Qualification of a Collaborative Human-robot Welding Cell

This work is focused on evaluating performance of a collaborative robot welding cell developed in our previous research. Such a cell is based on an interactive cooperation between a human supervisor and a welding robot. This approach to organizing a workstation allows to employ robots even in the case of prototypes or small productions. Research on collaborative robots usually focuses on safety issues and on the programming techniques. Present work deals with a complementary problem crucial to industrial applications: the qualification of the welding cell performance in terms of accuracy, repeatability and dependability. In this application, the human worker is responsible for handling of the parts to be assembled and for teaching the robot. The robot is in charge of actual welding. Teaching is executed by demonstration: the teacher shows the welding trajectories with a pointer observed by a motion capture system. The program is generated automatically and executed by the robot. Robots and humans share the same workspace in different times therefore human risk exposure is minimal. Industrial applications of this or similar technology require that the process reliability and capability be assessed. We describe and analyze error accumulation along the entire data flow from the measurement tool, through the reference system transformations, to the actual representation and execution of the robot program.