Abstract
The idea of using error models to predict the sources of the error in a machine is discussed. It is shown that the somewhat ad-hoc modeling methods used for robot calibration, although satisfactory for improving accuracy, are not useful for detecting the true causes of error. It is shown that even if only independent parameters are used in a model, they still may not represent the real error sources. If dependent parameters are used in the model the problem can be aggravated and these parameters may not even improve accuracy. To alleviate this problem, the error function is being proposed to represent errors. By proper attachment of the error function in a parametrically general calibration model, a complete model for machine error identification can be obtained.

This publication has 4 references indexed in Scilit: