Simplified autonomous object grasping in material handling process for human–robot collaboration

The application of Human–Robot Collaboration (HRC) in the manufacturing sector, especially in the material handling process, is aimed at improving productivity through robots actively working alongside humans. In this condition, the robots need to understand how to handle the objects by themselves according to user preferences with an autonomous system. However, there have been challenges in the aspect of teaching robots to autonomously identify object grasp positions only using an RGB camera due to the effect of camera perspective on object visualization for robots. Therefore, this study aimed to propose a simplified method on an RGB camera for autonomous object grasping in the material handling process and implement it for the HRC concept. The method used a prototype robot manipulator with a computer vision system for object detection. During the execution of object grasping, the robot achieved a success rate of 86% for a single object and 76% for multiple objects. In the HRC concept, the robot achieved a success rate of 92% for placing objects one by one and 84% for placing objects continuously. The result also showed fast inference time when the robot in real-time detected the object, which was even just running on the CPU and in the planning process without complexity and requiring additional equipment aside from an RGB camera.