Not Sure 2-dof Redundant Parallel Robot Control Method Research

Parallel robots have become the hotspot of the robot research field in recent years for its advantages such as high inflexibility, high accuracy, and high speed. However, the structure characteristic that parallel robot has multiple kinematic chains makes its kinematic analysis and dynamic modeling complex. And in the actual project application, parallel robots are usually affected by the uncertainties such as torque friction, modelling error. In this dissertation, based on a planar2-DOF parallel robot, researches are implemented from several aspects, including kinematic analysis, dynamic modeling and uncertainties analysis, controller design and so on. The main work is concluded as follows:(1) The kinematic position relation and velocity relation of parallel robot based on workspace and torque space are studied, and the workspace scope of parallel robot executer is acquired via theory calculate and simulation confirmation. Moreover, three kinds of motion planning of parallel robot executer are completed base on planar coordinate system built in the workspace of parallel robot. Then, combined with kinematic backwards-solution and motion planning, the motion and attitude of parallel robot are analyzed via MATLAB simulation.(2) The dynamic model of the planar2-DOF parallel robot is built by using Lagrange method, and the friction part is added to constitute the real dynamic model. The uncertainties consist of modelling error and friction compensation error are analyzed, by performing a path-tracking simulation experiment differently using PD controller and computed torque controller. The control effect of the kinematic controller and dynamic controller is compared, and it is stated that dynamic controller relies on the dynamic model and its control accuracy is limited by modelling error and friction compensation.(3) Considering that dynamic controller relies on the dynamic model which is difficult to obtain in actual project application, Radial Basis Function (RBF) Neural Network controller is designed. Using neural network to approximate to dynamic model entirely, meanwhile, adding a robust term into control law to compensate neural network approximation error. This control method has a good robustness, which doesn’t need the dynamic model of parallel robot. Via simulation, the method is confirmed available.(4) Considering the uncertainties of planar2-DOF parallel robot, sliding mode control method is studied. Firstly, based on the nominal dynamic model of parallel robot, sliding mode robust controller is designed with using sliding mode switchover term to compensate the uncertain term in system. On basis of this, sign function in control law is replaced with a kind of function to weaken chattering. By comparison simulation, improved controller greatly weakens the chattering. In order to obtain better control accuracy, sliding mode control method based on approaching law is studied and corresponding controller is designed. The method makes use of RBF neural network to approach the uncertain term in system, which decreases the switchover gain so that the chattering is weakened at a certain extent. Finally, error analysis of three kinds of control method in this dissertation is completed to compare the control accuracy.