Robotic Deep Rolling With Iterative Learning Motion and Force Control

Large industrial robots offer an attractive option for deep rolling in terms of cost and flexibility. These robots are typically designed for fast and precise motion, but may be commanded to perform force control by adjusting the position setpoint based on the measurements from a wrist-mounted force/torque sensor. Contact force during deep rolling may be as high as 2000 N. The force control performance is affected by robot dynamics, robot joint servo controllers, and motion-induced inertial force. In this letter, we compare three deep rolling force control strategies: position-based rolling with open-loop force control, impedance control, and gradient-based iterative learning control (ILC). Open loop force control is easy to implement but does not correct for any force deviation. Impedance control uses force feedback, but does not track well non-constant force profiles. The ILC augments the impedance control by updating the commanded motion and force profiles based on the motion and force error trajectories in the previous iteration. The update is based on the gradient of the motion and force trajectories with respect to the commanded motion and force. We show that this gradient may be generated experimentally without the need of an explicit model. This is possible because the mapping from the commanded joint motion to the actual joint motion is nearly identical for all joints in industrial robots. We have evaluated the approach on the physical testbed using an ABB robot and demonstrated the convergence of the ILC scheme. The final ILC tracking performance of a trapezoidal force profile improves by over 70% in terms of the RMS error compared with the impedance controller.