Partial and Whole Model Segmentation for Robotic Interference Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing random bin-picking techniques face challenges in accurately determining interference between objects and pallets during robotic gripping, particularly when using partial models for position and orientation calculations, as they may not account for the whole object shape effectively.
Innovation Solution
An information processing apparatus that acquires distance information and uses both partial and whole models to determine interference, where the partial model is used for recognition and the whole model for interference determination, ensuring accurate detection even if the object's position and orientation are derived using a partial shape.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a partial model is used to derive position and orientation of the measurement target object, then the recognition process can be simplified and processing speed improved, but the interference determination accuracy deteriorates because the partial model does not represent the whole object shape
Solution Approach 1:
The patent divides the model usage into two segments: a partial model (first model) is used for position and orientation derivation to simplify processing, while a whole model (second model) is used for interference determination to ensure accuracy. This segmentation allows each process to use the appropriate level of model detail.
Solution Approach 2:
Different quality levels of models are applied to different processing stages: a simplified partial model is used locally for position/orientation calculation where speed is critical, while a complete whole model is used locally for interference determination where accuracy is critical. Each stage receives the appropriate model quality for its specific purpose.
2Measurement precision
If a whole model is used for both position/orientation derivation and interference determination, then interference determination accuracy is maintained, but processing time increases due to handling larger model data
Solution Approach 1:
The patent segments the model data usage by creating two separate models: a partial model containing only the features necessary for position and orientation derivation, and a whole model containing complete object geometry for interference determination. This avoids using the entire whole model data in the time-sensitive position derivation step.
Solution Approach 2:
The patent applies partial action by using only the necessary portion of the object model (partial model) for position and orientation derivation, rather than processing the complete whole model. This reduces computational load in the time-critical recognition phase while still enabling accurate interference determination later using the whole model.
3Device complexity
If only distance information is used without three-dimensional CAD data, then system complexity is reduced, but the ability to determine interference between objects and pallets is lost
Solution Approach 1:
The patent performs preliminary action by pre-acquiring three-dimensional CAD data of the measurement target object and storing it as a whole model before the actual picking operation. This preparation enables subsequent interference determination without adding complexity to the real-time processing system, as the CAD data is prepared in advance.
Solution Approach 2:
The patent creates a digital copy (three-dimensional CAD model) of the measurement target object that can be used for interference determination. This virtual model allows the system to simulate and check for interference without physically manipulating the actual object, maintaining reliability while keeping the physical system simple.
Data Source
AI summary
An information processing apparatus includes a distance information acquisition unit configured to acquire distance information with respect to a measurement target object to determine interference between the object and a container for storing the object when the object is to be gripped, a derivation unit configured to derive a position and an orientation of the measurement target object based on the distance information and a partial model representing a partial shape of the measurement target object, and an interference determination unit configured, based on the position and the orientation derived by the derivation unit and a whole model representing a whole shape of the measurement target object, to determine whether interference has occurred between the measurement target object and the container for storing the measurement target object.