Vision-Based Object Handling in Collaborative Robotics

Abstract

The use of collaborative robots (cobots) is slowly being implemented in human-to-robot workplaces owing to the flexibility, safety, and versatility. Object manipulation Object handle Object handle Vision handled objects, Vision-based object handling as an enabling technology has become a critical inhibition technology for cobots enabling them to perceive, identify, localize and further manipulate objects in dynamic and unstructured environments. Compared to the old-fashioned industrial robotic systems, which use pre-set paths and absolute positioning of the objects, the collaborative robotic systems have to work under uncertainty, different lighting regimes, obstructions and human interference. This paper will give a deep examination of the vision-based object manipulation through collaborative robotics with respect to perception pipelines, object detection and recognition, pose estimation, grasp planning, and real time control integration. A literature review is carried out to examine both classical and deep learning-based vision methods implemented in co-operation manipulation problems. The suggested methodology includes a stepwise vision-based object manipulation system that combines the use of RGB-D sensing, the convolutional neural net-based object identification, and visual servo control in executing a closed-loop object manipulation. The performance measures that are discussed include accuracy of detection, success rate of grasp and latency to perform the grasping action. Results of the experiment with typical collaborative tasks prove the efficiency of vision-based systems in enhancing the adaptability and safety. Lastly, the paper identifies the major issues, among them real-time limits, robustness, and human conscious perception and presents research directions in future towards intelligent, autonomous, and reliable collaborative robotic systems.