Abstract
In this paper, the robot grasping for stacked objects is studied based on object detection and grasping order planning. Firstly, a novel stacked object classification network (SOCN) is proposed to realize stacked object recognition. The network takes into account the visible volume of the objects to further adjust its inverse density parameters, which makes the training process faster and smoother. At the same time, SOCN adopts the transformer architecture and has a self-attention mechanism for feature learning. Subsequently, a grasping order planning method is investigated, which depends on the security score and extracts the geometric relations and dependencies between stacked objects, it calculates the security score based on object relation, classification, and size. The proposed method is evaluated by using a depth camera and a UR-10 robot to complete grasping tasks. The results show that our method has high accuracy for stacked object classification, and the grasping order effectively and successfully executes safely.
Highlights
ROBOTS are expected to perceive objects, plan action sequences, and complete tasks independently when they are manipulating in complex and diverse environments
Numerous studies have been conducted on object detection and grasping order planning in industrial robots, and there are some key points in grasping tasks: (1) A stable recognition algorithm is essential
(2) For stacked objects, it is important to extract geometric relations and dependencies between entities in the scene, so that the robot can plan the motion to avoid object falling or collision issues. Focusing on these two points, a robot grasping approach based on stacked object classification network (SOCN) and grasping order planning method for stacked objects is proposed in this paper
Summary
ROBOTS are expected to perceive objects, plan action sequences, and complete tasks independently when they are manipulating in complex and diverse environments. (2) For stacked objects, it is important to extract geometric relations and dependencies between entities in the scene, so that the robot can plan the motion to avoid object falling or collision issues Focusing on these two points, a robot grasping approach based on SOCN and grasping order planning method for stacked objects is proposed in this paper. In [16], the author uses only the depth information and its geometric features to segment each object and estimate the position and orientation in a stacked scene. We propose a novel deep-learning-based SOCN which includes a transformer architecture for stacked objects detection. The proposed network employs the visible volume of objects to further adjust the inverse density parameters and has the self-attention mechanism to improve the accuracy of object detection for stacked objects
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
