The article introduces Look at Robot Base Once (LRBO), a novel method for hand-eye calibration in robotic systems. This process is crucial in estimating the transformation matrix between a camera and the robot flange. Unlike traditional methods, LRBO doesn’t rely on external markers or human assistance but uses the robot base instead. It leverages learning-based 3D detection and registration algorithms to estimate the robot base’s location and orientation. Experimental results show a translation deviation of 0.930 mm and a rotation deviation of 0.265 degrees, outperforming other 3D vision-based calibration methods. The method can be completed in 1 second, making it the fastest among its counterparts.
Publication date: 3 Nov 2023
Project Page: github.com/leihui6/LRBO
Paper: https://arxiv.org/pdf/2311.01335
