Issues in Trajectory planning and controlling an industrial Robot's Tool position for accurate passage through sharp Corners of a nominal Path

Abstract

Robot replaces manual labor all over the world every day. The demand for robots is extremely high in the ever growing automated society. The purpose of this master thesis is to improve the performance of an industrial robot manipulator, the ABB IRB 1600, at paths including sharp corners. In an industrial perspective there are several scenarios where precise robots at sharp corners are needed. An example is laser cutting of squares, where a deviation of millimeters gives poor result. In this project the precision is far behind what the human eye is capable to see. Therefore, the Nikon K610 camera equipment is used to track the position of the robot. This is a high precision measurement device that could measure positions very accurate. A dynamic model of the robot is developed using estimation tools and the data available from ABB. In addition to the dynamic model, a friction model is developed from experiments. This friction model gives the torque needed to overcome friction given the joint velocity. Optimization is used to find a motion generator that, given different start perturbations, results in the smallest error along the path. In most robotics scenarios, classical asymptotic stabilization, known as reference tracking control, is used in order to follow a predefined motion. In this project orbital stabilization is used, which means convergence of solutions of a closed-loop system to an orbit of a nominal motion. This means that instead of using the classical tracking error in the controller, the error is redefined to be the error to the nominal orbit. This could probably improve the absolute path accuracy of the robot. To perform orbital stabilization, the Inverse Dynamics Controller is redesigned. All the outcome of this report is tested at the NTNU Industrial Robotics Lab. The performance of the robot taking advantage of the developed friction model resulted in an improvement of 30 to 45 percent compared to a standard proportional-derivative controller. Compared to a good designed proportional-integral-derivative controller it is hard to see any improvement from the friction model. The new controller together with the optimized motion generator is tested at the lab, performing a square. The path accuracy performance of the robot measured by the camera equipment shows an improvement of 36 percent compared with the state-of-the-art ABB controller. The maximum path error from the ABB controller is measured to be 2.2 millimeters, while the new controller and motion design gives an error of 1.4 millimeters.