Robot Picking Grip Position Estimation Without Shape Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing picking systems face challenges in accurately estimating grip positions for picking robots, especially when shape model collation is difficult or impossible, leading to inefficiencies and increased manual adjustments, which affect handling conditions such as accuracy, safety, and efficiency.
Innovation Solution
A picking system that estimates grip positions using a controller with an image input unit, element reliability evaluation, grip position estimation, and recognition parameter estimation units, based on handling information and image data, to ensure accurate and efficient gripping operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If shape model collation is used to estimate grip positions, then grip position estimation accuracy is improved, but system complexity and preparation work increase
Solution Approach 1:
The patent extracts the essential function of shape model collation by using image data to directly identify grip positions without requiring complex shape model preparation. The system extracts key features from images and uses them to determine grip positions, eliminating the need for shape model collation while maintaining accuracy.
Solution Approach 2:
The patent creates a simplified copy of the shape model collation process by using image-based feature extraction as a substitute. Instead of working with complex shape models, the system uses visual copies of the article characteristics directly from captured images to determine grip positions.
2Reliability
If post-processing is added to satisfy handling conditions, then picking operation quality is improved, but manual adjustment time increases
Solution Approach 1:
The patent performs preliminary action by automatically determining grip positions that satisfy handling conditions during the initial image processing stage. The system pre-calculates appropriate grip positions based on article characteristics and handling requirements, eliminating the need for subsequent manual adjustments.
Solution Approach 2:
The patent implements self-service by enabling the system to automatically adjust and optimize grip positions without human intervention. The image processing system autonomously determines suitable grip positions that meet handling conditions, making the system self-sufficient and eliminating manual adjustment time.
3Productivity
If grip positions are preliminarily set with shape model collation, then picking efficiency is improved, but adaptability to different articles decreases
Solution Approach 1:
The patent applies dynamics by making the grip position determination adaptive and flexible rather than fixed. The system dynamically adjusts grip positions based on real-time image analysis of different articles, allowing it to adapt to various article shapes and sizes while maintaining high picking efficiency.
Solution Approach 2:
The patent uses parameter changes by modifying the approach from fixed shape model collation to variable image-based feature extraction. The system changes its grip position parameters based on the specific characteristics of each article as detected from images, enabling both efficiency and adaptability.
Data Source
AI summary
The present invention estimates a gripping position that is more likely to satisfy a handling condition even when it is difficult to perform shape model comparison. A picking system 1 comprises a picking robot 3 that grips and moves an article 7, and a control device 2 that controls the picking robot 3. The control device 2 comprises: an image input unit that receives image data of the article 7 acquired by a sensor 4; an element reliability evaluation unit that evaluates a plurality of element reliabilities indicating an index of grippability for each physical element of the article 7, on the basis of the image data; a gripping position estimation unit that estimates a gripping position of the picking robot 3 with respect to the article 7 on the basis of the plurality of element reliabilities; a handling information input unit that accepts handling information indicating a handling condition concerning movement and gripping of the article 7 to be gripped in the gripping position; and a recognition parameter estimation unit that estimates a recognition parameter that is used for the estimation of the gripping position by the gripping position estimation unit, on the basis of the handling information.