Tracking a feature point in profile data using an articulated object model and tuning parameters using the genetic algorithm-applied to an arc-welding robot system

A method is proposed for tracking a feature point in profile data measured with a laser range finder. Such tracking is indispensable for automating tracking tasks. Even when the profile data is complicated due to geometrical changes in the structure, robust tracking is achieved by using an articulated object model that adjusts to the changes. To apply this method to an arc welding robot system, two parameter tuning techniques that use the genetic algorithm have been developed. One tunes the parameters automatically by using variables in the system; the other uses teaching data given by cooperation between the operator and the computer. Experiments show that the performance of a robot system was improved by using these parameter tunings.