Handling Robot Dynamic Performance Control Based On Neural Network Algorithm

Industrial robot as a typical representative in the field of modern manufacturing, integrates numerous advanced manufacturing technologies, such as high accuracy, artificial intelligence, mechanical and electrical control, multi-sensor fusion technology, flexible manufacturing and computer applications, it becomes the indispensable mainstream equipment on industrial automation production line, it is of great significance to improve labor productivity, reduce costs, save energy and increase production and improve the product quality. With the economic system reform and the manufacturing upgrading, the performance requirement of industrial robots also has been improved, so improving the industrial robot ability of high speed with high precision and anti-interference has become an important topic in the field of robot control. In this paper, domestic handling robot named ER20-C10 is a research object, the dynamic simulation model of robot is established to analyze and study the robot dynamics, new solution about the study of the robot dynamics control theory is provided by designing controllers which combine neural network algorithm with other algorithms.Firstly, according to the parameterized model of ER20-C10 handing robot, the rigid robot dynamics simulation model is set up in Matlab/Simmechanics. The space simulation trajectory is also established and used by the D-H parameters method for inverse kinematics solution, the robot model involved in this paper conforms to the principle of Pieper, the inverse kinematics can be used by separation variable method, then the optimum solution is selected and each joint trajectory is optimized through cubic polynomial spline curve. Finally the dynamic simulation model is simulated in Matlab to get and analyze the position, velocity, acceleration and torque curve, the robot dynamic characteristics is validated under different load. Through the establishment of simulation model validation robot trajectory and torque characteristics, provide prediction and evaluation for robot research and development experiment, establishment of simulation model to validate and analyze robot trajectory and torque characteristics has an excellent promotional value, which provides prediction and evaluation for robot research and development experiment.Controller as the core of robots is one of the important factors to robot dynamics performance, the control system with high intelligent algorithm is the developing direction of the robot in the future. In this paper, the controller is designed to optimize the robot dynamic performance, which is based on the neural network algorithm combined with other intelligent algorithm. First of all, the neural network algorithm combines with particle swarm optimization (PSO) algorithm, because randomness of the weights of neural network algorithm has had shortcoming of control performance instability and slow convergence speed, the particle swarm optimization (PSO) algorithm is used for real-time on-line optimization of the weights of neural network. Second, the neural network algorithm also combines with nonsingular terminal sliding mode control, the sliding mode control is suitable for robot control because of its good characteristics of good quickness, no overshoot and strong real-time performance. Its chattering problem has been the important factors affecting the development of application, so the neural network is part of uncertain robot sliding dynamic compensation to reduce chattering, and the Lyapunov equation is used to determine the stability of the controller. Finally, above two kinds of controller, the control algorithms are wrote and simulated in Matlab to validated trajectory tracking and anti-interference ability of the optimized robot, the effectiveness of the proposed improved algorithm is proved through the simulation results.Finally, the ER20-C10 robot control system which includes hardware and software is introduced in this paper, and the experiment platform is set up. According to the teaching process, the robot trajectory is programmed and teached, and test for robot is under different load According to the experimental results, different load effects on handling robot position error and torque are analyzed, comparing to the results of robot dynamics simulation in chapter 2,the dynamics simulation model is proved to be correct.