An End-to-End Calibration Method for Welding Robot Laser Vision Systems With Deep Reinforcement Learning

Yanbiao Zou,Rui Lan

DOI: https://doi.org/10.1109/tim.2019.2942533

IF: 5.6

2020-07-01

IEEE Transactions on Instrumentation and Measurement

Abstract:Structured light calibration and robot hand-eye calibration are two of the most crucial parts in welding robot laser vision systems. The main aspect of this calibration is accuracy. To reduce the impact of errors in the calibration process and improve the accuracy, an end-to-end calibration method for laser vision systems based on deep reinforcement learning is proposed. The proposed method involves two dual learning tasks: a pixel-to-point module and a point-to-wrist transformation. The pixel-to-point module predicts the locations of the feature points in the welding image, while the point-to-wrist transformation establishes an accurate conversion relationship from the pixel frame to the wrist frame of the robot. Point-to-wrist transformation consists of two major components: "actor" and "critic." The "actor" model aims to infer the coordinates of the feature points in the wrist frame of the robot. In offline training, the "critic" model is introduced to guide the learning process of the "actor" model by maintaining the geometric consistency between the image coordinate and the robot coordinate. According to experimental results from the welding robot with continuous motion and changing poses, the proposed method significantly reduces the impact of errors in the calibration process and establishes a more accurate conversion relationship.

engineering, electrical & electronic,instruments & instrumentation