Simulation Study On Neural Network Adaptive Control Of Serial Robot

In the recent decades, robot control problems have been paid much attention in the research of robot. Robot is a highly complicated time-varied, strong-coupled, highly nonlinear system, and also subjected to various kinds of uncertainties, such as frictions, varying load, random disturbances, etc. So the robot control system needs high adaptive ability. According to the control problem of serial robot, this thesis uses the method of neural network adaptive control to have the robot moving on the prospective trajectory, in order to meet the performance requirements of the high speed, high precision robot.Firstly, solve the kinematics forward and inverse solution of a 4 degree of freedom robot, use the Matlab/Toolbox to make the three-dimensional kinematics simulation of this robot, to drive the slider in the slider-controlling picture, to drive the robot moving, like to control the robot itself, so the movement of the robot can be observed directly. Then, according to the comparison of the Newton-Euler equation, the Lagrange equation and the Kane equation by their difference on robot dynamic modeling, establish the dynamic model of serial robot based on the Kane equation.PD control theory is the basic of robot control, this thesis use the 2-joints robot system to make simulation experiment on the PD control of robot. Combine the neural network and adaptive control to apply to the control of the serial robot, and make the controller design. Use the 2-joints robot system to make simulation experiment by Matlab/Simulink, then compare the result of the simulation and the PD+feedforward control method, to verify that the neural network adaptive control method can perform well in the robot control aspect.