Comparison of the efficiency of five observability indices for robot calibration

The purpose of this work is to evaluate the efficiency of five observability indices commonly used in robot calibration. For illustrative purposes, the study is performed in the case of an actual five-bar parallel robot and another three-degrees-of-freedom planar parallel robot, which are modeled with twelve geometric parameters each. A two-degrees-of-freedom planar serial robot is also considered. For each of the observability indices, a simulation study is performed for one hundred robot units (i.e. one hundred different sets of errors imposed on the parameters), four different numbers of calibration configurations, and four different levels of measurement noise. Results show that, in the case of low levels of measurement noise, all five observability indices yield excellent improvement in robot position accuracy. The interest of choosing an appropriate observability index becomes evident for high levels of measurement noises, and the comparison of the robot accuracy obtained by using each observability index becomes interesting. The results show that there is an observability index, which can be considered as the best choice (i.e. it is the best index in the most cases of our simulated robots).