A mechanical arm grabbing method in a cluttered scene based on a hybrid gripper
By combining a hybrid gripper structure with a dual-model visual approach, the problem of stable gripping of multi-material and multi-shaped objects by a robotic arm in cluttered scenes was solved, achieving efficient and collision-free object handling.
CN122323192APending Publication Date: 2026-07-03CHONGQING UNIV +1
View PDF 0 Cites 0 Cited by
Patent Information
- Application Number
- CN202610676401.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-17
- Publication Date
- 2026-07-03
AI Technical Summary
Technical Problem
Existing robotic arm grasping technology struggles to effectively grasp complex objects of various materials and shapes in cluttered environments, and is prone to grasping failures or damage to the robotic arm due to visual obstruction and collisions.
Method used
Employing a hybrid gripper structure that combines parallel grippers, an air pump suction cup, and an electromagnet, the system calculates object pose and metal segmentation mask information using a dual-model visual approach, and combines depth map visibility and geometric collision filtering to achieve stable gripping.
Benefits of technology
It significantly improves the success rate and physical safety of object grasping in cluttered environments, avoiding grasping failures and damage to the robotic arm.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN122323192A_ABST
Abstract
The application discloses a mechanical arm grabbing method in a cluttered scene based on a mixed gripper and belongs to the technical field of robot grabbing. The method comprises the following steps: acquiring a target scene image, predicting the pose and circumscribed cube of a non-metallic object by using a pose estimation model FFB6D, and predicting a segmentation mask of an object containing a metal part by using an instance segmentation model YOLO11n-seg. A candidate grabbing pose of a parallel gripper and an air pump suction disc is generated according to the upward surface distribution of the circumscribed cube, and a candidate grabbing pose of an electromagnet is generated according to the three-dimensional position of the segmentation mask. Finally, the grabbing pose is eliminated based on the visibility of a rendered image and collision detection, and the optimal pose and the corresponding end tool are selected to perform grabbing. The application overcomes the problems of low object visibility and poor applicability of a single gripper in a cluttered and stacked scene, and significantly improves the grabbing success rate in the state of multiple object scattering and stacking.
Need to check novelty before this filing date? Find Prior Art