Modification and Optimization of the Trajectory of an Industrial Robot to Scan a 3D-Surface for Quality Inspection

An industrial robot equipped with vision sensors can be used to address the unavoidable uncertainties occurring in the process of optical quality inspection. One of these uncertainties is the randomness in the positioning of the workpiece under inspection. It might not always be possible to accurately position the workpiece. In such cases, an industrial robot being solely able to follow a programmed trajectory will not be helpful. In this paper a solution is proposed, with which a robot can modify and optimize its scan trajectory with respect to the actual position of the workpiece, while the only sensor used was the camera mounted on the end effector of the robot, which is also used for the inspection. The metric used by the optimization algorithm acquired by the camera was the focus measure. With this assumption, the case where using additional sensors is not possible has been addressed. The proof-of-concept of this solution have been shown by simulations and also in a real-world experimental environment. In the simulations, the Nelder-Mead algorithm was used to simultaneously optimize the translational and rotational variations in the robotic trajectory. In the experiments, a simple line search method was used to address only the translational variation in the workpiece’s position.