Implementation and evaluation of an obstacle avoidance system in a collaborative robot

Abstract

This thesis aims to study and implement an obstacle avoidance system for a collaborative robot application. This application was part of a previous Master's thesis, "Implementation and evaluation of movement primitives and a graphical user interface for a collaborative robot", developed by Roy Ove Eriksen. Which was developed in ROS to generate controls to follow a demonstrated trajectory with a Universal Robot 3 CB-series. A camera feedback pipeline has been developed to place the objects of the environment inside a virtual environment. With the objects in the virtual environment, a collision detection system can detect if any object is in a collision course with the robot. If a collision with the end-effector is detected, the system produces an alternative trajectory to avoid the collision using the implemented obstacle avoidance pipeline. The obstacle avoidance pipeline extends from the original Dynamic Movement Primitives (DMPs) code and uses Artificial Potential Fields (APFs) to generate the avoiding trajectory while following the demonstrated trajectory. Promising results have been obtained, showing the system has accurate workplace feedback and capabilities to detect collisions and act in consonance to avoid them.