Grasp-Point Labeling for Robot Manipulation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current methods for robots to detect objects and determine grasping points are time-consuming and often result in ineffective, damaging, or inappropriate solutions, failing to consider physical properties of objects, environments, and robot capabilities.

Innovation Solution

A label-based manipulation point determination tool that uses an AI algorithm to identify ideal grasping points by creating a training set of labeled object data through sensor observations of human manipulation tasks, considering object properties, environment, and robot capabilities.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If three-dimensional computer aided design and optimization algorithms are used to compute grasping points, then the robot can determine grasping points based on object geometry and physical properties, but the computation is time-consuming and may produce multiple solutions that are not always appropriate for the task

Engineering Contradiction:

Improvegrasping point accuracyVSAvoidcomputation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system pre-computes and stores grasping points for various object types in a database during the labeling phase. When a robot needs to grasp an object, it queries the database for pre-determined grasping points rather than performing real-time optimization calculations, significantly reducing computation time while maintaining accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system creates labeled datasets that copy and store successful grasping configurations from human demonstrations. These labeled examples serve as templates that can be directly applied to similar objects, avoiding the need to re-compute grasping points from scratch for each new object

Inventive Principle:

Principle #26Copying

2Reliability

If optimization algorithms evaluate robot kinematics and object physical properties to determine grasping points, then the robot can find mathematically valid grasping solutions, but the solutions may be inappropriate for the task (e.g., contaminating the interior of a cup)

Engineering Contradiction:

Improvegrasping solution validityVSAvoidtask appropriateness

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system uses labeled data from human demonstrations as feedback to train the grasping model. Human labels indicate which grasping points are task-appropriate, and this feedback is used to adjust the model's predictions, ensuring that the robot selects grasping points that are not only mechanically valid but also appropriate for the specific task

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system learns from human expertise and uses this learned knowledge to automatically determine task-appropriate grasping points without requiring real-time human intervention. The labeled database serves as a self-service resource that guides the robot's grasping decisions

Inventive Principle:

Principle #25Self-service

3Ease of operation

If the robot uses computed grasping points from optimization algorithms, then it can perform object manipulation, but the process lacks efficiency and may produce undesirable or damaging solutions

Engineering Contradiction:

Improveobject manipulation capabilityVSAvoidmanipulation efficiency

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

Grasping points are pre-computed and stored in a database during the offline labeling phase. During actual manipulation tasks, the robot simply queries this database for pre-determined grasping points, eliminating the need for time-consuming real-time optimization calculations and significantly improving manipulation efficiency

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system replaces the mechanical optimization computation process with a data-driven approach using labeled examples and machine learning models. This substitution transforms the grasping determination from a complex computational mechanics problem into a simpler pattern recognition and database query task

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS20250042022A1Grasp-point identifying and labeling of objects for robot manipulation

Publication Date: 2025.02.06 INTEL CORP

AI summary

Disclosed herein are systems, devices, and methods for labelling images with grasp points and/or task-related information. The system may receive sensor data of an observed grasping of an object by a hand. The system may also determine, based on the sensor data of the observed grasping of the object, a manipulation point in relation to the object. The system may also create a data label for the object, wherein the data label indicates the manipulation point for the object. The system may also control a robot to grasp an item at a grasping point based on the manipulation point in the data label.