The estimation of the mass of payloads, which are grasped by a vacuum cup gripper, may benefit several robotics applications, e.g.: object recognition, force-guided motions, hard failure detection. These applications require accurate, robust and fast estimation of the mass. Existing methods use standard kinetic batch least-squares techniques and exciting trajectories that cannot be implemented in the context of time-optimized industrial operations. Besides, no comparable projects have tackled the mechanical behavior of suction cup gripper.

    This project proposes a method to measure the mass of the payload grasped by a robotic manipulator by estimating its inertial parameters. A mechanical model of the vacuum cup is derived to approximate the inertial parameters with a Recursive Total Least Squares algorithm. A calibration process, based on the dynamic model, is developed to determine the inner properties of the tool and the vacuum cup.

    The approach was evaluated on a database of a pick-and-place robot operating in an e-commerce warehouse. The results show that the method can accurately estimate the mass for typical scenarios, although the performance decreases for higher speed tasks and shorter data recording duration.