Robotic Picking Training for Optimal Package Grasp Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated storage and retrieval systems face inefficiencies in picking packages with varying sizes, weights, and orientations due to suboptimal picking location selection, leading to reduced throughput.
Innovation Solution
A robotic package retrieval system creates and utilizes a database of article models, including 3D and 2D images, and center of gravity data to determine an optimal picking location, allowing for efficient picking of articles regardless of their weight distribution or orientation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional picking methods are used without article models, then the system is simpler to operate, but picking efficiency decreases and throughput is reduced
Solution Approach 1:
The system creates and stores article models (including 3D models, 2D images, and center of gravity data) in advance before the actual picking operation. This preliminary action allows the robotic system to quickly determine optimal picking locations without real-time complex calculations, thereby improving picking efficiency while managing system complexity through pre-processing
Solution Approach 2:
The system creates digital copies (models) of physical articles that capture essential geometric and physical properties. These models are stored in a database and used to simulate and determine optimal picking strategies without requiring physical trial-and-error, thus improving productivity while keeping the physical system relatively simple
2Measurement precision
If optimal picking locations are determined using article models, then picking accuracy improves, but the time required for picking increases
Solution Approach 1:
The system replaces complex real-time mechanical sensing and trial-and-error physical adjustments with pre-computed digital models and algorithms. The article models containing geometric and center of gravity data allow the system to calculate optimal picking locations through software rather than time-consuming physical experimentation, achieving high precision without excessive time loss
Solution Approach 2:
The system changes the state of article information from physical-only to digital-model-based, storing key parameters (dimensions, center of gravity, surface characteristics) in structured formats. This parameter transformation enables rapid computational determination of optimal picking locations, balancing accuracy requirements with time constraints
3Adaptability or versatility
If the system handles articles with varying weight distributions, then adaptability improves, but determining optimal picking locations becomes more difficult
Solution Approach 1:
The system creates digital models that include center of gravity information for articles with varying weight distributions. By capturing this physical property in the digital model, the system can adapt to different articles without physical measurement during operation, improving versatility while reducing the difficulty of real-time detection
Solution Approach 2:
The system determines and stores center of gravity data as part of the article model creation process before actual picking operations. This preliminary characterization of weight distribution properties allows the system to adapt to various articles without difficulty during runtime, as all necessary information is pre-captured in the digital models
Data Source
AI summary
A method of training a robot to pick a plurality of articles includes determining an identity of an article and determining if the article has a uniform distribution of weight. If the article has a uniform distribution of weight, a three-dimensional (3D) model of the article is created and stored in a database record associated with the identity. If the article does not have a uniform distribution of weight, a 3D model is created, two-dimensional (2D) images of each side of the article are created, and a 3D center of gravity of the article is determined. The 3D model, the 2D images, and the 3D center of gravity of the article are stored in a database record associated with the identity of the article.