Robot Object Detector Training Using Iterative Partial Mesh Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detector training methods for grasping robots require separate training systems and are typically completed offline, limiting their adaptability and efficiency in handling unknown objects.
Innovation Solution
A method for automatically generating object representations by grasping objects, determining measurements, and iteratively refining these measurements to create a complete object representation, which is then used to train an object detector online.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If separate training systems are used to generate training data and determine object information, then training can be completed offline with structured processes, but the system loses adaptability and efficiency in handling unknown objects during deployment
Solution Approach 1:
The patent merges the training system and deployment system into a unified object detector that can perform both training and inference functions. The detector is trained to directly process measurements from unknown objects and output grasping parameters, eliminating the need for separate training pipelines and enabling adaptive handling of novel objects during deployment.
Solution Approach 2:
The object detector is designed with multi-functionality, serving both as a training tool and a deployment system. It can process various object types, handle different measurement inputs, and adapt to unknown objects while maintaining a single unified system architecture rather than requiring specialized separate systems.
2Productivity
If training is completed offline before grasping, then the training process can be structured and controlled, but the system cannot adapt to new object types without retraining and loses time efficiency during operation
Solution Approach 1:
The object detector implements dynamic adaptability through mechanisms that allow it to adjust to new object types during deployment. The detector can process measurements from unseen objects and generate appropriate grasping parameters without requiring offline retraining, enabling real-time adaptation while maintaining efficient operation.
3Measurement precision
If manual input and prior training data are required for each new object type, then training accuracy can be maintained, but the system loses scalability and requires excessive manual effort
Solution Approach 1:
The object detector implements self-service capability by automatically adapting to new object types during deployment without requiring manual input or prior training data. The system processes measurements from unknown objects and generates grasping parameters autonomously, eliminating the need for manual annotation and enabling scalable operation across diverse object types.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for automatically generating object representations. One of the methods includes grasping, by a robot, an object at a first grasp point and generating a first partial object mesh based on one or more first sensor measurements of the object when held by the robot at the first grasp point. A second grasp point is identified for the object that is located in a region captured by the one or more first sensor measurements. A second partial object mesh is generated based on one or more second sensor measurements of the object when held by the robot at the second grasp point.