Abstract:Many manipulation tasks require precise following of a path while ensuring a desired contact force. Several approaches have been proposed for this task, spanning from impedance control with trajectory tracking, to direct force control based on nonlinear predictive control. We combine nonlinear model predictive path following with admittance control. The admittance dynamics, which describe the desired dynamic interaction between two objects, are hereby added to the optimal control problem. They are considered as virtual states which update the reference path to achieve the desired force. We study the stability of the proposed approach and validate the applicability considering the control of a lightweight industrial robot.