Abstract
Robots are commonly used for automated welding in many industries such as automotive manufacturing. The complexity and time required for programming presents an obstacle in using robotic automation in welding or other tasks for small to medium enterprises that lack resources for training or expertise in traditional robot programming strategies. It also dictates a high level of repeated parts to offset the cost of weld programming. Collaborative Robots or Cobots are robots designed for more collaborative operations with humans. Cobots permit new methods of task instruction (programming) through direct interaction between the operator and robot. This paper presents a model and model calibration strategy for a collaborative robot methods to aid teaching for and monitoring of welding tasks. The method makes use of a torque estimation model based on robot momentum to create an observer to evaluate external forces. The torque observer is used to characterize the friction that exists within the robot joints. This data is used to define the parameters of a friction model that combines static, Coulomb and viscous friction properties with a sigmoid function to represent transition between motion states. With an updated friction model, the torque observer is then used for collaborative robotic welding, first to provide a mode in which the robot can be taught weld paths through physical lead through, and second a mode to monitor the weld process from expected motion/force characteristics. The method is demonstrated on a commercial robot.