AI Model Training for Robotic Aircraft Fastener Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in industrial robotics is the variability in the arrangement of aircraft fasteners, which complicates the design of automated systems for their removal, as each aircraft part has a unique set of fasteners with different orientations, making it difficult to use one set as a template for another, and requires precise path and trajectory planning to avoid damage.
Innovation Solution
A system that uses dynamic path and trajectory planning guided by a user interface (UI) to direct a robot, incorporating point cloud data, feature extraction, and artificial intelligence (AI) for precise identification and classification of fasteners, enabling the generation of multi-dimensional representations and updated paths for robotic operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed template is used for fastener arrangement, then device complexity is reduced, but adaptability deteriorates due to variability in fastener positions and orientations
Solution Approach 1:
The system transitions from static pre-programmed paths to dynamic real-time path planning. The robot uses AI models to detect fastener positions and generates adaptive trajectories on-the-fly, allowing the system to accommodate variable fastener arrangements without requiring complex reprogramming or multiple fixed templates.
Solution Approach 2:
The system creates a digital replica (point cloud model) of the aircraft part and fastener arrangement through 3D scanning. This virtual copy allows the system to analyze and plan paths for any configuration without physical trial-and-error, reducing device complexity while maintaining high adaptability to different fastener patterns.
2Loss of time
If AI models are trained only on synthetic data, then training speed is improved, but measurement precision deteriorates due to domain gap between synthetic and real images
Solution Approach 1:
The system performs preliminary training on synthetic data to establish baseline model performance quickly. This preliminary action provides a good starting point that captures general fastener characteristics, which is then refined using real data to close the domain gap and achieve high precision.
Solution Approach 2:
The system implements a feedback loop where real-world detection results are used to fine-tune the AI model. Detection errors and mismatches between synthetic and real images provide feedback signals that guide additional training iterations, progressively improving measurement precision while leveraging the time efficiency of synthetic data training.
Data Source
AI summary
Aspects of the disclosure are directed towards artificial intelligence-based modeling of target objects, such as aircraft parts. In an example, a system initially trains a machine learning (ML) model based on synthetic images generated based on multi-dimensional representation of target objects. The same system or a different system subsequently further trains the ML model based on actual images generated by cameras positioned by robots relative to target objects. The ML model can be used to process an image generated by a camera positioned by a robot relative to a target object based on a multi-dimensional representation of the target object. The output of the ML model can indicate, for a detected target, position data, a target type, and/or a visual inspection property. This output can then be used to update the multi-dimensional representation, which is then used to perform robotics operations on the target object.