Self-learning in the inverse kinematics of robotic arm

摘要

This work makes a study in the usage of a learning machine and computer vision for the control of the movements of an robotic arm. The technique chosen for the learning process was the Map of Kohonen, because it is a classic technique with a self-learning capacity. The solution investigated was part of a huge problem: the learning process of tasks by articulated robots, for example the pick-and-place task. The results that were found indicated that the proposed solution was able to learn in two scenarios: data of movements of the end effector in line and data of end effector realizing movements in a plane. In a future work, it should address more complex scenarios and other stages of the task learning process by Robotic manipulators with computer vision.