Vision based robot manipulator error compensation

Currently one of the most important obstacles to a wider utilization of robots in industry is their accuracy. This is a critical issue for applications that allow no margin of error such as medical applications or precision machining. A possibility is to solve the problem at its root by manufacturing better robot, but that would require substantial economical and time investments. A viable alternative is to improve the robot accuracy through a calibration process, either by creating a model of the robot that closely represents the real robot or by studying the error model of the robot to compensate it directly. The aim of this study is to develop a correction algorithm to improve the robot accuracy by remapping the robot workspace through the use of computer vision and machine leaning techniques. Different correction techniques were tested against each other to find the one that would bring the robot accuracy closer to its repeatability, together with some important additions to seamlessly include the correction algorithm into an existing pipeline.