Identification of the Inverse Dynamics of a Serial Robot for Robot Drive Control

Inhalt:
This paper presents a methodology for the experimental identification of the inverse dynamics of serial robots based on a statistical analysis. The aim is to identify all parameters of the dynamic model in an unbiased and robust way by using one appropriate measurement process. For this purpose a Fourier series is arranged for each joint as desired trajectory. The Fourier series is optimized with a statistical criterion to guarantee optimal excitation of all parameters. With the identified parameters and the analytical inverse dynamics model, feedforward control torques can be calculated for each drive. Measurements of the serial industrial robot KUKAKR6-2 show that the presented trajectory generation and the linear parameter estimation methodology is suitable for robust identification of the inverse dynamics model. The strong impact of temperature dependent friction conditions is minimized by an identification in steady-state with quasi-constant temperature. The integration of feedforward torque-control on the developed test bench shows the expected improvements of control accuracy. An improvement of tracking accuracy compared to the original KUKAKRC4 controller is achieved.