Development and comparison of algorithms for beam stabilization in ultrashort pulsed laser equipped on a six-axis robot

Abstract

We demonstrate an innovative beam stabilization concept consisting of complementary metal-oxide semiconductor cameras and piezo actuators for a six-axis articulated ultrashort pulsed laser robot system. The beam stabilization system is fixed on robot axes 4 and 5. Moving robot axis 5 in an angular range between 0° and 90°, the laser beam position coupled to the actual robot position is monitored by two cameras integrated on robot axis 5 and used for laser beam characterization and model generation. A mathematical description and models generated with machine learning methods, namely, linear regression and neural network, are compared for predicting the beam position drift as a function of robot axis motion, where the neural network model shows a low prediction tolerance of about 7 pixels. In addition, a stand-alone time-triggered beam correction algorithm is developed and implemented on the system, which shows an excellent correction performance for large beam position drifts (below 500 pixels).