Autonomous Pick-and-Place Procedure with an Industrial Robot Using Multiple 3D Sensors for Object Detection and Obstacle Avoidance

Abstract

This thesis proposes a full pipeline autonomous pick-and-place procedure, integrating perception, planning, grasping and control for execution of tasks towards long term industrial automation. Within perception, we demonstrate the detection of a large object (target) including position and orientation (pose) estimation in 3D world. Further on, obstacles in the work area are mapped with proposed filtering prior to motion planning and navigation of an industrial robot to the target’s pose. The target is then picked using a custom built motorized 3D printed end gripper, and placed at a desired location in the robot’s reachable environment. Point cloud based model-free obstacle avoidance is performed throughout the whole process. The complete pipeline is targeted towards typical tasks in various industries including offshore, logistics and warehouse domain with scanning of the scene, picking and placing of a bulky object from one position to another without or with minimal human intervention.