Flexible Object Grasping via Surface Deformation for CNN Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Determining a suitable grasping position for flexible objects in a folded and stacked state is challenging due to the similarity in regular stacked patterns, making it difficult for machine learning to distinguish between suitable and unsuitable grasping positions using image data.
Innovation Solution
The learning apparatus and method utilize deformed image data of flexible objects, where ends combined into one end differ significantly from those not combined, allowing for the use of CNNs to learn and determine suitable grasping positions by generating and analyzing first and second image data with large differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine learning using image data is used for determining a grasping position of flexible objects in a folded and stacked state, then automation is improved, but measurement precision deteriorates because the difference between images of suitable and unsuitable grasping positions is insufficient
Solution Approach 1:
The system performs preliminary deformation of the flexible object by the grasping unit before image acquisition. This preliminary action creates distinct deformed forms at different ends of the stacked objects, which then serve as clear visual cues for the machine learning model to accurately determine grasping positions.
Solution Approach 2:
The system changes the physical state parameter of the flexible object by deforming it through the grasping unit. This parameter change transforms the object from a regular stacked pattern to a deformed state, creating sufficient visual difference between suitable and unsuitable grasping positions for accurate machine learning classification.
2Stability of the object's composition
If regular stacked pattern is maintained for flexible objects, then stability is improved, but ease of operation deteriorates because it is difficult to determine suitable grasping positions
Solution Approach 1:
The grasping unit applies preliminary deformation to the flexible object in a controlled manner before the grasping operation. This preliminary action temporarily disrupts the regular stacked pattern to create distinguishable deformed forms, enabling easy determination of grasping positions while the object remains otherwise stable in its stacked configuration.
Data Source
Figure 1
Figure 2A~2B
Figure 2C
AI summary
A grasping apparatus 1 includes: an image data acquisition unit 16; a robot arm 11; a control unit 12; and a grasping position determination unit 18 configured to determine, using the image data of photographed flexible objects in a folded and stacked state that is acquired by the image data acquisition unit 16, whether or not a part of the flexible objects in that image data is suitable for being grasped. The control unit 12 controls the robot arm 11 so as to deform ends of a top surface of the flexible object at the top of the stacked flexible objects. The grasping position determination unit 18 determines whether or not a part is suitable for being grasped using the image data of the photographed flexible object the ends of the top surface of which have been deformed.