Robotic Wire Contact Pose Estimation via Edge Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for automating the assembly of wire bundles in aircraft manufacturing lack precision in inserting wire contacts into connectors, particularly when dealing with varying wire shapes, colors, and reflectivity, and require complex machine learning models for accurate pose determination.

Innovation Solution

A robotic system equipped with a camera system and controller that generates a sequence of images of the wire contact as it moves, detects edges, removes background edges, and identifies the wire contact to determine its pose without prior modeling, allowing for precise insertion into connectors.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If machine learning models are used to determine wire contact pose, then accuracy in inserting wire contacts into connectors is improved, but device complexity increases

Engineering Contradiction:
Improvepose determination accuracyVSAvoidmodel complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces complex machine learning models with a simplified geometric modeling approach. Instead of using trained neural networks to determine wire contact pose, the system uses analytical geometry methods with pre-defined wire contact models, achieving high precision while significantly reducing computational complexity and processing time.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent creates simplified geometric copies or representations of wire contacts with known parameters. By using pre-defined geometric models that replicate the essential features of wire contacts, the system can determine pose through geometric calculations rather than complex learning models, maintaining accuracy while reducing complexity.

Inventive Principle:
Principle #26Copying

2Adaptability or versatility

If machine learning models are trained for various connector types and wire insulation settings, then adaptability is improved, but loss of time in training and deployment increases

Engineering Contradiction:
Improvecompatibility with various connectors and settingsVSAvoidtraining time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent implements a universal geometric modeling framework that can handle various connector types and wire insulation settings without requiring separate training for each case. The analytical approach uses general geometric principles that apply across different scenarios, providing adaptability while eliminating training time requirements.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system achieves adaptability by changing geometric parameters in the pre-defined models rather than retraining machine learning models. By adjusting wire contact dimensions, connector geometries, and material properties in the analytical models, the system adapts to different connector types and wire settings instantly without time loss.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4411640A1Robotic wire contact manipulation and pose estimation system
Publication Date: 2024.08.07 THE BOEING CO
  • EP4411640A1 patent drawingFigure 1
  • EP4411640A1 patent drawingFigure 2
  • EP4411640A1 patent drawingFigure 3

AI summary

A method, apparatus, system, and computer program product for positions a wire contact (138, 224, 342, 400). A sequence of images (232, 300) of a wire contact (138, 224, 342, 400) is generated while the wire contact (138, 224, 342, 400) moves from a first position (236) to a second position (238). The sequence of images (232, 300) is generated by the camera system (209) connected to the end effector (136, 208, 400) and the wire contact (138, 224, 342, 400) is held by the end effector (136, 208, 400). Edges (240) are detected in the sequence of images (232, 300) to form edge images (142, 241, 320). Background edges (242) are removed from the edges (240) in the edge images (142, 241, 320) leaving contact edges (244) in the edges (240) for the wire contact (138, 224, 342, 400) to form a contact edge image (246, 330). The wire contact (138, 224, 342, 400) is identified using the contact edges (244) in the contact edge image (246, 330). A pose (150, 226) of the wire contact (138, 224, 342, 400) is determined from the wire contact (138, 224, 342, 400) identified in the contact edge image (246, 330).