Passivity based iterative learning control for mechanical systems subject to dry friction

This paper considers iterative learning control applied to a mass-damper system subject to dry friction. The dry friction nonlinearity is discontinuous, and therefore poses challenges to the conventional learning control methods. We apply the passivity based analysis in learning control and show that it is applicable to the case with velocity output. In the case of combined position and velocity output, the passivity approach is not directly applicable. We derive a modified update law based on the 2D system perspective and adjust the combination coefficient in every iteration step to ensure passivity. Asymptotic convergence is shown under the condition that the combination coefficient does not asymptotically vanish. Simulation results are included to demonstrate the performance of these algorithms.