In the AI-SerVo project, the potential of machine learning methods to improve feedforward control in cascaded control structures of electric servo drives is being investigated.

The aim is to compensate for individual, drive-specific disturbances—such as friction, torque ripple, or manufacturing tolerances—through AI-based feedforward control, thereby reducing positioning time and increasing the efficiency of the drive systems. A particular focus is on designing the developed ML models in such a way that they can be executed in real time on highly resource-constrained embedded systems at control frequencies above 20 kHz.