Object Pose Selection Using Uncertainty for Robotic Grasping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robotic systems face challenges in determining the pose of objects for grasping in scenarios where complete 3D models are not available, especially with noisy camera data, leading to limitations in grasp determination and increased uncertainty.
Innovation Solution
The use of two neural networks trained on different data sources to generate and compare poses of an object from multiple viewpoints, with a pose comparison processor determining the most accurate set of poses for grasp determination, allowing robots to grasp objects effectively even in cluttered environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complete 3D models are used for pose generation, then grasp determination accuracy is improved, but system complexity and data requirements increase
Solution Approach 1:
The patent uses multiple neural networks as virtual copies that generate pose estimates from 2D image data without requiring complete 3D models. Each neural network acts as an independent estimator that processes image inputs to produce pose predictions, eliminating the need for complex 3D modeling while maintaining accuracy through ensemble comparison of multiple network outputs.
2Reliability
If multiple neural networks are used to reduce pose estimation errors, then grasp success rate is improved, but computational resources and processing time increase
Solution Approach 1:
The system implements feedback by comparing pose estimates from multiple neural networks and using the agreement level among them to determine reliability. The comparison processor analyzes discrepancies between network outputs and uses this feedback to identify the most reliable pose estimate, improving grasp success without requiring all networks to process every input equally, thus optimizing computational resource usage.
Data Source
AI summary
Apparatuses, systems, and techniques generate poses of an object based on data of the object observed from a first viewpoint and a second viewpoint. The poses can be evaluated to determine a portion of the data usable by an estimator to generate a pose of the object.