Ensuring solvability and analyzing results of the nonlinear robot calibration problem

Presented in this paper are methods of applying known mathematical techniques to greatly improve the identifiability and reliability of robot model parameters, and then better interpret the significance of the resulting identified parameter values. The results of these techniques can be very valuable to the robot manufacturer in analyzing the quality of robot production. They are also valuable to the robot user for not only improving robot accuracy, but also for determining robot maintenance and repair needs. Tools of model-based scaling, singular-value decomposition, variance analysis, pose accuracy improvement verification, and parameter reliability check are presented and applied to eliminate rank deficiencies and to improve numerical condition for finding an optimal model for robot calibration.