Observer-supported increase of the path accuracy of industrial robots under consideration of joint elasticities
Despite stiff construction, articulated arm robots are excited to vibrations in highly dynamic applications primarily due to joint elasticities and looseness. This leads to greatly reduced path accuracy. For control compensation, these influences must be measured and quantitatively analysed on the output side. The internal robot sensor system (angle encoder on the drive side) of most industrial manipulators cannot exactly reproduce the real path. High-performance 3D camera measuring systems that allow high-frequency scanning of the TCP path (tool centre point) are a possibility for precise metrological recording of highly dynamic trajectories. A cascaded methodology is used to improve the output-side path accuracy of serial robots for highly dynamic robot paths. This consists of three sub-steps: a kinematic and dynamic modelling and calibration, the compensation of the non-linearly coupled dynamics by means of model-based torque pre-control and the reduction of remaining path errors by iterative learning methods based on external camera measurements. This results in an overall effective concept for automated calibration and optimisation of the accuracy of highly dynamic trajectories of industrial robots by a kind of automated teach-in.
