Model-Free Control for Constrained Mechanical Systems

This paper presents a model-free motion control technique for constrained mechanical systems. First, we exploit the inherent positive-definiteness of an unconstrained system's dynamics matrix to achieve local input-to-state stability (LISS) with respect to state estimation errors. Then, we impose holonomic constraints on our system and show that we can preserve the (LISS) property if the derivatives of the constraint equations are known. This suggests applications to robotics (particularly for robot manipulators) where the plant's dynamics may change abruptly or where interactions with the environment are difficult to model.