Ultrashort pulsed lasers are widely utilized in industrial precision micromachining. A new concept of a dynamic ultrashort pulsed laser system implemented on a 6-axis industry robot is constructed in this research to allow for a flexible and largescale 3D manufacturing. However, according to the unique beam propagation characteristics of ultrashort pulsed lasers, a feedback laser beam stabilization system is highly demandable to ensure the laser beam stability after the movement of the robot arms.

Here, we demonstrate such a beam stabilization system, consisting of cameras and motorized mirrors. Particular attention of this contribution is drawn to the development of algorithms for laser beam status prediction based on the robot movement and beam position correction for improving the speed and precision of the laser beam stabilization process. We use conventional methods to generate the relation between the laser beam system and dynamic axes, represented by mathematic equations and implemented in the beam position prediction algorithm. In addition, machine learning as black box algorithm simplifies the data parsing of high complex and instable systems, which is therefore applied as the alternative model tuning tool in this research. Specifically, we demonstrate laser beam position prediction models with linear regression and neural networks after robot movement, which are compared with models from conventional algorithms. Finally, a live correction algorithm is developed and implemented in the system which allows for high correction accuracy in the micrometer range on short time scales.