Multi-point estimation weldment recognition and estimation of pose with data-driven robotics design

Abstract In robotic welding systems, weldment recognition and pose estimation play crucial roles in achieving precision and efficiency. Weldment recognition involves identifying and classifying different types of weld joints and components with high accuracy, often employing computer vision techniques and machine learning algorithms trained on diverse datasets. Concurrently, pose estimation determines the precise position and orientation of the welding torch relative to the weldment, which is crucial for ensuring proper alignment and execution of welding tasks. Hence, this study proposed a multi-point entropy estimation (MPEE) model for the pose estimation. The proposed MPEE model computes the multi-point in the weldment design with the data-driven model for the estimation of the welding points. The MPEE model estimates the multi-point in the weldment design and the estimation of the features. With the estimated points in the Weldmart, entropy points are tracked and estimated for fault estimation and fault detection. Through the data-driven approach, machine learning model is employed for the recognition and estimation of weldment with the robotics. The proposed MPEE model specifically addresses the challenge of pose estimation in welding tasks. The MPEE model focuses on estimating the position and orientation of multiple points within the weldment design. By leveraging a data-driven approach, which integrates machine learning models, the MPEE model enhances the accuracy and reliability of welding point estimation. The results stated that in a dataset comprising diverse weld joint variations, the system achieves a recognition accuracy of over 95% in real-time applications. Concurrently, employing geometric hashing and iterative closest point algorithms enables precise pose estimation of the welding torch with an average error margin of less than 1 mm.