Effective visual calibration system for parallel robot using decision tree with cooperative coevolution network approach

The objective of this paper is to present an effective visual calibration system for parallel robot. We propose a new hybrid algorithm to improve the weakness of traditional calibration system. The method is auto-calibrated, non-parametric, and has ability of adaptive learning for different environments. The proposed algorithm considers the accuracy of entire workspace, to ensure that every point in the workspace is well-calibrated. An improved Neural Network system model combines cooperative coevolutionary and decision tree, which is built to transform the nominal position to the correct end effector position. Experimental verification has been conducted that it can successfully reduce the average error 99.98% accuracy.