Iterative optimization of time-variant kinematic model for dynamic error compensation of robot vision measurement system

The industrial robot combined with a 3D vision sensor can be flexibly applied to many measurement fields, and repeatability is one of the key performances. When the robot runs for a long period, internal heating and environmental interference are important factors to decrease the repeatability. The traditional solution is to calibrate the robot dynamically and compensate the error based on kinematic model. However, a high redundancy model will increase the measuring time and dynamic error. In this paper, a time-variant model is applied for dynamic error compensation of robot vision measurement system based on an iterative optimization method. Firstly, the initial kinematic model is constructed and calibration spheres are scanned from different poses at regular intervals. Then, the model is extended to full-redundant 42 parameters. After that, a final model with low redundancy is obtained based on a time-variant identification criterion and an iterative optimization method. Finally, the experimental results on the FANUC LR mate 200iD demonstrate that the proposed method can maintain the repeatability within 0.069 mm on an automotive sheet-metal part.