Deep Learning-Based Robot Calibration for Aerospace Assembly

Machine learning approach for calibrating dual-arm robotic systems in precision aerospace manufacturing.

Stage
Research Prototype
Published
Who made it?
Ali Maghami
Tags
Deep LearningRoboticsAerospaceResearchMachine Learning
Share

Overview

Developed a deep-learning method for estimating and compensating errors in dual-arm cooperative robotic systems used in aerospace assembly applications.

Contributions

  • Trained neural networks to learn calibration residuals from robotic measurement data
  • Resolved data-quality issues through rigorous statistical analysis and filtering
  • Iteratively refined model architecture for accuracy and computational efficiency
  • Tested across multiple dual-arm configurations and assembly tasks

Key Results

  • Achieved error reduction in relative positioning between cooperating robots
  • Provided practical calibration procedures suitable for production environments
  • Documented methodology for application to other robotic systems

Technologies

PyTorch, Deep Neural Networks, Regression Modeling, NumPy, Pandas, Cross-validation