3D Gripper Pose Estimation for Flexible Robot Grasping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for determining the grasping posture of a robot gripper are inefficient, particularly in high-dimensional scenarios, and require costly CAD models, while they often only accommodate grasping from directly above and lack flexibility in learning various grasping methods.
Innovation Solution
A learning device that uses a neural network model to detect the location and posture of a gripper for grasping objects by converting image data from an RGB-D camera into supervised data, allowing the gripper to learn and adapt to different grasping postures and orientations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional methods using neural networks or deep learning are used to estimate gripper location and orientation, then the method can be applied widely, but it is difficult to find high-dimensional grasping posture and predict information that is difficult to annotate in images
Solution Approach 1:
The patent transitions from 2D image annotation to 3D point cloud representation, adding a spatial dimension to the data. By using depth information from RGB-D cameras, the system creates three-dimensional point clouds that preserve spatial relationships, enabling accurate prediction of high-dimensional grasping postures that cannot be captured in traditional 2D images.
Solution Approach 2:
The patent creates simplified 3D geometric models (CAD-like representations) from point cloud data, which serve as computationally efficient copies of complex objects. These models retain essential geometric features needed for grasping prediction while reducing computational complexity, allowing accurate posture prediction without requiring full detailed CAD models.
2Measurement precision
If CAD models are used to recognize three-dimensional objects, then object recognition can be achieved, but it requires high economical and temporal costs because a CAD model is required and the grasping posture has to be determined after the object is recognized
Solution Approach 1:
The patent extracts only the essential geometric features needed for grasping prediction directly from point cloud data, rather than requiring complete CAD models. By taking out only the necessary spatial and geometric information, the system achieves accurate object recognition and grasping posture determination without the overhead of full CAD model processing.
Solution Approach 2:
The patent performs grasping posture prediction simultaneously with object recognition in a single integrated process, rather than determining posture after recognition. The neural network predicts both object properties and grasping postures from the point cloud data in one operation, eliminating the sequential processing time of traditional methods.
3Adaptability or versatility
If conventional methods are used, then grasping from directly above can be learned, but examples of learning grasping methods other than that cannot be found
Solution Approach 1:
The patent creates a universal grasping prediction system that handles multiple grasping directions and styles through a single neural network model. The system can predict various types of grasps (lateral, under, over, pinch) and orientations from the same architecture, making it multi-functional without requiring separate specialized models for each grasping type.
Data Source
AI summary
An estimation device includes a memory and at least one processor. The at least one processor is configured to acquire information regarding a target object. The at least one processor is configured to estimate information regarding a location and a posture of a gripper relating to where the gripper is able to grasp the target object. The estimation is based on an output of a neural model having as an input the information regarding the target object. The estimated information regarding the posture includes information capable of expressing a rotation angle around a plurality of axes.