reaches to good precision of solution. The computation speed can meet the requirements for the manipulators'real time control system.The basic principle of fuzzy control is introduced and analyzed in detail, and a fuzzy control method is proposed. It does not depend on the accurate mathematics model, can overcome the effects of the nonlinear, coupling and uncertain factors. Then the abilities of fuzzy inductive inference and self-learning are realized by constructing fuzzy neural network (FNN) that combines fuzzy control and neural network together. FNN uses neural network to realize fuzzy inference. This make it has ability of fuzzy inductive inference. Since the construction of FNN has clear physical meanings, the design and initialization of FNN are also very easy, and a fuzzy Gaussian basis neural network is given in this paper. FNN control system is designed in this paper. The neural network is used to compensate the uncertainty of the system. Based on Lyapunov theory, the structure of controller and the learning algorithm of the neural network are designed. So the stability of the system is guaranteed, and the dynamic performance of the system is improved.