Research On The Control Of Manipulator Based On Neural Networks

Robot is the products of the contemporary science and technology, and it is the representative of high technology. Since the 1960s, along with the development of micro computer technology, the robotic scientific technology has developed rapidly. After more than 30 years, robot research has always focused on the study of manipulator. In this thesis, the object of reseach is the manipulator with six DOF. And this paper makes systemic research on the forward kinematics equation and inverse kinematics equation of the manipulator with six DOF. The main research contents are listed as followed:Firstlly, the kinematic model of the manipulator with six DOF is analysed and established. According to motion parameters of the joints of a manipulator, the forward kinematics equation can be solved. In addition, according to the given position and attitude parameters, the inverse kinematics equations can be solved too. The inverse kinematics of six DOF manipulator is one of the main difficult problem which researchers concern about in the study of robotics. On the one hand, the input-output equation derived from inverse kinematics algorithm has high value on theoretic research, many mechanism kinematics can be studied based on it. On the other hand, the inverse kinematics is a central problem for motion planning and path controlling.Secondly, based on the ability of neural network's function of approximating a nonlinear, BP neural network are applied to solve the inverse kinematic problem of manipulator in this thesis. BP neural network is the typical feed-forward neural network. But the conventional BP algorithm has the defects of slow convergent speed and easy convergence to a local minimum point of error function. In this thesis, the traditional BP algorithm is improved by adding items of the momentum.Finally, the BP neural network is set up and trained. The inverse kinematics model of six DOF manipulator is developed using BP neural network. And the inverse kinematics model of six DOF manipulator is simulated with Matlab. The simulation shows that BP neural network has a higher accuracy. The capacity of network generalization is better.