High Accuracy Hand-Eye Calibration Method Using Pose Graph Constraints and Sequential Quadratic Programming in Robotic Inspection

A high-accuracy hand-eye calibration is vital in robotic inspection (RI). The inaccurate robotic kinematic matrix and sensors extrinsic matrix decrease the accuracy of the hand-eye calibration matrix. Previous works focus on the parametrization of the transformation matrices, such as using dual quaternion, rotation axis, and 4-D Procrustes to increase the robustness of errors in the kinematic and extrinsic matrices. To further increase the accuracy of the above methods, we propose a novel hand-eye calibration framework with pose graph constraints to optimize the kinematic and extrinsic matrices simultaneously. Firstly, we introduce intermediate variables and optimize them via pose graph optimization (PGO) methods. Then, a constraint optimization problem with the constraints of optimized intermediate variables and LASSO penalty functions is proposed to optimize kinematic and hand-eye matrices simultaneously. The constrained problem is solved by sequential quadratic programming (SQP). The simulations and experiments verify that our framework can increase the accuracy of the hand-eye matrix compared to many classic hand-eye calibration methods.